mirror of
https://github.com/exo-explore/exo.git
synced 2026-02-19 15:27:02 -05:00
Compare commits
2 Commits
gh-screens
...
splitting-
| Author | SHA1 | Date | |
|---|---|---|---|
|
5ddc2574ec | ||
|
|
d5e56999de |
@@ -200,7 +200,7 @@ class Module(dict):
|
||||
) -> mx.MX_ARRAY_TREE: # -> dict[Any, Any | dict[Any, Any | dict[Any, Any] | list[Any]] | dict[Any, Any] | list[Any]]:
|
||||
"""Return the submodules that do not contain other modules."""
|
||||
|
||||
def update(self, parameters: dict[str, Any], strict: bool = ...) -> Module:
|
||||
def update(self, parameters: dict, strict: bool = ...) -> Module:
|
||||
"""Replace the parameters of this Module with the provided ones in the
|
||||
dict of dicts and lists.
|
||||
|
||||
|
||||
@@ -7,10 +7,7 @@ from typing import Any, Callable, Dict, List, Optional, Tuple, Union
|
||||
from mlx.core import MX_ARRAY_TREE
|
||||
|
||||
def tree_map(
|
||||
fn: Callable[..., Any],
|
||||
tree: Any,
|
||||
*rest: Any,
|
||||
is_leaf: Callable[..., bool] | None = ...,
|
||||
fn: Callable, tree: Any, *rest: Any, is_leaf: Optional[Callable] = ...
|
||||
) -> Any:
|
||||
"""Applies ``fn`` to the leaves of the Python tree ``tree`` and
|
||||
returns a new collection with the results.
|
||||
@@ -47,11 +44,11 @@ def tree_map(
|
||||
"""
|
||||
|
||||
def tree_map_with_path(
|
||||
fn: Callable[..., Any],
|
||||
fn: Callable,
|
||||
tree: Any,
|
||||
*rest: Any,
|
||||
is_leaf: Callable[..., bool] | None = ...,
|
||||
path: str | None = ...,
|
||||
is_leaf: Optional[Callable] = ...,
|
||||
path: Optional[Any] = ...,
|
||||
) -> Any:
|
||||
"""Applies ``fn`` to the path and leaves of the Python tree ``tree`` and
|
||||
returns a new collection with the results.
|
||||
@@ -83,9 +80,9 @@ def tree_map_with_path(
|
||||
def tree_flatten(
|
||||
tree: Any,
|
||||
prefix: str = ...,
|
||||
is_leaf: Callable[..., bool] | None = ...,
|
||||
destination: list[tuple[str, Any]] | dict[str, Any] | None = ...,
|
||||
) -> list[tuple[str, Any]] | dict[str, Any]:
|
||||
is_leaf: Optional[Callable] = ...,
|
||||
destination: Optional[Union[List[Tuple[str, Any]], Dict[str, Any]]] = ...,
|
||||
) -> Union[List[Tuple[str, Any]], Dict[str, Any]]:
|
||||
"""Flattens a Python tree to a list of key, value tuples.
|
||||
|
||||
The keys are using the dot notation to define trees of arbitrary depth and
|
||||
@@ -121,7 +118,7 @@ def tree_flatten(
|
||||
the Python tree.
|
||||
"""
|
||||
|
||||
def tree_unflatten(tree: list[tuple[str, Any]] | dict[str, Any]) -> Any:
|
||||
def tree_unflatten(tree: Union[List[Tuple[str, Any]], Dict[str, Any]]) -> Any:
|
||||
"""Recreate a Python tree from its flat representation.
|
||||
|
||||
.. code-block:: python
|
||||
|
||||
@@ -1,46 +0,0 @@
|
||||
"""Type stubs for mlx_lm.models.glm_moe_dsa"""
|
||||
|
||||
from dataclasses import dataclass
|
||||
from typing import Any, Dict, Optional
|
||||
|
||||
from .base import BaseModelArgs
|
||||
from .deepseek_v32 import Model as DSV32Model
|
||||
|
||||
@dataclass
|
||||
class ModelArgs(BaseModelArgs):
|
||||
model_type: str
|
||||
vocab_size: int
|
||||
hidden_size: int
|
||||
index_head_dim: int
|
||||
index_n_heads: int
|
||||
index_topk: int
|
||||
intermediate_size: int
|
||||
moe_intermediate_size: int
|
||||
num_hidden_layers: int
|
||||
num_attention_heads: int
|
||||
num_key_value_heads: int
|
||||
n_shared_experts: Optional[int]
|
||||
n_routed_experts: Optional[int]
|
||||
routed_scaling_factor: float
|
||||
kv_lora_rank: int
|
||||
q_lora_rank: int
|
||||
qk_rope_head_dim: int
|
||||
v_head_dim: int
|
||||
qk_nope_head_dim: int
|
||||
topk_method: str
|
||||
scoring_func: str
|
||||
norm_topk_prob: bool
|
||||
n_group: int
|
||||
topk_group: int
|
||||
num_experts_per_tok: int
|
||||
moe_layer_freq: int
|
||||
first_k_dense_replace: int
|
||||
max_position_embeddings: int
|
||||
rms_norm_eps: float
|
||||
rope_parameters: Dict[str, Any]
|
||||
attention_bias: bool
|
||||
rope_scaling: Dict[str, Any] | None
|
||||
rope_theta: float | None
|
||||
|
||||
class Model(DSV32Model):
|
||||
def __init__(self, config: ModelArgs) -> None: ...
|
||||
152
Cargo.lock
generated
152
Cargo.lock
generated
@@ -125,9 +125,9 @@ dependencies = [
|
||||
|
||||
[[package]]
|
||||
name = "anyhow"
|
||||
version = "1.0.100"
|
||||
version = "1.0.101"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "a23eb6b1614318a8071c9b2521f36b424b2c83db5eb3a0fead4a6c0809af6e61"
|
||||
checksum = "5f0e0fee31ef5ed1ba1316088939cea399010ed7731dba877ed44aeb407a75ea"
|
||||
|
||||
[[package]]
|
||||
name = "arc-swap"
|
||||
@@ -165,7 +165,7 @@ checksum = "3109e49b1e4909e9db6515a30c633684d68cdeaa252f215214cb4fa1a5bfee2c"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
"synstructure",
|
||||
]
|
||||
|
||||
@@ -177,7 +177,7 @@ checksum = "7b18050c2cd6fe86c3a76584ef5e0baf286d038cda203eb6223df2cc413565f7"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
@@ -224,7 +224,7 @@ checksum = "9035ad2d096bed7955a320ee7e2230574d28fd3c3a0f186cbea1ff3c7eed5dbb"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
@@ -421,9 +421,9 @@ dependencies = [
|
||||
|
||||
[[package]]
|
||||
name = "chrono"
|
||||
version = "0.4.42"
|
||||
version = "0.4.43"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "145052bdd345b87320e369255277e3fb5152762ad123a901ef5c262dd38fe8d2"
|
||||
checksum = "fac4744fb15ae8337dc853fee7fb3f4e48c0fbaa23d0afe49c447b4fab126118"
|
||||
dependencies = [
|
||||
"iana-time-zone",
|
||||
"js-sys",
|
||||
@@ -644,7 +644,7 @@ checksum = "f46882e17999c6cc590af592290432be3bce0428cb0d5f8b6715e4dc7b383eb3"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
@@ -670,7 +670,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "8d162beedaa69905488a8da94f5ac3edb4dd4788b732fadb7bd120b2625c1976"
|
||||
dependencies = [
|
||||
"data-encoding",
|
||||
"syn 2.0.111",
|
||||
"syn 1.0.109",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
@@ -681,7 +681,7 @@ checksum = "780eb241654bf097afb00fc5f054a09b687dad862e485fdcf8399bb056565370"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
@@ -738,7 +738,7 @@ checksum = "97369cbbc041bc366949bc74d34658d6cda5621039731c6310521892a3a20ae0"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
@@ -820,7 +820,7 @@ dependencies = [
|
||||
"heck",
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
@@ -890,7 +890,7 @@ dependencies = [
|
||||
"delegate",
|
||||
"env_logger",
|
||||
"extend",
|
||||
"futures",
|
||||
"futures-lite",
|
||||
"libp2p",
|
||||
"log",
|
||||
"networking",
|
||||
@@ -911,9 +911,15 @@ checksum = "311a6d2f1f9d60bff73d2c78a0af97ed27f79672f15c238192a5bbb64db56d00"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "fastrand"
|
||||
version = "2.3.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "37909eebbb50d72f9059c3b6d82c0463f2ff062c9e95845c43a6c9c0355411be"
|
||||
|
||||
[[package]]
|
||||
name = "ff"
|
||||
version = "0.13.1"
|
||||
@@ -1022,7 +1028,10 @@ version = "2.6.1"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "f78e10609fe0e0b3f4157ffab1876319b5b0db102a2c60dc4626306dc46b44ad"
|
||||
dependencies = [
|
||||
"fastrand",
|
||||
"futures-core",
|
||||
"futures-io",
|
||||
"parking",
|
||||
"pin-project-lite",
|
||||
]
|
||||
|
||||
@@ -1034,7 +1043,7 @@ checksum = "162ee34ebcb7c64a8abebc059ce0fee27c2262618d7b60ed8faf72fef13c3650"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
@@ -1648,7 +1657,7 @@ dependencies = [
|
||||
"heck",
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
@@ -1702,7 +1711,7 @@ checksum = "980af8b43c3ad5d8d349ace167ec8170839f753a42d233ba19e08afe1850fa69"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
@@ -2291,7 +2300,7 @@ checksum = "dd297cf53f0cb3dee4d2620bb319ae47ef27c702684309f682bdb7e55a18ae9c"
|
||||
dependencies = [
|
||||
"heck",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
@@ -2753,7 +2762,7 @@ dependencies = [
|
||||
"delegate",
|
||||
"either",
|
||||
"extend",
|
||||
"futures",
|
||||
"futures-lite",
|
||||
"futures-timer",
|
||||
"keccak-const",
|
||||
"libp2p",
|
||||
@@ -2829,9 +2838,9 @@ dependencies = [
|
||||
|
||||
[[package]]
|
||||
name = "num-conv"
|
||||
version = "0.1.0"
|
||||
version = "0.2.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "51d515d32fb182ee37cda2ccdcb92950d6a3c2893aa280e540671c2cd0f3b1d9"
|
||||
checksum = "cf97ec579c3c42f953ef76dbf8d55ac91fb219dde70e49aa4a6b7d74e9919050"
|
||||
|
||||
[[package]]
|
||||
name = "num-integer"
|
||||
@@ -3044,7 +3053,7 @@ checksum = "6e918e4ff8c4549eb882f14b3a4bc8c8bc93de829416eacf579f1207a8fbf861"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
@@ -3156,9 +3165,9 @@ dependencies = [
|
||||
|
||||
[[package]]
|
||||
name = "proc-macro2"
|
||||
version = "1.0.103"
|
||||
version = "1.0.106"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "5ee95bc4ef87b8d5ba32e8b7714ccc834865276eab0aed5c9958d00ec45f49e8"
|
||||
checksum = "8fd00f0bb2e90d81d1044c2b32617f68fcb9fa3bb7640c23e9c748e53fb30934"
|
||||
dependencies = [
|
||||
"unicode-ident",
|
||||
]
|
||||
@@ -3183,7 +3192,7 @@ checksum = "440f724eba9f6996b75d63681b0a92b06947f1457076d503a4d2e2c8f56442b8"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
@@ -3227,7 +3236,7 @@ checksum = "bcd7d70ee0ca1661c40407e6f84e4463ef2658c90a9e2fbbd4515b2bcdfcaeca"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
@@ -3269,7 +3278,7 @@ dependencies = [
|
||||
"proc-macro2",
|
||||
"pyo3-macros-backend",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
@@ -3282,14 +3291,14 @@ dependencies = [
|
||||
"proc-macro2",
|
||||
"pyo3-build-config",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "pyo3-stub-gen"
|
||||
version = "0.17.2"
|
||||
version = "0.19.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "398b833826a83ca72c1e26d1b2c7c71f9ca7c3bfc74eacc663901895c362ae33"
|
||||
checksum = "b159f7704044f57d058f528a6f1f22a0a0a327dcb595c5fb38beae658e0338d6"
|
||||
dependencies = [
|
||||
"anyhow",
|
||||
"chrono",
|
||||
@@ -3304,22 +3313,25 @@ dependencies = [
|
||||
"ordered-float",
|
||||
"pyo3",
|
||||
"pyo3-stub-gen-derive",
|
||||
"rustpython-parser",
|
||||
"serde",
|
||||
"serde_json",
|
||||
"time",
|
||||
"toml",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "pyo3-stub-gen-derive"
|
||||
version = "0.17.2"
|
||||
version = "0.19.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "2426ba759d848787239d80f9fdb1f223786976f87fb6c3da8188ca7c17744b28"
|
||||
checksum = "a8c79e7c5b1fcec7c39ab186594658a971c59911eb6fbab5a5932cf2318534be"
|
||||
dependencies = [
|
||||
"heck",
|
||||
"indexmap",
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"rustpython-parser",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
@@ -3402,9 +3414,9 @@ dependencies = [
|
||||
|
||||
[[package]]
|
||||
name = "quote"
|
||||
version = "1.0.42"
|
||||
version = "1.0.44"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "a338cc41d27e6cc6dce6cefc13a0729dfbb81c262b1f519331575dd80ef3067f"
|
||||
checksum = "21b2ebcf727b7760c461f091f9f0f539b77b8e87f2fd88131e7f1b433b3cece4"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
]
|
||||
@@ -3875,7 +3887,7 @@ checksum = "d540f220d3187173da220f885ab66608367b6574e925011a9353e4badda91d79"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
@@ -3893,9 +3905,9 @@ dependencies = [
|
||||
|
||||
[[package]]
|
||||
name = "serde_spanned"
|
||||
version = "1.0.3"
|
||||
version = "1.0.4"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "e24345aa0fe688594e73770a5f6d1b216508b4f93484c0026d521acd30134392"
|
||||
checksum = "f8bbf91e5a4d6315eee45e704372590b30e260ee83af6639d64557f51b067776"
|
||||
dependencies = [
|
||||
"serde_core",
|
||||
]
|
||||
@@ -4083,9 +4095,9 @@ dependencies = [
|
||||
|
||||
[[package]]
|
||||
name = "syn"
|
||||
version = "2.0.111"
|
||||
version = "2.0.116"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "390cc9a294ab71bdb1aa2e99d13be9c753cd2d7bd6560c77118597410c4d2e87"
|
||||
checksum = "3df424c70518695237746f84cede799c9c58fcb37450d7b23716568cc8bc69cb"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
@@ -4100,7 +4112,7 @@ checksum = "728a70f3dbaf5bab7f0c4b1ac8d7ae5ea60a4b5549c8a5914361c99147a709d2"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
@@ -4176,7 +4188,7 @@ checksum = "4fee6c4efc90059e10f81e6d42c60a18f76588c3d74cb83a0b242a2b6c7504c1"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
@@ -4187,7 +4199,7 @@ checksum = "3ff15c8ecd7de3849db632e14d18d2571fa09dfc5ed93479bc4485c7a517c913"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
@@ -4201,30 +4213,30 @@ dependencies = [
|
||||
|
||||
[[package]]
|
||||
name = "time"
|
||||
version = "0.3.44"
|
||||
version = "0.3.47"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "91e7d9e3bb61134e77bde20dd4825b97c010155709965fedf0f49bb138e52a9d"
|
||||
checksum = "743bd48c283afc0388f9b8827b976905fb217ad9e647fae3a379a9283c4def2c"
|
||||
dependencies = [
|
||||
"deranged",
|
||||
"itoa",
|
||||
"num-conv",
|
||||
"powerfmt",
|
||||
"serde",
|
||||
"serde_core",
|
||||
"time-core",
|
||||
"time-macros",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "time-core"
|
||||
version = "0.1.6"
|
||||
version = "0.1.8"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "40868e7c1d2f0b8d73e4a8c7f0ff63af4f6d19be117e90bd73eb1d62cf831c6b"
|
||||
checksum = "7694e1cfe791f8d31026952abf09c69ca6f6fa4e1a1229e18988f06a04a12dca"
|
||||
|
||||
[[package]]
|
||||
name = "time-macros"
|
||||
version = "0.2.24"
|
||||
version = "0.2.27"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "30cfb0125f12d9c277f35663a0a33f8c30190f4e4574868a330595412d34ebf3"
|
||||
checksum = "2e70e4c5a0e0a8a4823ad65dfe1a6930e4f4d756dcd9dd7939022b5e8c501215"
|
||||
dependencies = [
|
||||
"num-conv",
|
||||
"time-core",
|
||||
@@ -4300,7 +4312,7 @@ checksum = "af407857209536a95c8e56f8231ef2c2e2aff839b22e07a1ffcbc617e9db9fa5"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
@@ -4318,9 +4330,9 @@ dependencies = [
|
||||
|
||||
[[package]]
|
||||
name = "toml"
|
||||
version = "0.9.8"
|
||||
version = "1.0.2+spec-1.1.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "f0dc8b1fb61449e27716ec0e1bdf0f6b8f3e8f6b05391e8497b8b6d7804ea6d8"
|
||||
checksum = "d1dfefef6a142e93f346b64c160934eb13b5594b84ab378133ac6815cb2bd57f"
|
||||
dependencies = [
|
||||
"indexmap",
|
||||
"serde_core",
|
||||
@@ -4333,27 +4345,27 @@ dependencies = [
|
||||
|
||||
[[package]]
|
||||
name = "toml_datetime"
|
||||
version = "0.7.3"
|
||||
version = "1.0.0+spec-1.1.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "f2cdb639ebbc97961c51720f858597f7f24c4fc295327923af55b74c3c724533"
|
||||
checksum = "32c2555c699578a4f59f0cc68e5116c8d7cabbd45e1409b989d4be085b53f13e"
|
||||
dependencies = [
|
||||
"serde_core",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "toml_parser"
|
||||
version = "1.0.4"
|
||||
version = "1.0.9+spec-1.1.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "c0cbe268d35bdb4bb5a56a2de88d0ad0eb70af5384a99d648cd4b3d04039800e"
|
||||
checksum = "702d4415e08923e7e1ef96cd5727c0dfed80b4d2fa25db9647fe5eb6f7c5a4c4"
|
||||
dependencies = [
|
||||
"winnow",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "toml_writer"
|
||||
version = "1.0.4"
|
||||
version = "1.0.6+spec-1.1.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "df8b2b54733674ad286d16267dcfc7a71ed5c776e4ac7aa3c3e2561f7c637bf2"
|
||||
checksum = "ab16f14aed21ee8bfd8ec22513f7287cd4a91aa92e44edfe2c17ddd004e92607"
|
||||
|
||||
[[package]]
|
||||
name = "tower-service"
|
||||
@@ -4380,7 +4392,7 @@ checksum = "7490cfa5ec963746568740651ac6781f701c9c5ea257c58e057f3ba8cf69e8da"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
@@ -4692,7 +4704,7 @@ dependencies = [
|
||||
"bumpalo",
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
"wasm-bindgen-shared",
|
||||
]
|
||||
|
||||
@@ -4834,7 +4846,7 @@ checksum = "9107ddc059d5b6fbfbffdfa7a7fe3e22a226def0b2608f72e9d552763d3e1ad7"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
@@ -4845,7 +4857,7 @@ checksum = "053e2e040ab57b9dc951b72c264860db7eb3b0200ba345b4e4c3b14f67855ddf"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
@@ -4856,7 +4868,7 @@ checksum = "29bee4b38ea3cde66011baa44dba677c432a78593e202392d1e9070cf2a7fca7"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
@@ -4867,7 +4879,7 @@ checksum = "3f316c4a2570ba26bbec722032c4099d8c8bc095efccdc15688708623367e358"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
@@ -5256,7 +5268,7 @@ checksum = "b659052874eb698efe5b9e8cf382204678a0086ebf46982b79d6ca3182927e5d"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
"synstructure",
|
||||
]
|
||||
|
||||
@@ -5277,7 +5289,7 @@ checksum = "d8a8d209fdf45cf5138cbb5a506f6b52522a25afccc534d1475dad8e31105c6a"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
@@ -5297,7 +5309,7 @@ checksum = "d71e5d6e06ab090c67b5e44993ec16b72dcbaabc526db883a360057678b48502"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
"synstructure",
|
||||
]
|
||||
|
||||
@@ -5318,7 +5330,7 @@ checksum = "ce36e65b0d2999d2aafac989fb249189a141aee1f53c612c1f37d72631959f69"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
@@ -5351,5 +5363,5 @@ checksum = "eadce39539ca5cb3985590102671f2567e659fca9666581ad3411d59207951f3"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn 2.0.111",
|
||||
"syn 2.0.116",
|
||||
]
|
||||
|
||||
@@ -37,6 +37,7 @@ keccak-const = "0.2"
|
||||
# Async dependencies
|
||||
tokio = "1.46"
|
||||
futures = "0.3"
|
||||
futures-lite = "2.6.1"
|
||||
futures-timer = "3.0"
|
||||
|
||||
# Data structures
|
||||
|
||||
File diff suppressed because it is too large
Load Diff
@@ -1,47 +1,29 @@
|
||||
# type: ignore
|
||||
#!/usr/bin/env python3
|
||||
"""Tool-calling eval for exo's OpenAI-compatible API.
|
||||
|
||||
Tests whether models correctly:
|
||||
- Trigger tool calls when appropriate
|
||||
- Return valid JSON arguments matching function schemas
|
||||
- Handle multi-turn tool use (call -> result -> final answer)
|
||||
- Avoid calling tools when unnecessary
|
||||
|
||||
Start exo with a model first, then run:
|
||||
uv run python tool_call_eval.py --model <model-id>
|
||||
uv run python tool_call_eval.py --model <model-id> --host 10.0.0.5 --port 52415
|
||||
uv run python tool_call_eval.py --model <model-id> --repeat 3
|
||||
uv run python tool_call_eval.py --model <model-id> --scenarios weather_simple calculator_multi_turn
|
||||
"""
|
||||
|
||||
# pyright: reportAny=false, reportUnknownMemberType=false, reportUnknownVariableType=false, reportUnknownArgumentType=false
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import contextlib
|
||||
import http.client
|
||||
import itertools
|
||||
import json
|
||||
import os
|
||||
import sys
|
||||
import time
|
||||
from collections.abc import Callable
|
||||
from pathlib import Path
|
||||
from statistics import mean
|
||||
from typing import Any
|
||||
from urllib.parse import urlencode
|
||||
|
||||
from harness import (
|
||||
ExoClient,
|
||||
ExoHttpError,
|
||||
add_common_instance_args,
|
||||
instance_id_from_instance,
|
||||
nodes_used_in_instance,
|
||||
resolve_model_short_id,
|
||||
settle_and_fetch_placements,
|
||||
wait_for_instance_gone,
|
||||
wait_for_instance_ready,
|
||||
)
|
||||
from loguru import logger
|
||||
from transformers import AutoTokenizer
|
||||
|
||||
# Backoff constants for cluster settling retry
|
||||
_SETTLE_INITIAL_BACKOFF_S = 1.0
|
||||
_SETTLE_MAX_BACKOFF_S = 60.0
|
||||
_SETTLE_BACKOFF_MULTIPLIER = 2.0
|
||||
|
||||
# Monkey-patch for transformers 5.x compatibility
|
||||
# Kimi's tokenization_kimi.py imports bytes_to_unicode from the old location
|
||||
# which was moved in transformers 5.0.0rc2
|
||||
@@ -121,6 +103,154 @@ def load_tokenizer_for_bench(model_id: str) -> Any:
|
||||
return AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
|
||||
|
||||
|
||||
class ExoHttpError(RuntimeError):
|
||||
def __init__(self, status: int, reason: str, body_preview: str):
|
||||
super().__init__(f"HTTP {status} {reason}: {body_preview}")
|
||||
self.status = status
|
||||
|
||||
|
||||
class ExoClient:
|
||||
def __init__(self, host: str, port: int, timeout_s: float = 7200.0):
|
||||
self.host = host
|
||||
self.port = port
|
||||
self.timeout_s = timeout_s
|
||||
|
||||
def request_json(
|
||||
self,
|
||||
method: str,
|
||||
path: str,
|
||||
params: dict[str, Any] | None = None,
|
||||
body: dict[str, Any] | None = None,
|
||||
headers: dict[str, str] | None = None,
|
||||
) -> Any:
|
||||
if not path.startswith("/"):
|
||||
path = "/" + path
|
||||
if params:
|
||||
path = path + "?" + urlencode(params)
|
||||
|
||||
conn = http.client.HTTPConnection(self.host, self.port, timeout=self.timeout_s)
|
||||
try:
|
||||
payload: bytes | None = None
|
||||
hdrs: dict[str, str] = {"Accept": "application/json"}
|
||||
|
||||
if body is not None:
|
||||
payload = json.dumps(body).encode("utf-8")
|
||||
hdrs["Content-Type"] = "application/json"
|
||||
if headers:
|
||||
hdrs.update(headers)
|
||||
|
||||
conn.request(method.upper(), path, body=payload, headers=hdrs)
|
||||
resp = conn.getresponse()
|
||||
raw = resp.read()
|
||||
text = raw.decode("utf-8", errors="replace") if raw else ""
|
||||
|
||||
if resp.status >= 400:
|
||||
raise ExoHttpError(resp.status, resp.reason, text[:300])
|
||||
|
||||
if not text:
|
||||
return None
|
||||
return json.loads(text)
|
||||
finally:
|
||||
conn.close()
|
||||
|
||||
def post_bench_chat_completions(self, payload: dict[str, Any]) -> dict[str, Any]:
|
||||
return self.request_json("POST", "/bench/chat/completions", body=payload)
|
||||
|
||||
|
||||
def unwrap_instance(instance: dict[str, Any]) -> dict[str, Any]:
|
||||
if len(instance) != 1:
|
||||
raise KeyError(f"Expected 1 key, got keys={list(instance.keys())}")
|
||||
|
||||
tag = next(iter(instance))
|
||||
inner = instance[tag]
|
||||
if not isinstance(inner, dict):
|
||||
raise TypeError(f"payload for {tag} must be dict, got {type(inner)}")
|
||||
return inner
|
||||
|
||||
|
||||
def instance_id_from_instance(instance: dict[str, Any]) -> str:
|
||||
inner = unwrap_instance(instance)
|
||||
return str(inner["instanceId"])
|
||||
|
||||
|
||||
def nodes_used_in_instance(instance: dict[str, Any]) -> int:
|
||||
inner = unwrap_instance(instance)
|
||||
return len(inner["shardAssignments"]["nodeToRunner"])
|
||||
|
||||
|
||||
def runner_ids_from_instance(instance: dict[str, Any]) -> list[str]:
|
||||
inner = unwrap_instance(instance)
|
||||
runner_to_shard = inner["shardAssignments"]["runnerToShard"]
|
||||
return list(runner_to_shard.keys())
|
||||
|
||||
|
||||
def runner_ready(runner: dict[str, Any]) -> bool:
|
||||
return "RunnerReady" in runner
|
||||
|
||||
|
||||
def runner_failed(runner: dict[str, Any]) -> bool:
|
||||
return "RunnerFailed" in runner
|
||||
|
||||
|
||||
def get_runner_failed_message(runner: dict[str, Any]) -> str | None:
|
||||
if "RunnerFailed" in runner:
|
||||
return runner["RunnerFailed"].get("errorMessage")
|
||||
return None
|
||||
|
||||
|
||||
def wait_for_instance_ready(
|
||||
client: ExoClient, instance_id: str, timeout: float = 24000.0
|
||||
) -> None:
|
||||
start_time = time.time()
|
||||
instance_existed = False
|
||||
while time.time() - start_time < timeout:
|
||||
state = client.request_json("GET", "/state")
|
||||
instances = state.get("instances", {})
|
||||
|
||||
if instance_id not in instances:
|
||||
if instance_existed:
|
||||
# Instance was deleted after being created - likely due to runner failure
|
||||
raise RuntimeError(
|
||||
f"Instance {instance_id} was deleted (runner may have failed)"
|
||||
)
|
||||
time.sleep(0.1)
|
||||
continue
|
||||
|
||||
instance_existed = True
|
||||
instance = instances[instance_id]
|
||||
runner_ids = runner_ids_from_instance(instance)
|
||||
runners = state.get("runners", {})
|
||||
|
||||
# Check for failed runners first
|
||||
for rid in runner_ids:
|
||||
runner = runners.get(rid, {})
|
||||
if runner_failed(runner):
|
||||
error_msg = get_runner_failed_message(runner) or "Unknown error"
|
||||
raise RuntimeError(f"Runner {rid} failed: {error_msg}")
|
||||
|
||||
if all(runner_ready(runners.get(rid, {})) for rid in runner_ids):
|
||||
return
|
||||
|
||||
time.sleep(0.1)
|
||||
|
||||
raise TimeoutError(f"Instance {instance_id} did not become ready within {timeout=}")
|
||||
|
||||
|
||||
def wait_for_instance_gone(
|
||||
client: ExoClient, instance_id: str, timeout: float = 3.0
|
||||
) -> None:
|
||||
start_time = time.time()
|
||||
while time.time() - start_time < timeout:
|
||||
try:
|
||||
client.request_json("GET", f"/instance/{instance_id}")
|
||||
time.sleep(0.4)
|
||||
except ExoHttpError as e:
|
||||
if e.status == 404:
|
||||
return
|
||||
|
||||
raise TimeoutError(f"Instance {instance_id} did not get deleted within {timeout=}")
|
||||
|
||||
|
||||
def format_peak_memory(b: float) -> str:
|
||||
for unit in ["B", "KB", "MB", "GB", "TB"]:
|
||||
if b < 1024.0:
|
||||
@@ -139,6 +269,184 @@ def parse_int_list(values: list[str]) -> list[int]:
|
||||
return items
|
||||
|
||||
|
||||
def resolve_model_short_id(client: ExoClient, model_arg: str) -> tuple[str, str]:
|
||||
models = client.request_json("GET", "/models") or {}
|
||||
data = models.get("data") or []
|
||||
|
||||
for m in data:
|
||||
if m.get("name").lower() == model_arg.lower():
|
||||
short_id = str(m["name"])
|
||||
full_id = str(m.get("hugging_face_id") or m["name"])
|
||||
return short_id, full_id
|
||||
|
||||
for m in data:
|
||||
if m.get("hugging_face_id") == model_arg:
|
||||
short_id = str(m["name"])
|
||||
full_id = str(m["hugging_face_id"])
|
||||
return short_id, full_id
|
||||
|
||||
raise ValueError(f"Model not found in /models: {model_arg}")
|
||||
|
||||
|
||||
def run_planning_phase(
|
||||
client: ExoClient,
|
||||
full_model_id: str,
|
||||
preview: dict[str, Any],
|
||||
danger_delete: bool,
|
||||
timeout: float,
|
||||
settle_deadline: float | None,
|
||||
) -> None:
|
||||
"""Check disk space and ensure model is downloaded before benchmarking."""
|
||||
# Get model size from /models
|
||||
models = client.request_json("GET", "/models") or {}
|
||||
model_bytes = 0
|
||||
for m in models.get("data", []):
|
||||
if m.get("hugging_face_id") == full_model_id:
|
||||
model_bytes = m.get("storage_size_megabytes", 0) * 1024 * 1024
|
||||
break
|
||||
|
||||
if not model_bytes:
|
||||
logger.warning(
|
||||
f"Could not determine size for {full_model_id}, skipping disk check"
|
||||
)
|
||||
return
|
||||
|
||||
# Get nodes from preview
|
||||
inner = unwrap_instance(preview["instance"])
|
||||
node_ids = list(inner["shardAssignments"]["nodeToRunner"].keys())
|
||||
runner_to_shard = inner["shardAssignments"]["runnerToShard"]
|
||||
|
||||
state = client.request_json("GET", "/state")
|
||||
downloads = state.get("downloads", {})
|
||||
node_disk = state.get("nodeDisk", {})
|
||||
|
||||
for node_id in node_ids:
|
||||
node_downloads = downloads.get(node_id, [])
|
||||
|
||||
# Check if model already downloaded on this node
|
||||
already_downloaded = any(
|
||||
"DownloadCompleted" in p
|
||||
and unwrap_instance(p["DownloadCompleted"]["shardMetadata"])["modelCard"][
|
||||
"modelId"
|
||||
]
|
||||
== full_model_id
|
||||
for p in node_downloads
|
||||
)
|
||||
if already_downloaded:
|
||||
continue
|
||||
|
||||
# Wait for disk info if settle_deadline is set
|
||||
disk_info = node_disk.get(node_id, {})
|
||||
backoff = _SETTLE_INITIAL_BACKOFF_S
|
||||
while not disk_info and settle_deadline and time.monotonic() < settle_deadline:
|
||||
remaining = settle_deadline - time.monotonic()
|
||||
logger.info(
|
||||
f"Waiting for disk info on {node_id} ({remaining:.0f}s remaining)..."
|
||||
)
|
||||
time.sleep(min(backoff, remaining))
|
||||
backoff = min(backoff * _SETTLE_BACKOFF_MULTIPLIER, _SETTLE_MAX_BACKOFF_S)
|
||||
state = client.request_json("GET", "/state")
|
||||
node_disk = state.get("nodeDisk", {})
|
||||
disk_info = node_disk.get(node_id, {})
|
||||
|
||||
if not disk_info:
|
||||
logger.warning(f"No disk info for {node_id}, skipping space check")
|
||||
continue
|
||||
|
||||
avail = disk_info.get("available", {}).get("inBytes", 0)
|
||||
if avail >= model_bytes:
|
||||
continue
|
||||
|
||||
if not danger_delete:
|
||||
raise RuntimeError(
|
||||
f"Insufficient disk on {node_id}: need {model_bytes // (1024**3)}GB, "
|
||||
f"have {avail // (1024**3)}GB. Use --danger-delete-downloads to free space."
