DeepSeek/DeepSeek-Coder
1
0
Fork 0
mirror of https://github.com/deepseek-ai/DeepSeek-Coder synced 2024-12-04 18:14:44 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

fix add_generation_prompt in latest version

Browse Source
This commit is contained in:
Dejian Yang 2024-01-09 21:55:21 +08:00
parent d3bb741e25
commit c160d96860
2 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
Evaluation/HumanEval/eval_instruct.py
View File

@ -39,7 +39,7 @@ def generate_one(example, lang, tokenizer, model):
eos_token_id=stop_id eos_token_id=stop_id
) )
output = tokenizer.decode(outputs[0][len(inputs[0]):], skip_special_tokens=True) output = tokenizer.decode(outputs[0][len(inputs[0]):], skip_special_tokens=True, add_generation_prompt=True)
example['output'] = output example['output'] = output
return extract_generation_code(example, lang_code=lang) return extract_generation_code(example, lang_code=lang)

2
Evaluation/MBPP/eval_instruct.py
View File

@ -66,7 +66,7 @@ def generate_one(example, tokenizer, model):
prompt = example['prompt'] prompt = example['prompt']
inputs = tokenizer.apply_chat_template( inputs = tokenizer.apply_chat_template(
[{'role': 'user', 'content': prompt }], [{'role': 'user', 'content': prompt }],
return_tensors="pt" return_tensors="pt", add_generation_prompt=True
).to(model.device) ).to(model.device)
stop_id = tokenizer.convert_tokens_to_ids("<|EOT|>") stop_id = tokenizer.convert_tokens_to_ids("<|EOT|>")