StanfordShahLab

/

clmbr-t-base

like 31

Hello,

I'm trying to run CLMBR-T on a GPU-enabled Nero instance and can't get the demo code to work. I've entered it exactly as on the model card and am getting the following error:

TypeError Traceback (most recent call last)
Cell In[6], line 2
1 # Load model
----> 2 model = femr.models.transformer.FEMRModel.from_pretrained(model_name)
4 # Run model
5 with torch.no_grad():

File /opt/conda/envs/ehrshot_env/lib/python3.10/site-packages/transformers/modeling_utils.py:3236, in PreTrainedModel.from_pretrained(cls, pretrained_model_name_or_path, config, cache_dir, ignore_mismatched_sizes, force_download, local_files_only, token, revision, use_safetensors, *model_args, **kwargs)
3233 config = cls._check_and_enable_flash_attn_2(config, torch_dtype=torch_dtype, device_map=device_map)
3235 with ContextManagers(init_contexts):
-> 3236 model = cls(config, *model_args, **model_kwargs)
3238 # make sure we use the model's config since the init call might have copied it
3239 config = model.config

File /opt/conda/envs/ehrshot_env/lib/python3.10/site-packages/femr/models/transformer.py:234, in FEMRModel.init(self, config, **kwargs)
232 self.transformer = FEMRTransformer(self.config.transformer_config)
233 if self.config.task_config is not None:
--> 234 self.task_model = self.config.task_config.create_task_head(self.config.transformer_config.hidden_size)

File /opt/conda/envs/ehrshot_env/lib/python3.10/site-packages/femr/models/transformer.py:193, in FEMRTaskConfig.create_task_head(self, hidden_size)
191 def create_task_head(self, hidden_size: int):
192 if self.task_type == "clmbr":
--> 193 return CLMBRTaskHead(hidden_size, **self.kwargs)
194 elif self.task_type == "labeled_patients":
195 return LabeledPatientTaskHead(hidden_size, **self.kwargs)

TypeError: CLMBRTaskHead.init() got an unexpected keyword argument '_name_or_path'

Could you let me know how to fix this please? Thanks.