|
||||
)
|
||||
|
||||
# Delete from smallest to largest
|
||||
completed = [
|
||||
(
|
||||
unwrap_instance(p["DownloadCompleted"]["shardMetadata"])["modelCard"][
|
||||
"modelId"
|
||||
],
|
||||
p["DownloadCompleted"]["totalBytes"]["inBytes"],
|
||||
)
|
||||
for p in node_downloads
|
||||
if "DownloadCompleted" in p
|
||||
]
|
||||
for del_model, size in sorted(completed, key=lambda x: x[1]):
|
||||
logger.info(f"Deleting {del_model} from {node_id} ({size // (1024**2)}MB)")
|
||||
client.request_json("DELETE", f"/download/{node_id}/{del_model}")
|
||||
avail += size
|
||||
if avail >= model_bytes:
|
||||
break
|
||||
|
||||
if avail < model_bytes:
|
||||
raise RuntimeError(f"Could not free enough space on {node_id}")
|
||||
|
||||
# Start downloads (idempotent)
|
||||
for node_id in node_ids:
|
||||
runner_id = inner["shardAssignments"]["nodeToRunner"][node_id]
|
||||
shard = runner_to_shard[runner_id]
|
||||
client.request_json(
|
||||
"POST",
|
||||
"/download/start",
|
||||
body={
|
||||
"targetNodeId": node_id,
|
||||
"shardMetadata": shard,
|
||||
},
|
||||
)
|
||||
logger.info(f"Started download on {node_id}")
|
||||
|
||||
# Wait for downloads
|
||||
start = time.time()
|
||||
while time.time() - start < timeout:
|
||||
state = client.request_json("GET", "/state")
|
||||
downloads = state.get("downloads", {})
|
||||
all_done = True
|
||||
for node_id in node_ids:
|
||||
done = any(
|
||||
"DownloadCompleted" in p
|
||||
and unwrap_instance(p["DownloadCompleted"]["shardMetadata"])[
|
||||
"modelCard"
|
||||
]["modelId"]
|
||||
== full_model_id
|
||||
for p in downloads.get(node_id, [])
|
||||
)
|
||||
failed = [
|
||||
p["DownloadFailed"]["errorMessage"]
|
||||
for p in downloads.get(node_id, [])
|
||||
if "DownloadFailed" in p
|
||||
and unwrap_instance(p["DownloadFailed"]["shardMetadata"])["modelCard"][
|
||||
"modelId"
|
||||
]
|
||||
== full_model_id
|
||||
]
|
||||
if failed:
|
||||
raise RuntimeError(f"Download failed on {node_id}: {failed[0]}")
|
||||
if not done:
|
||||
all_done = False
|
||||
if all_done:
|
||||
return
|
||||
time.sleep(1)
|
||||
|
||||
raise TimeoutError("Downloads did not complete in time")
|
||||
|
||||
|
||||
def placement_filter(instance_meta: str, wanted: str) -> bool:
|
||||
s = (instance_meta or "").lower()
|
||||
if wanted == "both":
|
||||
return ("ring" in s) or ("jaccl" in s)
|
||||
return wanted in s
|
||||
|
||||
|
||||
def sharding_filter(sharding: str, wanted: str) -> bool:
|
||||
s = (sharding or "").lower()
|
||||
if wanted == "both":
|
||||
return ("pipeline" in s) or ("tensor" in s)
|
||||
return wanted in s
|
||||
|
||||
|
||||
def run_one_completion(
|
||||
client: ExoClient, model_id: str, pp_hint: int, tg: int, prompt_sizer: PromptSizer
|
||||
) -> tuple[dict[str, Any], int]:
|
||||
@@ -230,12 +538,76 @@ class PromptSizer:
|
||||
return content, tok
|
||||
|
||||
|
||||
def fetch_and_filter_placements(
|
||||
client: ExoClient, full_model_id: str, args: argparse.Namespace
|
||||
) -> list[dict[str, Any]]:
|
||||
previews_resp = client.request_json(
|
||||
"GET", "/instance/previews", params={"model_id": full_model_id}
|
||||
)
|
||||
previews = previews_resp.get("previews") or []
|
||||
|
||||
selected: list[dict[str, Any]] = []
|
||||
for p in previews:
|
||||
if p.get("error") is not None:
|
||||
continue
|
||||
if not placement_filter(str(p.get("instance_meta", "")), args.instance_meta):
|
||||
continue
|
||||
if not sharding_filter(str(p.get("sharding", "")), args.sharding):
|
||||
continue
|
||||
|
||||
instance = p.get("instance")
|
||||
if not isinstance(instance, dict):
|
||||
continue
|
||||
|
||||
n = nodes_used_in_instance(instance)
|
||||
# Skip tensor ring single node as it is pointless when pipeline ring
|
||||
if n == 1 and (
|
||||
(args.sharding == "both" and "tensor" in p.get("sharding", "").lower())
|
||||
or (
|
||||
args.instance_meta == "both"
|
||||
and "jaccl" in p.get("instance_meta", "").lower()
|
||||
)
|
||||
):
|
||||
continue
|
||||
|
||||
if (
|
||||
args.skip_pipeline_jaccl
|
||||
and (
|
||||
args.instance_meta == "both"
|
||||
and "jaccl" in p.get("instance_meta", "").lower()
|
||||
)
|
||||
and (
|
||||
args.sharding == "both" and "pipeline" in p.get("sharding", "").lower()
|
||||
)
|
||||
):
|
||||
continue
|
||||
|
||||
if (
|
||||
args.skip_tensor_ring
|
||||
and (
|
||||
args.instance_meta == "both"
|
||||
and "ring" in p.get("instance_meta", "").lower()
|
||||
)
|
||||
and (args.sharding == "both" and "tensor" in p.get("sharding", "").lower())
|
||||
):
|
||||
continue
|
||||
|
||||
if args.min_nodes <= n <= args.max_nodes:
|
||||
selected.append(p)
|
||||
|
||||
return selected
|
||||
|
||||
|
||||
def main() -> int:
|
||||
ap = argparse.ArgumentParser(
|
||||
prog="exo-bench",
|
||||
description="Benchmark exo model throughput across placement previews.",
|
||||
)
|
||||
add_common_instance_args(ap)
|
||||
ap.add_argument("--host", default=os.environ.get("EXO_HOST", "localhost"))
|
||||
ap.add_argument(
|
||||
"--port", type=int, default=int(os.environ.get("EXO_PORT", "52415"))
|
||||
)
|
||||
ap.add_argument("--model", required=True, help="Model short id or huggingface id")
|
||||
ap.add_argument(
|
||||
"--pp",
|
||||
nargs="+",
|
||||
@@ -248,6 +620,34 @@ def main() -> int:
|
||||
required=True,
|
||||
help="Generation lengths (ints). Accepts commas.",
|
||||
)
|
||||
ap.add_argument(
|
||||
"--max-nodes",
|
||||
type=int,
|
||||
default=4,
|
||||
help="Only consider placements using <= this many nodes.",
|
||||
)
|
||||
ap.add_argument(
|
||||
"--min-nodes",
|
||||
type=int,
|
||||
default=1,
|
||||
help="Only consider placements using >= this many nodes.",
|
||||
)
|
||||
ap.add_argument(
|
||||
"--instance-meta", choices=["ring", "jaccl", "both"], default="both"
|
||||
)
|
||||
ap.add_argument(
|
||||
"--sharding", choices=["pipeline", "tensor", "both"], default="both"
|
||||
)
|
||||
ap.add_argument(
|
||||
"--skip-pipeline-jaccl",
|
||||
action="store_true",
|
||||
help="Skip pipeline+jaccl placements, as it's often pointless.",
|
||||
)
|
||||
ap.add_argument(
|
||||
"--skip-tensor-ring",
|
||||
action="store_true",
|
||||
help="Skip tensor+ring placements, as it's so slow.",
|
||||
)
|
||||
ap.add_argument(
|
||||
"--repeat", type=int, default=1, help="Repetitions per (pp,tg) pair."
|
||||
)
|
||||
@@ -257,6 +657,9 @@ def main() -> int:
|
||||
default=0,
|
||||
help="Warmup runs per placement (uses first pp/tg).",
|
||||
)
|
||||
ap.add_argument(
|
||||
"--timeout", type=float, default=7200.0, help="HTTP timeout (seconds)."
|
||||
)
|
||||
ap.add_argument(
|
||||
"--json-out",
|
||||
default="bench/results.json",
|
||||
@@ -271,6 +674,17 @@ def main() -> int:
|
||||
action="store_true",
|
||||
help="Force all pp×tg combinations (cartesian product) even when lists have equal length.",
|
||||
)
|
||||
ap.add_argument(
|
||||
"--settle-timeout",
|
||||
type=float,
|
||||
default=0,
|
||||
help="Max seconds to wait for the cluster to produce valid placements (0 = try once).",
|
||||
)
|
||||
ap.add_argument(
|
||||
"--danger-delete-downloads",
|
||||
action="store_true",
|
||||
help="Delete existing models from smallest to largest to make room for benchmark model.",
|
||||
)
|
||||
args = ap.parse_args()
|
||||
|
||||
pp_list = parse_int_list(args.pp)
|
||||
@@ -305,10 +719,24 @@ def main() -> int:
|
||||
logger.error("[exo-bench] tokenizer usable but prompt sizing failed")
|
||||
raise
|
||||
|
||||
selected = settle_and_fetch_placements(
|
||||
client, full_model_id, args, settle_timeout=args.settle_timeout
|
||||
settle_deadline = (
|
||||
time.monotonic() + args.settle_timeout if args.settle_timeout > 0 else None
|
||||
)
|
||||
|
||||
selected = fetch_and_filter_placements(client, full_model_id, args)
|
||||
|
||||
if not selected and settle_deadline:
|
||||
backoff = _SETTLE_INITIAL_BACKOFF_S
|
||||
while not selected and time.monotonic() < settle_deadline:
|
||||
remaining = settle_deadline - time.monotonic()
|
||||
logger.warning(
|
||||
f"No valid placements yet (cluster may still be settling). "
|
||||
f"Retrying in {backoff:.1f}s ({remaining:.0f}s remaining)..."
|
||||
)
|
||||
time.sleep(min(backoff, remaining))
|
||||
backoff = min(backoff * _SETTLE_BACKOFF_MULTIPLIER, _SETTLE_MAX_BACKOFF_S)
|
||||
selected = fetch_and_filter_placements(client, full_model_id, args)
|
||||
|
||||
if not selected:
|
||||
logger.error("No valid placements matched your filters.")
|
||||
return 1
|
||||
@@ -332,6 +760,16 @@ def main() -> int:
|
||||
if args.dry_run:
|
||||
return 0
|
||||
|
||||
logger.info("Planning phase: checking downloads...")
|
||||
run_planning_phase(
|
||||
client,
|
||||
full_model_id,
|
||||
selected[0],
|
||||
args.danger_delete_downloads,
|
||||
args.timeout,
|
||||
settle_deadline,
|
||||
)
|
||||
|
||||
all_rows: list[dict[str, Any]] = []
|
||||
|
||||
for preview in selected:
|
||||
|
||||
327
bench/harness.py
327
bench/harness.py
@@ -1,327 +0,0 @@
|
||||
# type: ignore
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import http.client
|
||||
import json
|
||||
import os
|
||||
import time
|
||||
from typing import Any
|
||||
from urllib.parse import urlencode
|
||||
|
||||
from loguru import logger
|
||||
|
||||
_SETTLE_INITIAL_BACKOFF_S = 1.0
|
||||
_SETTLE_MAX_BACKOFF_S = 60.0
|
||||
_SETTLE_BACKOFF_MULTIPLIER = 2.0
|
||||
|
||||
|
||||
class ExoHttpError(RuntimeError):
|
||||
def __init__(self, status: int, reason: str, body_preview: str):
|
||||
super().__init__(f"HTTP {status} {reason}: {body_preview}")
|
||||
self.status = status
|
||||
|
||||
|
||||
class ExoClient:
|
||||
def __init__(self, host: str, port: int, timeout_s: float = 7200.0):
|
||||
self.host = host
|
||||
self.port = port
|
||||
self.timeout_s = timeout_s
|
||||
|
||||
def request_json(
|
||||
self,
|
||||
method: str,
|
||||
path: str,
|
||||
params: dict[str, Any] | None = None,
|
||||
body: dict[str, Any] | None = None,
|
||||
headers: dict[str, str] | None = None,
|
||||
) -> Any:
|
||||
if not path.startswith("/"):
|
||||
path = "/" + path
|
||||
if params:
|
||||
path = path + "?" + urlencode(params)
|
||||
|
||||
conn = http.client.HTTPConnection(self.host, self.port, timeout=self.timeout_s)
|
||||
try:
|
||||
payload: bytes | None = None
|
||||
hdrs: dict[str, str] = {"Accept": "application/json"}
|
||||
|
||||
if body is not None:
|
||||
payload = json.dumps(body).encode("utf-8")
|
||||
hdrs["Content-Type"] = "application/json"
|
||||
if headers:
|
||||
hdrs.update(headers)
|
||||
|
||||
conn.request(method.upper(), path, body=payload, headers=hdrs)
|
||||
resp = conn.getresponse()
|
||||
raw = resp.read()
|
||||
text = raw.decode("utf-8", errors="replace") if raw else ""
|
||||
|
||||
if resp.status >= 400:
|
||||
raise ExoHttpError(resp.status, resp.reason, text[:300])
|
||||
|
||||
if not text:
|
||||
return None
|
||||
return json.loads(text)
|
||||
finally:
|
||||
conn.close()
|
||||
|
||||
def post_bench_chat_completions(self, payload: dict[str, Any]) -> dict[str, Any]:
|
||||
return self.request_json("POST", "/bench/chat/completions", body=payload)
|
||||
|
||||
|
||||
def unwrap_instance(instance: dict[str, Any]) -> dict[str, Any]:
|
||||
if len(instance) != 1:
|
||||
raise KeyError(f"Expected 1 key, got keys={list(instance.keys())}")
|
||||
|
||||
tag = next(iter(instance))
|
||||
inner = instance[tag]
|
||||
if not isinstance(inner, dict):
|
||||
raise TypeError(f"payload for {tag} must be dict, got {type(inner)}")
|
||||
return inner
|
||||
|
||||
|
||||
def instance_id_from_instance(instance: dict[str, Any]) -> str:
|
||||
inner = unwrap_instance(instance)
|
||||
return str(inner["instanceId"])
|
||||
|
||||
|
||||
def nodes_used_in_instance(instance: dict[str, Any]) -> int:
|
||||
inner = unwrap_instance(instance)
|
||||
return len(inner["shardAssignments"]["nodeToRunner"])
|
||||
|
||||
|
||||
def runner_ids_from_instance(instance: dict[str, Any]) -> list[str]:
|
||||
inner = unwrap_instance(instance)
|
||||
runner_to_shard = inner["shardAssignments"]["runnerToShard"]
|
||||
return list(runner_to_shard.keys())
|
||||
|
||||
|
||||
def runner_ready(runner: dict[str, Any]) -> bool:
|
||||
return "RunnerReady" in runner
|
||||
|
||||
|
||||
def runner_failed(runner: dict[str, Any]) -> bool:
|
||||
return "RunnerFailed" in runner
|
||||
|
||||
|
||||
def get_runner_failed_message(runner: dict[str, Any]) -> str | None:
|
||||
if "RunnerFailed" in runner:
|
||||
return runner["RunnerFailed"].get("errorMessage")
|
||||
return None
|
||||
|
||||
|
||||
def wait_for_instance_ready(
|
||||
client: ExoClient, instance_id: str, timeout: float = 24000.0
|
||||
) -> None:
|
||||
start_time = time.time()
|
||||
instance_existed = False
|
||||
while time.time() - start_time < timeout:
|
||||
state = client.request_json("GET", "/state")
|
||||
instances = state.get("instances", {})
|
||||
|
||||
if instance_id not in instances:
|
||||
if instance_existed:
|
||||
# Instance was deleted after being created - likely due to runner failure
|
||||
raise RuntimeError(
|
||||
f"Instance {instance_id} was deleted (runner may have failed)"
|
||||
)
|
||||
time.sleep(0.1)
|
||||
continue
|
||||
|
||||
instance_existed = True
|
||||
instance = instances[instance_id]
|
||||
runner_ids = runner_ids_from_instance(instance)
|
||||
runners = state.get("runners", {})
|
||||
|
||||
# Check for failed runners first
|
||||
for rid in runner_ids:
|
||||
runner = runners.get(rid, {})
|
||||
if runner_failed(runner):
|
||||
error_msg = get_runner_failed_message(runner) or "Unknown error"
|
||||
raise RuntimeError(f"Runner {rid} failed: {error_msg}")
|
||||
|
||||
if all(runner_ready(runners.get(rid, {})) for rid in runner_ids):
|
||||
return
|
||||
|
||||
time.sleep(0.1)
|
||||
|
||||
raise TimeoutError(f"Instance {instance_id} did not become ready within {timeout=}")
|
||||
|
||||
|
||||
def wait_for_instance_gone(
|
||||
client: ExoClient, instance_id: str, timeout: float = 3.0
|
||||
) -> None:
|
||||
start_time = time.time()
|
||||
while time.time() - start_time < timeout:
|
||||
try:
|
||||
client.request_json("GET", f"/instance/{instance_id}")
|
||||
time.sleep(0.4)
|
||||
except ExoHttpError as e:
|
||||
if e.status == 404:
|
||||
return
|
||||
raise
|
||||
|
||||
raise TimeoutError(f"Instance {instance_id} did not get deleted within {timeout=}")
|
||||
|
||||
|
||||
def resolve_model_short_id(client: ExoClient, model_arg: str) -> tuple[str, str]:
|
||||
models = client.request_json("GET", "/models") or {}
|
||||
data = models.get("data") or []
|
||||
|
||||
for m in data:
|
||||
if (m.get("name") or "").lower() == model_arg.lower():
|
||||
short_id = str(m["name"])
|
||||
full_id = str(m.get("hugging_face_id") or m["name"])
|
||||
return short_id, full_id
|
||||
|
||||
for m in data:
|
||||
if m.get("hugging_face_id") == model_arg:
|
||||
short_id = str(m["name"])
|
||||
full_id = str(m["hugging_face_id"])
|
||||
return short_id, full_id
|
||||
|
||||
raise ValueError(f"Model not found in /models: {model_arg}")
|
||||
|
||||
|
||||
def placement_filter(instance_meta: str, wanted: str) -> bool:
|
||||
s = (instance_meta or "").lower()
|
||||
if wanted == "both":
|
||||
return ("ring" in s) or ("jaccl" in s)
|
||||
return wanted in s
|
||||
|
||||
|
||||
def sharding_filter(sharding: str, wanted: str) -> bool:
|
||||
s = (sharding or "").lower()
|
||||
if wanted == "both":
|
||||
return ("pipeline" in s) or ("tensor" in s)
|
||||
return wanted in s
|
||||
|
||||
|
||||
def fetch_and_filter_placements(
|
||||
client: ExoClient, full_model_id: str, args: argparse.Namespace
|
||||
) -> list[dict[str, Any]]:
|
||||
previews_resp = client.request_json(
|
||||
"GET", "/instance/previews", params={"model_id": full_model_id}
|
||||
)
|
||||
previews = previews_resp.get("previews") or []
|
||||
|
||||
selected: list[dict[str, Any]] = []
|
||||
for p in previews:
|
||||
if p.get("error") is not None:
|
||||
continue
|
||||
if not placement_filter(str(p.get("instance_meta", "")), args.instance_meta):
|
||||
continue
|
||||
if not sharding_filter(str(p.get("sharding", "")), args.sharding):
|
||||
continue
|
||||
|
||||
instance = p.get("instance")
|
||||
if not isinstance(instance, dict):
|
||||
continue
|
||||
|
||||
n = nodes_used_in_instance(instance)
|
||||
# Skip tensor ring single node as it is pointless when pipeline ring
|
||||
if n == 1 and (
|
||||
(args.sharding == "both" and "tensor" in p.get("sharding", "").lower())
|
||||
or (
|
||||
args.instance_meta == "both"
|
||||
and "jaccl" in p.get("instance_meta", "").lower()
|
||||
)
|
||||
):
|
||||
continue
|
||||
|
||||
if (
|
||||
args.skip_pipeline_jaccl
|
||||
and (
|
||||
args.instance_meta == "both"
|
||||
and "jaccl" in p.get("instance_meta", "").lower()
|
||||
)
|
||||
and (
|
||||
args.sharding == "both" and "pipeline" in p.get("sharding", "").lower()
|
||||
)
|
||||
):
|
||||
continue
|
||||
|
||||
if (
|
||||
args.skip_tensor_ring
|
||||
and (
|
||||
args.instance_meta == "both"
|
||||
and "ring" in p.get("instance_meta", "").lower()
|
||||
)
|
||||
and (args.sharding == "both" and "tensor" in p.get("sharding", "").lower())
|
||||
):
|
||||
continue
|
||||
|
||||
if args.min_nodes <= n <= args.max_nodes:
|
||||
selected.append(p)
|
||||
|
||||
return selected
|
||||
|
||||
|
||||
def settle_and_fetch_placements(
|
||||
client: ExoClient,
|
||||
full_model_id: str,
|
||||
args: argparse.Namespace,
|
||||
settle_timeout: float = 0,
|
||||
) -> list[dict[str, Any]]:
|
||||
selected = fetch_and_filter_placements(client, full_model_id, args)
|
||||
|
||||
if not selected and settle_timeout > 0:
|
||||
backoff = _SETTLE_INITIAL_BACKOFF_S
|
||||
deadline = time.monotonic() + settle_timeout
|
||||
while not selected and time.monotonic() < deadline:
|
||||
remaining = deadline - time.monotonic()
|
||||
logger.warning(
|
||||
f"No valid placements yet (cluster may still be settling). "
|
||||
f"Retrying in {backoff:.1f}s ({remaining:.0f}s remaining)..."
|
||||
)
|
||||
time.sleep(min(backoff, remaining))
|
||||
backoff = min(backoff * _SETTLE_BACKOFF_MULTIPLIER, _SETTLE_MAX_BACKOFF_S)
|
||||
selected = fetch_and_filter_placements(client, full_model_id, args)
|
||||
|
||||
return selected
|
||||
|
||||
|
||||
def add_common_instance_args(ap: argparse.ArgumentParser) -> None:
|
||||
ap.add_argument("--host", default=os.environ.get("EXO_HOST", "localhost"))
|
||||
ap.add_argument(
|
||||
"--port", type=int, default=int(os.environ.get("EXO_PORT", "52415"))
|
||||
)
|
||||
ap.add_argument("--model", required=True, help="Model short id or huggingface id")
|
||||
ap.add_argument(
|
||||
"--max-nodes",
|
||||
type=int,
|
||||
default=4,
|
||||
help="Only consider placements using <= this many nodes.",
|
||||
)
|
||||
ap.add_argument(
|
||||
"--min-nodes",
|
||||
type=int,
|
||||
default=1,
|
||||
help="Only consider placements using >= this many nodes.",
|
||||
)
|
||||
ap.add_argument(
|
||||
"--instance-meta", choices=["ring", "jaccl", "both"], default="both"
|
||||
)
|
||||
ap.add_argument(
|
||||
"--sharding", choices=["pipeline", "tensor", "both"], default="both"
|
||||
)
|
||||
ap.add_argument(
|
||||
"--skip-pipeline-jaccl",
|
||||
action="store_true",
|
||||
help="Skip pipeline+jaccl placements, as it's often pointless.",
|
||||
)
|
||||
ap.add_argument(
|
||||
"--skip-tensor-ring",
|
||||
action="store_true",
|
||||
help="Skip tensor+ring placements, as it's so slow.",
|
||||
)
|
||||
ap.add_argument(
|
||||
"--timeout", type=float, default=7200.0, help="HTTP timeout (seconds)."
|
||||
)
|
||||
ap.add_argument(
|
||||
"--settle-timeout",
|
||||
type=float,
|
||||
default=0,
|
||||
help="Max seconds to wait for the cluster to produce valid placements (0 = try once).",
|
||||
)
|
||||
@@ -4,7 +4,6 @@ version = "0.1.0"
|
||||
description = "Benchmarking tool for exo distributed inference"
|
||||
requires-python = ">=3.13"
|
||||
dependencies = [
|
||||
"httpx>=0.27.0",
|
||||
"loguru>=0.7.3",
|
||||
"transformers>=5.0.0",
|
||||
"huggingface-hub>=0.33.4",
|
||||
|
||||
@@ -1,240 +0,0 @@
|
||||
# Tool definitions — each becomes an OpenAI function tool.
|
||||
# All scenarios get all tools unless they specify a `tools` list.
|
||||
|
||||
[tools.get_current_weather]
|
||||
description = "Get the current weather in a given location"
|
||||
required = ["location"]
|
||||
|
||||
[tools.get_current_weather.properties.location]
|
||||
type = "string"
|
||||
description = "City and state, e.g. San Francisco, CA"
|
||||
|
||||
[tools.get_current_weather.properties.unit]
|
||||
type = "string"
|
||||
enum = ["celsius", "fahrenheit"]
|
||||
description = "Temperature unit"
|
||||
|
||||
[tools.calculate]
|
||||
description = "Evaluate a mathematical expression and return the numeric result"
|
||||
required = ["expression"]
|
||||
|
||||
[tools.calculate.properties.expression]
|
||||
type = "string"
|
||||
description = "The math expression to evaluate, e.g. '2 + 3 * 4'"
|
||||
|
||||
[tools.search_products]
|
||||
description = "Search for products in a catalog by query, category, and price"
|
||||
required = ["query"]
|
||||
|
||||
[tools.search_products.properties.query]
|
||||
type = "string"
|
||||
description = "Search query string"
|
||||
|
||||
[tools.search_products.properties.category]
|
||||
type = "string"
|
||||
enum = ["electronics", "clothing", "food", "books"]
|
||||
description = "Product category to filter by"
|
||||
|
||||
[tools.search_products.properties.max_price]
|
||||
type = "number"
|
||||
description = "Maximum price in USD"
|
||||
|
||||
# -- Should call a tool --
|
||||
|
||||
[[scenarios]]
|
||||
name = "weather_simple"
|
||||
description = "Basic weather query -> get_current_weather"
|
||||
expect_tool_call = true
|
||||
expected_function = "get_current_weather"
|
||||
required_arg_keys = ["location"]
|
||||
|
||||
[[scenarios.messages]]
|
||||
role = "user"
|
||||
content = "What's the weather like in Tokyo right now?"
|
||||
|
||||
[[scenarios]]
|
||||
name = "calculator_simple"
|
||||
description = "Math question -> calculate"
|
||||
expect_tool_call = true
|
||||
expected_function = "calculate"
|
||||
required_arg_keys = ["expression"]
|
||||
|
||||
[[scenarios.messages]]
|
||||
role = "user"
|
||||
content = "Use the calculator to compute 3847 * 926 + 17293"
|
||||
|
||||
[[scenarios]]
|
||||
name = "search_with_filters"
|
||||
description = "Product search with category and price filter"
|
||||
expect_tool_call = true
|
||||
expected_function = "search_products"
|
||||
required_arg_keys = ["query"]
|
||||
|
||||
[[scenarios.messages]]
|
||||
role = "user"
|
||||
content = "Find me electronics under $50"
|
||||
|
||||
# -- Multi-turn: tool call then follow-up --
|
||||
|
||||
[[scenarios]]
|
||||
name = "weather_multi_turn"
|
||||
description = "Weather query -> tool result -> natural language summary"
|
||||
expect_tool_call = true
|
||||
expected_function = "get_current_weather"
|
||||
required_arg_keys = ["location"]
|
||||
|
||||
[scenarios.tool_result]
|
||||
temperature = "18C"
|
||||
condition = "partly cloudy"
|
||||
humidity = "65%"
|
||||
wind = "12 km/h NW"
|
||||
|
||||
[[scenarios.messages]]
|
||||
role = "user"
|
||||
content = "What's the weather in Paris?"
|
||||
|
||||
[[scenarios]]
|
||||
name = "calculator_multi_turn"
|
||||
description = "Math query -> tool result -> model reports the answer"
|
||||
expect_tool_call = true
|
||||
expected_function = "calculate"
|
||||
required_arg_keys = ["expression"]
|
||||
|
||||
[scenarios.tool_result]
|
||||
result = 491682
|
||||
|
||||
[[scenarios.messages]]
|
||||
role = "user"
|
||||
content = "Use the calculator to compute 1847 * 263 + 5921"
|
||||
|
||||
[[scenarios]]
|
||||
name = "search_multi_turn"
|
||||
description = "Search query -> tool result -> model summarizes products"
|
||||
expect_tool_call = true
|
||||
expected_function = "search_products"
|
||||
required_arg_keys = ["query"]
|
||||
|
||||
[[scenarios.tool_result.results]]
|
||||
name = "Hands-On Machine Learning"
|
||||
price = 45.99
|
||||
rating = 4.8
|
||||
|
||||
[[scenarios.tool_result.results]]
|
||||
name = "Deep Learning with Python"
|
||||
price = 39.99
|
||||
rating = 4.6
|
||||
|
||||
[[scenarios.messages]]
|
||||
role = "user"
|
||||
content = "Search for books about machine learning"
|
||||
|
||||
# -- Sequential tool calls --
|
||||
|
||||
[[scenarios]]
|
||||
name = "chained_tool_calls_same"
|
||||
description = "Thinking + weather(Tokyo) -> result -> model must call weather(London)"
|
||||
expect_tool_call = true
|
||||
expected_function = "get_current_weather"
|
||||
required_arg_keys = ["location"]
|
||||
|
||||
[[scenarios.messages]]
|
||||
role = "user"
|
||||
content = "Compare the weather in Tokyo and London."
|
||||
|
||||
[[scenarios.messages]]
|
||||
role = "assistant"
|
||||
content = "I'll check both cities. Let me start with Tokyo."
|
||||
|
||||
[[scenarios.messages.tool_calls]]
|
||||
id = "call_1"
|
||||
name = "get_current_weather"
|
||||
arguments = { location = "Tokyo" }
|
||||
|
||||
[[scenarios.messages]]
|
||||
role = "tool"
|
||||
tool_call_id = "call_1"
|
||||
content = '{"temperature": "25C", "condition": "sunny"}'
|
||||
|
||||
[[scenarios]]
|
||||
name = "chained_tool_calls_different"
|
||||
description = "Thinking + weather(Berlin) -> result -> model must call calculator"
|
||||
expect_tool_call = true
|
||||
expected_function = "calculate"
|
||||
required_arg_keys = ["expression"]
|
||||
|
||||
[[scenarios.messages]]
|
||||
role = "user"
|
||||
content = "What's the weather in Berlin, and also use the calculator to compute 4819 * 37 + 291."
|
||||
|
||||
[[scenarios.messages]]
|
||||
role = "assistant"
|
||||
content = "I'll handle both. Let me check Berlin's weather first."
|
||||
|
||||
[[scenarios.messages.tool_calls]]
|
||||
id = "call_2"
|
||||
name = "get_current_weather"
|
||||
arguments = { location = "Berlin" }
|
||||
|
||||
[[scenarios.messages]]
|
||||
role = "tool"
|
||||
tool_call_id = "call_2"
|
||||
content = '{"temperature": "12C", "condition": "rainy"}'
|
||||
|
||||
[[scenarios]]
|
||||
name = "chained_tool_calls_three"
|
||||
description = "Two prior thinking+tool calls -> results -> model must make a third"
|
||||
expect_tool_call = true
|
||||
expected_function = "get_current_weather"
|
||||
required_arg_keys = ["location"]
|
||||
|
||||
[[scenarios.messages]]
|
||||
role = "user"
|
||||
content = "Compare weather in Tokyo, Paris, and London."
|
||||
|
||||
[[scenarios.messages]]
|
||||
role = "assistant"
|
||||
content = "I'll check all three cities. Starting with Tokyo."
|
||||
|
||||
[[scenarios.messages.tool_calls]]
|
||||
id = "call_3"
|
||||
name = "get_current_weather"
|
||||
arguments = { location = "Tokyo" }
|
||||
|
||||
[[scenarios.messages]]
|
||||
role = "tool"
|
||||
tool_call_id = "call_3"
|
||||
content = '{"temperature": "25C", "condition": "sunny"}'
|
||||
|
||||
[[scenarios.messages]]
|
||||
role = "assistant"
|
||||
content = "Got Tokyo. Now checking Paris."
|
||||
|
||||
[[scenarios.messages.tool_calls]]
|
||||
id = "call_4"
|
||||
name = "get_current_weather"
|
||||
arguments = { location = "Paris" }
|
||||
|
||||
[[scenarios.messages]]
|
||||
role = "tool"
|
||||
tool_call_id = "call_4"
|
||||
content = '{"temperature": "18C", "condition": "cloudy"}'
|
||||
|
||||
# -- Should NOT call a tool --
|
||||
|
||||
[[scenarios]]
|
||||
name = "no_tool_joke"
|
||||
description = "Joke request should NOT trigger any tool"
|
||||
expect_tool_call = false
|
||||
|
||||
[[scenarios.messages]]
|
||||
role = "user"
|
||||
content = "Tell me a funny joke about cats."
|
||||
|
||||
[[scenarios]]
|
||||
name = "no_tool_factual"
|
||||
description = "Factual question answerable from training data"
|
||||
expect_tool_call = false
|
||||
|
||||
[[scenarios.messages]]
|
||||
role = "user"
|
||||
content = "What is the capital of Japan?"
