[snip] File "/mnt/venvs/rocm6.4/lib64/python3.12/site-packages/transformers/cli/serving/chat_completion.py", line 174, in _streaming queue, streamer = gen_manager.generate_streaming(model, processor, inputs, gen_config, request_id=request_id) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/mnt/venvs/rocm6.4/lib64/python3.12/site-packages/transformers/cli/serving/utils.py", line 565, in generate_streaming streamer = DirectStreamer(processor._tokenizer, loop, queue, skip_special_tokens=True) ^^^^^^^^^^^^^^^^^^^^ AttributeError: 'Qwen3VLProcessor' object has no attribute '_tokenizer'. Did you mean: 'tokenizer'?

TL;DR

The most likely fix is to update the transformers library to a version that is compatible with the Qwen3.5-35B model, as the error suggests an attribute mismatch.

Guidance

• Verify that the Qwen3.5-35B model is compatible with the current version of transformers (5.5.0) by checking the model's documentation or release notes.
• Check if there are any known issues or fixes related to the AttributeError in the transformers library's issue tracker or documentation.
• Consider downgrading the transformers library to a version that is known to work with the Qwen3.5-35B model, such as the 4.57 dev version mentioned in the model's readme.
• Test the transformers chat functionality with other models to see if the issue is specific to the Qwen3.5-35B model.

Example

No code snippet is provided as the issue seems to be related to a compatibility problem between the transformers library and the Qwen3.5-35B model.

Notes

The issue may be related to a change in the transformers library's API or a bug in the Qwen3.5-35B model's implementation. Further investigation is needed to determine the root cause of the issue.

Recommendation

Apply workaround: Try downgrading the transformers library to a version that is known to work with the Qwen3.5-35B model, such as the 4.57 dev version, to see if it resolves the issue.