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.
Hi @infwinston @Mearman @zhisbug @jegonzal @Shawnlu25 I'm trying to generate meeting minutes using vicuna-13b, using a chunk from my meeting transcript (due to context size restrictions I'm creating chunks of...
command: ``` torchrun --nnodes=1 --nproc_per_node=1 --master_port=20001 fastchat/train/train_mem.py \ > --model_name_or_path /home/work/virtual-venv/fastchat-env/data/transformer_model_7b \ > --data_path playground/data/dummy.json \ > --fp16 True \ > --output_dir /home/work/virtual-venv/fastchat-env/data/vicuna-dummy \ > --num_train_epochs 2 \ > --per_device_train_batch_size...
Traceback (most recent call last): File "/data/app/FastChat/fastChat/train/train.py", line 335, in train() File "/data/app/FastChat/fastChat/train/train.py", line 328, in train trainer.train(resume_from_checkpoint=True) File "/data/app/install/transformers/src/transformers/trainer.py", line 1651, in train self._load_from_checkpoint(resume_from_checkpoint) File "/data/app/install/transformers/src/transformers/trainer.py", line 2159, in...
I trained the model using the 33B architecture and the train.py file with deepspeed , but when I saved the model using the safe_save_model_for_hf_trainer function, it was only 400M. the...
There isn't any documentation of [Cacheflow worker ](https://github.com/lm-sys/FastChat/blob/main/fastchat/serve/cacheflow_worker.py) using ray cluster. Can you please provide the updated dependency even though under optional for using the cacheflow worker to serve the...
I download the vicuna-13b-v1.1 and run ``` python3 -m fastchat.model.apply_delta \ --base llama-13b/ \ --target output/vicuna-13b \ --delta vicuna-13b-v1.1/ ``` Then I run the following code ``` CUDA_VISIBLE_DEVICES=2,3 python3 -m...
transformers 4.28.1 fschat 0.2.3 i use vicuna-7b-delta-v1.1 python fastchat/model/apply_delta.py --base-model ../llama-7b-hf-v1.1/ --target vicuna-7b --delta vicuna-7b-delta-v1.1/ python fastchat/serve/cli.py --model_path ./vicuna-7b ll ./vicuna-7b total 13161628 drwxr-xr-x 2 root root 279 Apr 22...
Duplicate of [#540](https://github.com/lm-sys/FastChat/issues/540) transformers 4.28.0 fschat 0.2.2 Successfuly finetuned llamma-13b using the following arguments: `CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6 torchrun --nproc_per_node=7 --master_port=2001 fastchat/train/train_mem.py --model_name_or_path huggyllama/llama-13b --data_path /root/autodl-data/code/hjw/FastChat/fastchat/data/temp_1k.json --bf16 True --output_dir output --num_train_epochs 1 --per_device_train_batch_size...
Input: 请将一个笑话并解释为什么他为什么好笑 Output: Please explain one joke and why he is a good comedian. Input: 请用中文来回答我的问题，我的问题是今天天气怎么样 Output: Please answer my questions with Chinese, my question is today's weather. 
I was wondering if it's possible to freeze some layers in vicuna models, and if that will have smaller memory footprint to fine-tune the model?