|
||||
@@ -103,7 +103,7 @@
|
||||
const modelSupportsThinking = $derived(() => {
|
||||
if (!currentModel) return false;
|
||||
const caps = modelCapabilities[currentModel] || [];
|
||||
return caps.includes("thinking_toggle") && caps.includes("text");
|
||||
return caps.includes("thinking") && caps.includes("text");
|
||||
});
|
||||
|
||||
const isEditOnlyWithoutImage = $derived(
|
||||
|
||||
@@ -59,14 +59,13 @@
|
||||
}
|
||||
|
||||
const sizeOptions: ImageGenerationParams["size"][] = [
|
||||
"auto",
|
||||
"512x512",
|
||||
"768x768",
|
||||
"1024x1024",
|
||||
"1024x768",
|
||||
"768x1024",
|
||||
"1024x1536",
|
||||
"1536x1024",
|
||||
"1024x1365",
|
||||
"1365x1024",
|
||||
];
|
||||
|
||||
const qualityOptions: ImageGenerationParams["quality"][] = [
|
||||
@@ -177,90 +176,92 @@
|
||||
<div class="border-b border-exo-medium-gray/30 px-3 py-2">
|
||||
<!-- Basic params row -->
|
||||
<div class="flex items-center gap-3 flex-wrap">
|
||||
<!-- Size -->
|
||||
<div class="flex items-center gap-1.5">
|
||||
<span class="text-xs text-exo-light-gray uppercase tracking-wider"
|
||||
>SIZE:</span
|
||||
>
|
||||
<div class="relative">
|
||||
<button
|
||||
bind:this={sizeButtonRef}
|
||||
type="button"
|
||||
onclick={() => (isSizeDropdownOpen = !isSizeDropdownOpen)}
|
||||
class="bg-exo-medium-gray/50 border border-exo-yellow/30 rounded pl-2 pr-6 py-1 text-xs font-mono text-exo-yellow cursor-pointer transition-all duration-200 hover:border-exo-yellow/50 focus:outline-none focus:border-exo-yellow/70 {isSizeDropdownOpen
|
||||
? 'border-exo-yellow/70'
|
||||
: ''}"
|
||||
<!-- Size (hidden in edit mode - output size comes from input image) -->
|
||||
{#if !isEditMode}
|
||||
<div class="flex items-center gap-1.5">
|
||||
<span class="text-xs text-exo-light-gray uppercase tracking-wider"
|
||||
>SIZE:</span
|
||||
>
|
||||
{params.size.toUpperCase()}
|
||||
</button>
|
||||
<div
|
||||
class="absolute right-1.5 top-1/2 -translate-y-1/2 pointer-events-none transition-transform duration-200 {isSizeDropdownOpen
|
||||
? 'rotate-180'
|
||||
: ''}"
|
||||
>
|
||||
<svg
|
||||
class="w-3 h-3 text-exo-yellow/60"
|
||||
fill="none"
|
||||
viewBox="0 0 24 24"
|
||||
stroke="currentColor"
|
||||
<div class="relative">
|
||||
<button
|
||||
bind:this={sizeButtonRef}
|
||||
type="button"
|
||||
onclick={() => (isSizeDropdownOpen = !isSizeDropdownOpen)}
|
||||
class="bg-exo-medium-gray/50 border border-exo-yellow/30 rounded pl-2 pr-6 py-1 text-xs font-mono text-exo-yellow cursor-pointer transition-all duration-200 hover:border-exo-yellow/50 focus:outline-none focus:border-exo-yellow/70 {isSizeDropdownOpen
|
||||
? 'border-exo-yellow/70'
|
||||
: ''}"
|
||||
>
|
||||
<path
|
||||
stroke-linecap="round"
|
||||
stroke-linejoin="round"
|
||||
stroke-width="2"
|
||||
d="M19 9l-7 7-7-7"
|
||||
/>
|
||||
</svg>
|
||||
</div>
|
||||
</div>
|
||||
|
||||
{#if isSizeDropdownOpen}
|
||||
<!-- Backdrop to close dropdown -->
|
||||
<button
|
||||
type="button"
|
||||
class="fixed inset-0 z-[9998] cursor-default"
|
||||
onclick={() => (isSizeDropdownOpen = false)}
|
||||
aria-label="Close dropdown"
|
||||
></button>
|
||||
|
||||
<!-- Dropdown Panel - fixed positioning to escape overflow:hidden -->
|
||||
<div
|
||||
class="fixed bg-exo-dark-gray border border-exo-yellow/30 rounded shadow-lg shadow-black/50 z-[9999] max-h-48 overflow-y-auto overflow-x-hidden min-w-max"
|
||||
style="bottom: calc(100vh - {sizeDropdownPosition()
|
||||
.top}px + 4px); left: {sizeDropdownPosition().left}px;"
|
||||
>
|
||||
<div class="py-1">
|
||||
{#each sizeOptions as size}
|
||||
<button
|
||||
type="button"
|
||||
onclick={() => selectSize(size)}
|
||||
class="w-full px-3 py-1.5 text-left text-xs font-mono tracking-wide transition-colors duration-100 flex items-center gap-2 {params.size ===
|
||||
size
|
||||
? 'bg-transparent text-exo-yellow'
|
||||
: 'text-exo-light-gray hover:text-exo-yellow'}"
|
||||
>
|
||||
{#if params.size === size}
|
||||
<svg
|
||||
class="w-3 h-3 flex-shrink-0"
|
||||
fill="currentColor"
|
||||
viewBox="0 0 20 20"
|
||||
>
|
||||
<path
|
||||
fill-rule="evenodd"
|
||||
d="M16.707 5.293a1 1 0 010 1.414l-8 8a1 1 0 01-1.414 0l-4-4a1 1 0 011.414-1.414L8 12.586l7.293-7.293a1 1 0 011.414 0z"
|
||||
clip-rule="evenodd"
|
||||
/>
|
||||
</svg>
|
||||
{:else}
|
||||
<span class="w-3"></span>
|
||||
{/if}
|
||||
<span>{size.toUpperCase()}</span>
|
||||
</button>
|
||||
{/each}
|
||||
{params.size}
|
||||
</button>
|
||||
<div
|
||||
class="absolute right-1.5 top-1/2 -translate-y-1/2 pointer-events-none transition-transform duration-200 {isSizeDropdownOpen
|
||||
? 'rotate-180'
|
||||
: ''}"
|
||||
>
|
||||
<svg
|
||||
class="w-3 h-3 text-exo-yellow/60"
|
||||
fill="none"
|
||||
viewBox="0 0 24 24"
|
||||
stroke="currentColor"
|
||||
>
|
||||
<path
|
||||
stroke-linecap="round"
|
||||
stroke-linejoin="round"
|
||||
stroke-width="2"
|
||||
d="M19 9l-7 7-7-7"
|
||||
/>
|
||||
</svg>
|
||||
</div>
|
||||
</div>
|
||||
{/if}
|
||||
</div>
|
||||
|
||||
{#if isSizeDropdownOpen}
|
||||
<!-- Backdrop to close dropdown -->
|
||||
<button
|
||||
type="button"
|
||||
class="fixed inset-0 z-[9998] cursor-default"
|
||||
onclick={() => (isSizeDropdownOpen = false)}
|
||||
aria-label="Close dropdown"
|
||||
></button>
|
||||
|
||||
<!-- Dropdown Panel - fixed positioning to escape overflow:hidden -->
|
||||
<div
|
||||
class="fixed bg-exo-dark-gray border border-exo-yellow/30 rounded shadow-lg shadow-black/50 z-[9999] max-h-48 overflow-y-auto min-w-max"
|
||||
style="bottom: calc(100vh - {sizeDropdownPosition()
|
||||
.top}px + 4px); left: {sizeDropdownPosition().left}px;"
|
||||
>
|
||||
<div class="py-1">
|
||||
{#each sizeOptions as size}
|
||||
<button
|
||||
type="button"
|
||||
onclick={() => selectSize(size)}
|
||||
class="w-full px-3 py-1.5 text-left text-xs font-mono tracking-wide transition-colors duration-100 flex items-center gap-2 {params.size ===
|
||||
size
|
||||
? 'bg-transparent text-exo-yellow'
|
||||
: 'text-exo-light-gray hover:text-exo-yellow'}"
|
||||
>
|
||||
{#if params.size === size}
|
||||
<svg
|
||||
class="w-3 h-3 flex-shrink-0"
|
||||
fill="currentColor"
|
||||
viewBox="0 0 20 20"
|
||||
>
|
||||
<path
|
||||
fill-rule="evenodd"
|
||||
d="M16.707 5.293a1 1 0 010 1.414l-8 8a1 1 0 01-1.414 0l-4-4a1 1 0 011.414-1.414L8 12.586l7.293-7.293a1 1 0 011.414 0z"
|
||||
clip-rule="evenodd"
|
||||
/>
|
||||
</svg>
|
||||
{:else}
|
||||
<span class="w-3"></span>
|
||||
{/if}
|
||||
<span>{size}</span>
|
||||
</button>
|
||||
{/each}
|
||||
</div>
|
||||
</div>
|
||||
{/if}
|
||||
</div>
|
||||
{/if}
|
||||
|
||||
<!-- Quality -->
|
||||
<div class="flex items-center gap-1.5">
|
||||
@@ -310,7 +311,7 @@
|
||||
|
||||
<!-- Dropdown Panel - fixed positioning to escape overflow:hidden -->
|
||||
<div
|
||||
class="fixed bg-exo-dark-gray border border-exo-yellow/30 rounded shadow-lg shadow-black/50 z-[9999] max-h-48 overflow-y-auto overflow-x-hidden min-w-max"
|
||||
class="fixed bg-exo-dark-gray border border-exo-yellow/30 rounded shadow-lg shadow-black/50 z-[9999] max-h-48 overflow-y-auto min-w-max"
|
||||
style="bottom: calc(100vh - {qualityDropdownPosition()
|
||||
.top}px + 4px); left: {qualityDropdownPosition().left}px;"
|
||||
>
|
||||
|
||||
@@ -306,14 +306,13 @@ const IMAGE_PARAMS_STORAGE_KEY = "exo-image-generation-params";
|
||||
export interface ImageGenerationParams {
|
||||
// Basic params
|
||||
size:
|
||||
| "auto"
|
||||
| "512x512"
|
||||
| "768x768"
|
||||
| "1024x1024"
|
||||
| "1024x768"
|
||||
| "768x1024"
|
||||
| "1024x1536"
|
||||
| "1536x1024";
|
||||
| "1024x1365"
|
||||
| "1365x1024";
|
||||
quality: "low" | "medium" | "high";
|
||||
outputFormat: "png" | "jpeg";
|
||||
numImages: number;
|
||||
@@ -337,7 +336,7 @@ export interface EditingImage {
|
||||
}
|
||||
|
||||
const DEFAULT_IMAGE_PARAMS: ImageGenerationParams = {
|
||||
size: "auto",
|
||||
size: "1024x1024",
|
||||
quality: "medium",
|
||||
outputFormat: "png",
|
||||
numImages: 1,
|
||||
|
||||
@@ -41,7 +41,7 @@ let
|
||||
|
||||
mlx = stdenv.mkDerivation rec {
|
||||
pname = "mlx";
|
||||
version = let v = "0.30.7.dev20260218+14841977"; in
|
||||
version = let v = "0.30.7.dev20260217+50487b41"; in
|
||||
assert v == uvLockMlxVersion || throw "MLX version mismatch: nix/mlx.nix has ${v} but uv.lock has ${uvLockMlxVersion}. Update both the version and hash in nix/mlx.nix.";
|
||||
v;
|
||||
pyproject = true;
|
||||
@@ -49,8 +49,8 @@ let
|
||||
src = fetchFromGitHub {
|
||||
owner = "rltakashige";
|
||||
repo = "mlx-jaccl-fix-small-recv";
|
||||
rev = "1484197707f35186ad3bd614357c7c47fdf86ebc";
|
||||
hash = "sha256-FupCMoK/SF/ldfKuvMSAKECcOP8c+ANgkQlPZttDsLk=";
|
||||
rev = "50487b4141f3c951122655db3b83df5146c1fbeb";
|
||||
hash = "sha256-IL4a9vMX5nocgJU1WG4zE8hArHkHJtnh4sdYh3od5zU=";
|
||||
};
|
||||
|
||||
patches = [
|
||||
|
||||
@@ -158,7 +158,6 @@
|
||||
exo-test-env = testVenv;
|
||||
} // {
|
||||
exo-bench = mkBenchScript "exo-bench" (inputs.self + /bench/exo_bench.py);
|
||||
exo-eval-tool-calls = mkBenchScript "exo-eval-tool-calls" (inputs.self + /bench/eval_tool_calls.py);
|
||||
exo-get-all-models-on-cluster = mkSimplePythonScript "exo-get-all-models-on-cluster" (inputs.self + /tests/get_all_models_on_cluster.py);
|
||||
};
|
||||
|
||||
|
||||
@@ -6,7 +6,7 @@ tasks = ["TextGeneration"]
|
||||
family = "deepseek"
|
||||
quantization = "4bit"
|
||||
base_model = "DeepSeek V3.1"
|
||||
capabilities = ["text", "thinking", "thinking_toggle"]
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 405874409472
|
||||
|
||||
@@ -6,7 +6,7 @@ tasks = ["TextGeneration"]
|
||||
family = "deepseek"
|
||||
quantization = "8bit"
|
||||
base_model = "DeepSeek V3.1"
|
||||
capabilities = ["text", "thinking", "thinking_toggle"]
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 765577920512
|
||||
|
||||
@@ -6,7 +6,7 @@ tasks = ["TextGeneration"]
|
||||
family = "glm"
|
||||
quantization = "8bit"
|
||||
base_model = "GLM 4.5 Air"
|
||||
capabilities = ["text", "thinking", "thinking_toggle"]
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 122406567936
|
||||
|
||||
@@ -6,7 +6,7 @@ tasks = ["TextGeneration"]
|
||||
family = "glm"
|
||||
quantization = "bf16"
|
||||
base_model = "GLM 4.5 Air"
|
||||
capabilities = ["text", "thinking", "thinking_toggle"]
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 229780750336
|
||||
|
||||
@@ -6,7 +6,7 @@ tasks = ["TextGeneration"]
|
||||
family = "glm"
|
||||
quantization = "4bit"
|
||||
base_model = "GLM 4.7"
|
||||
capabilities = ["text", "thinking", "thinking_toggle"]
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 198556925568
|
||||
|
||||
@@ -6,7 +6,7 @@ tasks = ["TextGeneration"]
|
||||
family = "glm"
|
||||
quantization = "6bit"
|
||||
base_model = "GLM 4.7"
|
||||
capabilities = ["text", "thinking", "thinking_toggle"]
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 286737579648
|
||||
|
||||
@@ -6,7 +6,7 @@ tasks = ["TextGeneration"]
|
||||
family = "glm"
|
||||
quantization = "8bit"
|
||||
base_model = "GLM 4.7"
|
||||
capabilities = ["text", "thinking", "thinking_toggle"]
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 396963397248
|
||||
|
||||
@@ -6,7 +6,7 @@ tasks = ["TextGeneration"]
|
||||
family = "glm"
|
||||
quantization = "4bit"
|
||||
base_model = "GLM 4.7 Flash"
|
||||
capabilities = ["text", "thinking", "thinking_toggle"]
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 19327352832
|
||||
|
||||
@@ -6,7 +6,7 @@ tasks = ["TextGeneration"]
|
||||
family = "glm"
|
||||
quantization = "5bit"
|
||||
base_model = "GLM 4.7 Flash"
|
||||
capabilities = ["text", "thinking", "thinking_toggle"]
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 22548578304
|
||||
|
||||
@@ -6,7 +6,7 @@ tasks = ["TextGeneration"]
|
||||
family = "glm"
|
||||
quantization = "6bit"
|
||||
base_model = "GLM 4.7 Flash"
|
||||
capabilities = ["text", "thinking", "thinking_toggle"]
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 26843545600
|
||||
|
||||
@@ -6,7 +6,7 @@ tasks = ["TextGeneration"]
|
||||
family = "glm"
|
||||
quantization = "8bit"
|
||||
base_model = "GLM 4.7 Flash"
|
||||
capabilities = ["text", "thinking", "thinking_toggle"]
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 34359738368
|
||||
|
||||
@@ -1,12 +0,0 @@
|
||||
model_id = "mlx-community/GLM-5-8bit-MXFP8"
|
||||
n_layers = 78
|
||||
hidden_size = 6144
|
||||
supports_tensor = true
|
||||
tasks = ["TextGeneration"]
|
||||
family = "glm"
|
||||
quantization = "8bit"
|
||||
base_model = "GLM-5"
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 790517400864
|
||||
@@ -1,12 +0,0 @@
|
||||
model_id = "mlx-community/GLM-5-MXFP4-Q8"
|
||||
n_layers = 78
|
||||
hidden_size = 6144
|
||||
supports_tensor = true
|
||||
tasks = ["TextGeneration"]
|
||||
family = "glm"
|
||||
quantization = "MXFP4-Q8"
|
||||
base_model = "GLM-5"
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 405478939008
|
||||
@@ -1,12 +0,0 @@
|
||||
model_id = "mlx-community/GLM-5"
|
||||
n_layers = 78
|
||||
hidden_size = 6144
|
||||
supports_tensor = true
|
||||
tasks = ["TextGeneration"]
|
||||
family = "glm"
|
||||
quantization = "bf16"
|
||||
base_model = "GLM-5"
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 1487822475264
|
||||
@@ -6,7 +6,7 @@ tasks = ["TextGeneration"]
|
||||
family = "kimi"
|
||||
quantization = ""
|
||||
base_model = "Kimi K2"
|
||||
capabilities = ["text", "thinking", "thinking_toggle"]
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 706522120192
|
||||
|
||||
@@ -6,7 +6,7 @@ tasks = ["TextGeneration"]
|
||||
family = "kimi"
|
||||
quantization = ""
|
||||
base_model = "Kimi K2.5"
|
||||
capabilities = ["text", "thinking", "thinking_toggle"]
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 662498705408
|
||||
|
||||
@@ -6,7 +6,7 @@ tasks = ["TextGeneration"]
|
||||
family = "minimax"
|
||||
quantization = "3bit"
|
||||
base_model = "MiniMax M2.1"
|
||||
capabilities = ["text", "thinking", "thinking_toggle"]
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 100086644736
|
||||
|
||||
@@ -6,7 +6,7 @@ tasks = ["TextGeneration"]
|
||||
family = "minimax"
|
||||
quantization = "8bit"
|
||||
base_model = "MiniMax M2.1"
|
||||
capabilities = ["text", "thinking", "thinking_toggle"]
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 242986745856
|
||||
|
||||
@@ -1,12 +0,0 @@
|
||||
model_id = "mlx-community/MiniMax-M2.5-4bit"
|
||||
n_layers = 62
|
||||
hidden_size = 3072
|
||||
supports_tensor = true
|
||||
tasks = ["TextGeneration"]
|
||||
family = "minimax"
|
||||
quantization = "4bit"
|
||||
base_model = "MiniMax M2.5"
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 128666664960
|
||||
@@ -1,12 +0,0 @@
|
||||
model_id = "mlx-community/MiniMax-M2.5-6bit"
|
||||
n_layers = 62
|
||||
hidden_size = 3072
|
||||
supports_tensor = true
|
||||
tasks = ["TextGeneration"]
|
||||
family = "minimax"
|
||||
quantization = "6bit"
|
||||
base_model = "MiniMax M2.5"
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 185826705408
|
||||
@@ -1,12 +0,0 @@
|
||||
model_id = "mlx-community/MiniMax-M2.5-8bit"
|
||||
n_layers = 62
|
||||
hidden_size = 3072
|
||||
supports_tensor = true
|
||||
tasks = ["TextGeneration"]
|
||||
family = "minimax"
|
||||
quantization = "8bit"
|
||||
base_model = "MiniMax M2.5"
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 242986745856
|
||||
@@ -6,7 +6,7 @@ tasks = ["TextGeneration"]
|
||||
family = "qwen"
|
||||
quantization = "4bit"
|
||||
base_model = "Qwen3 0.6B"
|
||||
capabilities = ["text", "thinking", "thinking_toggle"]
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 342884352
|
||||
|
||||
@@ -6,7 +6,7 @@ tasks = ["TextGeneration"]
|
||||
family = "qwen"
|
||||
quantization = "8bit"
|
||||
base_model = "Qwen3 0.6B"
|
||||
capabilities = ["text", "thinking", "thinking_toggle"]
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 698351616
|
||||
|
||||
@@ -6,7 +6,7 @@ tasks = ["TextGeneration"]
|
||||
family = "qwen"
|
||||
quantization = "4bit"
|
||||
base_model = "Qwen3 235B"
|
||||
capabilities = ["text", "thinking", "thinking_toggle"]
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 141733920768
|
||||
|
||||
@@ -6,7 +6,7 @@ tasks = ["TextGeneration"]
|
||||
family = "qwen"
|
||||
quantization = "8bit"
|
||||
base_model = "Qwen3 235B"
|
||||
capabilities = ["text", "thinking", "thinking_toggle"]
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 268435456000
|
||||
|
||||
@@ -6,7 +6,7 @@ tasks = ["TextGeneration"]
|
||||
family = "qwen"
|
||||
quantization = "4bit"
|
||||
base_model = "Qwen3 30B"
|
||||
capabilities = ["text", "thinking", "thinking_toggle"]
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 17612931072
|
||||
|
||||
@@ -6,7 +6,7 @@ tasks = ["TextGeneration"]
|
||||
family = "qwen"
|
||||
quantization = "8bit"
|
||||
base_model = "Qwen3 30B"
|
||||
capabilities = ["text", "thinking", "thinking_toggle"]
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 33279705088
|
||||
|
||||
@@ -6,7 +6,7 @@ tasks = ["TextGeneration"]
|
||||
family = "qwen"
|
||||
quantization = "4bit"
|
||||
base_model = "Qwen3 Next 80B"
|
||||
capabilities = ["text", "thinking", "thinking_toggle"]
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 47080074240
|
||||
|
||||
@@ -6,7 +6,7 @@ tasks = ["TextGeneration"]
|
||||
family = "qwen"
|
||||
quantization = "8bit"
|
||||
base_model = "Qwen3 Next 80B"
|
||||
capabilities = ["text", "thinking", "thinking_toggle"]
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 88814387200
|
||||
|
||||
@@ -6,7 +6,7 @@ tasks = ["TextGeneration"]
|
||||
family = "step"
|
||||
quantization = "4bit"
|
||||
base_model = "Step 3.5 Flash"
|
||||
capabilities = ["text", "thinking", "thinking_toggle"]
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 114572190076
|
||||
|
||||
@@ -6,7 +6,7 @@ tasks = ["TextGeneration"]
|
||||
family = "step"
|
||||
quantization = "6bit"
|
||||
base_model = "Step 3.5 Flash"
|
||||
capabilities = ["text", "thinking", "thinking_toggle"]
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 159039627774
|
||||
|
||||
@@ -6,7 +6,7 @@ tasks = ["TextGeneration"]
|
||||
family = "step"
|
||||
quantization = "8bit"
|
||||
base_model = "Step 3.5 Flash"
|
||||
capabilities = ["text", "thinking", "thinking_toggle"]
|
||||
capabilities = ["text", "thinking"]
|
||||
|
||||
[storage_size]
|
||||
in_bytes = 209082699847
|
||||
|
||||
@@ -27,7 +27,7 @@ networking = { workspace = true }
|
||||
# interop
|
||||
pyo3 = { version = "0.27.2", features = [
|
||||
# "abi3-py313", # tells pyo3 (and maturin) to build using the stable ABI with minimum Python version 3.13
|
||||
"nightly", # enables better-supported GIL integration
|
||||
# "nightly", # enables better-supported GIL integration
|
||||
"experimental-async", # async support in #[pyfunction] & #[pymethods]
|
||||
#"experimental-inspect", # inspection of generated binary => easier to automate type-hint generation
|
||||
#"py-clone", # adding Clone-ing of `Py<T>` without GIL (may cause panics - remove if panics happen)
|
||||
@@ -38,7 +38,7 @@ pyo3 = { version = "0.27.2", features = [
|
||||
# "ordered-float", "rust_decimal", "smallvec",
|
||||
# "anyhow", "chrono", "chrono-local", "chrono-tz", "eyre", "jiff-02", "lock_api", "parking-lot", "time", "serde",
|
||||
] }
|
||||
pyo3-stub-gen = { version = "0.17.2" }
|
||||
pyo3-stub-gen = { version = "0.19.0" }
|
||||
pyo3-async-runtimes = { version = "0.27.0", features = ["attributes", "tokio-runtime", "testing"] }
|
||||
pyo3-log = "0.13.2"
|
||||
|
||||
@@ -49,7 +49,7 @@ pin-project = { workspace = true }
|
||||
|
||||
# async runtime
|
||||
tokio = { workspace = true, features = ["full", "tracing"] }
|
||||
futures = { workspace = true }
|
||||
futures-lite = { workspace = true }
|
||||
|
||||
# utility dependencies
|
||||
util = { workspace = true }
|
||||
|
||||
@@ -1,155 +1,85 @@
|
||||
# This file is automatically generated by pyo3_stub_gen
|
||||
# ruff: noqa: E501, F401
|
||||
# ruff: noqa: E501, F401, F403, F405
|
||||
|
||||
import builtins
|
||||
import enum
|
||||
import typing
|
||||
__all__ = [
|
||||
"AllQueuesFullError",
|
||||
"Keypair",
|
||||
"NoPeersSubscribedToTopicError",
|
||||
"PyMessage",
|
||||
"PySwarm",
|
||||
]
|
||||
|
||||
@typing.final
|
||||
class AllQueuesFullError(builtins.Exception):
|
||||
def __new__(cls, *args: typing.Any) -> AllQueuesFullError: ...
|
||||
def __repr__(self) -> builtins.str: ...
|
||||
def __new__(cls, *_a: typing.Any) -> AllQueuesFullError: ...
|
||||
def __str__(self) -> builtins.str: ...
|
||||
|
||||
@typing.final
|
||||
class ConnectionUpdate:
|
||||
@property
|
||||
def update_type(self) -> ConnectionUpdateType:
|
||||
r"""
|
||||
Whether this is a connection or disconnection event
|
||||
"""
|
||||
@property
|
||||
def peer_id(self) -> PeerId:
|
||||
r"""
|
||||
Identity of the peer that we have connected to or disconnected from.
|
||||
"""
|
||||
@property
|
||||
def remote_ipv4(self) -> builtins.str:
|
||||
r"""
|
||||
Remote connection's IPv4 address.
|
||||
"""
|
||||
@property
|
||||
def remote_tcp_port(self) -> builtins.int:
|
||||
r"""
|
||||
Remote connection's TCP port.
|
||||
"""
|
||||
|
||||
@typing.final
|
||||
class Keypair:
|
||||
r"""
|
||||
Identity keypair of a node.
|
||||
"""
|
||||
@staticmethod
|
||||
def generate_ed25519() -> Keypair:
|
||||
def generate() -> Keypair:
|
||||
r"""
|
||||
Generate a new Ed25519 keypair.
|
||||
"""
|
||||
@staticmethod
|
||||
def generate_ecdsa() -> Keypair:
|
||||
r"""
|
||||
Generate a new ECDSA keypair.
|
||||
"""
|
||||
@staticmethod
|
||||
def generate_secp256k1() -> Keypair:
|
||||
r"""
|
||||
Generate a new Secp256k1 keypair.
|
||||
"""
|
||||
@staticmethod
|
||||
def from_protobuf_encoding(bytes: bytes) -> Keypair:
|
||||
def deserialize(bytes: bytes) -> Keypair:
|
||||
r"""
|
||||
Decode a private key from a protobuf structure and parse it as a `Keypair`.
|
||||
"""
|
||||
@staticmethod
|
||||
def rsa_from_pkcs8(bytes: bytes) -> Keypair:
|
||||
r"""
|
||||
Decode an keypair from a DER-encoded secret key in PKCS#8 `PrivateKeyInfo`
|
||||
format (i.e. unencrypted) as defined in [RFC5208].
|
||||
|
||||
[RFC5208]: https://tools.ietf.org/html/rfc5208#section-5
|
||||
"""
|
||||
@staticmethod
|
||||
def secp256k1_from_der(bytes: bytes) -> Keypair:
|
||||
r"""
|
||||
Decode a keypair from a DER-encoded Secp256k1 secret key in an `ECPrivateKey`
|
||||
structure as defined in [RFC5915].
|
||||
|
||||
[RFC5915]: https://tools.ietf.org/html/rfc5915
|
||||
"""
|
||||
@staticmethod
|
||||
def ed25519_from_bytes(bytes: bytes) -> Keypair: ...
|
||||
def to_protobuf_encoding(self) -> bytes:
|
||||
def serialize(self) -> bytes:
|
||||
r"""
|
||||
Encode a private key as protobuf structure.
|
||||
"""
|
||||
def to_peer_id(self) -> PeerId:
|
||||
def to_string(self) -> builtins.str:
|
||||
r"""
|
||||
Convert the `Keypair` into the corresponding `PeerId`.
|
||||
"""
|
||||
|
||||
@typing.final
|
||||
class Multiaddr:
|
||||
r"""
|
||||
Representation of a Multiaddr.
|
||||
"""
|
||||
@staticmethod
|
||||
def empty() -> Multiaddr:
|
||||
r"""
|
||||
Create a new, empty multiaddress.
|
||||
"""
|
||||
@staticmethod
|
||||
def with_capacity(n: builtins.int) -> Multiaddr:
|
||||
r"""
|
||||
Create a new, empty multiaddress with the given capacity.
|
||||
"""
|
||||
@staticmethod
|
||||
def from_bytes(bytes: bytes) -> Multiaddr:
|
||||
r"""
|
||||
Parse a `Multiaddr` value from its byte slice representation.
|
||||
"""
|
||||
@staticmethod
|
||||
def from_string(string: builtins.str) -> Multiaddr:
|
||||
r"""
|
||||
Parse a `Multiaddr` value from its string representation.
|
||||
"""
|
||||
def len(self) -> builtins.int:
|
||||
r"""
|
||||
Return the length in bytes of this multiaddress.
|
||||
"""
|
||||
def is_empty(self) -> builtins.bool:
|
||||
r"""
|
||||
Returns true if the length of this multiaddress is 0.
|
||||
"""
|
||||
def to_bytes(self) -> bytes:
|
||||
r"""
|
||||
Return a copy of this [`Multiaddr`]'s byte representation.
|
||||
"""
|
||||
def to_string(self) -> builtins.str:
|
||||
r"""
|
||||
Convert a Multiaddr to a string.
|
||||
"""
|
||||
class NoPeersSubscribedToTopicError(builtins.Exception):
|
||||
def __new__(cls, *_a: typing.Any) -> NoPeersSubscribedToTopicError: ...
|
||||
def __str__(self) -> builtins.str: ...
|
||||
|
||||
class PyMessage:
|
||||
@typing.final
|
||||
class Connection(PyMessage):
|
||||
__match_args__ = ("node_id", "connected",)
|
||||
@property
|
||||
def node_id(self) -> builtins.str: ...
|
||||
@property
|
||||
def connected(self) -> builtins.bool: ...
|
||||
def __new__(cls, node_id: builtins.str, connected: builtins.bool) -> PyMessage.Connection: ...
|
||||
|
||||
@typing.final
|
||||
class Gossip(PyMessage):
|
||||
__match_args__ = ("node_id", "topic", "data",)
|
||||
@property
|
||||
def node_id(self) -> builtins.str: ...
|
||||
@property
|
||||
def topic(self) -> builtins.str: ...
|
||||
@property
|
||||
def data(self) -> bytes: ...
|
||||
def __new__(cls, node_id: builtins.str, topic: builtins.str, data: bytes) -> PyMessage.Gossip: ...
|
||||
|
||||
...
|
||||
|
||||
@typing.final
|
||||
class NetworkingHandle:
|
||||
def __new__(cls, identity: Keypair) -> NetworkingHandle: ...
|
||||
async def connection_update_recv(self) -> ConnectionUpdate:
|
||||
class PySwarm:
|
||||
def __new__(cls, identity: Keypair) -> PySwarm: ...
|
||||
async def recv(self) -> PyMessage:
|
||||
r"""
|
||||
Receives the next `ConnectionUpdate` from networking.
|
||||
Receives the next message from networking.
|
||||
"""
|
||||
async def connection_update_recv_many(self, limit: builtins.int) -> builtins.list[ConnectionUpdate]:
|
||||
r"""
|
||||
Receives at most `limit` `ConnectionUpdate`s from networking and returns them.
|
||||
|
||||
For `limit = 0`, an empty collection of `ConnectionUpdate`s will be returned immediately.
|
||||
For `limit > 0`, if there are no `ConnectionUpdate`s in the channel's queue this method
|
||||
will sleep until a `ConnectionUpdate`s is sent.
|
||||
"""
|
||||
async def gossipsub_subscribe(self, topic: builtins.str) -> builtins.bool:
|
||||
async def gossipsub_subscribe(self, topic: builtins.str) -> None:
|
||||
r"""
|
||||
Subscribe to a `GossipSub` topic.
|
||||
|
||||
Returns `True` if the subscription worked. Returns `False` if we were already subscribed.
|
||||
"""
|
||||
async def gossipsub_unsubscribe(self, topic: builtins.str) -> builtins.bool:
|
||||
async def gossipsub_unsubscribe(self, topic: builtins.str) -> None:
|
||||
r"""
|
||||
Unsubscribes from a `GossipSub` topic.
|
||||
|
||||
@@ -157,65 +87,6 @@ class NetworkingHandle:
|
||||
"""
|
||||
async def gossipsub_publish(self, topic: builtins.str, data: bytes) -> None:
|
||||
r"""
|
||||
Publishes a message with multiple topics to the `GossipSub` network.
|
||||
|
||||
If no peers are found that subscribe to this topic, throws `NoPeersSubscribedToTopicError` exception.
|
||||
"""
|
||||
async def gossipsub_recv(self) -> tuple[builtins.str, bytes]:
|
||||
r"""
|
||||
Receives the next message from the `GossipSub` network.
|
||||
"""
|
||||
async def gossipsub_recv_many(self, limit: builtins.int) -> builtins.list[tuple[builtins.str, bytes]]:
|
||||
r"""
|
||||
Receives at most `limit` messages from the `GossipSub` network and returns them.
|
||||
|
||||
For `limit = 0`, an empty collection of messages will be returned immediately.
|
||||
For `limit > 0`, if there are no messages in the channel's queue this method
|
||||
will sleep until a message is sent.
|
||||
Publishes a message to the network on a specific topic.
|
||||
"""
|
||||
|
||||
@typing.final
|
||||
class NoPeersSubscribedToTopicError(builtins.Exception):
|
||||
def __new__(cls, *args: typing.Any) -> NoPeersSubscribedToTopicError: ...
|
||||
def __repr__(self) -> builtins.str: ...
|
||||
def __str__(self) -> builtins.str: ...
|
||||
|
||||
@typing.final
|
||||
class PeerId:
|
||||
r"""
|
||||
Identifier of a peer of the network.
|
||||
|
||||
The data is a `CIDv0` compatible multihash of the protobuf encoded public key of the peer
|
||||
as specified in [specs/peer-ids](https://github.com/libp2p/specs/blob/master/peer-ids/peer-ids.md).
|
||||
"""
|
||||
@staticmethod
|
||||
def random() -> PeerId:
|
||||
r"""
|
||||
Generates a random peer ID from a cryptographically secure PRNG.
|
||||
|
||||
This is useful for randomly walking on a DHT, or for testing purposes.
|
||||
"""
|
||||
@staticmethod
|
||||
def from_bytes(bytes: bytes) -> PeerId:
|
||||
r"""
|
||||
Parses a `PeerId` from bytes.
|
||||
"""
|
||||
def to_bytes(self) -> bytes:
|
||||
r"""
|
||||
Returns a raw bytes representation of this `PeerId`.
|
||||
"""
|
||||
def to_base58(self) -> builtins.str:
|
||||
r"""
|
||||
Returns a base-58 encoded string of this `PeerId`.
|
||||
"""
|
||||
def __repr__(self) -> builtins.str: ...
|
||||
def __str__(self) -> builtins.str: ...
|
||||
|
||||
@typing.final
|
||||
class ConnectionUpdateType(enum.Enum):
|
||||
r"""
|
||||
Connection or disconnection event discriminant type.
|
||||
"""
|
||||
Connected = ...
|
||||
Disconnected = ...
|
||||
|
||||
|
||||
@@ -1,38 +1,22 @@
|
||||
//! SEE: https://pyo3.rs/v0.26.0/async-await.html#detaching-from-the-interpreter-across-await
|
||||
//!
|
||||
|
||||
use pin_project::pin_project;
|
||||
use pyo3::marker::Ungil;
|
||||
//! See: <https://pyo3.rs/v0.27.2/async-await.html#detaching-from-the-interpreter-across-await>
|
||||
use pyo3::prelude::*;
|
||||
use std::{
|
||||
future::Future,
|
||||
pin::Pin,
|
||||
pin::{Pin, pin},
|
||||
task::{Context, Poll},
|
||||
};
|
||||
|
||||
/// SEE: https://pyo3.rs/v0.26.0/async-await.html#detaching-from-the-interpreter-across-await
|
||||
#[pin_project]
|
||||
#[repr(transparent)]
|
||||
pub(crate) struct AllowThreads<F>(#[pin] F);
|
||||
|
||||
impl<F> AllowThreads<F>
|
||||
where
|
||||
Self: Future,
|
||||
{
|
||||
pub fn new(f: F) -> Self {
|
||||
Self(f)
|
||||
}
|
||||
}
|
||||
pub struct AllowThreads<F>(pub(crate) F);
|
||||
|
||||
impl<F> Future for AllowThreads<F>
|
||||
where
|
||||
F: Future + Ungil,
|
||||
F::Output: Ungil,
|
||||
F: Future + Unpin + Send,
|
||||
F::Output: Send,
|
||||
{
|
||||
type Output = F::Output;
|
||||
|
||||
fn poll(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Self::Output> {
|
||||
fn poll(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Self::Output> {
|
||||
let waker = cx.waker();
|
||||
Python::attach(|py| py.detach(|| self.project().0.poll(&mut Context::from_waker(waker))))
|
||||
Python::attach(|py| py.detach(|| pin!(&mut self.0).poll(&mut Context::from_waker(waker))))
|
||||
}
|
||||
}
|
||||
|
||||
47
rust/exo_pyo3_bindings/src/ident.rs
Normal file
47
rust/exo_pyo3_bindings/src/ident.rs
Normal file
@@ -0,0 +1,47 @@
|
||||
use crate::ext::ResultExt as _;
|
||||
use libp2p::identity::Keypair;
|
||||
use pyo3::prelude::{PyBytesMethods as _, PyModule, PyModuleMethods as _};
|
||||
use pyo3::types::PyBytes;
|
||||
use pyo3::{Bound, PyResult, Python, pyclass, pymethods};
|
||||
use pyo3_stub_gen::derive::{gen_stub_pyclass, gen_stub_pymethods};
|
||||
|
||||
/// Identity keypair of a node.
