🐛 [Bug] slice_scatter: `ValueError: all elements of broadcast shape must be non-negative`

[zewenli98]

Bug Description

While compiling AM1.5 Language and Diffusion models, I got the following error:

[2026-07-28 00:22:00,452][torch_tensorrt.dynamo._compiler][WARNING] - Dynamic dimension 's13': Input range [1, 6] vs exported program range [2, 6]. Input.min_shape=1 but the exported program's min is 2 (PyTorch 0/1 specialization -- Dim(min=1) is recorded as min=2). The compiled engine's min will be 2.
[2026-07-28 00:22:00,452][torch_tensorrt.dynamo._compiler][WARNING] - Dynamic dimension 's15': Input range [1, 3358] vs exported program range [2, 3358]. Input.min_shape=1 but the exported program's min is 2 (PyTorch 0/1 specialization -- Dim(min=1) is recorded as min=2). The compiled engine's min will be 2.
Traceback (most recent call last):
  File "/home/scratch.zewenl_sw/docker_workspace/alpamayo-recipes/recipes/alpamayo1_5_quant/eval.py", line 245, in <module>
    main()
  File "/home/scratch.zewenl_sw/docker_workspace/alpamayo-recipes/recipes/alpamayo1_5_quant/eval.py", line 171, in main
    compiled_vision, compiled_language, compiled_diffusion = compile_trt(model, processor, clip_ids, device=device)
                                                             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/scratch.zewenl_sw/docker_workspace/alpamayo-recipes/recipes/alpamayo1_5_quant/compile_trt.py", line 133, in compile_trt
    compile_diffusion_step_no_cache(
  File "/home/scratch.zewenl_sw/docker_workspace/alpamayo-recipes/recipes/alpamayo1_5_quant/diffusion.py", line 305, in compile_diffusion_step_no_cache
    trt_step = torch_tensorrt.dynamo.compile(exported, inputs=trt_input_specs, **trt_settings)
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/scratch.zewenl_sw/docker_workspace/alpamayo-recipes/recipes/alpamayo1_5_quant/am15_quant/lib/python3.12/site-packages/torch_tensorrt/dynamo/_compiler.py", line 826, in compile
    trt_gm = compile_module(
             ^^^^^^^^^^^^^^^
  File "/home/scratch.zewenl_sw/docker_workspace/alpamayo-recipes/recipes/alpamayo1_5_quant/am15_quant/lib/python3.12/site-packages/torch_tensorrt/dynamo/_compiler.py", line 1416, in compile_module
    trt_module = convert_module(
                 ^^^^^^^^^^^^^^^
  File "/home/scratch.zewenl_sw/docker_workspace/alpamayo-recipes/recipes/alpamayo1_5_quant/am15_quant/lib/python3.12/site-packages/torch_tensorrt/dynamo/conversion/_conversion.py", line 348, in convert_module
    serialized_interpreter_result = interpret_module_to_result(
                                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/scratch.zewenl_sw/docker_workspace/alpamayo-recipes/recipes/alpamayo1_5_quant/am15_quant/lib/python3.12/site-packages/torch_tensorrt/dynamo/conversion/_conversion.py", line 281, in interpret_module_to_result
    interpreter_result = interpreter.run()
                         ^^^^^^^^^^^^^^^^^
  File "/home/scratch.zewenl_sw/docker_workspace/alpamayo-recipes/recipes/alpamayo1_5_quant/am15_quant/lib/python3.12/site-packages/torch_tensorrt/dynamo/conversion/_TRTInterpreter.py", line 442, in run
    self._construct_trt_network_def()
  File "/home/scratch.zewenl_sw/docker_workspace/alpamayo-recipes/recipes/alpamayo1_5_quant/am15_quant/lib/python3.12/site-packages/torch_tensorrt/dynamo/conversion/_TRTInterpreter.py", line 424, in _construct_trt_network_def
    super().run()
  File "/home/scratch.zewenl_sw/docker_workspace/alpamayo-recipes/recipes/alpamayo1_5_quant/am15_quant/lib/python3.12/site-packages/torch/fx/interpreter.py", line 197, in run
    self.env[node] = self.run_node(node)
                     ^^^^^^^^^^^^^^^^^^^
  File "/home/scratch.zewenl_sw/docker_workspace/alpamayo-recipes/recipes/alpamayo1_5_quant/am15_quant/lib/python3.12/site-packages/torch_tensorrt/dynamo/conversion/_TRTInterpreter.py", line 528, in run_node
    trt_node: torch.fx.Node = super().run_node(n)
                              ^^^^^^^^^^^^^^^^^^^
  File "/home/scratch.zewenl_sw/docker_workspace/alpamayo-recipes/recipes/alpamayo1_5_quant/am15_quant/lib/python3.12/site-packages/torch/fx/interpreter.py", line 294, in run_node
    return getattr(self, n.op)(n.target, args, kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/scratch.zewenl_sw/docker_workspace/alpamayo-recipes/recipes/alpamayo1_5_quant/am15_quant/lib/python3.12/site-packages/torch_tensorrt/dynamo/conversion/_TRTInterpreter.py", line 686, in call_function
    return converter(self.ctx, target, args, kwargs, self._cur_node_name)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/scratch.zewenl_sw/docker_workspace/alpamayo-recipes/recipes/alpamayo1_5_quant/am15_quant/lib/python3.12/site-packages/torch_tensorrt/dynamo/conversion/converter_utils.py", line 723, in convert_with_type_enforcement
    return func(ctx, target, new_args, new_kwargs, name)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/scratch.zewenl_sw/docker_workspace/alpamayo-recipes/recipes/alpamayo1_5_quant/am15_quant/lib/python3.12/site-packages/torch_tensorrt/dynamo/conversion/aten_ops_converters.py", line 1183, in aten_ops_slice_scatter
    return impl.slice_scatter.slice_scatter(
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/scratch.zewenl_sw/docker_workspace/alpamayo-recipes/recipes/alpamayo1_5_quant/am15_quant/lib/python3.12/site-packages/torch_tensorrt/dynamo/conversion/impl/slice_scatter.py", line 218, in slice_scatter
    indices_np = np.broadcast_to(indices_np, src_shape).astype(np.int64)
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/scratch.zewenl_sw/docker_workspace/alpamayo-recipes/recipes/alpamayo1_5_quant/am15_quant/lib/python3.12/site-packages/numpy/lib/_stride_tricks_impl.py", line 443, in broadcast_to
    return _broadcast_to(array, shape, subok=subok, readonly=True)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/scratch.zewenl_sw/docker_workspace/alpamayo-recipes/recipes/alpamayo1_5_quant/am15_quant/lib/python3.12/site-packages/numpy/lib/_stride_tricks_impl.py", line 379, in _broadcast_to
    raise ValueError('all elements of broadcast shape must be non-'
ValueError: all elements of broadcast shape must be non-negative

