Closed MaxHuber888 closed 7 months ago
Trained and uploaded to huggingface (private, ask for access) - https://huggingface.co/jeffzyu/llama-2-7b-chat-profit
2300 examples in training data. Trained on A6000 on Paperspace.
Updated notebook in repo (llama_v2.ipynb)
Hallucinates a lot, trying to fix with system prompt formatting, model parameters (temp, top_k, top_p).
Training full data set now, not sure how many examples we need.
Trained and uploaded to huggingface (private, ask for access) - https://huggingface.co/jeffzyu/llama-2-7b-chat-profit
2300 examples in training data. Trained on A6000 on Paperspace.
Updated notebook in repo (llama_v2.ipynb)
Hallucinates a lot, trying to fix with system prompt formatting, model parameters (temp, top_k, top_p).
Training full data set now, not sure how many examples we need.