FineTuning AutoModelForSequenceClassification.from_pretrained(meta-llama/Llama-3.2-1B) Bug:RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:1 and cuda:0! (when checking argument for argument target in method wrapper_CUDA_nll_loss_forward) and awq importing

src/transformers/modeling_utils.py

@@ -43,7 +43,8 @@
 from .activations import get_activation
 from .configuration_utils import PretrainedConfig
 from .dynamic_module_utils import custom_object_save
-from .generation import CompileConfig, GenerationConfig, GenerationMixin
+from .generation.configuration_utils import CompileConfig, GenerationConfig
+from .generation import GenerationMixin
 from .integrations import PeftAdapterMixin, deepspeed_config, is_deepspeed_zero3_enabled
 from .integrations.flash_attention import flash_attention_forward
 from .integrations.flex_attention import flex_attention_forward

src/transformers/models/llama/modeling_llama.py

@@ -958,6 +958,7 @@ def forward(
 
         loss = None
         if labels is not None:
+            labels = labels.to(logits.device)
             loss = self.loss_function(logits=logits, labels=labels, pooled_logits=pooled_logits, config=self.config)
 
         if not return_dict:

src/transformers/utils/import_utils.py

@@ -46,6 +46,7 @@ def _is_package_available(pkg_name: str, return_version: bool = False) -> Union[
     if package_exists:
         try:
             # Primary method to get the package version
+            if pkg_name == 'awq' : pkg_name='autoawq'
             package_version = importlib.metadata.version(pkg_name)
         except importlib.metadata.PackageNotFoundError:
             # Fallback method: Only for "torch" and versions containing "dev"