While executing %slice_scatter : [num_users=1] = call_function[target=torch.ops.aten.slice_scatter.default](args = (%select_5, %copy, 2, 1, 60, 3), kwargs = {})
Original traceback:
File "/home/scratch.zewenl_sw/docker_workspace/alpamayo-recipes/recipes/alpamayo1_5_quant/diffusion.py", line 51, in forward
    expert_out = self.expert(
  File "/home/scratch.zewenl_sw/docker_workspace/alpamayo-recipes/recipes/alpamayo1_5_quant/am15_quant/lib/python3.12/site-packages/transformers/models/qwen3_vl/modeling_qwen3_vl.py", line 846, in forward
    position_embeddings = self.rotary_emb(hidden_states, position_ids)
  File "/home/scratch.zewenl_sw/docker_workspace/alpamayo-recipes/recipes/alpamayo1_5_quant/am15_quant/lib/python3.12/site-packages/transformers/models/qwen3_vl/modeling_qwen3_vl.py", line 329, in forward
    freqs = self.apply_interleaved_mrope(freqs, self.mrope_section)
Use tlparse to see full graph. (https://github.com/pytorch/tlparse?tab=readme-ov-file#tlparse-parse-structured-pt2-logs)

To Reproduce

Compile AM1.5 Language and Diffusion models.

Expected behavior

pass w/o error

Environment

Build information about Torch-TensorRT can be found by turning on debug messages

• Torch-TensorRT Version (e.g. 1.0.0):
• PyTorch Version (e.g. 1.0):
• CPU Architecture:
• OS (e.g., Linux):
• How you installed PyTorch (conda, pip, libtorch, source):
• Build command you used (if compiling from source):
• Are you using local sources or building from archives:
• Python version:
• CUDA version:
• GPU models and configuration:
• Any other relevant information:

Additional context