Files
LocalAI/backend/python/tinygrad/tool_parsers/base.py
Ettore Di Giacinto 6f0051301b feat(backend): add tinygrad multimodal backend (experimental) (#9364)
* feat(backend): add tinygrad multimodal backend

Wire tinygrad as a new Python backend covering LLM text generation with
native tool-call extraction, embeddings, Stable Diffusion 1.x image
generation, and Whisper speech-to-text from a single self-contained
container.

Backend (`backend/python/tinygrad/`):
- `backend.py` gRPC servicer with LLM Predict/PredictStream (auto-detects
  Llama / Qwen2 / Mistral architecture from `config.json`, supports
  safetensors and GGUF), Embedding via mean-pooled last hidden state,
  GenerateImage via the vendored SD1.x pipeline, AudioTranscription +
  AudioTranscriptionStream via the vendored Whisper inference loop, plus
  Tokenize / ModelMetadata / Status / Free.
- Vendored upstream model code under `vendor/` (MIT, headers preserved):
  llama.py with an added `qkv_bias` flag for Qwen2-family bias support
  and an `embed()` method that returns the last hidden state, plus
  clip.py, unet.py, stable_diffusion.py (trimmed to drop the MLPerf
  training branch that pulls `mlperf.initializers`), audio_helpers.py
  and whisper.py (trimmed to drop the pyaudio listener).
- Pluggable tool-call parsers under `tool_parsers/`: hermes (Qwen2.5 /
  Hermes), llama3_json (Llama 3.1+), qwen3_xml (Qwen 3), mistral
  (Mistral / Mixtral). Auto-selected from model architecture or `Options`.
- `install.sh` pins Python 3.11.14 (tinygrad >=0.12 needs >=3.11; the
  default portable python is 3.10).
- `package.sh` bundles libLLVM.so.1 + libedit/libtinfo/libgomp/libsndfile
  into the scratch image. `run.sh` sets `CPU_LLVM=1` and `LLVM_PATH` so
  tinygrad's CPU device uses the in-process libLLVM JIT instead of
  shelling out to the missing `clang` binary.
- Local unit tests for Health and the four parsers in `test.py`.

Build wiring:
- Root `Makefile`: `.NOTPARALLEL`, `prepare-test-extra`, `test-extra`,
  `BACKEND_TINYGRAD = tinygrad|python|.|false|true`,
  docker-build-target eval, and `docker-build-backends` aggregator.
- `.github/workflows/backend.yml`: cpu / cuda12 / cuda13 build matrix
  entries (mirrors the transformers backend placement).
- `backend/index.yaml`: `&tinygrad` meta + cpu/cuda12/cuda13 image
  entries (latest + development).

E2E test wiring:
- `tests/e2e-backends/backend_test.go` gains an `image` capability that
  exercises GenerateImage and asserts a non-empty PNG is written to
  `dst`. New `BACKEND_TEST_IMAGE_PROMPT` / `BACKEND_TEST_IMAGE_STEPS`
  knobs.
- Five new make targets next to `test-extra-backend-vllm`:
  - `test-extra-backend-tinygrad` — Qwen2.5-0.5B-Instruct + hermes,
    mirrors the vllm target 1:1 (5/9 specs in ~57s).
  - `test-extra-backend-tinygrad-embeddings` — same model, embeddings
    via LLM hidden state (3/9 in ~10s).
  - `test-extra-backend-tinygrad-sd` — stable-diffusion-v1-5 mirror,
    health/load/image (3/9 in ~10min, 4 diffusion steps on CPU).
  - `test-extra-backend-tinygrad-whisper` — openai/whisper-tiny.en
    against jfk.wav from whisper.cpp samples (4/9 in ~49s).
  - `test-extra-backend-tinygrad-all` aggregate.

All four targets land green on the first MVP pass: 15 specs total, 0
failures across LLM+tools, embeddings, image generation, and speech
transcription.

