[Bug] qwen35 running with pp error

Bug Description

  File ".../slime/slime/backends/megatron_utils/actor.py", line 368, in train
    return self.train_actor(rollout_id, rollout_data)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File ".../slime/slime/backends/megatron_utils/actor.py", line 440, in train_actor
    self.compute_log_prob(
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File ".../slime/slime/backends/megatron_utils/actor.py", line 346, in compute_log_prob
    return forward_only(
           ^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/dist-packages/torch/utils/_contextlib.py", line 120, in decorate_context
    return func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File ".../slime/slime/backends/megatron_utils/model.py", line 263, in forward_only
    forward_data_store += forward_backward_func(
                          ^^^^^^^^^^^^^^^^^^^^^^
  File "/root/Megatron-LM/megatron/core/pipeline_parallel/schedules.py", line 2221, in forward_backward_pipelining_without_interleaving
    output_tensor, num_tokens = forward_step(
                                ^^^^^^^^^^^^^
  File "/root/Megatron-LM/megatron/core/pipeline_parallel/schedules.py", line 417, in forward_step
    output_tensor, loss_func = forward_step_func(data_iterator, model)
                               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File ".../slime/slime/backends/megatron_utils/model.py", line 227, in forward_step
    output_tensor = model(
                    ^^^^^^
  File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1786, in _call_impl
    return forward_call(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/root/Megatron-LM/megatron/core/distributed/data_parallel_base.py", line 22, in forward
    return self.module(*inputs, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1786, in _call_impl
    return forward_call(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/root/Megatron-LM/megatron/core/transformer/module.py", line 456, in forward
    outputs = self.module(*inputs, **kwargs)
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1786, in _call_impl
    return forward_call(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File ".../Megatron-Bridge-slime/src/megatron/bridge/models/qwen_vl/modelling_qwen3_vl/model.py", line 484, in forward
    output = self.language_model(
             ^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1786, in _call_impl
    return forward_call(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File ".../Megatron-Bridge-slime/src/megatron/bridge/models/qwen_vl/modelling_qwen3_vl/text_model.py", line 161, in forward
    hidden_states = self.decoder(
                    ^^^^^^^^^^^^^
  File "/root/Megatron-LM/megatron/core/transformer/transformer_block.py", line 586, in __call__
    return super().__call__(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/root/Megatron-LM/megatron/core/transformer/module.py", line 319, in __call__
    return super().__call__(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1786, in _call_impl
    return forward_call(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File ".../Megatron-Bridge-slime/src/megatron/bridge/models/qwen_vl/modelling_qwen3_vl/transformer_block.py", line 721, in forward
    hidden_states, context = layer(
                             ^^^^^^
  File "/root/Megatron-LM/megatron/core/transformer/transformer_layer.py", line 1044, in __call__
    return super().__call__(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/root/Megatron-LM/megatron/core/transformer/module.py", line 319, in __call__
    return super().__call__(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1786, in _call_impl
    return forward_call(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/root/Megatron-LM/megatron/core/transformer/transformer_layer.py", line 475, in forward
    hidden_states, context = self._forward_attention(*args, **kwargs)
                             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/root/Megatron-LM/megatron/core/transformer/transformer_layer.py", line 549, in _forward_attention
    attention_output_with_bias = self.self_attention(
                                 ^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1786, in _call_impl
    return forward_call(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File ".../Megatron-Bridge-slime/src/megatron/bridge/models/qwen_vl/modelling_qwen3_vl/attention.py", line 187, in forward
    query = apply_rotary_pos_emb_absolute(
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File ".../Megatron-Bridge-slime/src/megatron/bridge/models/qwen_vl/modelling_qwen3_vl/rope.py", line 324, in apply_rotary_pos_emb_absolute
    result = _apply_rotary_pos_emb_bshd(t, freqs, rotary_interleaved=config.rotary_interleaved)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/root/Megatron-LM/megatron/core/models/common/embeddings/rope_utils.py", line 125, in _apply_rotary_pos_emb_bshd
    t = (t * cos_) + (_rotate_half(t, rotary_interleaved) * sin_)
         ~~^~~~~~
RuntimeError: The size of tensor a (4096) must match the size of tensor b (1280) at non-singleton dimension 0

Steps to Reproduce

examples/geo3k_vlm/run_geo3k_qwen35.sh
with image dataset
with setting:
--tensor-model-parallel-size 2
--sequence-parallel
--pipeline-model-parallel-size 2
--context-parallel-size 1
--expert-model-parallel-size 4
--expert-tensor-parallel-size 1

Expected Behavior

no

Actual Behavior

no

Environment

• slime version:
• Python version:
• PyTorch version:
• CUDA/ROCm version:
• GPU type and count:
• OS:
• SGLang version (if relevant):
• Megatron-LM version (if relevant):

Logs

Additional Context

No response

Pre-submission Checklist

• I have read the CONTRIBUTING.md and understand the collaboration scope.
• I have read the documentation and my issue is not addressed there.
• I have searched for existing issues and this is not a duplicate.
• I have provided a minimal, reproducible example.

Start by reproducing examples/geo3k_vlm/run_geo3k_qwen35.sh with the image dataset and the listed parallelism settings. Trace the reported size mismatch through slime/backends/megatron_utils/model.py and the Qwen files under Megatron-Bridge-slime/src/megatron/bridge/models/qwen_vl/modelling_qwen3_vl/, especially rope.py and attention.py. Done means the run no longer fails in apply_rotary_pos_emb_absolute with those settings.