|
||||
#[gen_stub_pyclass]
|
||||
#[pyclass(name = "Keypair", frozen)]
|
||||
#[repr(transparent)]
|
||||
pub struct PyKeypair(pub Keypair);
|
||||
|
||||
#[gen_stub_pymethods]
|
||||
#[pymethods]
|
||||
#[allow(clippy::needless_pass_by_value)]
|
||||
impl PyKeypair {
|
||||
/// Generate a new Ed25519 keypair.
|
||||
#[staticmethod]
|
||||
fn generate() -> Self {
|
||||
Self(Keypair::generate_ed25519())
|
||||
}
|
||||
|
||||
/// Decode a private key from a protobuf structure and parse it as a `Keypair`.
|
||||
#[staticmethod]
|
||||
fn deserialize(bytes: Bound<'_, PyBytes>) -> PyResult<Self> {
|
||||
let bytes = Vec::from(bytes.as_bytes());
|
||||
Ok(Self(Keypair::from_protobuf_encoding(&bytes).pyerr()?))
|
||||
}
|
||||
|
||||
/// Encode a private key as protobuf structure.
|
||||
fn serialize<'py>(&self, py: Python<'py>) -> PyResult<Bound<'py, PyBytes>> {
|
||||
let bytes = self.0.to_protobuf_encoding().pyerr()?;
|
||||
Ok(PyBytes::new(py, &bytes))
|
||||
}
|
||||
|
||||
/// Convert the `Keypair` into the corresponding `PeerId`.
|
||||
fn to_string(&self) -> String {
|
||||
self.0.public().to_peer_id().to_base58()
|
||||
}
|
||||
}
|
||||
|
||||
pub fn ident_submodule(m: &Bound<'_, PyModule>) -> PyResult<()> {
|
||||
m.add_class::<PyKeypair>()?;
|
||||
|
||||
Ok(())
|
||||
}
|
||||
@@ -4,25 +4,12 @@
|
||||
//!
|
||||
//!
|
||||
|
||||
// enable Rust-unstable features for convenience
|
||||
#![feature(trait_alias)]
|
||||
#![feature(tuple_trait)]
|
||||
#![feature(unboxed_closures)]
|
||||
// #![feature(stmt_expr_attributes)]
|
||||
// #![feature(assert_matches)]
|
||||
// #![feature(async_fn_in_dyn_trait)]
|
||||
// #![feature(async_for_loop)]
|
||||
// #![feature(auto_traits)]
|
||||
// #![feature(negative_impls)]
|
||||
|
||||
extern crate core;
|
||||
mod allow_threading;
|
||||
pub(crate) mod networking;
|
||||
pub(crate) mod pylibp2p;
|
||||
mod ident;
|
||||
mod networking;
|
||||
|
||||
use crate::ident::ident_submodule;
|
||||
use crate::networking::networking_submodule;
|
||||
use crate::pylibp2p::ident::ident_submodule;
|
||||
use crate::pylibp2p::multiaddr::multiaddr_submodule;
|
||||
use pyo3::prelude::PyModule;
|
||||
use pyo3::{Bound, PyResult, pyclass, pymodule};
|
||||
use pyo3_stub_gen::define_stub_info_gatherer;
|
||||
@@ -32,25 +19,13 @@ pub(crate) mod r#const {
|
||||
pub const MPSC_CHANNEL_SIZE: usize = 1024;
|
||||
}
|
||||
|
||||
/// Namespace for all the type/trait aliases used by this crate.
|
||||
pub(crate) mod alias {
|
||||
use std::marker::Tuple;
|
||||
|
||||
pub trait SendFn<Args: Tuple + Send + 'static, Output> =
|
||||
Fn<Args, Output = Output> + Send + 'static;
|
||||
}
|
||||
|
||||
/// Namespace for crate-wide extension traits/methods
|
||||
pub(crate) mod ext {
|
||||
use crate::allow_threading::AllowThreads;
|
||||
use extend::ext;
|
||||
use pyo3::exceptions::{PyConnectionError, PyRuntimeError};
|
||||
use pyo3::exceptions::PyRuntimeError;
|
||||
use pyo3::types::PyBytes;
|
||||
use pyo3::{Py, PyErr, PyResult, Python};
|
||||
use tokio::runtime::Runtime;
|
||||
use tokio::sync::mpsc;
|
||||
use tokio::sync::mpsc::error::TryRecvError;
|
||||
use tokio::task::JoinHandle;
|
||||
use pyo3::{Py, PyResult, Python};
|
||||
|
||||
#[ext(pub, name = ByteArrayExt)]
|
||||
impl [u8] {
|
||||
@@ -70,102 +45,16 @@ pub(crate) mod ext {
|
||||
}
|
||||
|
||||
pub trait FutureExt: Future + Sized {
|
||||
/// SEE: https://pyo3.rs/v0.26.0/async-await.html#detaching-from-the-interpreter-across-await
|
||||
/// SEE: https://pyo3.rs/v0.27.2/async-await.html#detaching-from-the-interpreter-across-await
|
||||
fn allow_threads_py(self) -> AllowThreads<Self>
|
||||
where
|
||||
AllowThreads<Self>: Future,
|
||||
{
|
||||
AllowThreads::new(self)
|
||||
AllowThreads(self)
|
||||
}
|
||||
}
|
||||
|
||||
impl<T: Future> FutureExt for T {}
|
||||
|
||||
#[ext(pub, name = PyErrExt)]
|
||||
impl PyErr {
|
||||
fn receiver_channel_closed() -> Self {
|
||||
PyConnectionError::new_err("Receiver channel closed unexpectedly")
|
||||
}
|
||||
}
|
||||
|
||||
#[ext(pub, name = PyResultExt)]
|
||||
impl<T> PyResult<T> {
|
||||
fn write_unraisable(self) -> Option<T> {
|
||||
Python::attach(|py| self.write_unraisable_with(py))
|
||||
}
|
||||
|
||||
fn write_unraisable_with(self, py: Python<'_>) -> Option<T> {
|
||||
match self {
|
||||
Ok(v) => Some(v),
|
||||
Err(e) => {
|
||||
// write error back to python
|
||||
e.write_unraisable(py, None);
|
||||
None
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
#[ext(pub, name = TokioRuntimeExt)]
|
||||
impl Runtime {
|
||||
fn spawn_with_scope<F>(&self, py: Python<'_>, future: F) -> PyResult<JoinHandle<F::Output>>
|
||||
where
|
||||
F: Future + Send + 'static,
|
||||
F::Output: Send + 'static,
|
||||
{
|
||||
let locals = pyo3_async_runtimes::tokio::get_current_locals(py)?;
|
||||
Ok(self.spawn(pyo3_async_runtimes::tokio::scope(locals, future)))
|
||||
}
|
||||
}
|
||||
|
||||
#[ext(pub, name = TokioMpscSenderExt)]
|
||||
impl<T> mpsc::Sender<T> {
|
||||
/// Sends a value, waiting until there is capacity.
|
||||
///
|
||||
/// A successful send occurs when it is determined that the other end of the
|
||||
/// channel has not hung up already. An unsuccessful send would be one where
|
||||
/// the corresponding receiver has already been closed.
|
||||
async fn send_py(&self, value: T) -> PyResult<()> {
|
||||
self.send(value)
|
||||
.await
|
||||
.map_err(|_| PyErr::receiver_channel_closed())
|
||||
}
|
||||
}
|
||||
|
||||
#[ext(pub, name = TokioMpscReceiverExt)]
|
||||
impl<T> mpsc::Receiver<T> {
|
||||
/// Receives the next value for this receiver.
|
||||
async fn recv_py(&mut self) -> PyResult<T> {
|
||||
self.recv().await.ok_or_else(PyErr::receiver_channel_closed)
|
||||
}
|
||||
|
||||
/// Receives at most `limit` values for this receiver and returns them.
|
||||
///
|
||||
/// For `limit = 0`, an empty collection of messages will be returned immediately.
|
||||
/// For `limit > 0`, if there are no messages in the channel's queue this method
|
||||
/// will sleep until a message is sent.
|
||||
async fn recv_many_py(&mut self, limit: usize) -> PyResult<Vec<T>> {
|
||||
// get updates from receiver channel
|
||||
let mut updates = Vec::with_capacity(limit);
|
||||
let received = self.recv_many(&mut updates, limit).await;
|
||||
|
||||
// if we received zero items, then the channel was unexpectedly closed
|
||||
if limit != 0 && received == 0 {
|
||||
return Err(PyErr::receiver_channel_closed());
|
||||
}
|
||||
|
||||
Ok(updates)
|
||||
}
|
||||
|
||||
/// Tries to receive the next value for this receiver.
|
||||
fn try_recv_py(&mut self) -> PyResult<Option<T>> {
|
||||
match self.try_recv() {
|
||||
Ok(v) => Ok(Some(v)),
|
||||
Err(TryRecvError::Empty) => Ok(None),
|
||||
Err(TryRecvError::Disconnected) => Err(PyErr::receiver_channel_closed()),
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/// A Python module implemented in Rust. The name of this function must match
|
||||
@@ -176,16 +65,9 @@ fn main_module(m: &Bound<'_, PyModule>) -> PyResult<()> {
|
||||
// install logger
|
||||
pyo3_log::init();
|
||||
|
||||
// TODO: for now this is all NOT a submodule, but figure out how to make the submodule system
|
||||
// work with maturin, where the types generate correctly, in the right folder, without
|
||||
// too many importing issues...
|
||||
ident_submodule(m)?;
|
||||
multiaddr_submodule(m)?;
|
||||
networking_submodule(m)?;
|
||||
|
||||
// top-level constructs
|
||||
// TODO: ...
|
||||
|
||||
Ok(())
|
||||
}
|
||||
|
||||
|
||||
@@ -1,27 +1,20 @@
|
||||
#![allow(
|
||||
clippy::multiple_inherent_impl,
|
||||
clippy::unnecessary_wraps,
|
||||
clippy::unused_self,
|
||||
clippy::needless_pass_by_value
|
||||
)]
|
||||
|
||||
use crate::r#const::MPSC_CHANNEL_SIZE;
|
||||
use crate::ext::{ByteArrayExt as _, FutureExt, PyErrExt as _};
|
||||
use crate::ext::{ResultExt as _, TokioMpscReceiverExt as _, TokioMpscSenderExt as _};
|
||||
use crate::ext::ResultExt as _;
|
||||
use crate::ext::{ByteArrayExt as _, FutureExt as _};
|
||||
use crate::ident::PyKeypair;
|
||||
use crate::networking::exception::{PyAllQueuesFullError, PyNoPeersSubscribedToTopicError};
|
||||
use crate::pyclass;
|
||||
use crate::pylibp2p::ident::{PyKeypair, PyPeerId};
|
||||
use libp2p::futures::StreamExt as _;
|
||||
use libp2p::gossipsub;
|
||||
use libp2p::gossipsub::{IdentTopic, Message, MessageId, PublishError};
|
||||
use libp2p::swarm::SwarmEvent;
|
||||
use networking::discovery;
|
||||
use networking::swarm::create_swarm;
|
||||
use pyo3::prelude::{PyModule, PyModuleMethods as _};
|
||||
use futures_lite::FutureExt as _;
|
||||
use networking::swarm::{FromSwarm, Swarm, ToSwarm};
|
||||
use pyo3::coroutine::CancelHandle;
|
||||
use pyo3::exceptions::{PyConnectionError, PyRuntimeError};
|
||||
use pyo3::prelude::*;
|
||||
use pyo3::types::PyBytes;
|
||||
use pyo3::{Bound, Py, PyErr, PyResult, PyTraverseError, PyVisit, Python, pymethods};
|
||||
use pyo3_stub_gen::derive::{gen_stub_pyclass, gen_stub_pyclass_enum, gen_stub_pymethods};
|
||||
use std::net::IpAddr;
|
||||
use tokio::sync::{Mutex, mpsc, oneshot};
|
||||
use pyo3_async_runtimes::tokio::get_runtime;
|
||||
use pyo3_stub_gen::derive::{gen_stub_pyclass, gen_stub_pyclass_complex_enum, gen_stub_pymethods};
|
||||
use std::pin::pin;
|
||||
use std::sync::Arc;
|
||||
use tokio::sync::{Mutex, mpsc};
|
||||
|
||||
mod exception {
|
||||
use pyo3::types::PyTuple;
|
||||
@@ -49,16 +42,11 @@ mod exception {
|
||||
#[pymethods]
|
||||
impl PyNoPeersSubscribedToTopicError {
|
||||
#[new]
|
||||
#[pyo3(signature = (*args))]
|
||||
#[allow(unused_variables)]
|
||||
pub(crate) fn new(args: &Bound<'_, PyTuple>) -> Self {
|
||||
#[pyo3(signature = (*_a))]
|
||||
pub(crate) fn new(_a: &Bound<'_, PyTuple>) -> Self {
|
||||
Self {}
|
||||
}
|
||||
|
||||
fn __repr__(&self) -> String {
|
||||
format!("PeerId(\"{}\")", Self::MSG)
|
||||
}
|
||||
|
||||
fn __str__(&self) -> String {
|
||||
Self::MSG.to_string()
|
||||
}
|
||||
@@ -84,488 +72,179 @@ mod exception {
|
||||
#[pymethods]
|
||||
impl PyAllQueuesFullError {
|
||||
#[new]
|
||||
#[pyo3(signature = (*args))]
|
||||
#[allow(unused_variables)]
|
||||
pub(crate) fn new(args: &Bound<'_, PyTuple>) -> Self {
|
||||
#[pyo3(signature = (*_a))]
|
||||
pub(crate) fn new(_a: &Bound<'_, PyTuple>) -> Self {
|
||||
Self {}
|
||||
}
|
||||
|
||||
fn __repr__(&self) -> String {
|
||||
format!("PeerId(\"{}\")", Self::MSG)
|
||||
}
|
||||
|
||||
fn __str__(&self) -> String {
|
||||
Self::MSG.to_string()
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/// Connection or disconnection event discriminant type.
|
||||
#[gen_stub_pyclass_enum]
|
||||
#[pyclass(eq, eq_int, name = "ConnectionUpdateType")]
|
||||
#[derive(Debug, Clone, PartialEq)]
|
||||
enum PyConnectionUpdateType {
|
||||
Connected = 0,
|
||||
Disconnected,
|
||||
}
|
||||
|
||||
#[gen_stub_pyclass]
|
||||
#[pyclass(frozen, name = "ConnectionUpdate")]
|
||||
#[derive(Debug, Clone)]
|
||||
struct PyConnectionUpdate {
|
||||
/// Whether this is a connection or disconnection event
|
||||
#[pyo3(get)]
|
||||
update_type: PyConnectionUpdateType,
|
||||
|
||||
/// Identity of the peer that we have connected to or disconnected from.
|
||||
#[pyo3(get)]
|
||||
peer_id: PyPeerId,
|
||||
|
||||
/// Remote connection's IPv4 address.
|
||||
#[pyo3(get)]
|
||||
remote_ipv4: String,
|
||||
|
||||
/// Remote connection's TCP port.
|
||||
#[pyo3(get)]
|
||||
remote_tcp_port: u16,
|
||||
#[pyclass]
|
||||
struct PySwarm {
|
||||
swarm: Arc<Mutex<Swarm>>,
|
||||
from_swarm: Mutex<mpsc::Receiver<FromSwarm>>,
|
||||
to_swarm: Mutex<mpsc::Sender<ToSwarm>>,
|
||||
}
|
||||
|
||||
enum ToTask {
|
||||
GossipsubSubscribe {
|
||||
topic: String,
|
||||
result_tx: oneshot::Sender<PyResult<bool>>,
|
||||
#[gen_stub_pyclass_complex_enum]
|
||||
#[pyclass]
|
||||
pub enum PyMessage {
|
||||
Connection {
|
||||
node_id: String,
|
||||
connected: bool,
|
||||
},
|
||||
GossipsubUnsubscribe {
|
||||
Gossip {
|
||||
node_id: String,
|
||||
topic: String,
|
||||
result_tx: oneshot::Sender<bool>,
|
||||
},
|
||||
GossipsubPublish {
|
||||
topic: String,
|
||||
data: Vec<u8>,
|
||||
result_tx: oneshot::Sender<PyResult<MessageId>>,
|
||||
data: Py<PyBytes>,
|
||||
},
|
||||
}
|
||||
|
||||
#[allow(clippy::enum_glob_use)]
|
||||
async fn networking_task(
|
||||
mut swarm: networking::swarm::Swarm,
|
||||
mut to_task_rx: mpsc::Receiver<ToTask>,
|
||||
connection_update_tx: mpsc::Sender<PyConnectionUpdate>,
|
||||
gossipsub_message_tx: mpsc::Sender<(String, Vec<u8>)>,
|
||||
) {
|
||||
use SwarmEvent::*;
|
||||
use ToTask::*;
|
||||
use networking::swarm::BehaviourEvent::*;
|
||||
|
||||
log::info!("RUST: networking task started");
|
||||
|
||||
loop {
|
||||
tokio::select! {
|
||||
message = to_task_rx.recv() => {
|
||||
// handle closed channel
|
||||
let Some(message) = message else {
|
||||
log::info!("RUST: channel closed");
|
||||
break;
|
||||
};
|
||||
|
||||
// dispatch incoming messages
|
||||
match message {
|
||||
GossipsubSubscribe { topic, result_tx } => {
|
||||
// try to subscribe
|
||||
let result = swarm.behaviour_mut()
|
||||
.gossipsub.subscribe(&IdentTopic::new(topic));
|
||||
|
||||
// send response oneshot
|
||||
if let Err(e) = result_tx.send(result.pyerr()) {
|
||||
log::error!("RUST: could not subscribe to gossipsub topic since channel already closed: {e:?}");
|
||||
continue;
|
||||
}
|
||||
}
|
||||
GossipsubUnsubscribe { topic, result_tx } => {
|
||||
// try to unsubscribe from the topic
|
||||
let result = swarm.behaviour_mut()
|
||||
.gossipsub.unsubscribe(&IdentTopic::new(topic));
|
||||
|
||||
// send response oneshot (or exit if connection closed)
|
||||
if let Err(e) = result_tx.send(result) {
|
||||
log::error!("RUST: could not unsubscribe from gossipsub topic since channel already closed: {e:?}");
|
||||
continue;
|
||||
}
|
||||
}
|
||||
GossipsubPublish { topic, data, result_tx } => {
|
||||
// try to publish the data -> catch NoPeersSubscribedToTopic error & convert to correct exception
|
||||
let result = swarm.behaviour_mut().gossipsub.publish(
|
||||
IdentTopic::new(topic), data);
|
||||
let pyresult: PyResult<MessageId> = if let Err(PublishError::NoPeersSubscribedToTopic) = result {
|
||||
Err(exception::PyNoPeersSubscribedToTopicError::new_err())
|
||||
} else if let Err(PublishError::AllQueuesFull(_)) = result {
|
||||
Err(exception::PyAllQueuesFullError::new_err())
|
||||
} else {
|
||||
result.pyerr()
|
||||
};
|
||||
|
||||
// send response oneshot (or exit if connection closed)
|
||||
if let Err(e) = result_tx.send(pyresult) {
|
||||
log::error!("RUST: could not publish gossipsub message since channel already closed: {e:?}");
|
||||
continue;
|
||||
}
|
||||
}
|
||||
impl TryFrom<FromSwarm> for PyMessage {
|
||||
type Error = PyErr;
|
||||
fn try_from(value: FromSwarm) -> Result<Self, Self::Error> {
|
||||
match value {
|
||||
FromSwarm::Discovered(nid) => Ok(PyMessage::Connection {
|
||||
node_id: nid.to_base58(),
|
||||
connected: true,
|
||||
}),
|
||||
FromSwarm::Expired(nid) => Ok(PyMessage::Connection {
|
||||
node_id: nid.to_base58(),
|
||||
connected: false,
|
||||
}),
|
||||
FromSwarm::Message(nid, topic, data) => Ok(PyMessage::Gossip {
|
||||
node_id: nid.to_base58(),
|
||||
topic,
|
||||
data: data.pybytes(),
|
||||
}),
|
||||
FromSwarm::PublishError(e) => match e {
|
||||
libp2p::gossipsub::PublishError::NoPeersSubscribedToTopic => {
|
||||
Err(PyNoPeersSubscribedToTopicError::new_err())
|
||||
}
|
||||
}
|
||||
|
||||
// architectural solution to this problem:
|
||||
// create keep_alive behavior who's job it is to dial peers discovered by mDNS (and drop when expired)
|
||||
// -> it will emmit TRUE connected/disconnected events consumable elsewhere
|
||||
//
|
||||
// gossipsub will feed off-of dial attempts created by networking, and that will bootstrap its' peers list
|
||||
// then for actual communication it will dial those peers if need-be
|
||||
swarm_event = swarm.select_next_some() => {
|
||||
match swarm_event {
|
||||
Behaviour(Gossipsub(gossipsub::Event::Message {
|
||||
message: Message {
|
||||
topic,
|
||||
data,
|
||||
..
|
||||
},
|
||||
..
|
||||
})) => {
|
||||
// topic-ID is just the topic hash!!! (since we used identity hasher)
|
||||
let message = (topic.into_string(), data);
|
||||
|
||||
// send incoming message to channel (or exit if connection closed)
|
||||
if let Err(e) = gossipsub_message_tx.send(message).await {
|
||||
log::error!("RUST: could not send incoming gossipsub message since channel already closed: {e}");
|
||||
continue;
|
||||
}
|
||||
},
|
||||
Behaviour(Discovery(discovery::Event::ConnectionEstablished { peer_id, remote_ip, remote_tcp_port, .. })) => {
|
||||
// grab IPv4 string
|
||||
let remote_ipv4 = match remote_ip {
|
||||
IpAddr::V4(ip) => ip.to_string(),
|
||||
IpAddr::V6(ip) => {
|
||||
log::warn!("RUST: ignoring connection to IPv6 address: {ip}");
|
||||
continue;
|
||||
}
|
||||
};
|
||||
|
||||
// send connection event to channel (or exit if connection closed)
|
||||
if let Err(e) = connection_update_tx.send(PyConnectionUpdate {
|
||||
update_type: PyConnectionUpdateType::Connected,
|
||||
peer_id: PyPeerId(peer_id),
|
||||
remote_ipv4,
|
||||
remote_tcp_port,
|
||||
}).await {
|
||||
log::error!("RUST: could not send connection update since channel already closed: {e}");
|
||||
continue;
|
||||
}
|
||||
},
|
||||
Behaviour(Discovery(discovery::Event::ConnectionClosed { peer_id, remote_ip, remote_tcp_port, .. })) => {
|
||||
// grab IPv4 string
|
||||
let remote_ipv4 = match remote_ip {
|
||||
IpAddr::V4(ip) => ip.to_string(),
|
||||
IpAddr::V6(ip) => {
|
||||
log::warn!("RUST: ignoring disconnection from IPv6 address: {ip}");
|
||||
continue;
|
||||
}
|
||||
};
|
||||
|
||||
// send disconnection event to channel (or exit if connection closed)
|
||||
if let Err(e) = connection_update_tx.send(PyConnectionUpdate {
|
||||
update_type: PyConnectionUpdateType::Disconnected,
|
||||
peer_id: PyPeerId(peer_id),
|
||||
remote_ipv4,
|
||||
remote_tcp_port,
|
||||
}).await {
|
||||
log::error!("RUST: could not send connection update since channel already closed: {e}");
|
||||
continue;
|
||||
}
|
||||
},
|
||||
e => {
|
||||
log::info!("RUST: other event {e:?}");
|
||||
}
|
||||
libp2p::gossipsub::PublishError::AllQueuesFull(_) => {
|
||||
Err(PyAllQueuesFullError::new_err())
|
||||
}
|
||||
}
|
||||
e => Err(PyRuntimeError::new_err(e.to_string())),
|
||||
},
|
||||
}
|
||||
}
|
||||
|
||||
log::info!("RUST: networking task stopped");
|
||||
}
|
||||
|
||||
#[gen_stub_pyclass]
|
||||
#[pyclass(name = "NetworkingHandle")]
|
||||
#[derive(Debug)]
|
||||
struct PyNetworkingHandle {
|
||||
// channels
|
||||
to_task_tx: Option<mpsc::Sender<ToTask>>,
|
||||
connection_update_rx: Mutex<mpsc::Receiver<PyConnectionUpdate>>,
|
||||
gossipsub_message_rx: Mutex<mpsc::Receiver<(String, Vec<u8>)>>,
|
||||
}
|
||||
|
||||
impl Drop for PyNetworkingHandle {
|
||||
fn drop(&mut self) {
|
||||
// TODO: may or may not need to await a "kill-signal" oneshot channel message,
|
||||
// to ensure that the networking task is done BEFORE exiting the clear function...
|
||||
// but this may require GIL?? and it may not be safe to call GIL here??
|
||||
self.to_task_tx = None; // Using Option<T> as a trick to force channel to be dropped
|
||||
}
|
||||
}
|
||||
|
||||
#[allow(clippy::expect_used)]
|
||||
impl PyNetworkingHandle {
|
||||
fn new(
|
||||
to_task_tx: mpsc::Sender<ToTask>,
|
||||
connection_update_rx: mpsc::Receiver<PyConnectionUpdate>,
|
||||
gossipsub_message_rx: mpsc::Receiver<(String, Vec<u8>)>,
|
||||
) -> Self {
|
||||
Self {
|
||||
to_task_tx: Some(to_task_tx),
|
||||
connection_update_rx: Mutex::new(connection_update_rx),
|
||||
gossipsub_message_rx: Mutex::new(gossipsub_message_rx),
|
||||
}
|
||||
}
|
||||
|
||||
const fn to_task_tx(&self) -> &mpsc::Sender<ToTask> {
|
||||
self.to_task_tx
|
||||
.as_ref()
|
||||
.expect("The sender should only be None after de-initialization.")
|
||||
}
|
||||
}
|
||||
|
||||
#[gen_stub_pymethods]
|
||||
#[pymethods]
|
||||
impl PyNetworkingHandle {
|
||||
// NOTE: `async fn`s here that use `.await` will wrap the future in `.allow_threads_py()`
|
||||
// immediately beforehand to release the interpreter.
|
||||
// SEE: https://pyo3.rs/v0.26.0/async-await.html#detaching-from-the-interpreter-across-await
|
||||
|
||||
// ---- Lifecycle management methods ----
|
||||
|
||||
impl PySwarm {
|
||||
#[new]
|
||||
fn py_new(identity: Bound<'_, PyKeypair>) -> PyResult<Self> {
|
||||
use pyo3_async_runtimes::tokio::get_runtime;
|
||||
|
||||
// create communication channels
|
||||
let (to_task_tx, to_task_rx) = mpsc::channel(MPSC_CHANNEL_SIZE);
|
||||
let (connection_update_tx, connection_update_rx) = mpsc::channel(MPSC_CHANNEL_SIZE);
|
||||
let (gossipsub_message_tx, gossipsub_message_rx) = mpsc::channel(MPSC_CHANNEL_SIZE);
|
||||
|
||||
// get identity
|
||||
let identity = identity.borrow().0.clone();
|
||||
|
||||
let (to_swarm, from_client) = mpsc::channel(MPSC_CHANNEL_SIZE);
|
||||
let (to_client, from_swarm) = mpsc::channel(MPSC_CHANNEL_SIZE);
|
||||
// create networking swarm (within tokio context!! or it crashes)
|
||||
let swarm = get_runtime()
|
||||
.block_on(async { create_swarm(identity) })
|
||||
.block_on(async { Swarm::new(identity, from_client, to_client) })
|
||||
.pyerr()?;
|
||||
|
||||
// spawn tokio task running the networking logic
|
||||
get_runtime().spawn(async move {
|
||||
networking_task(
|
||||
swarm,
|
||||
to_task_rx,
|
||||
connection_update_tx,
|
||||
gossipsub_message_tx,
|
||||
)
|
||||
.await;
|
||||
Ok(Self {
|
||||
swarm: Arc::new(Mutex::new(swarm)),
|
||||
from_swarm: Mutex::new(from_swarm),
|
||||
to_swarm: Mutex::new(to_swarm),
|
||||
})
|
||||
}
|
||||
|
||||
#[gen_stub(skip)]
|
||||
async fn run(&self, #[pyo3(cancel_handle)] mut cancel: CancelHandle) -> PyResult<()> {
|
||||
let copy = Arc::clone(&self.swarm);
|
||||
let jh = get_runtime().spawn(async move {
|
||||
copy.try_lock()
|
||||
.expect("tried to run swarm twice")
|
||||
.run()
|
||||
.await
|
||||
});
|
||||
Ok(Self::new(
|
||||
to_task_tx,
|
||||
connection_update_rx,
|
||||
gossipsub_message_rx,
|
||||
))
|
||||
}
|
||||
|
||||
#[gen_stub(skip)]
|
||||
const fn __traverse__(&self, _visit: PyVisit<'_>) -> Result<(), PyTraverseError> {
|
||||
Ok(()) // This is needed purely so `__clear__` can work
|
||||
}
|
||||
|
||||
#[gen_stub(skip)]
|
||||
fn __clear__(&mut self) {
|
||||
// TODO: may or may not need to await a "kill-signal" oneshot channel message,
|
||||
// to ensure that the networking task is done BEFORE exiting the clear function...
|
||||
// but this may require GIL?? and it may not be safe to call GIL here??
|
||||
self.to_task_tx = None; // Using Option<T> as a trick to force channel to be dropped
|
||||
jh.or(async {
|
||||
cancel.cancelled().await;
|
||||
Ok(())
|
||||
})
|
||||
.await
|
||||
.map_err(|e| PyRuntimeError::new_err(e.to_string()))
|
||||
}
|
||||
|
||||
// ---- Connection update receiver methods ----
|
||||
|
||||
/// Receives the next `ConnectionUpdate` from networking.
|
||||
async fn connection_update_recv(&self) -> PyResult<PyConnectionUpdate> {
|
||||
self.connection_update_rx
|
||||
.lock()
|
||||
.allow_threads_py() // allow-threads-aware async call
|
||||
.await
|
||||
.recv_py()
|
||||
.allow_threads_py() // allow-threads-aware async call
|
||||
.await
|
||||
/// Receives the next message from networking.
|
||||
async fn recv(&self) -> PyResult<PyMessage> {
|
||||
let msg = pin!(
|
||||
self.from_swarm
|
||||
.try_lock()
|
||||
.expect("called recv concurrently")
|
||||
.recv()
|
||||
)
|
||||
.allow_threads_py()
|
||||
.await;
|
||||
match msg {
|
||||
None => Err(PyConnectionError::new_err("swarm closed")),
|
||||
Some(msg) => msg.try_into(),
|
||||
}
|
||||
}
|
||||
|
||||
/// Receives at most `limit` `ConnectionUpdate`s from networking and returns them.
|
||||
///
|
||||
/// For `limit = 0`, an empty collection of `ConnectionUpdate`s will be returned immediately.
|
||||
/// For `limit > 0`, if there are no `ConnectionUpdate`s in the channel's queue this method
|
||||
/// will sleep until a `ConnectionUpdate`s is sent.
|
||||
async fn connection_update_recv_many(&self, limit: usize) -> PyResult<Vec<PyConnectionUpdate>> {
|
||||
self.connection_update_rx
|
||||
.lock()
|
||||
.allow_threads_py() // allow-threads-aware async call
|
||||
.await
|
||||
.recv_many_py(limit)
|
||||
.allow_threads_py() // allow-threads-aware async call
|
||||
.await
|
||||
}
|
||||
|
||||
// TODO: rn this blocks main thread if anything else is awaiting the channel (bc its a mutex)
|
||||
// so its too dangerous to expose just yet. figure out a better semantics for handling this,
|
||||
// so things don't randomly block
|
||||
// /// Tries to receive the next `ConnectionUpdate` from networking.
|
||||
// fn connection_update_try_recv(&self) -> PyResult<Option<PyConnectionUpdate>> {
|
||||
// self.connection_update_rx.blocking_lock().try_recv_py()
|
||||
// }
|
||||
//
|
||||
// /// Checks if the `ConnectionUpdate` channel is empty.
|
||||
// fn connection_update_is_empty(&self) -> bool {
|
||||
// self.connection_update_rx.blocking_lock().is_empty()
|
||||
// }
|
||||
//
|
||||
// /// Returns the number of `ConnectionUpdate`s in the channel.
|
||||
// fn connection_update_len(&self) -> usize {
|
||||
// self.connection_update_rx.blocking_lock().len()
|
||||
// }
|
||||
|
||||
// ---- Gossipsub management methods ----
|
||||
|
||||
/// Subscribe to a `GossipSub` topic.
|
||||
///
|
||||
/// Returns `True` if the subscription worked. Returns `False` if we were already subscribed.
|
||||
async fn gossipsub_subscribe(&self, topic: String) -> PyResult<bool> {
|
||||
let (tx, rx) = oneshot::channel();
|
||||
|
||||
async fn gossipsub_subscribe(&self, topic: String) -> PyResult<()> {
|
||||
// send off request to subscribe
|
||||
self.to_task_tx()
|
||||
.send_py(ToTask::GossipsubSubscribe {
|
||||
topic,
|
||||
result_tx: tx,
|
||||
})
|
||||
.allow_threads_py() // allow-threads-aware async call
|
||||
.await?;
|
||||
|
||||
// wait for response & return any errors
|
||||
rx.allow_threads_py() // allow-threads-aware async call
|
||||
.await
|
||||
.map_err(|_| PyErr::receiver_channel_closed())?
|
||||
pin!(
|
||||
self.to_swarm
|
||||
.try_lock()
|
||||
.expect("called send concurrently")
|
||||
.send(ToSwarm::Subscribe(topic))
|
||||
)
|
||||
.allow_threads_py() // allow-threads-aware async call
|
||||
.await
|
||||
.map_err(|_| PyConnectionError::new_err("swarm closed"))
|
||||
}
|
||||
|
||||
/// Unsubscribes from a `GossipSub` topic.
|
||||
///
|
||||
/// Returns `True` if we were subscribed to this topic. Returns `False` if we were not subscribed.
|
||||
async fn gossipsub_unsubscribe(&self, topic: String) -> PyResult<bool> {
|
||||
let (tx, rx) = oneshot::channel();
|
||||
|
||||
async fn gossipsub_unsubscribe(&self, topic: String) -> PyResult<()> {
|
||||
// send off request to unsubscribe
|
||||
self.to_task_tx()
|
||||
.send_py(ToTask::GossipsubUnsubscribe {
|
||||
topic,
|
||||
result_tx: tx,
|
||||
})
|
||||
.allow_threads_py() // allow-threads-aware async call
|
||||
.await?;
|
||||
|
||||
// wait for response & convert any errors
|
||||
rx.allow_threads_py() // allow-threads-aware async call
|
||||
.await
|
||||
.map_err(|_| PyErr::receiver_channel_closed())
|
||||
pin!(
|
||||
self.to_swarm
|
||||
.try_lock()
|
||||
.expect("called send concurrently")
|
||||
.send(ToSwarm::Unsubscribe(topic))
|
||||
)
|
||||
.allow_threads_py() // allow-threads-aware async call
|
||||
.await
|
||||
.map_err(|_| PyConnectionError::new_err("swarm closed"))
|
||||
}
|
||||
|
||||
/// Publishes a message with multiple topics to the `GossipSub` network.
