GPTQ-for-LLaMa - Get "No module named 'llama'" when trying to run GPTQ models

python3 -m fastchat.serve.model_worker --model-path /workspace/models/LongChat-13B-GPTQ/ --gptq-wbits 4 --gptq-groupsize 128 \

results in:

2023-07-21 22:44:26 | INFO | model_worker | args: Namespace(host='localhost', port=21002, worker_address='http://localhost:21002', controller_address='http://localhost:21001', model_path='/workspace/models/LongChat-13B-GPTQ/', revision='main', device='cuda', gpus=None, num_gpus=1, max_gpu_memory=None, load_8bit=False, cpu_offloading=False, gptq_ckpt=None, gptq_wbits=4, gptq_groupsize=128, gptq_act_order=False, model_names=None, limit_worker_concurrency=5, stream_interval=2, no_register=False)
2023-07-21 22:44:26 | INFO | model_worker | Loading the model ['LongChat-13B-GPTQ'] on worker c689fab6 ...
2023-07-21 22:44:26 | INFO | stdout | Loading GPTQ quantized model...
**2023-07-21 22:44:26 | INFO | stdout | Error: Failed to load GPTQ-for-LLaMa. No module named 'llama'**
2023-07-21 22:44:26 | INFO | stdout | See https://github.com/lm-sys/FastChat/blob/main/docs/gptq.md

Same error in the cli too. Tried to pip install llama just to see and got strange error.

Thanks

Start with the model_worker command and the GPTQ setup described in docs/gptq.md. Reproduce the failure for the LongChat-13B-GPTQ model, then trace why the llama module cannot be imported; done means GPTQ model loading works in both the worker and CLI without this error.