lmsys

/

vicuna-13b-v1.1

like 98

Follow

Large Model Systems Organization 562

Hi Bloke,

Hope you are well. I am trying to use this model and get this error:

>>> model = AutoModelForCausalLM.from_pretrained(model_name, quantization_config=quant_config, device_map={"":0})
Loading checkpoint shards:   0%|                                 | 0/3 [00:06<?, ?it/s]
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/mnt/disks/sdb/finetuning-with-qlora/.env/lib/python3.11/site-packages/transformers/models/auto/auto_factory.py", line 484, in from_pretrained
    return model_class.from_pretrained(
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/disks/sdb/finetuning-with-qlora/.env/lib/python3.11/site-packages/transformers/modeling_utils.py", line 2897, in from_pretrained
    ) = cls._load_pretrained_model(
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/disks/sdb/finetuning-with-qlora/.env/lib/python3.11/site-packages/transformers/modeling_utils.py", line 3236, in _load_pretrained_model
    new_error_msgs, offload_index, state_dict_index = _load_state_dict_into_meta_model(
                                                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/disks/sdb/finetuning-with-qlora/.env/lib/python3.11/site-packages/transformers/modeling_utils.py", line 718, in _load_state_dict_into_meta_model
    set_module_quantized_tensor_to_device(
  File "/mnt/disks/sdb/finetuning-with-qlora/.env/lib/python3.11/site-packages/transformers/utils/bitsandbytes.py", line 91, in set_module_quantized_tensor_to_device
    new_value = bnb.nn.Params4bit(new_value, requires_grad=False, **kwargs).to(device)
                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/disks/sdb/finetuning-with-qlora/.env/lib/python3.11/site-packages/bitsandbytes/nn/modules.py", line 176, in to
    return self.cuda(device)
           ^^^^^^^^^^^^^^^^^
  File "/mnt/disks/sdb/finetuning-with-qlora/.env/lib/python3.11/site-packages/bitsandbytes/nn/modules.py", line 153, in cuda
    w = self.data.contiguous().half().cuda(device)
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
RuntimeError: CUDA error: an illegal memory access was encountered
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1.
Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.

I have

>>> torch.__version__
'2.0.1+cu117'

could you please help? Is there another model I can use which does not give this error? from here:

If you get an this issue ("illegal memory access") then you should use a newer HF LLaMA conversion or downgrade your PyTorch version.