|
||||
///
|
||||
/// If no peers are found that subscribe to this topic, throws `NoPeersSubscribedToTopicError` exception.
|
||||
/// Publishes a message to the network on a specific topic.
|
||||
async fn gossipsub_publish(&self, topic: String, data: Py<PyBytes>) -> PyResult<()> {
|
||||
let (tx, rx) = oneshot::channel();
|
||||
|
||||
// send off request to subscribe
|
||||
let data = Python::attach(|py| Vec::from(data.as_bytes(py)));
|
||||
self.to_task_tx()
|
||||
.send_py(ToTask::GossipsubPublish {
|
||||
topic,
|
||||
data,
|
||||
result_tx: tx,
|
||||
})
|
||||
.allow_threads_py() // allow-threads-aware async call
|
||||
.await?;
|
||||
|
||||
// wait for response & return any errors => ignore messageID for now!!!
|
||||
let _ = rx
|
||||
.allow_threads_py() // allow-threads-aware async call
|
||||
.await
|
||||
.map_err(|_| PyErr::receiver_channel_closed())??;
|
||||
Ok(())
|
||||
pin!(
|
||||
self.to_swarm
|
||||
.try_lock()
|
||||
.expect("called send concurrently")
|
||||
.send(ToSwarm::Message(topic, data))
|
||||
)
|
||||
.allow_threads_py() // allow-threads-aware async call
|
||||
.await
|
||||
.map_err(|_| PyConnectionError::new_err("swarm closed"))
|
||||
}
|
||||
|
||||
// ---- Gossipsub message receiver methods ----
|
||||
|
||||
/// Receives the next message from the `GossipSub` network.
|
||||
async fn gossipsub_recv(&self) -> PyResult<(String, Py<PyBytes>)> {
|
||||
self.gossipsub_message_rx
|
||||
.lock()
|
||||
.allow_threads_py() // allow-threads-aware async call
|
||||
.await
|
||||
.recv_py()
|
||||
.allow_threads_py() // allow-threads-aware async call
|
||||
.await
|
||||
.map(|(t, d)| (t, d.pybytes()))
|
||||
}
|
||||
|
||||
/// Receives at most `limit` messages from the `GossipSub` network and returns them.
|
||||
///
|
||||
/// For `limit = 0`, an empty collection of messages will be returned immediately.
|
||||
/// For `limit > 0`, if there are no messages in the channel's queue this method
|
||||
/// will sleep until a message is sent.
|
||||
async fn gossipsub_recv_many(&self, limit: usize) -> PyResult<Vec<(String, Py<PyBytes>)>> {
|
||||
Ok(self
|
||||
.gossipsub_message_rx
|
||||
.lock()
|
||||
.allow_threads_py() // allow-threads-aware async call
|
||||
.await
|
||||
.recv_many_py(limit)
|
||||
.allow_threads_py() // allow-threads-aware async call
|
||||
.await?
|
||||
.into_iter()
|
||||
.map(|(t, d)| (t, d.pybytes()))
|
||||
.collect())
|
||||
}
|
||||
|
||||
// TODO: rn this blocks main thread if anything else is awaiting the channel (bc its a mutex)
|
||||
// so its too dangerous to expose just yet. figure out a better semantics for handling this,
|
||||
// so things don't randomly block
|
||||
// /// Tries to receive the next message from the `GossipSub` network.
|
||||
// fn gossipsub_try_recv(&self) -> PyResult<Option<(String, Py<PyBytes>)>> {
|
||||
// Ok(self
|
||||
// .gossipsub_message_rx
|
||||
// .blocking_lock()
|
||||
// .try_recv_py()?
|
||||
// .map(|(t, d)| (t, d.pybytes())))
|
||||
// }
|
||||
//
|
||||
// /// Checks if the `GossipSub` message channel is empty.
|
||||
// fn gossipsub_is_empty(&self) -> bool {
|
||||
// self.gossipsub_message_rx.blocking_lock().is_empty()
|
||||
// }
|
||||
//
|
||||
// /// Returns the number of `GossipSub` messages in the channel.
|
||||
// fn gossipsub_len(&self) -> usize {
|
||||
// self.gossipsub_message_rx.blocking_lock().len()
|
||||
// }
|
||||
}
|
||||
|
||||
pub fn networking_submodule(m: &Bound<'_, PyModule>) -> PyResult<()> {
|
||||
m.add_class::<exception::PyNoPeersSubscribedToTopicError>()?;
|
||||
m.add_class::<exception::PyAllQueuesFullError>()?;
|
||||
|
||||
m.add_class::<PyConnectionUpdateType>()?;
|
||||
m.add_class::<PyConnectionUpdate>()?;
|
||||
m.add_class::<PyConnectionUpdateType>()?;
|
||||
m.add_class::<PyNetworkingHandle>()?;
|
||||
m.add_class::<PySwarm>()?;
|
||||
m.add_class::<PyMessage>()?;
|
||||
|
||||
Ok(())
|
||||
}
|
||||
|
||||
@@ -1,159 +0,0 @@
|
||||
use crate::ext::ResultExt as _;
|
||||
use libp2p::PeerId;
|
||||
use libp2p::identity::Keypair;
|
||||
use pyo3::prelude::{PyBytesMethods as _, PyModule, PyModuleMethods as _};
|
||||
use pyo3::types::PyBytes;
|
||||
use pyo3::{Bound, PyResult, Python, pyclass, pymethods};
|
||||
use pyo3_stub_gen::derive::{gen_stub_pyclass, gen_stub_pymethods};
|
||||
|
||||
/// Identity keypair of a node.
|
||||
#[gen_stub_pyclass]
|
||||
#[pyclass(name = "Keypair", frozen)]
|
||||
#[repr(transparent)]
|
||||
pub struct PyKeypair(pub Keypair);
|
||||
|
||||
#[gen_stub_pymethods]
|
||||
#[pymethods]
|
||||
#[allow(clippy::needless_pass_by_value)]
|
||||
impl PyKeypair {
|
||||
/// Generate a new Ed25519 keypair.
|
||||
#[staticmethod]
|
||||
fn generate_ed25519() -> Self {
|
||||
Self(Keypair::generate_ed25519())
|
||||
}
|
||||
|
||||
/// Generate a new ECDSA keypair.
|
||||
#[staticmethod]
|
||||
fn generate_ecdsa() -> Self {
|
||||
Self(Keypair::generate_ecdsa())
|
||||
}
|
||||
|
||||
/// Generate a new Secp256k1 keypair.
|
||||
#[staticmethod]
|
||||
fn generate_secp256k1() -> Self {
|
||||
Self(Keypair::generate_secp256k1())
|
||||
}
|
||||
|
||||
/// Decode a private key from a protobuf structure and parse it as a `Keypair`.
|
||||
#[staticmethod]
|
||||
fn from_protobuf_encoding(bytes: Bound<'_, PyBytes>) -> PyResult<Self> {
|
||||
let bytes = Vec::from(bytes.as_bytes());
|
||||
Ok(Self(Keypair::from_protobuf_encoding(&bytes).pyerr()?))
|
||||
}
|
||||
|
||||
/// Decode an keypair from a DER-encoded secret key in PKCS#8 `PrivateKeyInfo`
|
||||
/// format (i.e. unencrypted) as defined in [RFC5208].
|
||||
///
|
||||
/// [RFC5208]: https://tools.ietf.org/html/rfc5208#section-5
|
||||
#[staticmethod]
|
||||
fn rsa_from_pkcs8(bytes: Bound<'_, PyBytes>) -> PyResult<Self> {
|
||||
let mut bytes = Vec::from(bytes.as_bytes());
|
||||
Ok(Self(Keypair::rsa_from_pkcs8(&mut bytes).pyerr()?))
|
||||
}
|
||||
|
||||
/// Decode a keypair from a DER-encoded Secp256k1 secret key in an `ECPrivateKey`
|
||||
/// structure as defined in [RFC5915].
|
||||
///
|
||||
/// [RFC5915]: https://tools.ietf.org/html/rfc5915
|
||||
#[staticmethod]
|
||||
fn secp256k1_from_der(bytes: Bound<'_, PyBytes>) -> PyResult<Self> {
|
||||
let mut bytes = Vec::from(bytes.as_bytes());
|
||||
Ok(Self(Keypair::secp256k1_from_der(&mut bytes).pyerr()?))
|
||||
}
|
||||
|
||||
#[staticmethod]
|
||||
fn ed25519_from_bytes(bytes: Bound<'_, PyBytes>) -> PyResult<Self> {
|
||||
let mut bytes = Vec::from(bytes.as_bytes());
|
||||
Ok(Self(Keypair::ed25519_from_bytes(&mut bytes).pyerr()?))
|
||||
}
|
||||
|
||||
/// Encode a private key as protobuf structure.
|
||||
fn to_protobuf_encoding<'py>(&self, py: Python<'py>) -> PyResult<Bound<'py, PyBytes>> {
|
||||
let bytes = self.0.to_protobuf_encoding().pyerr()?;
|
||||
Ok(PyBytes::new(py, &bytes))
|
||||
}
|
||||
|
||||
/// Convert the `Keypair` into the corresponding `PeerId`.
|
||||
fn to_peer_id(&self) -> PyPeerId {
|
||||
PyPeerId(self.0.public().to_peer_id())
|
||||
}
|
||||
|
||||
// /// Hidden constructor for pickling support. TODO: figure out how to do pickling...
|
||||
// #[gen_stub(skip)]
|
||||
// #[new]
|
||||
// fn py_new(bytes: Bound<'_, PyBytes>) -> PyResult<Self> {
|
||||
// Self::from_protobuf_encoding(bytes)
|
||||
// }
|
||||
//
|
||||
// #[gen_stub(skip)]
|
||||
// fn __setstate__(&mut self, state: Bound<'_, PyBytes>) -> PyResult<()> {
|
||||
// *self = Self::from_protobuf_encoding(state)?;
|
||||
// Ok(())
|
||||
// }
|
||||
//
|
||||
// #[gen_stub(skip)]
|
||||
// fn __getstate__<'py>(&self, py: Python<'py>) -> PyResult<Bound<'py, PyBytes>> {
|
||||
// self.to_protobuf_encoding(py)
|
||||
// }
|
||||
//
|
||||
// #[gen_stub(skip)]
|
||||
// pub fn __getnewargs__<'py>(&self, py: Python<'py>) -> PyResult<(Bound<'py, PyBytes>,)> {
|
||||
// Ok((self.to_protobuf_encoding(py)?,))
|
||||
// }
|
||||
}
|
||||
|
||||
/// Identifier of a peer of the network.
|
||||
///
|
||||
/// The data is a `CIDv0` compatible multihash of the protobuf encoded public key of the peer
|
||||
/// as specified in [specs/peer-ids](https://github.com/libp2p/specs/blob/master/peer-ids/peer-ids.md).
|
||||
#[gen_stub_pyclass]
|
||||
#[pyclass(name = "PeerId", frozen)]
|
||||
#[derive(Debug, Clone)]
|
||||
#[repr(transparent)]
|
||||
pub struct PyPeerId(pub PeerId);
|
||||
|
||||
#[gen_stub_pymethods]
|
||||
#[pymethods]
|
||||
#[allow(clippy::needless_pass_by_value)]
|
||||
impl PyPeerId {
|
||||
/// Generates a random peer ID from a cryptographically secure PRNG.
|
||||
///
|
||||
/// This is useful for randomly walking on a DHT, or for testing purposes.
|
||||
#[staticmethod]
|
||||
fn random() -> Self {
|
||||
Self(PeerId::random())
|
||||
}
|
||||
|
||||
/// Parses a `PeerId` from bytes.
|
||||
#[staticmethod]
|
||||
fn from_bytes(bytes: Bound<'_, PyBytes>) -> PyResult<Self> {
|
||||
let bytes = Vec::from(bytes.as_bytes());
|
||||
Ok(Self(PeerId::from_bytes(&bytes).pyerr()?))
|
||||
}
|
||||
|
||||
/// Returns a raw bytes representation of this `PeerId`.
|
||||
fn to_bytes<'py>(&self, py: Python<'py>) -> Bound<'py, PyBytes> {
|
||||
let bytes = self.0.to_bytes();
|
||||
PyBytes::new(py, &bytes)
|
||||
}
|
||||
|
||||
/// Returns a base-58 encoded string of this `PeerId`.
|
||||
fn to_base58(&self) -> String {
|
||||
self.0.to_base58()
|
||||
}
|
||||
|
||||
fn __repr__(&self) -> String {
|
||||
format!("PeerId({})", self.to_base58())
|
||||
}
|
||||
|
||||
fn __str__(&self) -> String {
|
||||
self.to_base58()
|
||||
}
|
||||
}
|
||||
|
||||
pub fn ident_submodule(m: &Bound<'_, PyModule>) -> PyResult<()> {
|
||||
m.add_class::<PyKeypair>()?;
|
||||
m.add_class::<PyPeerId>()?;
|
||||
|
||||
Ok(())
|
||||
}
|
||||
@@ -1,8 +0,0 @@
|
||||
//! A module for exposing Rust's libp2p datatypes over Pyo3
|
||||
//!
|
||||
//! TODO: right now we are coupled to libp2p's identity, but eventually we want to create our own
|
||||
//! independent identity type of some kind or another. This may require handshaking.
|
||||
//!
|
||||
|
||||
pub mod ident;
|
||||
pub mod multiaddr;
|
||||
@@ -1,81 +0,0 @@
|
||||
use crate::ext::ResultExt as _;
|
||||
use libp2p::Multiaddr;
|
||||
use pyo3::prelude::{PyBytesMethods as _, PyModule, PyModuleMethods as _};
|
||||
use pyo3::types::PyBytes;
|
||||
use pyo3::{Bound, PyResult, Python, pyclass, pymethods};
|
||||
use pyo3_stub_gen::derive::{gen_stub_pyclass, gen_stub_pymethods};
|
||||
use std::str::FromStr as _;
|
||||
|
||||
/// Representation of a Multiaddr.
|
||||
#[gen_stub_pyclass]
|
||||
#[pyclass(name = "Multiaddr", frozen)]
|
||||
#[derive(Debug, Clone)]
|
||||
#[repr(transparent)]
|
||||
pub struct PyMultiaddr(pub Multiaddr);
|
||||
|
||||
#[gen_stub_pymethods]
|
||||
#[pymethods]
|
||||
#[allow(clippy::needless_pass_by_value)]
|
||||
impl PyMultiaddr {
|
||||
/// Create a new, empty multiaddress.
|
||||
#[staticmethod]
|
||||
fn empty() -> Self {
|
||||
Self(Multiaddr::empty())
|
||||
}
|
||||
|
||||
/// Create a new, empty multiaddress with the given capacity.
|
||||
#[staticmethod]
|
||||
fn with_capacity(n: usize) -> Self {
|
||||
Self(Multiaddr::with_capacity(n))
|
||||
}
|
||||
|
||||
/// Parse a `Multiaddr` value from its byte slice representation.
|
||||
#[staticmethod]
|
||||
fn from_bytes(bytes: Bound<'_, PyBytes>) -> PyResult<Self> {
|
||||
let bytes = Vec::from(bytes.as_bytes());
|
||||
Ok(Self(Multiaddr::try_from(bytes).pyerr()?))
|
||||
}
|
||||
|
||||
/// Parse a `Multiaddr` value from its string representation.
|
||||
#[staticmethod]
|
||||
fn from_string(string: String) -> PyResult<Self> {
|
||||
Ok(Self(Multiaddr::from_str(&string).pyerr()?))
|
||||
}
|
||||
|
||||
/// Return the length in bytes of this multiaddress.
|
||||
fn len(&self) -> usize {
|
||||
self.0.len()
|
||||
}
|
||||
|
||||
/// Returns true if the length of this multiaddress is 0.
|
||||
fn is_empty(&self) -> bool {
|
||||
self.0.is_empty()
|
||||
}
|
||||
|
||||
/// Return a copy of this [`Multiaddr`]'s byte representation.
|
||||
fn to_bytes<'py>(&self, py: Python<'py>) -> Bound<'py, PyBytes> {
|
||||
let bytes = self.0.to_vec();
|
||||
PyBytes::new(py, &bytes)
|
||||
}
|
||||
|
||||
/// Convert a Multiaddr to a string.
|
||||
fn to_string(&self) -> String {
|
||||
self.0.to_string()
|
||||
}
|
||||
|
||||
#[gen_stub(skip)]
|
||||
fn __repr__(&self) -> String {
|
||||
format!("Multiaddr({})", self.0)
|
||||
}
|
||||
|
||||
#[gen_stub(skip)]
|
||||
fn __str__(&self) -> String {
|
||||
self.to_string()
|
||||
}
|
||||
}
|
||||
|
||||
pub fn multiaddr_submodule(m: &Bound<'_, PyModule>) -> PyResult<()> {
|
||||
m.add_class::<PyMultiaddr>()?;
|
||||
|
||||
Ok(())
|
||||
}
|
||||
@@ -22,7 +22,7 @@ delegate = { workspace = true }
|
||||
|
||||
# async
|
||||
tokio = { workspace = true, features = ["full"] }
|
||||
futures = { workspace = true }
|
||||
futures-lite = { workspace = true }
|
||||
futures-timer = { workspace = true }
|
||||
|
||||
# utility dependencies
|
||||
|
||||
@@ -1,6 +1,6 @@
|
||||
use futures::stream::StreamExt as _;
|
||||
use libp2p::{gossipsub, identity, swarm::SwarmEvent};
|
||||
use networking::{discovery, swarm};
|
||||
use libp2p::identity;
|
||||
use networking::swarm::{FromSwarm, Swarm, ToSwarm};
|
||||
use tokio::sync::mpsc;
|
||||
use tokio::{io, io::AsyncBufReadExt as _, select};
|
||||
use tracing_subscriber::EnvFilter;
|
||||
use tracing_subscriber::filter::LevelFilter;
|
||||
@@ -11,60 +11,50 @@ async fn main() {
|
||||
.with_env_filter(EnvFilter::from_default_env().add_directive(LevelFilter::INFO.into()))
|
||||
.try_init();
|
||||
|
||||
let (to_swarm, from_client) = mpsc::channel(20);
|
||||
let (to_client, mut from_swarm) = mpsc::channel(20);
|
||||
// Configure swarm
|
||||
let mut swarm =
|
||||
swarm::create_swarm(identity::Keypair::generate_ed25519()).expect("Swarm creation failed");
|
||||
let mut swarm = Swarm::new(
|
||||
identity::Keypair::generate_ed25519(),
|
||||
from_client,
|
||||
to_client,
|
||||
)
|
||||
.expect("Swarm creation failed");
|
||||
|
||||
// Create a Gossipsub topic & subscribe
|
||||
let topic = gossipsub::IdentTopic::new("test-net");
|
||||
swarm
|
||||
.behaviour_mut()
|
||||
.gossipsub
|
||||
.subscribe(&topic)
|
||||
.expect("Subscribing to topic failed");
|
||||
_ = to_swarm
|
||||
.send(ToSwarm::Subscribe("test-net".to_owned()))
|
||||
.await;
|
||||
|
||||
// Read full lines from stdin
|
||||
let mut stdin = io::BufReader::new(io::stdin()).lines();
|
||||
println!("Enter messages via STDIN and they will be sent to connected peers using Gossipsub");
|
||||
|
||||
tokio::task::spawn(async move { swarm.run().await });
|
||||
|
||||
// Kick it off
|
||||
loop {
|
||||
select! {
|
||||
// on gossipsub outgoing
|
||||
Ok(Some(line)) = stdin.next_line() => {
|
||||
if let Err(e) = swarm
|
||||
.behaviour_mut().gossipsub
|
||||
.publish(topic.clone(), line.as_bytes()) {
|
||||
println!("Publish error: {e:?}");
|
||||
}
|
||||
_= to_swarm.send(ToSwarm::Message("test-net".to_owned(), line.into_bytes())).await;
|
||||
}
|
||||
event = swarm.select_next_some() => match event {
|
||||
event = from_swarm.recv() => match event {
|
||||
// on gossipsub incoming
|
||||
SwarmEvent::Behaviour(swarm::BehaviourEvent::Gossipsub(gossipsub::Event::Message {
|
||||
propagation_source: peer_id,
|
||||
message_id: id,
|
||||
message,
|
||||
})) => println!(
|
||||
"\n\nGot message: '{}' with id: {id} from peer: {peer_id}\n\n",
|
||||
String::from_utf8_lossy(&message.data),
|
||||
),
|
||||
Some(FromSwarm::Message(pid, topic, content)) => {
|
||||
assert_eq!(topic, "test-net");
|
||||
let fmt = String::from_utf8_lossy(&content);
|
||||
println!("{pid}: {fmt}");
|
||||
}
|
||||
|
||||
// on discovery
|
||||
SwarmEvent::Behaviour(swarm::BehaviourEvent::Discovery(e)) => match e {
|
||||
discovery::Event::ConnectionEstablished {
|
||||
peer_id, connection_id, remote_ip, remote_tcp_port
|
||||
} => {
|
||||
println!("\n\nConnected to: {peer_id}; connection ID: {connection_id}; remote IP: {remote_ip}; remote TCP port: {remote_tcp_port}\n\n");
|
||||
}
|
||||
discovery::Event::ConnectionClosed {
|
||||
peer_id, connection_id, remote_ip, remote_tcp_port
|
||||
} => {
|
||||
eprintln!("\n\nDisconnected from: {peer_id}; connection ID: {connection_id}; remote IP: {remote_ip}; remote TCP port: {remote_tcp_port}\n\n");
|
||||
Some(FromSwarm::Discovered(pid)) => {
|
||||
eprintln!("\n\nConnected to: {pid}\n\n");
|
||||
}
|
||||
Some(FromSwarm::Expired(pid)) => {
|
||||
eprintln!("\n\nDisconnected from: {pid}\n\n");
|
||||
}
|
||||
|
||||
// ignore outgoing errors: those are normal
|
||||
e@SwarmEvent::OutgoingConnectionError { .. } => { log::debug!("Outgoing connection error: {e:?}"); }
|
||||
None => break,
|
||||
|
||||
// otherwise log any other event
|
||||
e => { log::info!("Other event {e:?}"); }
|
||||
|
||||
@@ -1,127 +0,0 @@
|
||||
// Copyright 2018 Parity Technologies (UK) Ltd.
|
||||
//
|
||||
// Permission is hereby granted, free of charge, to any person obtaining a
|
||||
// copy of this software and associated documentation files (the "Software"),
|
||||
// to deal in the Software without restriction, including without limitation
|
||||
// the rights to use, copy, modify, merge, publish, distribute, sublicense,
|
||||
// and/or sell copies of the Software, and to permit persons to whom the
|
||||
// Software is furnished to do so, subject to the following conditions:
|
||||
//
|
||||
// The above copyright notice and this permission notice shall be included in
|
||||
// all copies or substantial portions of the Software.
|
||||
//
|
||||
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
|
||||
// OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
||||
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
|
||||
// FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
|
||||
// DEALINGS IN THE SOFTWARE.
|
||||
|
||||
use futures::stream::StreamExt;
|
||||
use libp2p::{
|
||||
gossipsub, mdns, noise,
|
||||
swarm::{NetworkBehaviour, SwarmEvent},
|
||||
tcp, yamux,
|
||||
};
|
||||
use std::error::Error;
|
||||
use std::time::Duration;
|
||||
use tokio::{io, io::AsyncBufReadExt, select};
|
||||
use tracing_subscriber::EnvFilter;
|
||||
|
||||
// We create a custom network behaviour that combines Gossipsub and Mdns.
|
||||
#[derive(NetworkBehaviour)]
|
||||
struct MyBehaviour {
|
||||
gossipsub: gossipsub::Behaviour,
|
||||
mdns: mdns::tokio::Behaviour,
|
||||
}
|
||||
|
||||
#[tokio::main]
|
||||
async fn main() -> Result<(), Box<dyn Error>> {
|
||||
let _ = tracing_subscriber::fmt()
|
||||
.with_env_filter(EnvFilter::from_default_env())
|
||||
.try_init();
|
||||
|
||||
let mut swarm = libp2p::SwarmBuilder::with_new_identity()
|
||||
.with_tokio()
|
||||
.with_tcp(
|
||||
tcp::Config::default(),
|
||||
noise::Config::new,
|
||||
yamux::Config::default,
|
||||
)?
|
||||
.with_behaviour(|key| {
|
||||
// Set a custom gossipsub configuration
|
||||
let gossipsub_config = gossipsub::ConfigBuilder::default()
|
||||
.heartbeat_interval(Duration::from_secs(10))
|
||||
.validation_mode(gossipsub::ValidationMode::Strict) // This sets the kind of message validation. The default is Strict (enforce message signing)
|
||||
.build()
|
||||
.map_err(io::Error::other)?; // Temporary hack because `build` does not return a proper `std::error::Error`.
|
||||
|
||||
// build a gossipsub network behaviour
|
||||
let gossipsub = gossipsub::Behaviour::new(
|
||||
gossipsub::MessageAuthenticity::Signed(key.clone()),
|
||||
gossipsub_config,
|
||||
)?;
|
||||
|
||||
let mdns =
|
||||
mdns::tokio::Behaviour::new(mdns::Config::default(), key.public().to_peer_id())?;
|
||||
Ok(MyBehaviour { gossipsub, mdns })
|
||||
})?
|
||||
.build();
|
||||
|
||||
println!("Running swarm with identity {}", swarm.local_peer_id());
|
||||
|
||||
// Create a Gossipsub topic
|
||||
let topic = gossipsub::IdentTopic::new("test-net");
|
||||
// subscribes to our topic
|
||||
swarm.behaviour_mut().gossipsub.subscribe(&topic)?;
|
||||
|
||||
// Read full lines from stdin
|
||||
let mut stdin = io::BufReader::new(io::stdin()).lines();
|
||||
|
||||
// Listen on all interfaces and whatever port the OS assigns
|
||||
swarm.listen_on("/ip4/0.0.0.0/tcp/0".parse()?)?;
|
||||
|
||||
println!("Enter messages via STDIN and they will be sent to connected peers using Gossipsub");
|
||||
|
||||
// Kick it off
|
||||
loop {
|
||||
select! {
|
||||
Ok(Some(line)) = stdin.next_line() => {
|
||||
if let Err(e) = swarm
|
||||
.behaviour_mut().gossipsub
|
||||
.publish(topic.clone(), line.as_bytes()) {
|
||||
println!("Publish error: {e:?}");
|
||||
}
|
||||
}
|
||||
event = swarm.select_next_some() => match event {
|
||||
SwarmEvent::Behaviour(MyBehaviourEvent::Mdns(mdns::Event::Discovered(list))) => {
|
||||
for (peer_id, multiaddr) in list {
|
||||
println!("mDNS discovered a new peer: {peer_id} on {multiaddr}");
|
||||
swarm.behaviour_mut().gossipsub.add_explicit_peer(&peer_id);
|
||||
}
|
||||
},
|
||||
SwarmEvent::Behaviour(MyBehaviourEvent::Mdns(mdns::Event::Expired(list))) => {
|
||||
for (peer_id, multiaddr) in list {
|
||||
println!("mDNS discover peer has expired: {peer_id} on {multiaddr}");
|
||||
swarm.behaviour_mut().gossipsub.remove_explicit_peer(&peer_id);
|
||||
}
|
||||
},
|
||||
SwarmEvent::Behaviour(MyBehaviourEvent::Gossipsub(gossipsub::Event::Message {
|
||||
propagation_source: peer_id,
|
||||
message_id: id,
|
||||
message,
|
||||
})) => println!(
|
||||
"Got message: '{}' with id: {id} from peer: {peer_id}",
|
||||
String::from_utf8_lossy(&message.data),
|
||||
),
|
||||
SwarmEvent::NewListenAddr { address, .. } => {
|
||||
println!("Local node is listening on {address}");
|
||||
}
|
||||
e => {
|
||||
println!("Other swarm event: {:?}", e);
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
@@ -1,10 +1,10 @@
|
||||
use crate::ext::MultiaddrExt;
|
||||
use delegate::delegate;
|
||||
use either::Either;
|
||||
use futures::FutureExt;
|
||||
use futures_timer::Delay;
|
||||
use libp2p::core::transport::PortUse;
|
||||
use libp2p::core::{ConnectedPoint, Endpoint};
|
||||
use libp2p::futures::FutureExt;
|
||||
use libp2p::swarm::behaviour::ConnectionEstablished;
|
||||
use libp2p::swarm::dial_opts::DialOpts;
|
||||
use libp2p::swarm::{
|
||||
|
||||
@@ -3,6 +3,7 @@
|
||||
//! this is here as a placeholder documentation
|
||||
//!
|
||||
//!
|
||||
|
||||
pub mod discovery;
|
||||
pub mod swarm;
|
||||
|
||||
|
||||
@@ -1,9 +1,30 @@
|
||||
use crate::alias;
|
||||
use crate::discovery;
|
||||
use crate::swarm::transport::tcp_transport;
|
||||
pub use behaviour::{Behaviour, BehaviourEvent};
|
||||
use libp2p::{SwarmBuilder, identity};
|
||||
use behaviour::{Behaviour, BehaviourEvent};
|
||||
use futures_lite::StreamExt;
|
||||
use libp2p::{PeerId, SwarmBuilder, gossipsub, identity, swarm::SwarmEvent};
|
||||
use tokio::sync::mpsc;
|
||||
|
||||
pub type Swarm = libp2p::Swarm<Behaviour>;
|
||||
pub struct Swarm {
|
||||
swarm: libp2p::Swarm<Behaviour>,
|
||||
from_client: mpsc::Receiver<ToSwarm>,
|
||||
to_client: mpsc::Sender<FromSwarm>,
|
||||
}
|
||||
|
||||
#[derive(Debug)]
|
||||
pub enum FromSwarm {
|
||||
PublishError(gossipsub::PublishError),
|
||||
Discovered(PeerId),
|
||||
Expired(PeerId),
|
||||
Message(PeerId, String, Vec<u8>),
|
||||
}
|
||||
#[derive(Debug)]
|
||||
pub enum ToSwarm {
|
||||
Message(String, Vec<u8>),
|
||||
Subscribe(String),
|
||||
Unsubscribe(String),
|
||||
}
|
||||
|
||||
/// The current version of the network: this prevents devices running different versions of the
|
||||
/// software from interacting with each other.
|
||||
@@ -15,23 +36,142 @@ pub type Swarm = libp2p::Swarm<Behaviour>;
|
||||
pub const NETWORK_VERSION: &[u8] = b"v0.0.1";
|
||||
pub const OVERRIDE_VERSION_ENV_VAR: &str = "EXO_LIBP2P_NAMESPACE";
|
||||
|
||||
/// Create and configure a swarm which listens to all ports on OS
|
||||
pub fn create_swarm(keypair: identity::Keypair) -> alias::AnyResult<Swarm> {
|
||||
let mut swarm = SwarmBuilder::with_existing_identity(keypair)
|
||||
.with_tokio()
|
||||
.with_other_transport(tcp_transport)?
|
||||
.with_behaviour(Behaviour::new)?
|
||||
.build();
|
||||
impl Swarm {
|
||||
/// Create and configure a swarm which listens to all ports on OS
|
||||
pub fn new(
|
||||
keypair: identity::Keypair,
|
||||
from_client: mpsc::Receiver<ToSwarm>,
|
||||
to_client: mpsc::Sender<FromSwarm>,
|
||||
) -> alias::AnyResult<Swarm> {
|
||||
let mut swarm = SwarmBuilder::with_existing_identity(keypair)
|
||||
.with_tokio()
|
||||
.with_other_transport(tcp_transport)?
|
||||
.with_behaviour(Behaviour::new)?
|
||||
.build();
|
||||
|
||||
// Listen on all interfaces and whatever port the OS assigns
|
||||
swarm.listen_on("/ip4/0.0.0.0/tcp/0".parse()?)?;
|
||||
Ok(swarm)
|
||||
// Listen on all interfaces and whatever port the OS assigns
|
||||
swarm.listen_on("/ip4/0.0.0.0/tcp/0".parse()?)?;
|
||||
Ok(Self {
|
||||
swarm,
|
||||
from_client,
|
||||
to_client,
|
||||
})
|
||||
}
|
||||
pub async fn run(&mut self) {
|
||||
log::info!("RUST: networking task started");
|
||||
|
||||
loop {
|
||||
tokio::select! {
|
||||
message = self.from_client.recv() => {
|
||||
// handle closed channel
|
||||
let Some(message) = message else {
|
||||
log::info!("RUST: channel closed");
|
||||
break;
|
||||
};
|
||||
|
||||
// dispatch incoming messages
|
||||
match message {
|
||||
ToSwarm::Subscribe(topic) => {
|
||||
// try to subscribe
|
||||
match self.swarm.behaviour_mut().gossipsub.subscribe(&gossipsub::IdentTopic::new(topic.clone())) {
|
||||
Err(e) => {
|
||||
let gossipsub::SubscriptionError::PublishError(e) = e else {
|
||||
unreachable!("topic filter used")
|
||||
};
|
||||
let Ok(()) = self.to_client.send(FromSwarm::PublishError(e)).await else {
|
||||
log::warn!("RUST: client connection closed");
|
||||
break
|
||||
};
|
||||
},
|
||||
Ok(false) => log::warn!("RUST: tried to subscribe to topic twice"),
|
||||
Ok(true) => {},
|
||||
}
|
||||
}
|
||||
ToSwarm::Unsubscribe(topic) => {
|
||||
// try to subscribe
|
||||
if !self.swarm.behaviour_mut().gossipsub.unsubscribe(&gossipsub::IdentTopic::new(topic)) {
|
||||
log::warn!("RUST: tried to unsubscribe from topic twice");
|
||||
}
|
||||
}
|
||||
ToSwarm::Message( topic, data ) => {
|
||||
// try to publish the data -> catch NoPeersSubscribedToTopic error & convert to correct exception
|
||||
match self.swarm.behaviour_mut().gossipsub.publish(
|
||||
gossipsub::IdentTopic::new(topic), data
|
||||
) {
|
||||
Ok(_) => {},
|
||||
Err(e) => {
|
||||
let Ok(()) = self.to_client.send(FromSwarm::PublishError(e)).await else {
|
||||
log::warn!("RUST: client connection closed");
|
||||
break
|
||||
};
|
||||
},
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// architectural solution to this problem:
|
||||
// create keep_alive behavior who's job it is to dial peers discovered by mDNS (and drop when expired)
|
||||
// -> it will emmit TRUE connected/disconnected events consumable elsewhere
|
||||
//
|
||||
// gossipsub will feed off-of dial attempts created by networking, and that will bootstrap its' peers list
|
||||
// then for actual communication it will dial those peers if need-be
|
||||
swarm_event = self.swarm.next() => {
|
||||
let Some(swarm_event) = swarm_event else {
|
||||
log::warn!("RUST: swarm closed communication");
|
||||
break
|
||||
};
|
||||
let SwarmEvent::Behaviour(behaviour_event) = swarm_event else {
|
||||
continue
|
||||
};
|
||||
match behaviour_event {
|
||||
BehaviourEvent::Gossipsub(gossipsub::Event::Message {
|
||||
message: gossipsub::Message {
|
||||
source,
|
||||
topic,
|
||||
data,
|
||||
..
