LongChat
LongChat copied to clipboard
DachengLi1
Official repository for LongChat and LongEval
Hi, the printed message I can not understand. I set ``` def __init__( self, dim, ratio, max_position_embeddings=2048, base=10000, device=None ): ``` ratio=2, max_position_embeddings=1024 Since my GPU can not fit minimal...
from the script provided, I think longchat is full sft rather than lora, but the equal batch size total is just 1 (batch_size * gradient_accum * num_gpus) But vicuna original...
Does your xformers monkey patch compatible for llama model with default context (without condensed rotary embedding)? And you said in another issue that xformers can be use in non-A100 gpu,...
What do you guys recommend as for the inference configuration for : temperature and repeat penalty etc ... ?
reply like this: 
AMD? CPU? Single GPU? Is this all possible via FastChat?
I have 9 V100 16G GPUs,but training CUDA out of memory. The specific errors are as follows: Formatting inputs...Skip in lazy mode /home/chat_glm6b/anaconda3/envs/longeval/lib/python3.10/site-packages/torch/distributed/fsdp/_init_utils.py:295: UserWarning: FSDP is switching to use `NO_SHARD`...
when i was run： python -m torch.distributed.run --nproc_per_node=2 longchat/train/fine_tune/train.py --model_name_or_path /mnt/yuchao/open_model/longchat/longchat-13b-16k --data_path /mnt/workspace/sft_data.json --bf16 --output_dir /mnt/yuchao/yuchao/longchat-13b-16k --num_train_epochs 3 --per_device_train_batch_size 1 --per_device_eval_batch_size 4 --gradient_accumulation_steps 1 --evaluation_strategy no --save_strategy steps --save_steps 1000...
Hi, thanks for the great job! But I'd like to know is there a web GUI for this longchat model, it seems that the model worker in Vicuna does not...
Hoping for support training on some models that customized not inside transformers like baicuhan