Qwen3-VL can't be loaded in transformers dev branch (5.0.0.dev0 at commit 24807bf):
AttributeError: 'Qwen3VLTextConfig' object has no attribute 'pad_token_id'
After investigation, the issue appeared after the merge of this PR:
As @zucchini-nlp pointed out, there are no slow tests for qwen3VL, which could have prevented this.
See related issue in TRL:
Reproduction
from transformers import AutoModel
AutoModel.from_pretrained('Qwen/Qwen3-VL-2B-Instruct')Traceback:
File "<string>", line 1, in <module>
File "transformers/src/transformers/models/auto/auto_factory.py", line 372, in from_pretrained
return model_class.from_pretrained(
File "transformers/src/transformers/modeling_utils.py", line 4020, in from_pretrained
model = cls(config, *model_args, **model_kwargs)
File "transformers/src/transformers/models/qwen3_vl/modeling_qwen3_vl.py", line 938, in __init__
self.language_model = Qwen3VLTextModel._from_config(config.text_config)
File "transformers/src/transformers/modeling_utils.py", line 1464, in _from_config
model = cls(config, **kwargs)
File "transformers/src/transformers/models/qwen3_vl/modeling_qwen3_vl.py", line 807, in __init__
self.padding_idx = config.pad_token_id
File "transformers/src/transformers/configuration_utils.py", line 164, in __getattribute__
return super().__getattribute__(key)
AttributeError: 'Qwen3VLTextConfig' object has no attribute 'pad_token_id'. Did you mean: 'bos_token_id'?Who can help?
@zucchini-nlp, who created the PR
Qwen3-VL can't be loaded in transformers dev branch (5.0.0.dev0 at commit 24807bf):
After investigation, the issue appeared after the merge of this PR:
As @zucchini-nlp pointed out, there are no slow tests for
qwen3VL, which could have prevented this.See related issue in TRL:
Reproduction
Traceback:
Who can help?
@zucchini-nlp, who created the PR