|
||||
},
|
||||
..
|
||||
}) => {
|
||||
let Some(peer_id) = source else {
|
||||
log::warn!("RUST: ignoring message with unknown source on {topic}");
|
||||
continue;
|
||||
};
|
||||
// send incoming message to channel (or exit if connection closed)
|
||||
if let Err(e) = self.to_client.send(FromSwarm::Message(peer_id, topic.into_string(), data)).await {
|
||||
log::warn!("RUST: could not send incoming gossipsub message since channel already closed: {e}");
|
||||
break
|
||||
};
|
||||
},
|
||||
BehaviourEvent::Discovery(discovery::Event::ConnectionEstablished { peer_id, .. }) => {
|
||||
// send connection event to channel (or exit if connection closed)
|
||||
if let Err(_) = self.to_client.send(FromSwarm::Discovered(peer_id)).await {
|
||||
log::warn!("RUST: swarm closed communication");
|
||||
};
|
||||
},
|
||||
BehaviourEvent::Discovery(discovery::Event::ConnectionClosed { peer_id, .. }) => {
|
||||
// send connection event to channel (or exit if connection closed)
|
||||
if let Err(_) = self.to_client.send(FromSwarm::Expired(peer_id)).await {
|
||||
log::warn!("RUST: swarm closed communication");
|
||||
};
|
||||
},
|
||||
e => {
|
||||
log::debug!("RUST: other event {e:?}");
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
log::info!("RUST: networking task stopped");
|
||||
}
|
||||
}
|
||||
|
||||
mod transport {
|
||||
use crate::alias;
|
||||
use crate::swarm::{NETWORK_VERSION, OVERRIDE_VERSION_ENV_VAR};
|
||||
use futures::{AsyncRead, AsyncWrite};
|
||||
use futures_lite::{AsyncRead, AsyncWrite};
|
||||
use keccak_const::Sha3_256;
|
||||
use libp2p::core::muxing;
|
||||
use libp2p::core::transport::Boxed;
|
||||
|
||||
@@ -47,7 +47,6 @@ class DownloadCoordinator:
|
||||
download_command_receiver: Receiver[ForwarderDownloadCommand]
|
||||
local_event_sender: Sender[ForwarderEvent]
|
||||
event_index_counter: Iterator[int]
|
||||
offline: bool = False
|
||||
|
||||
# Local state
|
||||
download_status: dict[ModelId, DownloadProgress] = field(default_factory=dict)
|
||||
@@ -63,8 +62,6 @@ class DownloadCoordinator:
|
||||
|
||||
def __post_init__(self) -> None:
|
||||
self.event_sender, self.event_receiver = channel[Event]()
|
||||
if self.offline:
|
||||
self.shard_downloader.set_internet_connection(False)
|
||||
self.shard_downloader.on_progress(self._download_progress_callback)
|
||||
|
||||
def _model_dir(self, model_id: ModelId) -> str:
|
||||
@@ -110,30 +107,23 @@ class DownloadCoordinator:
|
||||
self._last_progress_time[model_id] = current_time()
|
||||
|
||||
async def run(self) -> None:
|
||||
logger.info(
|
||||
f"Starting DownloadCoordinator{' (offline mode)' if self.offline else ''}"
|
||||
)
|
||||
if not self.offline:
|
||||
self._test_internet_connection()
|
||||
logger.info("Starting DownloadCoordinator")
|
||||
self._test_internet_connection()
|
||||
async with self._tg as tg:
|
||||
tg.start_soon(self._command_processor)
|
||||
tg.start_soon(self._forward_events)
|
||||
tg.start_soon(self._emit_existing_download_progress)
|
||||
if not self.offline:
|
||||
tg.start_soon(self._check_internet_connection)
|
||||
tg.start_soon(self._check_internet_connection)
|
||||
|
||||
def _test_internet_connection(self) -> None:
|
||||
# Try multiple endpoints since some ISPs/networks block specific IPs
|
||||
for host in ("1.1.1.1", "8.8.8.8", "1.0.0.1"):
|
||||
try:
|
||||
socket.create_connection((host, 443), timeout=3).close()
|
||||
self.shard_downloader.set_internet_connection(True)
|
||||
logger.debug(f"Internet connectivity: True (via {host})")
|
||||
return
|
||||
except OSError:
|
||||
continue
|
||||
self.shard_downloader.set_internet_connection(False)
|
||||
logger.debug("Internet connectivity: False")
|
||||
try:
|
||||
socket.create_connection(("1.1.1.1", 443), timeout=3).close()
|
||||
self.shard_downloader.set_internet_connection(True)
|
||||
except OSError:
|
||||
self.shard_downloader.set_internet_connection(False)
|
||||
logger.debug(
|
||||
f"Internet connectivity: {self.shard_downloader.internet_connection}"
|
||||
)
|
||||
|
||||
async def _check_internet_connection(self) -> None:
|
||||
first_connection = True
|
||||
@@ -212,20 +202,6 @@ class DownloadCoordinator:
|
||||
)
|
||||
return
|
||||
|
||||
if self.offline:
|
||||
logger.warning(
|
||||
f"Offline mode: model {model_id} is not fully available locally, cannot download"
|
||||
)
|
||||
failed = DownloadFailed(
|
||||
shard_metadata=shard,
|
||||
node_id=self.node_id,
|
||||
error_message=f"Model files not found locally in offline mode: {model_id}",
|
||||
model_directory=self._model_dir(model_id),
|
||||
)
|
||||
self.download_status[model_id] = failed
|
||||
await self.event_sender.send(NodeDownloadProgress(download_progress=failed))
|
||||
return
|
||||
|
||||
# Start actual download
|
||||
self._start_download_task(shard, initial_progress)
|
||||
|
||||
|
||||
@@ -448,13 +448,12 @@ async def download_file_with_retry(
|
||||
target_dir: Path,
|
||||
on_progress: Callable[[int, int, bool], None] = lambda _, __, ___: None,
|
||||
on_connection_lost: Callable[[], None] = lambda: None,
|
||||
skip_internet: bool = False,
|
||||
) -> Path:
|
||||
n_attempts = 3
|
||||
for attempt in range(n_attempts):
|
||||
try:
|
||||
return await _download_file(
|
||||
model_id, revision, path, target_dir, on_progress, skip_internet
|
||||
model_id, revision, path, target_dir, on_progress
|
||||
)
|
||||
except HuggingFaceAuthenticationError:
|
||||
raise
|
||||
@@ -488,14 +487,10 @@ async def _download_file(
|
||||
path: str,
|
||||
target_dir: Path,
|
||||
on_progress: Callable[[int, int, bool], None] = lambda _, __, ___: None,
|
||||
skip_internet: bool = False,
|
||||
) -> Path:
|
||||
target_path = target_dir / path
|
||||
|
||||
if await aios.path.exists(target_path):
|
||||
if skip_internet:
|
||||
return target_path
|
||||
|
||||
local_size = (await aios.stat(target_path)).st_size
|
||||
|
||||
# Try to verify against remote, but allow offline operation
|
||||
@@ -515,11 +510,6 @@ async def _download_file(
|
||||
)
|
||||
return target_path
|
||||
|
||||
if skip_internet:
|
||||
raise FileNotFoundError(
|
||||
f"File {path} not found locally and cannot download in offline mode"
|
||||
)
|
||||
|
||||
await aios.makedirs((target_dir / path).parent, exist_ok=True)
|
||||
length, etag = await file_meta(model_id, revision, path)
|
||||
remote_hash = etag[:-5] if etag.endswith("-gzip") else etag
|
||||
@@ -824,7 +814,6 @@ async def download_shard(
|
||||
file, curr_bytes, total_bytes, is_renamed
|
||||
),
|
||||
on_connection_lost=on_connection_lost,
|
||||
skip_internet=skip_internet,
|
||||
)
|
||||
|
||||
if not skip_download:
|
||||
|
||||
@@ -1,230 +0,0 @@
|
||||
"""Tests for offline/air-gapped mode."""
|
||||
|
||||
from collections.abc import AsyncIterator
|
||||
from pathlib import Path
|
||||
from unittest.mock import AsyncMock, patch
|
||||
|
||||
import aiofiles
|
||||
import aiofiles.os as aios
|
||||
import pytest
|
||||
|
||||
from exo.download.download_utils import (
|
||||
_download_file, # pyright: ignore[reportPrivateUsage]
|
||||
download_file_with_retry,
|
||||
fetch_file_list_with_cache,
|
||||
)
|
||||
from exo.shared.types.common import ModelId
|
||||
from exo.shared.types.worker.downloads import FileListEntry
|
||||
|
||||
|
||||
@pytest.fixture
|
||||
def model_id() -> ModelId:
|
||||
return ModelId("test-org/test-model")
|
||||
|
||||
|
||||
@pytest.fixture
|
||||
async def temp_models_dir(tmp_path: Path) -> AsyncIterator[Path]:
|
||||
models_dir = tmp_path / "models"
|
||||
await aios.makedirs(models_dir, exist_ok=True)
|
||||
with patch("exo.download.download_utils.EXO_MODELS_DIR", models_dir):
|
||||
yield models_dir
|
||||
|
||||
|
||||
class TestDownloadFileOffline:
|
||||
"""Tests for _download_file with skip_internet=True."""
|
||||
|
||||
async def test_returns_local_file_without_http_verification(
|
||||
self, model_id: ModelId, tmp_path: Path
|
||||
) -> None:
|
||||
"""When skip_internet=True and file exists locally, return it immediately
|
||||
without making any HTTP calls (no file_meta verification)."""
|
||||
target_dir = tmp_path / "downloads"
|
||||
await aios.makedirs(target_dir, exist_ok=True)
|
||||
|
||||
local_file = target_dir / "model.safetensors"
|
||||
async with aiofiles.open(local_file, "wb") as f:
|
||||
await f.write(b"model weights data")
|
||||
|
||||
with patch(
|
||||
"exo.download.download_utils.file_meta",
|
||||
new_callable=AsyncMock,
|
||||
) as mock_file_meta:
|
||||
result = await _download_file(
|
||||
model_id,
|
||||
"main",
|
||||
"model.safetensors",
|
||||
target_dir,
|
||||
skip_internet=True,
|
||||
)
|
||||
|
||||
assert result == local_file
|
||||
mock_file_meta.assert_not_called()
|
||||
|
||||
async def test_raises_file_not_found_for_missing_file(
|
||||
self, model_id: ModelId, tmp_path: Path
|
||||
) -> None:
|
||||
"""When skip_internet=True and file does NOT exist locally,
|
||||
raise FileNotFoundError instead of attempting download."""
|
||||
target_dir = tmp_path / "downloads"
|
||||
await aios.makedirs(target_dir, exist_ok=True)
|
||||
|
||||
with pytest.raises(FileNotFoundError, match="offline mode"):
|
||||
await _download_file(
|
||||
model_id,
|
||||
"main",
|
||||
"missing_model.safetensors",
|
||||
target_dir,
|
||||
skip_internet=True,
|
||||
)
|
||||
|
||||
async def test_returns_local_file_in_subdirectory(
|
||||
self, model_id: ModelId, tmp_path: Path
|
||||
) -> None:
|
||||
"""When skip_internet=True and file exists in a subdirectory,
|
||||
return it without HTTP calls."""
|
||||
target_dir = tmp_path / "downloads"
|
||||
subdir = target_dir / "transformer"
|
||||
await aios.makedirs(subdir, exist_ok=True)
|
||||
|
||||
local_file = subdir / "diffusion_pytorch_model.safetensors"
|
||||
async with aiofiles.open(local_file, "wb") as f:
|
||||
await f.write(b"weights")
|
||||
|
||||
with patch(
|
||||
"exo.download.download_utils.file_meta",
|
||||
new_callable=AsyncMock,
|
||||
) as mock_file_meta:
|
||||
result = await _download_file(
|
||||
model_id,
|
||||
"main",
|
||||
"transformer/diffusion_pytorch_model.safetensors",
|
||||
target_dir,
|
||||
skip_internet=True,
|
||||
)
|
||||
|
||||
assert result == local_file
|
||||
mock_file_meta.assert_not_called()
|
||||
|
||||
|
||||
class TestDownloadFileWithRetryOffline:
|
||||
"""Tests for download_file_with_retry with skip_internet=True."""
|
||||
|
||||
async def test_propagates_skip_internet_to_download_file(
|
||||
self, model_id: ModelId, tmp_path: Path
|
||||
) -> None:
|
||||
"""Verify skip_internet is passed through to _download_file."""
|
||||
target_dir = tmp_path / "downloads"
|
||||
await aios.makedirs(target_dir, exist_ok=True)
|
||||
|
||||
local_file = target_dir / "config.json"
|
||||
async with aiofiles.open(local_file, "wb") as f:
|
||||
await f.write(b'{"model_type": "qwen2"}')
|
||||
|
||||
with patch(
|
||||
"exo.download.download_utils.file_meta",
|
||||
new_callable=AsyncMock,
|
||||
) as mock_file_meta:
|
||||
result = await download_file_with_retry(
|
||||
model_id,
|
||||
"main",
|
||||
"config.json",
|
||||
target_dir,
|
||||
skip_internet=True,
|
||||
)
|
||||
|
||||
assert result == local_file
|
||||
mock_file_meta.assert_not_called()
|
||||
|
||||
async def test_file_not_found_does_not_retry(
|
||||
self, model_id: ModelId, tmp_path: Path
|
||||
) -> None:
|
||||
"""FileNotFoundError from offline mode should not trigger retries."""
|
||||
target_dir = tmp_path / "downloads"
|
||||
await aios.makedirs(target_dir, exist_ok=True)
|
||||
|
||||
with pytest.raises(FileNotFoundError):
|
||||
await download_file_with_retry(
|
||||
model_id,
|
||||
"main",
|
||||
"nonexistent.safetensors",
|
||||
target_dir,
|
||||
skip_internet=True,
|
||||
)
|
||||
|
||||
|
||||
class TestFetchFileListOffline:
|
||||
"""Tests for fetch_file_list_with_cache with skip_internet=True."""
|
||||
|
||||
async def test_uses_cached_file_list(
|
||||
self, model_id: ModelId, temp_models_dir: Path
|
||||
) -> None:
|
||||
"""When skip_internet=True and cache file exists, use it without network."""
|
||||
from pydantic import TypeAdapter
|
||||
|
||||
cache_dir = temp_models_dir / "caches" / model_id.normalize()
|
||||
await aios.makedirs(cache_dir, exist_ok=True)
|
||||
|
||||
cached_list = [
|
||||
FileListEntry(type="file", path="model.safetensors", size=1000),
|
||||
FileListEntry(type="file", path="config.json", size=200),
|
||||
]
|
||||
cache_file = cache_dir / f"{model_id.normalize()}--main--file_list.json"
|
||||
async with aiofiles.open(cache_file, "w") as f:
|
||||
await f.write(
|
||||
TypeAdapter(list[FileListEntry]).dump_json(cached_list).decode()
|
||||
)
|
||||
|
||||
with patch(
|
||||
"exo.download.download_utils.fetch_file_list_with_retry",
|
||||
new_callable=AsyncMock,
|
||||
) as mock_fetch:
|
||||
result = await fetch_file_list_with_cache(
|
||||
model_id, "main", skip_internet=True
|
||||
)
|
||||
|
||||
assert result == cached_list
|
||||
mock_fetch.assert_not_called()
|
||||
|
||||
async def test_falls_back_to_local_directory_scan(
|
||||
self, model_id: ModelId, temp_models_dir: Path
|
||||
) -> None:
|
||||
"""When skip_internet=True and no cache but local files exist,
|
||||
build file list from local directory."""
|
||||
import json
|
||||
|
||||
model_dir = temp_models_dir / model_id.normalize()
|
||||
await aios.makedirs(model_dir, exist_ok=True)
|
||||
|
||||
async with aiofiles.open(model_dir / "config.json", "w") as f:
|
||||
await f.write('{"model_type": "qwen2"}')
|
||||
|
||||
index_data = {
|
||||
"metadata": {},
|
||||
"weight_map": {"model.layers.0.weight": "model.safetensors"},
|
||||
}
|
||||
async with aiofiles.open(model_dir / "model.safetensors.index.json", "w") as f:
|
||||
await f.write(json.dumps(index_data))
|
||||
|
||||
async with aiofiles.open(model_dir / "model.safetensors", "wb") as f:
|
||||
await f.write(b"x" * 500)
|
||||
|
||||
with patch(
|
||||
"exo.download.download_utils.fetch_file_list_with_retry",
|
||||
new_callable=AsyncMock,
|
||||
) as mock_fetch:
|
||||
result = await fetch_file_list_with_cache(
|
||||
model_id, "main", skip_internet=True
|
||||
)
|
||||
|
||||
mock_fetch.assert_not_called()
|
||||
paths = {entry.path for entry in result}
|
||||
assert "config.json" in paths
|
||||
assert "model.safetensors" in paths
|
||||
|
||||
async def test_raises_when_no_cache_and_no_local_files(
|
||||
self, model_id: ModelId, temp_models_dir: Path
|
||||
) -> None:
|
||||
"""When skip_internet=True and neither cache nor local files exist,
|
||||
raise FileNotFoundError."""
|
||||
with pytest.raises(FileNotFoundError, match="No internet"):
|
||||
await fetch_file_list_with_cache(model_id, "main", skip_internet=True)
|
||||
@@ -39,13 +39,12 @@ class Node:
|
||||
|
||||
node_id: NodeId
|
||||
event_index_counter: Iterator[int]
|
||||
offline: bool
|
||||
_tg: TaskGroup = field(init=False, default_factory=anyio.create_task_group)
|
||||
|
||||
@classmethod
|
||||
async def create(cls, args: "Args") -> "Self":
|
||||
keypair = get_node_id_keypair()
|
||||
node_id = NodeId(keypair.to_peer_id().to_base58())
|
||||
node_id = NodeId(keypair.to_string())
|
||||
session_id = SessionId(master_node_id=node_id, election_clock=0)
|
||||
router = Router.create(keypair)
|
||||
await router.register_topic(topics.GLOBAL_EVENTS)
|
||||
@@ -69,7 +68,6 @@ class Node:
|
||||
download_command_receiver=router.receiver(topics.DOWNLOAD_COMMANDS),
|
||||
local_event_sender=router.sender(topics.LOCAL_EVENTS),
|
||||
event_index_counter=event_index_counter,
|
||||
offline=args.offline,
|
||||
)
|
||||
else:
|
||||
download_coordinator = None
|
||||
@@ -134,7 +132,6 @@ class Node:
|
||||
api,
|
||||
node_id,
|
||||
event_index_counter,
|
||||
args.offline,
|
||||
)
|
||||
|
||||
async def run(self):
|
||||
@@ -225,7 +222,6 @@ class Node:
|
||||
),
|
||||
local_event_sender=self.router.sender(topics.LOCAL_EVENTS),
|
||||
event_index_counter=self.event_index_counter,
|
||||
offline=self.offline,
|
||||
)
|
||||
self._tg.start_soon(self.download_coordinator.run)
|
||||
if self.worker:
|
||||
@@ -264,9 +260,6 @@ def main():
|
||||
logger.info("Starting EXO")
|
||||
logger.info(f"EXO_LIBP2P_NAMESPACE: {os.getenv('EXO_LIBP2P_NAMESPACE')}")
|
||||
|
||||
if args.offline:
|
||||
logger.info("Running in OFFLINE mode — no internet checks, local models only")
|
||||
|
||||
# Set FAST_SYNCH override env var for runner subprocesses
|
||||
if args.fast_synch is True:
|
||||
os.environ["EXO_FAST_SYNCH"] = "on"
|
||||
@@ -289,7 +282,6 @@ class Args(CamelCaseModel):
|
||||
tb_only: bool = False
|
||||
no_worker: bool = False
|
||||
no_downloads: bool = False
|
||||
offline: bool = False
|
||||
fast_synch: bool | None = None # None = auto, True = force on, False = force off
|
||||
|
||||
@classmethod
|
||||
@@ -337,11 +329,6 @@ class Args(CamelCaseModel):
|
||||
action="store_true",
|
||||
help="Disable the download coordinator (node won't download models)",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--offline",
|
||||
action="store_true",
|
||||
help="Run in offline/air-gapped mode: skip internet checks, use only pre-staged local models",
|
||||
)
|
||||
fast_synch_group = parser.add_mutually_exclusive_group()
|
||||
fast_synch_group.add_argument(
|
||||
"--fast-synch",
|
||||
|
||||
@@ -85,7 +85,6 @@ from exo.shared.types.api import (
|
||||
ImageGenerationTaskParams,
|
||||
ImageListItem,
|
||||
ImageListResponse,
|
||||
ImageSize,
|
||||
ModelList,
|
||||
ModelListModel,
|
||||
PlaceInstanceParams,
|
||||
@@ -101,7 +100,6 @@ from exo.shared.types.api import (
|
||||
TraceRankStats,
|
||||
TraceResponse,
|
||||
TraceStatsResponse,
|
||||
normalize_image_size,
|
||||
)
|
||||
from exo.shared.types.chunks import (
|
||||
ErrorChunk,
|
||||
@@ -145,7 +143,6 @@ from exo.shared.types.openai_responses import (
|
||||
ResponsesResponse,
|
||||
)
|
||||
from exo.shared.types.state import State
|
||||
from exo.shared.types.worker.downloads import DownloadCompleted
|
||||
from exo.shared.types.worker.instances import Instance, InstanceId, InstanceMeta
|
||||
from exo.shared.types.worker.shards import Sharding
|
||||
from exo.utils.banner import print_startup_banner
|
||||
@@ -754,11 +751,9 @@ class API:
|
||||
When stream=True and partial_images > 0, returns a StreamingResponse
|
||||
with SSE-formatted events for partial and final images.
|
||||
"""
|
||||
payload.model = await self._validate_image_model(ModelId(payload.model))
|
||||
payload = payload.model_copy(
|
||||
update={
|
||||
"model": await self._validate_image_model(ModelId(payload.model)),
|
||||
"advanced_params": _ensure_seed(payload.advanced_params),
|
||||
}
|
||||
update={"advanced_params": _ensure_seed(payload.advanced_params)}
|
||||
)
|
||||
|
||||
command = ImageGeneration(
|
||||
@@ -1014,13 +1009,12 @@ class API:
|
||||
async def bench_image_generations(
|
||||
self, request: Request, payload: BenchImageGenerationTaskParams
|
||||
) -> BenchImageGenerationResponse:
|
||||
payload.model = await self._validate_image_model(ModelId(payload.model))
|
||||
|
||||
payload.stream = False
|
||||
payload.partial_images = 0
|
||||
payload = payload.model_copy(
|
||||
update={
|
||||
"model": await self._validate_image_model(ModelId(payload.model)),
|
||||
"stream": False,
|
||||
"partial_images": 0,
|
||||
"advanced_params": _ensure_seed(payload.advanced_params),
|
||||
}
|
||||
update={"advanced_params": _ensure_seed(payload.advanced_params)}
|
||||
)
|
||||
|
||||
command = ImageGeneration(
|
||||
@@ -1041,7 +1035,7 @@ class API:
|
||||
prompt: str,
|
||||
model: ModelId,
|
||||
n: int,
|
||||
size: ImageSize,
|
||||
size: str,
|
||||
response_format: Literal["url", "b64_json"],
|
||||
input_fidelity: Literal["low", "high"],
|
||||
stream: bool,
|
||||
@@ -1111,7 +1105,7 @@ class API:
|
||||
prompt: str = Form(...),
|
||||
model: str = Form(...),
|
||||
n: int = Form(1),
|
||||
size: str | None = Form(None),
|
||||
size: str = Form("1024x1024"),
|
||||
response_format: Literal["url", "b64_json"] = Form("b64_json"),
|
||||
input_fidelity: Literal["low", "high"] = Form("low"),
|
||||
stream: str = Form("false"),
|
||||
@@ -1137,7 +1131,7 @@ class API:
|
||||
prompt=prompt,
|
||||
model=ModelId(model),
|
||||
n=n,
|
||||
size=normalize_image_size(size),
|
||||
size=size,
|
||||
response_format=response_format,
|
||||
input_fidelity=input_fidelity,
|
||||
stream=stream_bool,
|
||||
@@ -1173,7 +1167,7 @@ class API:
|
||||
prompt: str = Form(...),
|
||||
model: str = Form(...),
|
||||
n: int = Form(1),
|
||||
size: str | None = Form(None),
|
||||
size: str = Form("1024x1024"),
|
||||
response_format: Literal["url", "b64_json"] = Form("b64_json"),
|
||||
input_fidelity: Literal["low", "high"] = Form("low"),
|
||||
quality: Literal["high", "medium", "low"] = Form("medium"),
|
||||
@@ -1193,7 +1187,7 @@ class API:
|
||||
prompt=prompt,
|
||||
model=ModelId(model),
|
||||
n=n,
|
||||
size=normalize_image_size(size),
|
||||
size=size,
|
||||
response_format=response_format,
|
||||
input_fidelity=input_fidelity,
|
||||
stream=False,
|
||||
@@ -1293,18 +1287,8 @@ class API:
|
||||
|
||||
return total_available
|
||||
|
||||
async def get_models(self, status: str | None = Query(default=None)) -> ModelList:
|
||||
"""Returns list of available models, optionally filtered by being downloaded."""
|
||||
cards = await get_model_cards()
|
||||
|
||||
if status == "downloaded":
|
||||
downloaded_model_ids: set[str] = set()
|
||||
for node_downloads in self.state.downloads.values():
|
||||
for dl in node_downloads:
|
||||
if isinstance(dl, DownloadCompleted):
|
||||
downloaded_model_ids.add(dl.shard_metadata.model_card.model_id)
|
||||
cards = [c for c in cards if c.model_id in downloaded_model_ids]
|
||||
|
||||
async def get_models(self) -> ModelList:
|
||||
"""Returns list of available models."""
|
||||
return ModelList(
|
||||
data=[
|
||||
ModelListModel(
|
||||
@@ -1322,7 +1306,7 @@ class API:
|
||||
base_model=card.base_model,
|
||||
capabilities=card.capabilities,
|
||||
)
|
||||
for card in cards
|
||||
for card in await get_model_cards()
|
||||
]
|
||||
)
|
||||
|
||||
|
||||
@@ -42,7 +42,7 @@ from exo.utils.channels import channel
|
||||
@pytest.mark.asyncio
|
||||
async def test_master():
|
||||
keypair = get_node_id_keypair()
|
||||
node_id = NodeId(keypair.to_peer_id().to_base58())
|
||||
node_id = NodeId(keypair.to_string())
|
||||
session_id = SessionId(master_node_id=node_id, election_clock=0)
|
||||
|
||||
ge_sender, global_event_receiver = channel[ForwarderEvent]()
|
||||
@@ -75,7 +75,7 @@ async def test_master():
|
||||
async with anyio.create_task_group() as tg:
|
||||
tg.start_soon(master.run)
|
||||
|
||||
sender_node_id = NodeId(f"{keypair.to_peer_id().to_base58()}_sender")
|
||||
sender_node_id = NodeId(f"{keypair.to_string()}_sender")
|
||||
# inject a NodeGatheredInfo event
|
||||
logger.info("inject a NodeGatheredInfo event")
|
||||
await local_event_sender.send(
|
||||
|
||||
@@ -1,37 +0,0 @@
|
||||
from enum import Enum
|
||||
|
||||
from exo_pyo3_bindings import ConnectionUpdate, ConnectionUpdateType
|
||||
|
||||
from exo.shared.types.common import NodeId
|
||||
from exo.utils.pydantic_ext import CamelCaseModel
|
||||
|
||||
"""Serialisable types for Connection Updates/Messages"""
|
||||
|
||||
|
||||
class ConnectionMessageType(Enum):
|
||||
Connected = 0
|
||||
Disconnected = 1
|
||||
|
||||
@staticmethod
|
||||
def from_update_type(update_type: ConnectionUpdateType):
|
||||
match update_type:
|
||||
case ConnectionUpdateType.Connected:
|
||||
return ConnectionMessageType.Connected
|
||||
case ConnectionUpdateType.Disconnected:
|
||||
return ConnectionMessageType.Disconnected
|
||||
|
||||
|
||||
class ConnectionMessage(CamelCaseModel):
|
||||
node_id: NodeId
|
||||
connection_type: ConnectionMessageType
|
||||
remote_ipv4: str
|
||||
remote_tcp_port: int
|
||||
|
||||
@classmethod
|
||||
def from_update(cls, update: ConnectionUpdate) -> "ConnectionMessage":
|
||||
return cls(
|
||||
node_id=NodeId(update.peer_id.to_base58()),
|
||||
connection_type=ConnectionMessageType.from_update_type(update.update_type),
|
||||
remote_ipv4=update.remote_ipv4,
|
||||
remote_tcp_port=update.remote_tcp_port,
|
||||
)
|
||||
@@ -16,17 +16,19 @@ from anyio.abc import TaskGroup
|
||||
from exo_pyo3_bindings import (
|
||||
AllQueuesFullError,
|
||||
Keypair,
|
||||
NetworkingHandle,
|
||||
NoPeersSubscribedToTopicError,
|
||||
PyMessage,
|
||||
PySwarm,
|
||||
)
|
||||
from filelock import FileLock
|
||||
from loguru import logger
|
||||
|
||||
from exo.shared.constants import EXO_NODE_ID_KEYPAIR
|
||||
from exo.shared.election import ConnectionMessage
|
||||
from exo.shared.types.common import NodeId
|
||||
from exo.utils.channels import Receiver, Sender, channel
|
||||
from exo.utils.pydantic_ext import CamelCaseModel
|
||||
|
||||
from .connection_message import ConnectionMessage
|
||||
from .topics import CONNECTION_MESSAGES, PublishPolicy, TypedTopic
|
||||
|
||||
|
||||
@@ -102,13 +104,13 @@ class TopicRouter[T: CamelCaseModel]:
|
||||
class Router:
|
||||
@classmethod
|
||||
def create(cls, identity: Keypair) -> "Router":
|
||||
return cls(handle=NetworkingHandle(identity))
|
||||
return cls(handle=PySwarm(identity))
|
||||
|
||||
def __init__(self, handle: NetworkingHandle):
|
||||
def __init__(self, handle: PySwarm):
|
||||
self.topic_routers: dict[str, TopicRouter[CamelCaseModel]] = {}
|
||||
send, recv = channel[tuple[str, bytes]]()
|
||||
self.networking_receiver: Receiver[tuple[str, bytes]] = recv
|
||||
self._net: NetworkingHandle = handle
|
||||
self._net = handle
|
||||
self._tmp_networking_sender: Sender[tuple[str, bytes]] | None = send
|
||||
self._id_count = count()
|
||||
self._tg: TaskGroup | None = None
|
||||
@@ -154,7 +156,6 @@ class Router:
|
||||
router = self.topic_routers[topic]
|
||||
tg.start_soon(router.run)
|
||||
tg.start_soon(self._networking_recv)
|
||||
tg.start_soon(self._networking_recv_connection_messages)
|
||||
tg.start_soon(self._networking_publish)
|
||||
# Router only shuts down if you cancel it.
|
||||
await sleep_forever()
|
||||
@@ -179,38 +180,44 @@ class Router:
|
||||
|
||||
async def _networking_recv(self):
|
||||
while True:
|
||||
topic, data = await self._net.gossipsub_recv()
|
||||
logger.trace(f"Received message on {topic} with payload {data}")
|
||||
if topic not in self.topic_routers:
|
||||
logger.warning(f"Received message on unknown or inactive topic {topic}")
|
||||
try:
|
||||
msg = await self._net.recv()
|
||||
except NoPeersSubscribedToTopicError:
|
||||
continue
|
||||
except AllQueuesFullError:
|
||||
logger.warning("All peer queues full, messages have been lost")
|
||||
continue
|
||||
|
||||
router = self.topic_routers[topic]
|
||||
await router.publish_bytes(data)
|
||||
|
||||
async def _networking_recv_connection_messages(self):
|
||||
while True:
|
||||
update = await self._net.connection_update_recv()
|
||||
message = ConnectionMessage.from_update(update)
|
||||
logger.trace(
|
||||
f"Received message on connection_messages with payload {message}"
|
||||
)
|
||||
if CONNECTION_MESSAGES.topic in self.topic_routers:
|
||||
router = self.topic_routers[CONNECTION_MESSAGES.topic]
|
||||
assert router.topic.model_type == ConnectionMessage
|
||||
router = cast(TopicRouter[ConnectionMessage], router)
|
||||
await router.publish(message)
|
||||
match msg:
|
||||
case PyMessage.Connection():
|
||||
if CONNECTION_MESSAGES.topic in self.topic_routers:
|
||||
router = self.topic_routers[CONNECTION_MESSAGES.topic]
|
||||
assert router.topic.model_type == ConnectionMessage
|
||||
router = cast(TopicRouter[ConnectionMessage], router)
|
||||
await router.publish(
|
||||
ConnectionMessage(
|
||||
node_id=NodeId(msg.node_id), connected=msg.connected
|
||||
)
|
||||
)
|
||||
case PyMessage.Gossip():
|
||||
if msg.topic not in self.topic_routers:
|
||||
logger.warning(
|
||||
f"Received message on unknown or inactive topic {msg.topic}"
|
||||
)
|
||||
continue
|
||||
logger.trace(
|
||||
f"Received message on {msg.topic} with payload {msg.data}"
|
||||
)
|
||||
router = self.topic_routers[msg.topic]
|
||||
await router.publish_bytes(msg.data)
|
||||
case _:
|
||||
raise ValueError("net recv returned something impossible")
|
||||
|
||||
async def _networking_publish(self):
|
||||
with self.networking_receiver as networked_items:
|
||||
async for topic, data in networked_items:
|
||||
try:
|
||||
logger.trace(f"Sending message on {topic} with payload {data}")
|
||||
await self._net.gossipsub_publish(topic, data)
|
||||
except NoPeersSubscribedToTopicError:
|
||||
pass
|
||||
except AllQueuesFullError:
|
||||
logger.warning(f"All peer queues full, dropping message on {topic}")
|
||||
logger.trace(f"Sending message on {topic} with payload {data}")
|
||||
await self._net.gossipsub_publish(topic, data)
|
||||
|
||||
|
||||
def get_node_id_keypair(
|
||||
@@ -221,7 +228,7 @@ def get_node_id_keypair(
|
||||
Obtain the :class:`PeerId` by from it.
