starlight-apk/Qwen
1
0
Fork 0
mirror of https://github.com/QwenLM/Qwen.git synced 2026-05-20 16:35:47 +08:00
Code Issues Packages Projects Releases Wiki Activity

specify repetition penalty

Browse Source
This commit is contained in:
yangapku
2023-10-13 11:44:48 +08:00
parent 024146bc82
commit f076e2fa42
4 changed files with 4 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
eval/evaluate_chat_ceval.py
View File

@@ -31,6 +31,7 @@ def load_models_tokenizer(args):
args.checkpoint_path, trust_remote_code=True
)
model.generation_config.do_sample = False # use greedy decoding
model.generation_config.repetition_penalty = 1.0 # disable repetition penalty
return model, tokenizer
def process_before_extraction(gen, question, choice_dict):

1
eval/evaluate_chat_gsm8k.py
View File

@@ -129,6 +129,7 @@ if __name__ == "__main__":
args.checkpoint_path, trust_remote_code=True
)
model.generation_config.do_sample = False # use greedy decoding
model.generation_config.repetition_penalty = 1.0 # disable repetition penalty
test = dataset["test"]

1
eval/evaluate_chat_humaneval.py
View File

@@ -93,6 +93,7 @@ if __name__ == "__main__":
args.checkpoint_path, trust_remote_code=True
)
model.generation_config.do_sample = False # use greedy decoding
model.generation_config.repetition_penalty = 1.0 # disable repetition penalty
f_output = jsonlines.Writer(open(args.sample_output_file, "w", encoding="utf-8"))

1
eval/evaluate_chat_mmlu.py
View File

@@ -35,6 +35,7 @@ def load_models_tokenizer(args):
args.checkpoint_path, trust_remote_code=True
)
model.generation_config.do_sample = False # use greedy decoding
model.generation_config.repetition_penalty = 1.0 # disable repetition penalty
return model, tokenizer