FastChat
FastChat copied to clipboard
lm-sys
An open platform for training, serving, and evaluating large language models. Release repo for Vicuna and Chatbot Arena.
python -m fastchat.model.apply_delta --base-model-path LLaMA/13B_hf --target-model-path model/vicuna-13b-v1.1 --delta-path vicuna-13b-delta-v1.1 
## Why are these changes needed? The example prompt template in the docs does not reflect the prompt template used when training Vicuna. Other implementers may be mislead and this...
Hi, Amazing framework. Thanks a lot. I was comparing the inference speed (seconds) of `python -m fastchat.serve.cli` with and without `--load-8bit`. Both with the Vicuna 7B and 13B models, `--load-8bit`...
Would you suggest me some ways to debug that? By the way, this model can be successfully inferenced with fastchat cli. Error Log: 2023-06-13 15:19:44 | INFO | model_worker |...
I'm trying to host my arena, but I want to do so with models that have been finetuned using QLoRA. To test if LoRA models worked, I thought I'd first...
Hi, I'm trying to print "İ" and "ı" characters when im chatting from console. I'm using FastChatT5. Here is my output.  I think it might be encoding, but ü...
## Why are these changes needed? When we save the model checkpoint, it saves the entire model in `pytorch_model.bin` which is extremely large, but we often only want the adapter...
I have recently finetuned a model which is performing well overall. However, I have noticed that when I ask it generic questions such as "Hello, how are you?" it fails...
lcw99/polyglot-en-12.8b-chang-instruct-chat Add support for this model, which is based on EleutherAI's Polyglot 12.8b and is the largest publicly available Korean model. This model has been fine-tuned to follow Korean instructions.