|
||||
"""
|
||||
# TODO(evan): bring back node id persistence once we figure out how to deal with duplicates
|
||||
return Keypair.generate_ed25519()
|
||||
return Keypair.generate()
|
||||
|
||||
def lock_path(path: str | bytes | PathLike[str] | PathLike[bytes]) -> Path:
|
||||
return Path(str(path) + ".lock")
|
||||
@@ -235,12 +242,12 @@ def get_node_id_keypair(
|
||||
protobuf_encoded = f.read()
|
||||
|
||||
try: # if decoded successfully, save & return
|
||||
return Keypair.from_protobuf_encoding(protobuf_encoded)
|
||||
return Keypair.deserialize(protobuf_encoded)
|
||||
except ValueError as e: # on runtime error, assume corrupt file
|
||||
logger.warning(f"Encountered error when trying to get keypair: {e}")
|
||||
|
||||
# if no valid credentials, create new ones and persist
|
||||
with open(path, "w+b") as f:
|
||||
keypair = Keypair.generate_ed25519()
|
||||
f.write(keypair.to_protobuf_encoding())
|
||||
keypair = Keypair.generate()
|
||||
f.write(keypair.serialize())
|
||||
return keypair
|
||||
|
||||
@@ -1,8 +1,7 @@
|
||||
from dataclasses import dataclass
|
||||
from enum import Enum
|
||||
|
||||
from exo.routing.connection_message import ConnectionMessage
|
||||
from exo.shared.election import ElectionMessage
|
||||
from exo.shared.election import ConnectionMessage, ElectionMessage
|
||||
from exo.shared.types.commands import ForwarderCommand, ForwarderDownloadCommand
|
||||
from exo.shared.types.events import (
|
||||
ForwarderEvent,
|
||||
|
||||
@@ -10,7 +10,6 @@ from anyio import (
|
||||
from anyio.abc import TaskGroup
|
||||
from loguru import logger
|
||||
|
||||
from exo.routing.connection_message import ConnectionMessage
|
||||
from exo.shared.types.commands import ForwarderCommand
|
||||
from exo.shared.types.common import NodeId, SessionId
|
||||
from exo.utils.channels import Receiver, Sender
|
||||
@@ -19,6 +18,11 @@ from exo.utils.pydantic_ext import CamelCaseModel
|
||||
DEFAULT_ELECTION_TIMEOUT = 3.0
|
||||
|
||||
|
||||
class ConnectionMessage(CamelCaseModel):
|
||||
node_id: NodeId
|
||||
connected: bool
|
||||
|
||||
|
||||
class ElectionMessage(CamelCaseModel):
|
||||
clock: int
|
||||
seniority: int
|
||||
|
||||
@@ -44,8 +44,7 @@ async def _refresh_card_cache():
|
||||
async for toml_file in path.rglob("*.toml"):
|
||||
try:
|
||||
card = await ModelCard.load_from_path(toml_file)
|
||||
if card.model_id not in _card_cache:
|
||||
_card_cache[card.model_id] = card
|
||||
_card_cache[card.model_id] = card
|
||||
except (ValidationError, TOMLKitError):
|
||||
pass
|
||||
|
||||
@@ -183,7 +182,6 @@ class ConfigData(BaseModel):
|
||||
def supports_tensor(self) -> bool:
|
||||
return self.architectures in [
|
||||
["Glm4MoeLiteForCausalLM"],
|
||||
["GlmMoeDsaForCausalLM"],
|
||||
["DeepseekV32ForCausalLM"],
|
||||
["DeepseekV3ForCausalLM"],
|
||||
["Qwen3NextForCausalLM"],
|
||||
|
||||
@@ -1,7 +1,7 @@
|
||||
import pytest
|
||||
from anyio import create_task_group, fail_after, move_on_after
|
||||
|
||||
from exo.routing.connection_message import ConnectionMessage, ConnectionMessageType
|
||||
from exo.routing.router import ConnectionMessage
|
||||
from exo.shared.election import Election, ElectionMessage, ElectionResult
|
||||
from exo.shared.types.commands import ForwarderCommand, TestCommand
|
||||
from exo.shared.types.common import NodeId, SessionId
|
||||
@@ -330,9 +330,7 @@ async def test_connection_message_triggers_new_round_broadcast() -> None:
|
||||
await cm_tx.send(
|
||||
ConnectionMessage(
|
||||
node_id=NodeId(),
|
||||
connection_type=ConnectionMessageType.Connected,
|
||||
remote_ipv4="",
|
||||
remote_tcp_port=0,
|
||||
connected=True,
|
||||
)
|
||||
)
|
||||
|
||||
|
||||
@@ -23,7 +23,7 @@ def _get_keypair_concurrent_subprocess_task(
|
||||
sem.release()
|
||||
# wait to be told to begin simultaneous read
|
||||
ev.wait()
|
||||
queue.put(get_node_id_keypair().to_protobuf_encoding())
|
||||
queue.put(get_node_id_keypair().serialize())
|
||||
|
||||
|
||||
def _get_keypair_concurrent(num_procs: int) -> bytes:
|
||||
|
||||
@@ -1,9 +1,9 @@
|
||||
import time
|
||||
from collections.abc import Generator
|
||||
from typing import Annotated, Any, Literal, get_args
|
||||
from typing import Annotated, Any, Literal
|
||||
from uuid import uuid4
|
||||
|
||||
from pydantic import BaseModel, Field, field_validator
|
||||
from pydantic import BaseModel, Field
|
||||
|
||||
from exo.shared.models.model_cards import ModelCard, ModelId
|
||||
from exo.shared.types.common import CommandId, NodeId
|
||||
@@ -262,27 +262,6 @@ class DeleteInstanceResponse(BaseModel):
|
||||
instance_id: InstanceId
|
||||
|
||||
|
||||
ImageSize = Literal[
|
||||
"auto",
|
||||
"512x512",
|
||||
"768x768",
|
||||
"1024x768",
|
||||
"768x1024",
|
||||
"1024x1024",
|
||||
"1024x1536",
|
||||
"1536x1024",
|
||||
]
|
||||
|
||||
|
||||
def normalize_image_size(v: object) -> ImageSize:
|
||||
"""Shared validator for ImageSize fields: maps None → "auto" and rejects invalid values."""
|
||||
if v is None:
|
||||
return "auto"
|
||||
if v not in get_args(ImageSize):
|
||||
raise ValueError(f"Invalid size: {v!r}. Must be one of {get_args(ImageSize)}")
|
||||
return v # pyright: ignore[reportReturnType]
|
||||
|
||||
|
||||
class AdvancedImageParams(BaseModel):
|
||||
seed: Annotated[int, Field(ge=0)] | None = None
|
||||
num_inference_steps: Annotated[int, Field(ge=1, le=100)] | None = None
|
||||
@@ -302,7 +281,7 @@ class ImageGenerationTaskParams(BaseModel):
|
||||
partial_images: int | None = 0
|
||||
quality: Literal["high", "medium", "low"] | None = "medium"
|
||||
response_format: Literal["url", "b64_json"] | None = "b64_json"
|
||||
size: ImageSize = "auto"
|
||||
size: str | None = "1024x1024"
|
||||
stream: bool | None = False
|
||||
style: str | None = "vivid"
|
||||
user: str | None = None
|
||||
@@ -310,11 +289,6 @@ class ImageGenerationTaskParams(BaseModel):
|
||||
# Internal flag for benchmark mode - set by API, preserved through serialization
|
||||
bench: bool = False
|
||||
|
||||
@field_validator("size", mode="before")
|
||||
@classmethod
|
||||
def normalize_size(cls, v: object) -> ImageSize:
|
||||
return normalize_image_size(v)
|
||||
|
||||
|
||||
class BenchImageGenerationTaskParams(ImageGenerationTaskParams):
|
||||
bench: bool = True
|
||||
@@ -331,18 +305,13 @@ class ImageEditsTaskParams(BaseModel):
|
||||
quality: Literal["high", "medium", "low"] | None = "medium"
|
||||
output_format: Literal["png", "jpeg", "webp"] = "png"
|
||||
response_format: Literal["url", "b64_json"] | None = "b64_json"
|
||||
size: ImageSize = "auto"
|
||||
size: str | None = "1024x1024"
|
||||
image_strength: float | None = 0.7
|
||||
stream: bool = False
|
||||
partial_images: int | None = 0
|
||||
advanced_params: AdvancedImageParams | None = None
|
||||
bench: bool = False
|
||||
|
||||
@field_validator("size", mode="before")
|
||||
@classmethod
|
||||
def normalize_size(cls, v: object) -> ImageSize:
|
||||
return normalize_image_size(v)
|
||||
|
||||
def __repr_args__(self) -> Generator[tuple[str, Any], None, None]:
|
||||
for name, value in super().__repr_args__(): # pyright: ignore[reportAny]
|
||||
if name == "image_data":
|
||||
|
||||
@@ -4,13 +4,10 @@ from collections.abc import Sequence
|
||||
|
||||
from mlx_lm.models.cache import (
|
||||
ArraysCache,
|
||||
CacheList,
|
||||
KVCache,
|
||||
QuantizedKVCache,
|
||||
RotatingKVCache,
|
||||
)
|
||||
|
||||
# This list contains one cache entry per transformer layer
|
||||
KVCacheType = Sequence[
|
||||
KVCache | RotatingKVCache | QuantizedKVCache | ArraysCache | CacheList
|
||||
]
|
||||
KVCacheType = Sequence[KVCache | RotatingKVCache | QuantizedKVCache | ArraysCache]
|
||||
|
||||
@@ -1,4 +1,3 @@
|
||||
import contextlib
|
||||
import multiprocessing as mp
|
||||
from dataclasses import dataclass, field
|
||||
from math import inf
|
||||
@@ -133,8 +132,7 @@ class MpSender[T]:
|
||||
def close(self) -> None:
|
||||
if not self._state.closed.is_set():
|
||||
self._state.closed.set()
|
||||
with contextlib.suppress(Exception):
|
||||
self._state.buffer.put_nowait(_MpEndOfStream())
|
||||
self._state.buffer.put(_MpEndOfStream())
|
||||
self._state.buffer.close()
|
||||
|
||||
# == unique to Mp channels ==
|
||||
@@ -206,8 +204,6 @@ class MpReceiver[T]:
|
||||
def close(self) -> None:
|
||||
if not self._state.closed.is_set():
|
||||
self._state.closed.set()
|
||||
with contextlib.suppress(Exception):
|
||||
self._state.buffer.put_nowait(_MpEndOfStream())
|
||||
self._state.buffer.close()
|
||||
|
||||
# == unique to Mp channels ==
|
||||
|
||||
@@ -14,7 +14,6 @@ from exo.shared.types.api import (
|
||||
ImageEditsTaskParams,
|
||||
ImageGenerationStats,
|
||||
ImageGenerationTaskParams,
|
||||
ImageSize,
|
||||
)
|
||||
from exo.shared.types.memory import Memory
|
||||
from exo.shared.types.worker.runner_response import (
|
||||
@@ -24,9 +23,9 @@ from exo.shared.types.worker.runner_response import (
|
||||
from exo.worker.engines.image.distributed_model import DistributedImageModel
|
||||
|
||||
|
||||
def parse_size(size_str: ImageSize) -> tuple[int, int]:
|
||||
def parse_size(size_str: str | None) -> tuple[int, int]:
|
||||
"""Parse size parameter like '1024x1024' to (width, height) tuple."""
|
||||
if size_str == "auto":
|
||||
if not size_str:
|
||||
return (1024, 1024)
|
||||
|
||||
try:
|
||||
@@ -110,9 +109,6 @@ def generate_image(
|
||||
# Decode base64 image data and save to temp file
|
||||
image_path = Path(tmpdir) / "input.png"
|
||||
image_path.write_bytes(base64.b64decode(task.image_data))
|
||||
if task.size == "auto":
|
||||
with Image.open(image_path) as img:
|
||||
width, height = img.size
|
||||
|
||||
for image_num in range(num_images):
|
||||
# Increment seed for each image to ensure unique results
|
||||
|
||||
@@ -163,14 +163,11 @@ class PipelineLastLayer(CustomMlxLayer):
|
||||
output, (self.r + 1) % self.s, group=self.group
|
||||
)
|
||||
if cache is not None:
|
||||
# CacheList (used by MLA models like DeepSeekV32, GLM MoE DSA)
|
||||
# doesn't have .keys directly; access via first sub-cache.
|
||||
_cache = cache[0] if hasattr(cache, "caches") else cache # type: ignore
|
||||
_cache.keys = mx.depends(_cache.keys, output) # type: ignore
|
||||
cache.keys = mx.depends(cache.keys, output) # type: ignore[reportUnknownMemberType]
|
||||
if self.is_prefill:
|
||||
mx.eval(output)
|
||||
if cache is not None:
|
||||
mx.eval(_cache.keys) # type: ignore
|
||||
mx.eval(cache.keys) # type: ignore
|
||||
|
||||
if not self.is_prefill:
|
||||
output = mx.distributed.all_gather(output, group=self.group)[
|
||||
@@ -310,9 +307,7 @@ def patch_pipeline_model[T](model: T, group: mx.distributed.Group) -> T:
|
||||
|
||||
# Add dependency to last cache entry to ensure distributed ops are evaluated
|
||||
if cache is not None:
|
||||
last = cache[-1] # type: ignore
|
||||
dep_cache = last[0] if hasattr(last, "caches") else last # type: ignore
|
||||
dep_cache.keys = mx.depends(dep_cache.keys, logits) # type: ignore
|
||||
cache[-1].state = mx.depends(cache[-1].state, logits) # type: ignore
|
||||
|
||||
return logits
|
||||
|
||||
@@ -338,9 +333,7 @@ def patch_tensor_model[T](model: T) -> T:
|
||||
|
||||
# Add dependency to last cache entry to ensure distributed ops are evaluated
|
||||
if cache is not None and len(cache) > 0: # pyright: ignore[reportAny]
|
||||
last = cache[-1] # pyright: ignore[reportAny]
|
||||
dep_cache = last[0] if hasattr(last, "caches") else last # pyright: ignore[reportAny]
|
||||
dep_cache.keys = mx.depends(dep_cache.keys, logits) # pyright: ignore[reportAny,reportUnknownMemberType]
|
||||
cache[-1].state = mx.depends(cache[-1].state, logits) # pyright: ignore[reportAny,reportUnknownMemberType]
|
||||
|
||||
return logits
|
||||
|
||||
@@ -554,12 +547,10 @@ class DeepSeekShardingStrategy(TensorParallelShardingStrategy):
|
||||
on_timeout: TimeoutCallback | None,
|
||||
) -> nn.Module:
|
||||
model = cast(DeepseekV3Model, model)
|
||||
|
||||
for layer in model.layers:
|
||||
eval_with_timeout(
|
||||
layer.parameters(), timeout_seconds / len(model.layers), on_timeout
|
||||
)
|
||||
|
||||
# Shard the self attention
|
||||
if layer.self_attn.q_lora_rank is None:
|
||||
layer.self_attn.q_proj = self.all_to_sharded_linear(
|
||||
@@ -590,18 +581,12 @@ class DeepSeekShardingStrategy(TensorParallelShardingStrategy):
|
||||
layer.mlp.down_proj = self.sharded_to_all_linear(layer.mlp.down_proj)
|
||||
layer.mlp.up_proj = self.all_to_sharded_linear(layer.mlp.up_proj)
|
||||
|
||||
# Shard the MoE.
|
||||
# Shard the MoE. Shard in place since the MoE should be responsible
|
||||
# for aggregating the results.
|
||||
else:
|
||||
if getattr(layer.mlp, "shared_experts", None) is not None:
|
||||
self.all_to_sharded_linear_in_place(
|
||||
layer.mlp.shared_experts.gate_proj
|
||||
)
|
||||
self.sharded_to_all_linear_in_place(
|
||||
layer.mlp.shared_experts.down_proj
|
||||
)
|
||||
self.all_to_sharded_linear_in_place(
|
||||
layer.mlp.shared_experts.up_proj
|
||||
)
|
||||
self.all_to_sharded_linear_in_place(layer.mlp.shared_experts.gate_proj)
|
||||
self.sharded_to_all_linear_in_place(layer.mlp.shared_experts.down_proj)
|
||||
self.all_to_sharded_linear_in_place(layer.mlp.shared_experts.up_proj)
|
||||
self.all_to_sharded_linear_in_place(layer.mlp.switch_mlp.gate_proj)
|
||||
self.sharded_to_all_linear_in_place(layer.mlp.switch_mlp.down_proj)
|
||||
self.all_to_sharded_linear_in_place(layer.mlp.switch_mlp.up_proj)
|
||||
@@ -794,7 +779,8 @@ class MiniMaxShardingStrategy(TensorParallelShardingStrategy):
|
||||
|
||||
layer.self_attn = WrappedMiniMaxAttention(layer.self_attn, self.group) # pyright: ignore[reportAttributeAccessIssue,reportArgumentType]
|
||||
|
||||
# Shard the MoE.
|
||||
# Shard the MoE. Shard in place since the MoE should be responsible
|
||||
# for aggregating the results.
|
||||
self.all_to_sharded_linear_in_place(
|
||||
layer.block_sparse_moe.switch_mlp.gate_proj
|
||||
)
|
||||
@@ -907,7 +893,8 @@ class QwenShardingStrategy(TensorParallelShardingStrategy):
|
||||
layer.self_attn.num_attention_heads //= self.N
|
||||
layer.self_attn.num_key_value_heads //= self.N
|
||||
|
||||
# Shard the MoE.
|
||||
# Shard the MoE. Shard in place since the MoE should be responsible
|
||||
# for aggregating the results.
|
||||
if isinstance(layer.mlp, (Qwen3MoeSparseMoeBlock, Qwen3NextSparseMoeBlock)):
|
||||
self.all_to_sharded_linear_in_place(layer.mlp.switch_mlp.gate_proj)
|
||||
self.sharded_to_all_linear_in_place(layer.mlp.switch_mlp.down_proj)
|
||||
|
||||
@@ -5,7 +5,6 @@ import mlx.core as mx
|
||||
import psutil
|
||||
from mlx_lm.models.cache import (
|
||||
ArraysCache,
|
||||
CacheList,
|
||||
KVCache,
|
||||
QuantizedKVCache,
|
||||
RotatingKVCache,
|
||||
@@ -18,22 +17,10 @@ from exo.worker.engines.mlx import Model
|
||||
from exo.worker.engines.mlx.constants import CACHE_GROUP_SIZE, KV_CACHE_BITS
|
||||
from exo.worker.runner.bootstrap import logger
|
||||
|
||||
|
||||
# Fraction of device memory above which LRU eviction kicks in.
|
||||
# Smaller machines need more aggressive eviction.
|
||||
def _default_memory_threshold() -> float:
|
||||
total_gb = psutil.virtual_memory().total / (1024**3)
|
||||
if total_gb >= 128:
|
||||
return 0.85
|
||||
if total_gb >= 64:
|
||||
return 0.80
|
||||
if total_gb >= 32:
|
||||
return 0.75
|
||||
return 0.70
|
||||
|
||||
|
||||
# Fraction of device memory above which LRU eviction kicks in
|
||||
_DEFAULT_MEMORY_THRESHOLD = 0.9
|
||||
_MEMORY_THRESHOLD = float(
|
||||
os.environ.get("EXO_MEMORY_THRESHOLD", _default_memory_threshold())
|
||||
os.environ.get("EXO_MEMORY_THRESHOLD", _DEFAULT_MEMORY_THRESHOLD)
|
||||
)
|
||||
|
||||
|
||||
@@ -77,7 +64,7 @@ def has_non_kv_caches(cache: KVCacheType) -> bool:
|
||||
|
||||
|
||||
class KVPrefixCache:
|
||||
def __init__(self, group: mx.distributed.Group | None):
|
||||
def __init__(self, group: mx.distributed.Group | None = None):
|
||||
self.prompts: list[mx.array] = [] # mx array of tokens (ints)
|
||||
self.caches: list[KVCacheType] = []
|
||||
self._snapshots: list[list[CacheSnapshot] | None] = []
|
||||
@@ -169,15 +156,15 @@ class KVPrefixCache:
|
||||
best_length = 0
|
||||
is_exact = False
|
||||
|
||||
# Find best cache match
|
||||
# Find best cache
|
||||
for i, cached_prompt in enumerate(self.prompts):
|
||||
length = get_prefix_length(prompt_tokens, cached_prompt)
|
||||
if length >= max_length - 1:
|
||||
best_index, best_length = i, length
|
||||
is_exact = True
|
||||
break
|
||||
if length > best_length:
|
||||
best_index, best_length = i, length
|
||||
if length == max_length:
|
||||
is_exact = True
|
||||
best_index, best_length = i, length
|
||||
break
|
||||
|
||||
if best_index is None:
|
||||
return make_kv_cache(model), prompt_tokens, None
|
||||
@@ -185,12 +172,11 @@ class KVPrefixCache:
|
||||
# For exact match: trim to max_length-1 so remaining has the last token
|
||||
# For partial match: trim to best_length, remaining has suffix to prefill
|
||||
# This ensures stream_generate always has at least one token to start with
|
||||
has_ssm = has_non_kv_caches(self.caches[best_index])
|
||||
target = (max_length - 1) if is_exact and not has_ssm else best_length
|
||||
target = (max_length - 1) if is_exact else best_length
|
||||
restore_pos, restore_snap = self._get_snapshot(best_index, target)
|
||||
|
||||
# No usable snapshot — need fresh cache
|
||||
if restore_snap is None and has_ssm:
|
||||
if restore_snap is None and has_non_kv_caches(self.caches[best_index]):
|
||||
return make_kv_cache(model), prompt_tokens, None
|
||||
|
||||
prompt_cache = deepcopy(self.caches[best_index])
|
||||
@@ -271,21 +257,10 @@ def encode_prompt(tokenizer: TokenizerWrapper, prompt: str) -> mx.array:
|
||||
return mx.array(prompt_tokens)
|
||||
|
||||
|
||||
def _entry_length(
|
||||
c: KVCache | RotatingKVCache | QuantizedKVCache | ArraysCache | CacheList,
|
||||
) -> int:
|
||||
# Use .offset attribute which KVCache types have (len() not implemented in older QuantizedKVCache).
|
||||
if hasattr(c, "offset"):
|
||||
return c.offset
|
||||
# For CacheList
|
||||
if hasattr(c, "size"):
|
||||
return int(c.size()) # type: ignore
|
||||
return 0
|
||||
|
||||
|
||||
def cache_length(cache: KVCacheType) -> int:
|
||||
"""Get the number of tokens in a KV cache."""
|
||||
return max(_entry_length(c) for c in cache)
|
||||
# Use .offset attribute which KVCache types have (len() not implemented in older QuantizedKVCache).
|
||||
return max(getattr(c, "offset", 0) for c in cache)
|
||||
|
||||
|
||||
def get_prefix_length(prompt: mx.array, cached_prompt: mx.array) -> int:
|
||||
|
||||
@@ -48,7 +48,7 @@ from exo.worker.runner.bootstrap import logger
|
||||
|
||||
generation_stream = mx.new_stream(mx.default_device())
|
||||
|
||||
_MIN_PREFIX_HIT_RATIO_TO_UPDATE = 0.5
|
||||
_MIN_PREFIX_HIT_TO_UPDATE = 1000
|
||||
|
||||
|
||||
def prefill(
|
||||
@@ -57,7 +57,6 @@ def prefill(
|
||||
sampler: Callable[[mx.array], mx.array],
|
||||
prompt_tokens: mx.array,
|
||||
cache: KVCacheType,
|
||||
group: mx.distributed.Group | None,
|
||||
) -> tuple[float, int, list[CacheSnapshot]]:
|
||||
"""Prefill the KV cache with prompt tokens.
|
||||
|
||||
@@ -87,9 +86,6 @@ def prefill(
|
||||
|
||||
set_pipeline_prefill(model, is_prefill=True)
|
||||
|
||||
mx_barrier(group)
|
||||
logger.info("Starting prefill")
|
||||
|
||||
# Use max_tokens=1 because max_tokens=0 does not work.
|
||||
# We just throw away the generated token - we only care about filling the cache
|
||||
for _ in stream_generate(
|
||||
@@ -133,7 +129,7 @@ def prefill(
|
||||
def warmup_inference(
|
||||
model: Model,
|
||||
tokenizer: TokenizerWrapper,
|
||||
group: mx.distributed.Group | None,
|
||||
group: mx.distributed.Group | None = None,
|
||||
) -> int:
|
||||
content = "Prompt to warm up the inference engine. Repeat this."
|
||||
|
||||
@@ -255,8 +251,8 @@ def mlx_generate(
|
||||
tokenizer: TokenizerWrapper,
|
||||
task: TextGenerationTaskParams,
|
||||
prompt: str,
|
||||
kv_prefix_cache: KVPrefixCache | None,
|
||||
group: mx.distributed.Group | None,
|
||||
kv_prefix_cache: KVPrefixCache | None = None,
|
||||
group: mx.distributed.Group | None = None,
|
||||
) -> Generator[GenerationResponse]:
|
||||
# Ensure that generation stats only contains peak memory for this generation
|
||||
mx.reset_peak_memory()
|
||||
@@ -309,9 +305,16 @@ def mlx_generate(
|
||||
)
|
||||
max_stop_len = max((len(s) for s in stop_sequences), default=0)
|
||||
|
||||
mx_barrier(group)
|
||||
logger.info("Starting prefill")
|
||||
|
||||
# Prefill cache with all tokens except the last one
|
||||
prefill_tps, prefill_tokens, ssm_snapshots_list = prefill(
|
||||
model, tokenizer, sampler, prompt_tokens[:-1], caches, group
|
||||
model,
|
||||
tokenizer,
|
||||
sampler,
|
||||
prompt_tokens[:-1],
|
||||
caches,
|
||||
)
|
||||
cache_snapshots: list[CacheSnapshot] | None = ssm_snapshots_list or None
|
||||
|
||||
@@ -328,7 +331,6 @@ def mlx_generate(
|
||||
think_start = tokenizer.think_start
|
||||
think_end = tokenizer.think_end
|
||||
|
||||
logger.info("Starting decode")
|
||||
mx_barrier(group)
|
||||
|
||||
for completion_tokens, out in enumerate(
|
||||
@@ -436,14 +438,9 @@ def mlx_generate(
|
||||
full_prompt_tokens = mx.concatenate(
|
||||
[all_prompt_tokens, generated_tokens_array]
|
||||
)
|
||||
hit_ratio = (
|
||||
prefix_hit_length / len(all_prompt_tokens)
|
||||
if len(all_prompt_tokens) > 0
|
||||
else 0.0
|
||||
)
|
||||
if (
|
||||
matched_index is not None
|
||||
and hit_ratio >= _MIN_PREFIX_HIT_RATIO_TO_UPDATE
|
||||
and prefix_hit_length >= _MIN_PREFIX_HIT_TO_UPDATE
|
||||
):
|
||||
kv_prefix_cache.update_kv_cache(
|
||||
matched_index,
|
||||
|
||||
@@ -285,15 +285,11 @@ def get_eos_token_ids_for_model(model_id: ModelId) -> list[int] | None:
|
||||
model_id_lower = model_id.lower()
|
||||
if "kimi-k2" in model_id_lower:
|
||||
return [163586]
|
||||
elif "glm-5" in model_id_lower or "glm-4.7" in model_id_lower:
|
||||
# For GLM-5 and GLM-4.7
|
||||
elif "glm-4.7-flash" in model_id_lower:
|
||||
# 154820: <|endoftext|>, 154827: <|user|>, 154829: <|observation|>
|
||||
return [154820, 154827, 154829]
|
||||
elif "glm" in model_id_lower:
|
||||
# For GLM-4.5 and older
|
||||
return [151336, 151329, 151338]
|
||||
elif "gpt-oss" in model_id_lower:
|
||||
return [200002, 200012]
|
||||
return None
|
||||
|
||||
|
||||
|
||||
@@ -11,7 +11,6 @@ from mlx_lm.models.gpt_oss import Model as GptOssModel
|
||||
from mlx_lm.tokenizer_utils import TokenizerWrapper
|
||||
from openai_harmony import ( # pyright: ignore[reportMissingTypeStubs]
|
||||
HarmonyEncodingName,
|
||||
HarmonyError, # pyright: ignore[reportUnknownVariableType]
|
||||
Role,
|
||||
StreamableParser,
|
||||
load_harmony_encoding,
|
||||
@@ -589,11 +588,7 @@ def parse_gpt_oss(
|
||||
|
||||
for response in responses:
|
||||
assert isinstance(response, GenerationResponse)
|
||||
try:
|
||||
stream.process(response.token)
|
||||
except HarmonyError:
|
||||
logger.error("Encountered critical Harmony Error, returning early")
|
||||
return
|
||||
stream.process(response.token)
|
||||
|
||||
delta = stream.last_content_delta
|
||||
ch = stream.current_channel
|
||||
|
||||
@@ -103,7 +103,7 @@ class RunnerSupervisor:
|
||||
self._event_sender.close()
|
||||
self._cancel_sender.send(TaskId("CANCEL_CURRENT_TASK"))
|
||||
self._cancel_sender.close()
|
||||
self.runner_process.join(5)
|
||||
self.runner_process.join(1)
|
||||
if not self.runner_process.is_alive():
|
||||
logger.info("Runner process succesfully terminated")
|
||||
return
|
||||
@@ -191,7 +191,7 @@ class RunnerSupervisor:
|
||||
logger.info("Checking runner's status")
|
||||
if self.runner_process.is_alive():
|
||||
logger.info("Runner was found to be alive, attempting to join process")
|
||||
await to_thread.run_sync(self.runner_process.join, 5)
|
||||
await to_thread.run_sync(self.runner_process.join, 1)
|
||||
rc = self.runner_process.exitcode
|
||||
logger.info(f"RunnerSupervisor exited with exit code {rc}")
|
||||
if rc == 0:
|
||||
|
||||
@@ -123,12 +123,7 @@ def run_gpt_oss_pipeline_device(
|
||||
generated_text = ""
|
||||
|
||||
for response in mlx_generate(
|
||||
model=model,
|
||||
tokenizer=tokenizer,
|
||||
task=task,
|
||||
prompt=prompt,
|
||||
kv_prefix_cache=None,
|
||||
group=group,
|
||||
model=model, tokenizer=tokenizer, task=task, prompt=prompt
|
||||
):
|
||||
generated_text += response.text
|
||||
if response.finish_reason is not None:
|
||||
@@ -199,8 +194,6 @@ def run_gpt_oss_tensor_parallel_device(
|
||||
tokenizer=tokenizer,
|
||||
task=task,
|
||||
prompt=prompt,
|
||||
kv_prefix_cache=None,
|
||||
group=group,
|
||||
):
|
||||
generated_text += response.text
|
||||
if response.finish_reason is not None:
|
||||
|
||||
@@ -88,12 +88,12 @@ class TestKVPrefix:
|
||||
return tokenizer
|
||||
|
||||
def test_starts_empty(self, mock_tokenizer):
|
||||
cache = KVPrefixCache(None)
|
||||
cache = KVPrefixCache()
|
||||
assert len(cache.prompts) == 0
|
||||
assert len(cache.caches) == 0
|
||||
|
||||
def test_clear_empties_cache(self, mock_tokenizer):
|
||||
cache = KVPrefixCache(None)
|
||||
cache = KVPrefixCache()
|
||||
cache.prompts.append(mx.array([1, 2, 3]))
|
||||
cache.caches.append([KVCache()])
|
||||
cache.clear()
|
||||
@@ -101,7 +101,7 @@ class TestKVPrefix:
|
||||
assert len(cache.caches) == 0
|
||||
|
||||
def test_clear_on_empty_cache(self, mock_tokenizer):
|
||||
cache = KVPrefixCache(None)
|
||||
cache = KVPrefixCache()
|
||||
cache.clear()
|
||||
assert len(cache.prompts) == 0
|
||||
|
||||
@@ -142,9 +142,7 @@ class TestKVPrefixCacheWithModel:
|
||||
tokens = encode_prompt(tokenizer, prompt)
|
||||
cache = make_kv_cache(model)
|
||||
|
||||
_, _, snapshots = prefill(
|
||||
model, tokenizer, make_sampler(0.0), tokens, cache, group=None
|
||||
)
|
||||
_, _, snapshots = prefill(model, tokenizer, make_sampler(0.0), tokens, cache)
|
||||
|
||||
# Cache should now hold the prompt tokens minus one
|
||||
assert cache_length(cache) == len(tokens) - 1
|
||||
@@ -163,11 +161,9 @@ class TestKVPrefixCacheWithModel:
|
||||
tokens = encode_prompt(tokenizer, prompt)
|
||||
cache = make_kv_cache(model)
|
||||
|
||||
_, _, snapshots = prefill(
|
||||
model, tokenizer, make_sampler(0.0), tokens, cache, group=None
|
||||
)
|
||||
_, _, snapshots = prefill(model, tokenizer, make_sampler(0.0), tokens, cache)
|
||||
|
||||
kv_prefix_cache = KVPrefixCache(None)
|
||||
kv_prefix_cache = KVPrefixCache()
|
||||
kv_prefix_cache.add_kv_cache(tokens, cache, snapshots)
|
||||
|
||||
assert len(kv_prefix_cache.prompts) == 1
|
||||
@@ -180,11 +176,9 @@ class TestKVPrefixCacheWithModel:
|
||||
)
|
||||
assert matched_index == 0
|
||||
|
||||
# Exact match returns last token(s) — for models with SSM/rotating caches,
|
||||
# snapshot availability constrains how far back we can trim, so remaining
|
||||
# may be 1 or 2 tokens depending on the model.
|
||||
assert len(remaining_tokens) >= 1
|
||||
assert mx.array_equal(remaining_tokens, tokens[-len(remaining_tokens) :])
|
||||
# Exact match returns only last token
|
||||
assert len(remaining_tokens) == 1
|
||||
assert mx.array_equal(remaining_tokens, tokens[-1:])
|
||||
|
||||
def test_add_and_get_prefix_match(self, model_and_tokenizer):
|
||||
"""get_kv_cache with a longer prompt sharing prefix should return partial match."""
|
||||
@@ -200,10 +194,10 @@ class TestKVPrefixCacheWithModel:
|
||||
cache = make_kv_cache(model)
|
||||
|
||||
_, _, snapshots = prefill(
|
||||
model, tokenizer, make_sampler(0.0), short_tokens, cache, group=None
|
||||
model, tokenizer, make_sampler(0.0), short_tokens, cache
|
||||
)
|
||||
|
||||
kv_prefix_cache = KVPrefixCache(None)
|
||||
kv_prefix_cache = KVPrefixCache()
|
||||
kv_prefix_cache.add_kv_cache(short_tokens, cache, snapshots)
|
||||
|
||||
# Query with longer prompt that shares the chat template prefix
|
||||
@@ -244,11 +238,9 @@ class TestKVPrefixCacheWithModel:
|
||||
tokens = encode_prompt(tokenizer, prompt)
|
||||
cache = make_kv_cache(model)
|
||||
|
||||
_, _, snapshots = prefill(
|
||||
model, tokenizer, make_sampler(0.0), tokens, cache, group=None
|
||||
)
|
||||
_, _, snapshots = prefill(model, tokenizer, make_sampler(0.0), tokens, cache)
|
||||
|
||||
kv_prefix_cache = KVPrefixCache(None)
|
||||
kv_prefix_cache = KVPrefixCache()
|
||||
kv_prefix_cache.add_kv_cache(tokens, cache, snapshots)
|
||||
|
||||
stored_length = cache_length(kv_prefix_cache.caches[0])
|
||||
@@ -284,11 +276,9 @@ class TestKVPrefixCacheWithModel:
|
||||
tokens = encode_prompt(tokenizer, prompt)
|
||||
cache = make_kv_cache(model)
|
||||
|
||||
_, _, snapshots = prefill(
|
||||
model, tokenizer, make_sampler(0.0), tokens, cache, group=None
|
||||
)
|
||||
_, _, snapshots = prefill(model, tokenizer, make_sampler(0.0), tokens, cache)
|
||||
|
||||
kv_prefix_cache = KVPrefixCache(None)
|
||||
kv_prefix_cache = KVPrefixCache()
|
||||
kv_prefix_cache.add_kv_cache(tokens, cache, snapshots)
|
||||
|
||||
stored_length = cache_length(kv_prefix_cache.caches[0])
|
||||
@@ -311,7 +301,7 @@ class TestKVPrefixCacheWithModel:
|
||||
"""mlx_generate should save the cache after generation completes."""
