mirror/OpenLLM
1
0
Fork 0
mirror of https://github.com/bentoml/OpenLLM.git synced 2026-06-12 10:29:36 -04:00
Code Issues Packages Projects Releases Wiki Activity

chore: ignore new lines split [skip ci]

Browse Source 
Signed-off-by: aarnphm-ec2-dev <29749331+aarnphm@users.noreply.github.com>
This commit is contained in:
aarnphm-ec2-dev
2023-09-01 17:00:49 +00:00
parent 608de0b667
commit 7d893e6cd2
70 changed files with 575 additions and 950 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
openllm-python/src/openllm/models/opt/modeling_flax_opt.py
View File

@@ -16,8 +16,7 @@ class FlaxOPT(openllm.LLM['transformers.TFOPTForCausalLM', 'transformers.GPT2Tok
__openllm_internal__ = True
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])
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),
@@ -34,11 +33,7 @@ class FlaxOPT(openllm.LLM['transformers.TFOPTForCausalLM', 'transformers.GPT2Tok
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
'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]:

3
openllm-python/src/openllm/models/opt/modeling_tf_opt.py
View File

@@ -11,8 +11,7 @@ class TFOPT(openllm.LLM['transformers.TFOPTForCausalLM', 'transformers.GPT2Token
def import_model(self, *args: t.Any, trust_remote_code: bool = False, **attrs: t.Any) -> bentoml.Model:
import transformers
config, tokenizer = transformers.AutoConfig.from_pretrained(self.model_id), transformers.AutoTokenizer.from_pretrained(
self.model_id, **self.llm_parameters[-1])
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),

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

@@ -19,8 +19,5 @@ class VLLMOPT(openllm.LLM['vllm.LLMEngine', 'transformers.GPT2Tokenizer']):
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
'max_new_tokens': max_new_tokens, 'temperature': temperature, 'top_k': top_k, 'num_return_sequences': num_return_sequences
}, {}