A (possible) problem while running with deepseek

[JesseZZZZZ]

Cost approximation has an error? 'Could not automatically map cl100k_base to a tokeniser. Please use tiktoken.get_encoding to explicitly get the tokeniser you expect.'

I have no errors but this one, and it only appears sometimes. Is there any ways to fix this or can I just ignore it? Thank you!

[ice007yang]

I encountered the same error. After reviewing the code, I found that this error occurs when calculating the cost of each query. It does not affect the agent's functionality. I made a minor rewrite to line 194 in inferenc.py to fix this bug.

 elif model_str in ["deepseek-chat"]:
      # encoding = tiktoken.encoding_for_model("cl100k_base")
      encoding = AutoTokenizer.from_pretrained("deepseek-ai/deepseek-llm-7b-base")

[wangdapang815]

May I ask if you have ever encountered the problem of blank chapters in the paper? I also used deepseek-chat. There was no problem with the program, but the final generated paper always had some chapters without content.