|
||||
model, tokenizer = model_and_tokenizer
|
||||
|
||||
kv_prefix_cache = KVPrefixCache(None)
|
||||
kv_prefix_cache = KVPrefixCache()
|
||||
task = TextGenerationTaskParams(
|
||||
model=DEFAULT_GPT_OSS_MODEL_ID,
|
||||
input=[InputMessage(role="user", content="Hello")],
|
||||
@@ -328,7 +318,6 @@ class TestKVPrefixCacheWithModel:
|
||||
task=task,
|
||||
prompt=prompt,
|
||||
kv_prefix_cache=kv_prefix_cache,
|
||||
group=None,
|
||||
):
|
||||
generated_tokens += 1
|
||||
|
||||
@@ -342,7 +331,7 @@ class TestKVPrefixCacheWithModel:
|
||||
"""Second mlx_generate call with same prompt should get a prefix hit from stored cache."""
|
||||
model, tokenizer = model_and_tokenizer
|
||||
|
||||
kv_prefix_cache = KVPrefixCache(None)
|
||||
kv_prefix_cache = KVPrefixCache()
|
||||
task = TextGenerationTaskParams(
|
||||
model=DEFAULT_GPT_OSS_MODEL_ID,
|
||||
input=[InputMessage(role="user", content="Reuse test")],
|
||||
@@ -358,7 +347,6 @@ class TestKVPrefixCacheWithModel:
|
||||
task=task,
|
||||
prompt=prompt,
|
||||
kv_prefix_cache=kv_prefix_cache,
|
||||
group=None,
|
||||
):
|
||||
pass
|
||||
|
||||
@@ -380,7 +368,7 @@ class TestKVPrefixCacheWithModel:
|
||||
"""With a prompt > 1000 tokens, second generation should update the cache entry in-place."""
|
||||
model, tokenizer = model_and_tokenizer
|
||||
|
||||
kv_prefix_cache = KVPrefixCache(None)
|
||||
kv_prefix_cache = KVPrefixCache()
|
||||
|
||||
# Build a long user message (> 1000 tokens) to exceed _MIN_PREFIX_HIT_TO_UPDATE
|
||||
base_text = "The quick brown fox jumps over the lazy dog. "
|
||||
@@ -407,7 +395,6 @@ class TestKVPrefixCacheWithModel:
|
||||
task=task1,
|
||||
prompt=prompt1,
|
||||
kv_prefix_cache=kv_prefix_cache,
|
||||
group=None,
|
||||
):
|
||||
pass
|
||||
first_gen_time = time.perf_counter() - t0
|
||||
@@ -440,7 +427,6 @@ class TestKVPrefixCacheWithModel:
|
||||
task=task2,
|
||||
prompt=prompt2,
|
||||
kv_prefix_cache=kv_prefix_cache,
|
||||
group=None,
|
||||
):
|
||||
pass
|
||||
second_gen_time = time.perf_counter() - t0
|
||||
@@ -461,7 +447,7 @@ class TestKVPrefixCacheWithModel:
|
||||
"""After mlx_generate saves a cache, a second generation must not corrupt the stored copy."""
|
||||
model, tokenizer = model_and_tokenizer
|
||||
|
||||
kv_prefix_cache = KVPrefixCache(None)
|
||||
kv_prefix_cache = KVPrefixCache()
|
||||
task = TextGenerationTaskParams(
|
||||
model=DEFAULT_GPT_OSS_MODEL_ID,
|
||||
input=[InputMessage(role="user", content="Immutable test")],
|
||||
@@ -476,7 +462,6 @@ class TestKVPrefixCacheWithModel:
|
||||
task=task,
|
||||
prompt=prompt,
|
||||
kv_prefix_cache=kv_prefix_cache,
|
||||
group=None,
|
||||
):
|
||||
pass
|
||||
|
||||
@@ -489,7 +474,6 @@ class TestKVPrefixCacheWithModel:
|
||||
task=task,
|
||||
prompt=prompt,
|
||||
kv_prefix_cache=kv_prefix_cache,
|
||||
group=None,
|
||||
):
|
||||
pass
|
||||
|
||||
@@ -500,7 +484,7 @@ class TestKVPrefixCacheWithModel:
|
||||
"""Under memory pressure, adding a new cache entry evicts the least recently used one."""
|
||||
model, tokenizer = model_and_tokenizer
|
||||
|
||||
kv_prefix_cache = KVPrefixCache(None)
|
||||
kv_prefix_cache = KVPrefixCache()
|
||||
|
||||
# Add three cache entries with different prompts
|
||||
prompts = ["First entry", "Second entry", "Third entry"]
|
||||
@@ -513,7 +497,7 @@ class TestKVPrefixCacheWithModel:
|
||||
prompt = apply_chat_template(tokenizer, task)
|
||||
tokens = encode_prompt(tokenizer, prompt)
|
||||
cache = make_kv_cache(model)
|
||||
prefill(model, tokenizer, make_sampler(0.0), tokens, cache, group=None)
|
||||
prefill(model, tokenizer, make_sampler(0.0), tokens, cache)
|
||||
kv_prefix_cache.add_kv_cache(tokens, cache)
|
||||
# Stagger _last_used so LRU order is deterministic
|
||||
kv_prefix_cache._last_used[i] = float(i)
|
||||
@@ -538,7 +522,7 @@ class TestKVPrefixCacheWithModel:
|
||||
prompt = apply_chat_template(tokenizer, task)
|
||||
tokens = encode_prompt(tokenizer, prompt)
|
||||
cache = make_kv_cache(model)
|
||||
prefill(model, tokenizer, make_sampler(0.0), tokens, cache, group=None)
|
||||
prefill(model, tokenizer, make_sampler(0.0), tokens, cache)
|
||||
kv_prefix_cache.add_kv_cache(tokens, cache)
|
||||
|
||||
# LRU entries should have been evicted (entries 0, 1, 2 in order of _last_used)
|
||||
|
||||
@@ -1,297 +0,0 @@
|
||||
import copy
|
||||
import gc
|
||||
import importlib
|
||||
import json
|
||||
import shutil
|
||||
import tempfile
|
||||
from dataclasses import dataclass
|
||||
from pathlib import Path
|
||||
from typing import Any, cast
|
||||
|
||||
import mlx.core as mx
|
||||
import mlx.nn as nn
|
||||
import pytest
|
||||
from mlx.utils import tree_flatten, tree_unflatten
|
||||
from mlx_lm.tokenizer_utils import TokenizerWrapper
|
||||
|
||||
from exo.shared.types.common import ModelId
|
||||
from exo.shared.types.text_generation import InputMessage, TextGenerationTaskParams
|
||||
from exo.worker.engines.mlx import Model
|
||||
from exo.worker.engines.mlx.cache import KVPrefixCache
|
||||
from exo.worker.engines.mlx.generator.generate import mlx_generate
|
||||
from exo.worker.engines.mlx.utils_mlx import (
|
||||
apply_chat_template,
|
||||
load_tokenizer_for_model_id,
|
||||
)
|
||||
|
||||
HF_CACHE = Path.home() / ".cache" / "huggingface" / "hub"
|
||||
|
||||
# ── Config reduction ──────────────────────────────────────────────────────── #
|
||||
|
||||
_REDUCE = {
|
||||
"num_hidden_layers": 4,
|
||||
"hidden_size": 256,
|
||||
"num_attention_heads": 4,
|
||||
"num_key_value_heads": 4,
|
||||
"intermediate_size": 512,
|
||||
"moe_intermediate_size": 128,
|
||||
"num_experts": 4,
|
||||
"num_experts_per_tok": 2,
|
||||
"n_routed_experts": 4,
|
||||
"num_local_experts": 4,
|
||||
"num_nextn_predict_layers": 0,
|
||||
"first_k_dense_replace": 0,
|
||||
"linear_num_key_heads": 2,
|
||||
"linear_num_value_heads": 2,
|
||||
"num_attention_groups": 4,
|
||||
}
|
||||
|
||||
|
||||
def _reduce_dict(cfg: dict[str, Any]) -> dict[str, Any]:
|
||||
result = dict(cfg)
|
||||
for key, val in _REDUCE.items():
|
||||
if key in result:
|
||||
result[key] = val
|
||||
return result
|
||||
|
||||
|
||||
def _reduce_config(cfg: dict[str, Any]) -> dict[str, Any]:
|
||||
result = _reduce_dict(cfg)
|
||||
n_layers = cast(int, result.get("num_hidden_layers", 4))
|
||||
|
||||
if "text_config" in result and isinstance(result["text_config"], dict):
|
||||
result["text_config"] = _reduce_dict(
|
||||
cast(dict[str, Any], result["text_config"])
|
||||
)
|
||||
tc: dict[str, Any] = result["text_config"]
|
||||
if "num_nextn_predict_layers" in tc:
|
||||
tc["num_nextn_predict_layers"] = 0
|
||||
|
||||
if "layer_types" in result and isinstance(result["layer_types"], list):
|
||||
result["layer_types"] = result["layer_types"][:n_layers]
|
||||
|
||||
if "attention_other_setting" in result and isinstance(
|
||||
result["attention_other_setting"], dict
|
||||
):
|
||||
aos: dict[str, Any] = dict(
|
||||
cast(dict[str, Any], result["attention_other_setting"])
|
||||
)
|
||||
if "num_attention_heads" in aos:
|
||||
aos["num_attention_heads"] = result.get("num_attention_heads", 4)
|
||||
if "num_attention_groups" in aos:
|
||||
aos["num_attention_groups"] = result.get(
|
||||
"num_attention_groups", cast(int, aos["num_attention_groups"])
|
||||
)
|
||||
result["attention_other_setting"] = aos
|
||||
|
||||
if "moe_layers_enum" in result and isinstance(result["moe_layers_enum"], str):
|
||||
indices = [int(x) for x in result["moe_layers_enum"].split(",") if x.strip()]
|
||||
valid = [i for i in indices if i < n_layers]
|
||||
result["moe_layers_enum"] = ",".join(str(i) for i in valid) if valid else ""
|
||||
|
||||
return result
|
||||
|
||||
|
||||
# ── Helpers ───────────────────────────────────────────────────────────────── #
|
||||
|
||||
|
||||
def _find_snapshot(hub_name: str) -> Path | None:
|
||||
model_dir = HF_CACHE / f"models--mlx-community--{hub_name}"
|
||||
snaps = model_dir / "snapshots"
|
||||
if not snaps.exists():
|
||||
return None
|
||||
children = sorted(snaps.iterdir())
|
||||
return children[0] if children else None
|
||||
|
||||
|
||||
def _copy_tokenizer(src: Path, dst: Path) -> None:
|
||||
for f in src.iterdir():
|
||||
name = f.name
|
||||
if (
|
||||
"tokeniz" in name.lower()
|
||||
or "tiktoken" in name.lower()
|
||||
or name.startswith("vocab")
|
||||
or name.endswith(".jinja")
|
||||
or "tool_declaration" in name
|
||||
) and f.is_file():
|
||||
shutil.copy2(f, dst / name)
|
||||
|
||||
|
||||
def _build_model(module_name: str, cfg: dict[str, Any]) -> Model:
|
||||
mod = importlib.import_module(f"mlx_lm.models.{module_name}")
|
||||
args = mod.ModelArgs.from_dict(cfg) # pyright: ignore[reportAny]
|
||||
model: nn.Module = mod.Model(args) # pyright: ignore[reportAny]
|
||||
flat = cast(list[tuple[str, mx.array]], tree_flatten(model.parameters()))
|
||||
random_weights = [
|
||||
(k, mx.random.normal(shape=v.shape, dtype=mx.float16)) for k, v in flat
|
||||
]
|
||||
model.update(cast(dict[str, Any], tree_unflatten(random_weights)))
|
||||
mx.eval(model.parameters())
|
||||
return cast(Model, model)
|
||||
|
||||
|
||||
def _collect_tokens(
|
||||
model: Model,
|
||||
tokenizer: TokenizerWrapper,
|
||||
task: TextGenerationTaskParams,
|
||||
prompt: str,
|
||||
kv_prefix_cache: KVPrefixCache | None,
|
||||
) -> list[int]:
|
||||
tokens: list[int] = []
|
||||
for resp in mlx_generate(
|
||||
model=model,
|
||||
tokenizer=tokenizer,
|
||||
task=task,
|
||||
prompt=prompt,
|
||||
kv_prefix_cache=kv_prefix_cache,
|
||||
group=None,
|
||||
):
|
||||
tokens.append(resp.token)
|
||||
if resp.finish_reason is not None:
|
||||
break
|
||||
return tokens
|
||||
|
||||
|
||||
# ── Architecture definitions ──────────────────────────────────────────────── #
|
||||
|
||||
|
||||
@dataclass(frozen=True)
|
||||
class ArchSpec:
|
||||
name: str
|
||||
hub_name: str
|
||||
module: str
|
||||
tokenizer_hub: str | None = None # fallback for models without bundled tokenizer
|
||||
|
||||
|
||||
ARCHITECTURES: list[ArchSpec] = [
|
||||
ArchSpec("llama", "Llama-3.2-1B-Instruct-4bit", "llama"),
|
||||
ArchSpec("glm_moe_dsa", "GLM-5-MXFP4-Q8", "glm_moe_dsa"),
|
||||
ArchSpec(
|
||||
"glm4_moe", "GLM-4.5-Air-8bit", "glm4_moe", tokenizer_hub="GLM-4.7-8bit-gs32"
|
||||
),
|
||||
ArchSpec(
|
||||
"glm4_moe_lite",
|
||||
"GLM-4.7-Flash-8bit",
|
||||
"glm4_moe_lite",
|
||||
tokenizer_hub="GLM-4.7-8bit-gs32",
|
||||
),
|
||||
ArchSpec("glm4_moe_47", "GLM-4.7-8bit-gs32", "glm4_moe"),
|
||||
ArchSpec("qwen3", "Qwen3-4B-Instruct-2507-4bit", "qwen3"),
|
||||
ArchSpec("qwen3_moe", "Qwen3-30B-A3B-4bit", "qwen3_moe"),
|
||||
ArchSpec("qwen3_next", "Qwen3-Next-80B-A3B-Thinking-4bit", "qwen3_next"),
|
||||
ArchSpec("minimax", "MiniMax-M2.1-3bit", "minimax"),
|
||||
ArchSpec("gpt_oss", "gpt-oss-20b-MXFP4-Q8", "gpt_oss"),
|
||||
ArchSpec("step3p5", "Step-3.5-Flash-4bit", "step3p5"),
|
||||
ArchSpec("kimi_k25", "Kimi-K2.5", "kimi_k25"),
|
||||
]
|
||||
|
||||
|
||||
def _arch_available(spec: ArchSpec) -> bool:
|
||||
snap = _find_snapshot(spec.hub_name)
|
||||
if snap is None:
|
||||
return False
|
||||
if spec.tokenizer_hub is not None:
|
||||
return _find_snapshot(spec.tokenizer_hub) is not None
|
||||
return True
|
||||
|
||||
|
||||
def _make_task() -> TextGenerationTaskParams:
|
||||
return TextGenerationTaskParams(
|
||||
model=ModelId("test"),
|
||||
input=[
|
||||
InputMessage(
|
||||
role="user",
|
||||
content="Use the calculator to compute 1847 * 263 + 5921",
|
||||
)
|
||||
],
|
||||
max_output_tokens=20,
|
||||
temperature=0.0,
|
||||
tools=[
|
||||
{
|
||||
"type": "function",
|
||||
"function": {
|
||||
"name": "calculate",
|
||||
"description": "Evaluate a mathematical expression",
|
||||
"parameters": {
|
||||
"type": "object",
|
||||
"properties": {"expression": {"type": "string"}},
|
||||
"required": ["expression"],
|
||||
},
|
||||
},
|
||||
}
|
||||
],
|
||||
)
|
||||
|
||||
|
||||
# ── Test class ────────────────────────────────────────────────────────────── #
|
||||
|
||||
|
||||
@pytest.mark.slow
|
||||
class TestPrefixCacheArchitectures:
|
||||
"""Verify prefix cache produces identical output to fresh generation for every architecture."""
|
||||
|
||||
@pytest.fixture(autouse=True)
|
||||
def _cleanup(self):
|
||||
yield
|
||||
mx.clear_cache()
|
||||
gc.collect()
|
||||
|
||||
@pytest.mark.parametrize(
|
||||
"spec",
|
||||
ARCHITECTURES,
|
||||
ids=[a.name for a in ARCHITECTURES],
|
||||
)
|
||||
def test_prefix_cache_exact_hit(self, spec: ArchSpec) -> None:
|
||||
if not _arch_available(spec):
|
||||
pytest.skip(f"Model {spec.hub_name} not cached locally")
|
||||
|
||||
snapshot = _find_snapshot(spec.hub_name)
|
||||
assert snapshot is not None
|
||||
|
||||
tmpdir = Path(tempfile.mkdtemp(prefix=f"exo_test_{spec.name}_"))
|
||||
try:
|
||||
# Build reduced config
|
||||
with open(snapshot / "config.json") as f:
|
||||
cfg = cast(dict[str, Any], json.load(f))
|
||||
reduced = _reduce_config(copy.deepcopy(cfg))
|
||||
(tmpdir / "config.json").write_text(json.dumps(reduced))
|
||||
|
||||
# Copy tokenizer
|
||||
tok_src = snapshot
|
||||
if spec.tokenizer_hub is not None:
|
||||
alt = _find_snapshot(spec.tokenizer_hub)
|
||||
if alt is not None:
|
||||
tok_src = alt
|
||||
_copy_tokenizer(tok_src, tmpdir)
|
||||
|
||||
# Load tokenizer and model
|
||||
model_id = ModelId(f"mlx-community/{spec.hub_name}")
|
||||
tokenizer = load_tokenizer_for_model_id(model_id, tmpdir)
|
||||
mx.random.seed(0)
|
||||
model = _build_model(spec.module, reduced)
|
||||
|
||||
task = _make_task()
|
||||
prompt = apply_chat_template(tokenizer=tokenizer, task_params=task)
|
||||
|
||||
# Run 1: fresh
|
||||
mx.random.seed(42)
|
||||
fresh = _collect_tokens(model, tokenizer, task, prompt, None)
|
||||
assert len(fresh) > 0, "Fresh generation produced no tokens"
|
||||
|
||||
# Run 2: populate cache
|
||||
kv = KVPrefixCache(None)
|
||||
mx.random.seed(42)
|
||||
populate = _collect_tokens(model, tokenizer, task, prompt, kv)
|
||||
|
||||
# Run 3: exact cache hit
|
||||
mx.random.seed(42)
|
||||
cached = _collect_tokens(model, tokenizer, task, prompt, kv)
|
||||
|
||||
assert fresh == populate, (
|
||||
f"Fresh vs populate mismatch: {fresh[:5]} vs {populate[:5]}"
|
||||
)
|
||||
assert fresh == cached, (
|
||||
f"Fresh vs cached mismatch: {fresh[:5]} vs {cached[:5]}"
|
||||
)
|
||||
finally:
|
||||
shutil.rmtree(tmpdir, ignore_errors=True)
|
||||
@@ -343,16 +343,8 @@ async def test_kimi_tokenizer_specifically():
|
||||
@pytest.mark.asyncio
|
||||
async def test_glm_tokenizer_specifically():
|
||||
"""Test GLM tokenizer with its specific EOS tokens."""
|
||||
|
||||
def contains(card: ModelCard, x: str):
|
||||
return x in card.model_id.lower()
|
||||
|
||||
glm_model_cards = [
|
||||
card
|
||||
for card in await get_model_cards()
|
||||
if contains(card, "glm")
|
||||
and not contains(card, "-5")
|
||||
and not contains(card, "4.7")
|
||||
card for card in await get_model_cards() if "glm" in card.model_id.lower()
|
||||
]
|
||||
|
||||
if not glm_model_cards:
|
||||
|
||||
@@ -1,162 +0,0 @@
|
||||
from collections.abc import Generator
|
||||
|
||||
from exo.shared.types.worker.runner_response import (
|
||||
GenerationResponse,
|
||||
ToolCallResponse,
|
||||
)
|
||||
from exo.worker.runner.runner import parse_gpt_oss
|
||||
|
||||
# Token IDs from mlx-community/gpt-oss-20b-MXFP4-Q8 tokenizer.
|
||||
# These are stable since they come from the model's vocabulary.
|
||||
_CHANNEL = 200005 # <|channel|>
|
||||
_START = 200006 # <|start|>
|
||||
_MESSAGE = 200008 # <|message|>
|
||||
_CALL = 200012 # <|call|>
|
||||
_END = 200007 # <|end|>
|
||||
_ASSISTANT = 173781 # "assistant"
|
||||
|
||||
# fmt: off
|
||||
# " to=functions.get_current_weather<|channel|>commentary json<|message|>{\"location\": \"Tokyo\"}<|call|>"
|
||||
FORMAT_A_TOKENS: list[tuple[int, str]] = [
|
||||
(316, " to"),
|
||||
(28, "="),
|
||||
(44580, "functions"),
|
||||
(775, ".get"),
|
||||
(23981, "_current"),
|
||||
(170154, "_weather"),
|
||||
(_CHANNEL, "<|channel|>"),
|
||||
(12606, "comment"),
|
||||
(815, "ary"),
|
||||
(5701, " json"),
|
||||
(_MESSAGE, "<|message|>"),
|
||||
(10848, '{"'),
|
||||
(7693, "location"),
|
||||
(1243, '":'),
|
||||
(392, ' "'),
|
||||
(173844, "Tokyo"),
|
||||
(18583, '"}'),
|
||||
(_CALL, "<|call|>"),
|
||||
]
|
||||
|
||||
# "<|channel|>commentary to=functions.get_current_weather json<|message|>{\"location\": \"Tokyo\"}<|call|>"
|
||||
FORMAT_B_TOKENS: list[tuple[int, str]] = [
|
||||
(_CHANNEL, "<|channel|>"),
|
||||
(12606, "comment"),
|
||||
(815, "ary"),
|
||||
(316, " to"),
|
||||
(28, "="),
|
||||
(44580, "functions"),
|
||||
(775, ".get"),
|
||||
(23981, "_current"),
|
||||
(170154, "_weather"),
|
||||
(5701, " json"),
|
||||
(_MESSAGE, "<|message|>"),
|
||||
(10848, '{"'),
|
||||
(7693, "location"),
|
||||
(1243, '":'),
|
||||
(392, ' "'),
|
||||
(173844, "Tokyo"),
|
||||
(18583, '"}'),
|
||||
(_CALL, "<|call|>"),
|
||||
]
|
||||
|
||||
# "<|channel|>analysis<|message|>Let me think...<|end|><|start|>assistant<|channel|>commentary to=functions.X ..."
|
||||
# Full analysis-then-tool-call as the model actually generates it.
|
||||
THINKING_THEN_TOOL_TOKENS: list[tuple[int, str]] = [
|
||||
(_CHANNEL, "<|channel|>"),
|
||||
(35644, "analysis"),
|
||||
(_MESSAGE, "<|message|>"),
|
||||
(12845, "Let"),
|
||||
(668, " me"),
|
||||
(2411, " think"),
|
||||
(1078, " about"),
|
||||
(495, " this"),
|
||||
(13, "."),
|
||||
(_END, "<|end|>"),
|
||||
# Model generates a new message header for the tool call:
|
||||
(_START, "<|start|>"),
|
||||
(_ASSISTANT, "assistant"),
|
||||
*FORMAT_B_TOKENS,
|
||||
]
|
||||
# fmt: on
|
||||
|
||||
|
||||
def _make_gen_responses(
|
||||
tokens: list[tuple[int, str]],
|
||||
) -> list[GenerationResponse]:
|
||||
"""Build GenerationResponse list from (token_id, text) pairs."""
|
||||
responses: list[GenerationResponse] = []
|
||||
for i, (tid, text) in enumerate(tokens):
|
||||
is_last = i == len(tokens) - 1
|
||||
responses.append(
|
||||
GenerationResponse(
|
||||
text=text,
|
||||
token=tid,
|
||||
finish_reason="stop" if is_last else None,
|
||||
usage=None,
|
||||
)
|
||||
)
|
||||
return responses
|
||||
|
||||
|
||||
def _collect(
|
||||
tokens: list[tuple[int, str]],
|
||||
) -> list[GenerationResponse | ToolCallResponse]:
|
||||
"""Feed tokens through parse_gpt_oss and collect all yielded responses."""
|
||||
|
||||
def _gen() -> Generator[GenerationResponse, None, None]:
|
||||
yield from _make_gen_responses(tokens)
|
||||
|
||||
return list(parse_gpt_oss(_gen()))
|
||||
|
||||
|
||||
def _get_tool_call(
|
||||
results: list[GenerationResponse | ToolCallResponse],
|
||||
) -> ToolCallResponse:
|
||||
"""Extract the single ToolCallResponse from results."""
|
||||
tool_calls = [r for r in results if isinstance(r, ToolCallResponse)]
|
||||
assert len(tool_calls) == 1, f"Expected 1 ToolCallResponse, got {len(tool_calls)}"
|
||||
return tool_calls[0]
|
||||
|
||||
|
||||
class TestParseGptOssRecipientPlacement:
|
||||
"""Both Harmony recipient placements must produce identical tool calls."""
|
||||
|
||||
def test_format_a_yields_tool_call(self):
|
||||
results = _collect(FORMAT_A_TOKENS)
|
||||
tc = _get_tool_call(results)
|
||||
assert tc.tool_calls[0].name == "get_current_weather"
|
||||
assert '"location"' in tc.tool_calls[0].arguments
|
||||
assert "Tokyo" in tc.tool_calls[0].arguments
|
||||
|
||||
def test_format_b_yields_tool_call(self):
|
||||
results = _collect(FORMAT_B_TOKENS)
|
||||
tc = _get_tool_call(results)
|
||||
assert tc.tool_calls[0].name == "get_current_weather"
|
||||
assert '"location"' in tc.tool_calls[0].arguments
|
||||
assert "Tokyo" in tc.tool_calls[0].arguments
|
||||
|
||||
def test_both_formats_produce_identical_tool_calls(self):
|
||||
tc_a = _get_tool_call(_collect(FORMAT_A_TOKENS))
|
||||
tc_b = _get_tool_call(_collect(FORMAT_B_TOKENS))
|
||||
assert tc_a.tool_calls[0].name == tc_b.tool_calls[0].name
|
||||
assert tc_a.tool_calls[0].arguments == tc_b.tool_calls[0].arguments
|
||||
|
||||
|
||||
class TestParseGptOssThinkingThenToolCall:
|
||||
"""Analysis (thinking) followed by a tool call must yield both."""
|
||||
|
||||
def test_thinking_then_tool_call(self):
|
||||
results = _collect(THINKING_THEN_TOOL_TOKENS)
|
||||
|
||||
# Should have thinking tags + content + tool call
|
||||
text_parts = [r.text for r in results if isinstance(r, GenerationResponse)]
|
||||
combined = "".join(text_parts)
|
||||
assert "<think>" in combined
|
||||
assert "</think>" in combined
|
||||
assert "Let me think about this." in combined
|
||||
|
||||
# And the tool call
|
||||
tc = _get_tool_call(results)
|
||||
assert tc.tool_calls[0].name == "get_current_weather"
|
||||
assert "Tokyo" in tc.tool_calls[0].arguments
|
||||
@@ -1,55 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
|
||||
[ $# -lt 1 ] && {
|
||||
echo "Usage: $0 host1 [host2 ...]"
|
||||
exit 1
|
||||
}
|
||||
|
||||
[ -z "$(git status --porcelain)" ] || {
|
||||
echo "Uncommitted changes"
|
||||
exit 1
|
||||
}
|
||||
|
||||
commit=$(git rev-parse HEAD)
|
||||
git fetch -q origin
|
||||
git branch -r --contains "$commit" | grep -qE '^\s*origin/' || {
|
||||
echo "Not pushed to origin"
|
||||
exit 1
|
||||
}
|
||||
hosts=("$@")
|
||||
cleanup() {
|
||||
for host in "${hosts[@]}"; do
|
||||
ssh -T -o BatchMode=yes "$host@$host" "pkill -f bin/exo" &
|
||||
done
|
||||
sleep 1
|
||||
jobs -pr | xargs -r kill 2>/dev/null || true
|
||||
}
|
||||
trap 'cleanup' EXIT INT TERM
|
||||
|
||||
for host; do
|
||||
ssh -T -o BatchMode=yes -o ServerAliveInterval=30 "$host@$host" \
|
||||
"EXO_LIBP2P_NAMESPACE=$commit /nix/var/nix/profiles/default/bin/nix build github:exo-explore/exo/$commit" &
|
||||
done
|
||||
wait
|
||||
for host; do
|
||||
ssh -T -o BatchMode=yes -o ServerAliveInterval=30 "$host@$host" \
|
||||
"EXO_LIBP2P_NAMESPACE=$commit /nix/var/nix/profiles/default/bin/nix run github:exo-explore/exo/$commit" &>/dev/null &
|
||||
done
|
||||
|
||||
for host; do
|
||||
echo "Waiting for $host..." 1>&2
|
||||
until curl -sf "http://$host:52415/models" &>/dev/null; do sleep 1; done
|
||||
done
|
||||
|
||||
echo "Waiting 30s for cluster setup" 1>&2
|
||||
sleep 30
|
||||
echo "EXO loaded" 1>&2
|
||||
eval_runner="${hosts[0]}"
|
||||
mkdir -p "./bench/$commit"
|
||||
nix run .#exo-get-all-models-on-cluster -- "$eval_runner" | while IFS= read -r model; do
|
||||
echo "running eval for $model" 1>&2
|
||||
ssh -Tn -o BatchMode=yes -o ServerAliveInterval=30 "$eval_runner@$eval_runner" \
|
||||
"/nix/var/nix/profiles/default/bin/nix run github:exo-explore/exo/$commit#exo-eval-tool-calls -- --model $model --stdout" \
|
||||
>>"./bench/$commit/${model//\//--}-eval.json"
|
||||
echo
|
||||
done
|
||||
@@ -43,4 +43,5 @@ for host; do
|
||||
echo "Waiting for $host..."
|
||||
until curl -sf "http://$host:52415/models" &>/dev/null; do sleep 1; done
|
||||
done
|
||||
echo "all hosts alive!"
|
||||
wait
|
||||
|
||||
12
uv.lock
generated
12
uv.lock
generated
@@ -378,7 +378,7 @@ dependencies = [
|
||||
{ name = "loguru", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
|
||||
{ name = "mflux", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
|
||||
{ name = "mlx", version = "0.30.6", source = { registry = "https://pypi.org/simple" }, extra = ["cpu"], marker = "sys_platform == 'linux'" },
|
||||
{ name = "mlx", version = "0.30.7.dev20260218+14841977", source = { git = "https://github.com/rltakashige/mlx-jaccl-fix-small-recv.git?branch=address-rdma-gpu-locks#1484197707f35186ad3bd614357c7c47fdf86ebc" }, marker = "sys_platform == 'darwin'" },
|
||||
{ name = "mlx", version = "0.30.7.dev20260217+50487b41", source = { git = "https://github.com/rltakashige/mlx-jaccl-fix-small-recv.git?branch=address-rdma-gpu-locks#50487b4141f3c951122655db3b83df5146c1fbeb" }, marker = "sys_platform == 'darwin'" },
|
||||
{ name = "mlx-lm", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
|
||||
{ name = "msgspec", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
|
||||
{ name = "openai-harmony", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
|
||||
@@ -447,7 +447,6 @@ name = "exo-bench"
|
||||
version = "0.1.0"
|
||||
source = { editable = "bench" }
|
||||
dependencies = [
|
||||
{ name = "httpx", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
|
||||
{ name = "huggingface-hub", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
|
||||
{ name = "jinja2", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
|
||||
{ name = "loguru", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
|
||||
@@ -457,7 +456,6 @@ dependencies = [
|
||||
|
||||
[package.metadata]
|
||||
requires-dist = [
|
||||
{ name = "httpx", specifier = ">=0.27.0" },
|
||||
{ name = "huggingface-hub", specifier = ">=0.33.4" },
|
||||
{ name = "jinja2", specifier = ">=3.1.0" },
|
||||
{ name = "loguru", specifier = ">=0.7.3" },
|
||||
@@ -1023,7 +1021,7 @@ dependencies = [
|
||||
{ name = "huggingface-hub", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
|
||||
{ name = "matplotlib", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
|
||||
{ name = "mlx", version = "0.30.6", source = { registry = "https://pypi.org/simple" }, extra = ["cuda13"], marker = "sys_platform == 'linux'" },
|
||||
{ name = "mlx", version = "0.30.7.dev20260218+14841977", source = { git = "https://github.com/rltakashige/mlx-jaccl-fix-small-recv.git?branch=address-rdma-gpu-locks#1484197707f35186ad3bd614357c7c47fdf86ebc" }, marker = "sys_platform == 'darwin'" },
|
||||
{ name = "mlx", version = "0.30.7.dev20260217+50487b41", source = { git = "https://github.com/rltakashige/mlx-jaccl-fix-small-recv.git?branch=address-rdma-gpu-locks#50487b4141f3c951122655db3b83df5146c1fbeb" }, marker = "sys_platform == 'darwin'" },
|
||||
{ name = "numpy", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
|
||||
{ name = "opencv-python", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
|
||||
{ name = "piexif", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
|
||||
@@ -1070,8 +1068,8 @@ cuda13 = [
|
||||
|
||||
[[package]]
|
||||
name = "mlx"
|
||||
version = "0.30.7.dev20260218+14841977"
|
||||
source = { git = "https://github.com/rltakashige/mlx-jaccl-fix-small-recv.git?branch=address-rdma-gpu-locks#1484197707f35186ad3bd614357c7c47fdf86ebc" }
|
||||
version = "0.30.7.dev20260217+50487b41"
|
||||
source = { git = "https://github.com/rltakashige/mlx-jaccl-fix-small-recv.git?branch=address-rdma-gpu-locks#50487b4141f3c951122655db3b83df5146c1fbeb" }
|
||||
resolution-markers = [
|
||||
"sys_platform == 'darwin'",
|
||||
]
|
||||
@@ -1106,7 +1104,7 @@ version = "0.30.7"
|
||||
source = { registry = "https://pypi.org/simple" }
|
||||
dependencies = [
|
||||
{ name = "jinja2", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
|
||||
{ name = "mlx", version = "0.30.7.dev20260218+14841977", source = { git = "https://github.com/rltakashige/mlx-jaccl-fix-small-recv.git?branch=address-rdma-gpu-locks#1484197707f35186ad3bd614357c7c47fdf86ebc" }, marker = "sys_platform == 'darwin'" },
|
||||
{ name = "mlx", version = "0.30.7.dev20260217+50487b41", source = { git = "https://github.com/rltakashige/mlx-jaccl-fix-small-recv.git?branch=address-rdma-gpu-locks#50487b4141f3c951122655db3b83df5146c1fbeb" }, marker = "sys_platform == 'darwin'" },
|
||||
{ name = "numpy", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
|
||||
{ name = "protobuf", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
|
||||
{ name = "pyyaml", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
|
||||
|
||||
Reference in New Issue
Block a user