mirror/OpenLLM
1
0
Fork 0
mirror of https://github.com/bentoml/OpenLLM.git synced 2026-06-12 18:39:16 -04:00
Code Issues Packages Projects Releases Wiki Activity

chore(style): reduce line length and truncate compression

Browse Source 
Signed-off-by: aarnphm-ec2-dev <29749331+aarnphm@users.noreply.github.com>
This commit is contained in:
aarnphm-ec2-dev
2023-08-22 17:02:00 +00:00
parent bc851b1d13
commit eddbc06374
67 changed files with 1661 additions and 301 deletions
Download Patch File Download Diff File Expand all files Collapse all files

25
openllm-python/src/openllm/models/opt/modeling_flax_opt.py
View File

@@ -13,10 +13,27 @@ class FlaxOPT(openllm.LLM["transformers.TFOPTForCausalLM", "transformers.GPT2Tok
def import_model(self, *args: t.Any, trust_remote_code: bool = False, **attrs: t.Any) -> bentoml.Model:
config, tokenizer = transformers.AutoConfig.from_pretrained(self.model_id), transformers.AutoTokenizer.from_pretrained(self.model_id, **self.llm_parameters[-1])
tokenizer.pad_token_id = config.pad_token_id
return bentoml.transformers.save_model(self.tag, transformers.FlaxAutoModelForCausalLM.from_pretrained(self.model_id, **attrs), custom_objects={"tokenizer": tokenizer}, labels=generate_labels(self))
return bentoml.transformers.save_model(
self.tag, transformers.FlaxAutoModelForCausalLM.from_pretrained(self.model_id, **attrs), custom_objects={"tokenizer": tokenizer}, labels=generate_labels(self)
)
def sanitize_parameters(self, prompt: str, max_new_tokens: int | None = None, temperature: float | None = None, top_k: int | None = None, num_return_sequences: int | None = None, repetition_penalty: float | None = None, use_default_prompt_template: bool = False, **attrs: t.Any) -> tuple[str, dict[str, t.Any], dict[str, t.Any]]:
return process_prompt(prompt, DEFAULT_PROMPT_TEMPLATE, use_default_prompt_template, **attrs), {"max_new_tokens": max_new_tokens, "temperature": temperature, "top_k": top_k, "num_return_sequences": num_return_sequences, "repetition_penalty": repetition_penalty}, {}
def sanitize_parameters(
self,
prompt: str,
max_new_tokens: int | None = None,
temperature: float | None = None,
top_k: int | None = None,
num_return_sequences: int | None = None,
repetition_penalty: float | None = None,
use_default_prompt_template: bool = False,
**attrs: t.Any
) -> tuple[str, dict[str, t.Any], dict[str, t.Any]]:
return process_prompt(prompt, DEFAULT_PROMPT_TEMPLATE, use_default_prompt_template, **attrs), {
"max_new_tokens": max_new_tokens, "temperature": temperature, "top_k": top_k, "num_return_sequences": num_return_sequences, "repetition_penalty": repetition_penalty
}, {}
def generate(self, prompt: str, **attrs: t.Any) -> list[str]:
return self.tokenizer.batch_decode(self.model.generate(**self.tokenizer(prompt, return_tensors="np"), do_sample=True, generation_config=self.config.model_construct_env(**attrs).to_generation_config()).sequences, skip_special_tokens=True)
return self.tokenizer.batch_decode(
self.model.generate(**self.tokenizer(prompt, return_tensors="np"), do_sample=True, generation_config=self.config.model_construct_env(**attrs).to_generation_config()).sequences,
skip_special_tokens=True
)

5
openllm-python/src/openllm/models/opt/modeling_opt.py
View File

@@ -14,4 +14,7 @@ class OPT(openllm.LLM["transformers.OPTForCausalLM", "transformers.GPT2Tokenizer
def generate(self, prompt: str, **attrs: t.Any) -> list[str]:
import torch
with torch.inference_mode():
return self.tokenizer.batch_decode(self.model.generate(**self.tokenizer(prompt, return_tensors="pt").to(self.device), do_sample=True, generation_config=self.config.model_construct_env(**attrs).to_generation_config()), skip_special_tokens=True)
return self.tokenizer.batch_decode(
self.model.generate(**self.tokenizer(prompt, return_tensors="pt").to(self.device), do_sample=True, generation_config=self.config.model_construct_env(**attrs).to_generation_config()),
skip_special_tokens=True
)

12
openllm-python/src/openllm/models/opt/modeling_tf_opt.py
View File

@@ -9,7 +9,15 @@ class TFOPT(openllm.LLM["transformers.TFOPTForCausalLM", "transformers.GPT2Token
import transformers
config, tokenizer = transformers.AutoConfig.from_pretrained(self.model_id), transformers.AutoTokenizer.from_pretrained(self.model_id, **self.llm_parameters[-1])
tokenizer.pad_token_id = config.pad_token_id
return bentoml.transformers.save_model(self.tag, transformers.TFOPTForCausalLM.from_pretrained(self.model_id, trust_remote_code=trust_remote_code, **attrs), custom_objects={"tokenizer": tokenizer}, labels=generate_labels(self))
return bentoml.transformers.save_model(
self.tag,
transformers.TFOPTForCausalLM.from_pretrained(self.model_id, trust_remote_code=trust_remote_code, **attrs),
custom_objects={"tokenizer": tokenizer},
labels=generate_labels(self)
)
def generate(self, prompt: str, **attrs: t.Any) -> list[str]:
return self.tokenizer.batch_decode(self.model.generate(**self.tokenizer(prompt, return_tensors="tf"), do_sample=True, generation_config=self.config.model_construct_env(**attrs).to_generation_config()), skip_special_tokens=True)
return self.tokenizer.batch_decode(
self.model.generate(**self.tokenizer(prompt, return_tensors="tf"), do_sample=True, generation_config=self.config.model_construct_env(**attrs).to_generation_config()),
skip_special_tokens=True
)

15
openllm-python/src/openllm/models/opt/modeling_vllm_opt.py
View File

@@ -7,5 +7,16 @@ class VLLMOPT(openllm.LLM["vllm.LLMEngine", "transformers.GPT2Tokenizer"]):
__openllm_internal__ = True
tokenizer_id = "local"
def sanitize_parameters(self, prompt: str, max_new_tokens: int | None = None, temperature: float | None = None, top_k: int | None = None, num_return_sequences: int | None = None, use_default_prompt_template: bool = True, **attrs: t.Any) -> tuple[str, dict[str, t.Any], dict[str, t.Any]]:
return process_prompt(prompt, DEFAULT_PROMPT_TEMPLATE, use_default_prompt_template, **attrs), {"max_new_tokens": max_new_tokens, "temperature": temperature, "top_k": top_k, "num_return_sequences": num_return_sequences}, {}
def sanitize_parameters(
self,
prompt: str,
max_new_tokens: int | None = None,
temperature: float | None = None,
top_k: int | None = None,
num_return_sequences: int | None = None,
use_default_prompt_template: bool = True,
**attrs: t.Any
) -> tuple[str, dict[str, t.Any], dict[str, t.Any]]:
return process_prompt(prompt, DEFAULT_PROMPT_TEMPLATE, use_default_prompt_template, **attrs), {
"max_new_tokens": max_new_tokens, "temperature": temperature, "top_k": top_k, "num_return_sequences": num_return_sequences
}, {}