用最新的不修改是ok的,想测试以前的,修改文件latest对应的checkpoint,启动web_demo.py就卡着了:
python web_demo.py
[2023-10-29 08:23:41,107] [INFO] [real_accelerator.py:158:get_accelerator] Setting ds_accelerator to cuda (auto detect)
[2023-10-29 08:23:41,539] [WARNING] Failed to load bitsandbytes:No module named 'bitsandbytes'
[2023-10-29 08:23:41,540] [INFO] building FineTuneVisualGLMModel model ...
[2023-10-29 08:23:41,570] [INFO] [RANK 0] > initializing model parallel with size 1
[2023-10-29 08:23:41,570] [INFO] [RANK 0] You didn't pass in LOCAL_WORLD_SIZE environment variable. We use the guessed LOCAL_WORLD_SIZE=1. If this is wrong, please pass the LOCAL_WORLD_SIZE manually.
[2023-10-29 08:23:41,570] [INFO] [RANK 0] You are using model-only mode.
For torch.distributed users or loading model parallel models, set environment variables RANK, WORLD_SIZE and LOCAL_RANK.
/home/aifont/anaconda3/envs/vglm/lib/python3.10/site-packages/torch/nn/init.py:412: UserWarning: Initializing zero-element tensors is a no-op
warnings.warn("Initializing zero-element tensors is a no-op")
[2023-10-29 08:23:46,454] [INFO] [RANK 0] replacing layer 0 attention with lora
[2023-10-29 08:23:46,679] [INFO] [RANK 0] replacing layer 6 attention with lora
[2023-10-29 08:23:46,897] [INFO] [RANK 0] replacing layer 12 attention with lora
[2023-10-29 08:23:47,115] [INFO] [RANK 0] replacing layer 16 attention with lora
[2023-10-29 08:23:47,333] [INFO] [RANK 0] replacing layer 24 attention with lora
[2023-10-29 08:23:47,553] [INFO] [RANK 0] > number of parameters on model parallel rank 0: 7807108608
[2023-10-29 08:23:49,328] [INFO] [RANK 0] global rank 0 is loading checkpoint checkpoints/finetune-visualglm-6b-10-27-10-22/100000/mp_rank_00_model_states.pt
用最新的不修改是ok的,想测试以前的,修改文件latest对应的checkpoint,启动web_demo.py就卡着了: python web_demo.py [2023-10-29 08:23:41,107] [INFO] [real_accelerator.py:158:get_accelerator] Setting ds_accelerator to cuda (auto detect) [2023-10-29 08:23:41,539] [WARNING] Failed to load bitsandbytes:No module named 'bitsandbytes' [2023-10-29 08:23:41,540] [INFO] building FineTuneVisualGLMModel model ... [2023-10-29 08:23:41,570] [INFO] [RANK 0] > initializing model parallel with size 1 [2023-10-29 08:23:41,570] [INFO] [RANK 0] You didn't pass in LOCAL_WORLD_SIZE environment variable. We use the guessed LOCAL_WORLD_SIZE=1. If this is wrong, please pass the LOCAL_WORLD_SIZE manually. [2023-10-29 08:23:41,570] [INFO] [RANK 0] You are using model-only mode. For torch.distributed users or loading model parallel models, set environment variables RANK, WORLD_SIZE and LOCAL_RANK. /home/aifont/anaconda3/envs/vglm/lib/python3.10/site-packages/torch/nn/init.py:412: UserWarning: Initializing zero-element tensors is a no-op warnings.warn("Initializing zero-element tensors is a no-op") [2023-10-29 08:23:46,454] [INFO] [RANK 0] replacing layer 0 attention with lora [2023-10-29 08:23:46,679] [INFO] [RANK 0] replacing layer 6 attention with lora [2023-10-29 08:23:46,897] [INFO] [RANK 0] replacing layer 12 attention with lora [2023-10-29 08:23:47,115] [INFO] [RANK 0] replacing layer 16 attention with lora [2023-10-29 08:23:47,333] [INFO] [RANK 0] replacing layer 24 attention with lora [2023-10-29 08:23:47,553] [INFO] [RANK 0] > number of parameters on model parallel rank 0: 7807108608 [2023-10-29 08:23:49,328] [INFO] [RANK 0] global rank 0 is loading checkpoint checkpoints/finetune-visualglm-6b-10-27-10-22/100000/mp_rank_00_model_states.pt