cli.py: error: unrecognized arguments: --model-path /path/to/model/weights

[akdcelcopr77]

as title

when i want to run in gpu, i got the error.

[mpetruc]

replace /path/to/model/weights with actual path to the model weights, for example /mnt/llama

[wsfung2008]

I got the same error when running !python3 -m fastchat.serve.cli --model-path lmsys/fastchat-t5-3b-v1.0 on kaggle as the readme on https://github.com/lm-sys/FastChat#fastchat-t5 suggests.

usage: cli.py [-h] [--model-name MODEL_NAME] [--num-gpus NUM_GPUS] [--device {cuda,cpu}] [--conv-template CONV_TEMPLATE] [--temperature TEMPERATURE] [--max-new-tokens MAX_NEW_TOKENS] [--debug] cli.py: error: unrecognized arguments: --model-path lmsys/fastchat-t5-3b-v1.0

It doesn't accept --model-path argument. I also tried --model-name and got another error.

The weird thing is when I ran it on sagemaker studio, I didn't have this error.

Before that I installed fschat with pip3 install fschat

[merrymercy]

@wsfung2008 could you try installing from the source or fetching the latest version of fschat?

[5ky1s61ue]

I got the same error when running !python3 -m fastchat.serve.cli --model-path lmsys/fastchat-t5-3b-v1.0 on kaggle as the readme on https://github.com/lm-sys/FastChat#fastchat-t5 suggests.

usage: cli.py [-h] [--model-name MODEL_NAME] [--num-gpus NUM_GPUS] [--device {cuda,cpu}] [--conv-template CONV_TEMPLATE] [--temperature TEMPERATURE] [--max-new-tokens MAX_NEW_TOKENS] [--debug] cli.py: error: unrecognized arguments: --model-path lmsys/fastchat-t5-3b-v1.0

It doesn't accept --model-path argument. I also tried --model-name and got another error.

The weird thing is when I ran it on sagemaker studio, I didn't have this error.

Before that I installed fschat with pip3 install fschat

you can try to upgrade python3&pip to the latest version, it works when i met this issue