* refactor(tinygrad): collapse to a single backend image

tinygrad generates its own GPU kernels (PTX renderer for CUDA, the
autogen ctypes wrappers for HIP / Metal / WebGPU) and never links
against cuDNN, cuBLAS, or any toolkit-version-tied library. The only
runtime dependency that varies across hosts is the driver's libcuda.so.1
/ libamdhip64.so, which are injected into the container at run time by
the nvidia-container / rocm runtimes. So unlike torch- or vLLM-based
backends, there is no reason to ship per-CUDA-version images.

- Drop the cuda12-tinygrad and cuda13-tinygrad build-matrix entries
  from .github/workflows/backend.yml. The sole remaining entry is
  renamed to -tinygrad (from -cpu-tinygrad) since it is no longer
  CPU-only.
- Collapse backend/index.yaml to a single meta + development pair.
  The meta anchor carries the latest uri directly; the development
  entry points at the master tag.
- run.sh picks the tinygrad device at launch time by probing
  /usr/lib/... for libcuda.so.1 / libamdhip64.so. When libcuda is
  visible we set CUDA=1 + CUDA_PTX=1 so tinygrad uses its own PTX
  renderer (avoids any nvrtc/toolkit dependency); otherwise we fall
  back to HIP or CLANG. CPU_LLVM=1 + LLVM_PATH keep the in-process
  libLLVM JIT for the CLANG path.
- backend.py's _select_tinygrad_device() is trimmed to a CLANG-only
  fallback since production device selection happens in run.sh.

Re-ran test-extra-backend-tinygrad after the change:
  Ran 5 of 9 Specs in 56.541 seconds — 5 Passed, 0 Failed
2026-04-15 19:48:23 +02:00

86 lines
2.6 KiB
Python

"""Common types + parser registry for tool-call extraction."""
from __future__ import annotations
from dataclasses import dataclass, field
from typing import Optional
@dataclass
class ToolCall:
"""One extracted tool call — maps 1:1 to backend_pb2.ToolCallDelta."""
index: int
name: str
arguments: str # JSON string
id: str = ""
class ToolParser:
"""Parser interface.
Subclasses implement `parse` (full non-streaming pass) and optionally
`parse_stream` (incremental). The default `parse_stream` buffers until a
full response is available and then delegates to `parse`.
"""
name: str = "base"
def __init__(self) -> None:
self._stream_buffer = ""
self._stream_index = 0
def parse(self, text: str) -> tuple[str, list[ToolCall]]:
"""Return (content_for_user, tool_calls)."""
raise NotImplementedError
def parse_stream(self, delta: str, finished: bool = False) -> tuple[str, list[ToolCall]]:
"""Accumulate a streaming delta. Emits any tool calls that have closed.
Default behavior: buffer until `finished=True`, then parse once.
Subclasses can override to emit mid-stream.
"""
self._stream_buffer += delta
if not finished:
return "", []
content, calls = self.parse(self._stream_buffer)
# Re-index starting from whatever we've already emitted in this stream.
reindexed: list[ToolCall] = []
for i, c in enumerate(calls):
reindexed.append(ToolCall(
index=self._stream_index + i,
name=c.name,
arguments=c.arguments,
id=c.id,
))
self._stream_index += len(reindexed)
return content, reindexed
def reset(self) -> None:
self._stream_buffer = ""
self._stream_index = 0
_REGISTRY: dict[str, type[ToolParser]] = {}
def register(cls: type[ToolParser]) -> type[ToolParser]:
_REGISTRY[cls.name] = cls
return cls
def resolve_parser(name: Optional[str]) -> ToolParser:
"""Return a parser instance by name, falling back to a no-op passthrough."""
# Import for side effects — each module registers itself.
from . import hermes, llama3_json, mistral, qwen3_xml # noqa: F401
if name and name in _REGISTRY:
return _REGISTRY[name]()
return PassthroughToolParser()
class PassthroughToolParser(ToolParser):
"""No-op parser — used when no tool_parser is configured."""
name = "passthrough"
def parse(self, text: str) -> tuple[str, list[ToolCall]]:
return text, []