AMD quark class not updated

[debasisdwivedy]

System Info

• transformers version: 5.13.1
• Platform: Linux-6.8.0-1058-gcp-x86_64-with-glibc2.39
• Python version: 3.13.13
• Huggingface_hub version: 1.23.0
• Safetensors version: 0.8.0
• Accelerate version: 1.14.0
• Accelerate config: not found
• DeepSpeed version: not installed
• PyTorch version (accelerator?): 2.12.1+cu130 (CUDA)
• Using distributed or parallel set-up in script?:
• Using GPU in script?:
• GPU type: Tesla T4

Who can help?

Hi ,

The quantization config for AMD class points to an older release and needs to be updated.
File : src/transformers/utils/quantization_config.py

Previous versions of Quark had the config as from quark.torch.quantization.config.config import Config, which has been changed to from quark.torch.quantization.config.config import QConfig

Regards

Information

• The official example scripts
• My own modified scripts

Tasks

• An officially supported task in the examples folder (such as GLUE/SQuAD, ...)
• My own task or dataset (give details below)

Reproduction

def main():
evaluation_tracker = EvaluationTracker(
output_dir="./results",
save_details=True
)

pipeline_params = PipelineParameters(
    launcher_type=ParallelismManager.ACCELERATE,
    custom_tasks_directory=None,  # Set to path if using custom tasks
    # Remove the parameter below once your configuration is tested
    # max_samples=10
)

model_config = TransformersModelConfig(
    model_name="", #Quark quantized folder path
    device="cuda"
)

task = ""

pipeline = Pipeline(
    tasks=task,
    pipeline_parameters=pipeline_params,
    evaluation_tracker=evaluation_tracker,
    model_config=model_config,
)

pipeline.evaluate()
pipeline.save_and_push_results()
pipeline.show_results()

Expected behavior

Should load the quantized model, but we get error: "Unable to Import from quark.torch.quantization.config.config import Config"

[debasisdwivedy]

New quark install pip install amd-quark