Using Deepseek V3

[David-Hoa2023]

Nice project! I'd love to save the API cost by switching to DeekSeek V3. Could you please help!

[David-Hoa2023]

I tried using DeepSeek but had tokenizer issue:

Inference Exception: 'Could not automatically map deepseek-chat to a tokeniser. Please use tiktoken.get_encoding to explicitly get the tokeniser you expect.'

To replicate this, I used:
python ai_lab_repo.py --api-key "API_KEY_HERE" --research-topic "RESEARCH_IDEA" --llm-backend "deepseek-chat" --copilot-mode "true"

[SamuelSchmidgall]

Hello David. Currently only three model backends are supported: gpt-4o, o1-preview, and o1. Please feel free to add deepseek to the code if you're interested. See the inference.py file.

[brokenjade3000]

#10

[David-Hoa2023]

Hello David. Currently only three model backends are supported: gpt-4o, o1-preview, and o1. Please feel free to add deepseek to the code if you're interested. See the inference.py file.

@SamuelSchmidgall, thank you for your reply. I followed @vishwamartur and it is now working beautifully with DeepSeek V3

[SamuelSchmidgall]

Thank you @David-Hoa2023 and @vishwamartur. I have integrated deepseek with the code. Let me know if this works for you.