AssertionError: Unsupported function type position_ids

[JChunX]

Hi, I am trying to run the build example build.py but I am getting the error:

(webgpu-env) user@le-big-mac repos/web-stable-diffusion (main *) » python3 build.py
Automatically configuring target: metal -keys=metal,gpu -max_function_args=31 -max_num_threads=256 -max_shared_memory_per_block=32768 -max_threads_per_block=1024 -thread_warp_size=32
`text_config_dict` is provided which will be used to initialize `CLIPTextConfig`. The value `text_config["id2label"]` will be overriden.
`text_config_dict` is provided which will be used to initialize `CLIPTextConfig`. The value `text_config["bos_token_id"]` will be overriden.
`text_config_dict` is provided which will be used to initialize `CLIPTextConfig`. The value `text_config["eos_token_id"]` will be overriden.
Traceback (most recent call last):
  File "~Dropbox/repos/web-stable-diffusion/build.py", line 157, in <module>
    mod, params = trace_models(torch_dev_key)
  File "~Dropbox/repos/web-stable-diffusion/build.py", line 80, in trace_models
    clip = trace.clip_to_text_embeddings(pipe)
  File "~Dropbox/repos/web-stable-diffusion/web_stable_diffusion/trace/model_trace.py", line 27, in clip_to_text_embeddings
    mod = dynamo_capture_subgraphs(
  File "~miniconda3/envs/webgpu-env/lib/python3.10/site-packages/tvm/relax/frontend/torch/dynamo.py", line 198, in dynamo_capture_subgraphs
    compiled_model(*params, **kwargs)
  File "~miniconda3/envs/webgpu-env/lib/python3.10/site-packages/torch/_dynamo/eval_frame.py", line 408, in _fn
    return fn(*args, **kwargs)
  File "~miniconda3/envs/webgpu-env/lib/python3.10/site-packages/torch/_dynamo/eval_frame.py", line 569, in catch_errors
    return callback(frame, cache_entry, hooks, frame_state)
  File "~miniconda3/envs/webgpu-env/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 671, in _convert_frame
    result = inner_convert(frame, cache_entry, hooks, frame_state)
  File "~miniconda3/envs/webgpu-env/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 377, in _convert_frame_assert
    return _compile(
  File "~miniconda3/envs/webgpu-env/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 595, in _compile
    guarded_code = compile_inner(code, one_graph, hooks, transform)
  File "~miniconda3/envs/webgpu-env/lib/python3.10/site-packages/torch/_dynamo/utils.py", line 243, in time_wrapper
    r = func(*args, **kwargs)
  File "~miniconda3/envs/webgpu-env/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 512, in compile_inner
    out_code = transform_code_object(code, transform)
  File "~miniconda3/envs/webgpu-env/lib/python3.10/site-packages/torch/_dynamo/bytecode_transformation.py", line 1033, in transform_code_object
    transformations(instructions, code_options)
  File "~miniconda3/envs/webgpu-env/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 150, in _fn
    return fn(*args, **kwargs)
  File "~miniconda3/envs/webgpu-env/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 477, in transform
    tracer.run()
  File "~miniconda3/envs/webgpu-env/lib/python3.10/site-packages/torch/_dynamo/symbolic_convert.py", line 2120, in run
    super().run()
  File "~miniconda3/envs/webgpu-env/lib/python3.10/site-packages/torch/_dynamo/symbolic_convert.py", line 815, in run
    and self.step()
  File "~miniconda3/envs/webgpu-env/lib/python3.10/site-packages/torch/_dynamo/symbolic_convert.py", line 778, in step
    getattr(self, inst.opname)(inst)
  File "~miniconda3/envs/webgpu-env/lib/python3.10/site-packages/torch/_dynamo/symbolic_convert.py", line 2235, in RETURN_VALUE
    self.output.compile_subgraph(
  File "~miniconda3/envs/webgpu-env/lib/python3.10/site-packages/torch/_dynamo/output_graph.py", line 880, in compile_subgraph
    self.compile_and_call_fx_graph(tx, list(reversed(stack_values)), root)
  File "~miniconda3/envs/webgpu-env/lib/python3.10/contextlib.py", line 79, in inner
    return func(*args, **kwds)
  File "~miniconda3/envs/webgpu-env/lib/python3.10/site-packages/torch/_dynamo/output_graph.py", line 1025, in compile_and_call_fx_graph
    compiled_fn = self.call_user_compiler(gm)
  File "~miniconda3/envs/webgpu-env/lib/python3.10/site-packages/torch/_dynamo/utils.py", line 243, in time_wrapper
    r = func(*args, **kwargs)
  File "~miniconda3/envs/webgpu-env/lib/python3.10/site-packages/torch/_dynamo/output_graph.py", line 1096, in call_user_compiler
    raise BackendCompilerFailed(self.compiler_fn, e).with_traceback(
  File "~miniconda3/envs/webgpu-env/lib/python3.10/site-packages/torch/_dynamo/output_graph.py", line 1077, in call_user_compiler
    compiled_fn = compiler_fn(gm, self.example_inputs())
  File "~miniconda3/envs/webgpu-env/lib/python3.10/site-packages/torch/_dynamo/repro/after_dynamo.py", line 117, in debug_wrapper
    compiled_gm = compiler_fn(gm, example_inputs)
  File "~miniconda3/envs/webgpu-env/lib/python3.10/site-packages/torch/__init__.py", line 1655, in __call__
    return self.compiler_fn(model_, inputs_, **self.kwargs)
  File "~miniconda3/envs/webgpu-env/lib/python3.10/site-packages/tvm/relax/frontend/torch/dynamo.py", line 184, in _capture
    mod_ = from_fx(
  File "~miniconda3/envs/webgpu-env/lib/python3.10/site-packages/tvm/relax/frontend/torch/fx_translator.py", line 1635, in from_fx
    return TorchFXImporter().from_fx(
  File "~miniconda3/envs/webgpu-env/lib/python3.10/site-packages/tvm/relax/frontend/torch/fx_translator.py", line 1520, in from_fx
    func_name in self.convert_map
torch._dynamo.exc.BackendCompilerFailed: backend='_capture' raised:
AssertionError: Unsupported function type position_ids

Set TORCH_LOGS="+dynamo" and TORCHDYNAMO_VERBOSE=1 for more information


You can suppress this exception and fall back to eager by setting:
    import torch._dynamo
    torch._dynamo.config.suppress_errors = True

OS: macOS Sonoma 14.0
Device: M1 Mac
Python: 3.10

[yunho-c]

I encountered the same issue with the preview (nightly) branch of PyTorch and resolved it by installing stable.
However, I encounter another issue AssertionError: Unsupported module type <class 'diffusers.models.attention_processor.Attention' which seems to originate from breaking API change in diffusers library

[guoyaol]

This is mainly caused by the breaking API change diffusers library, we are working on the support and will keep you updated

[senlyu163]

Same problem, can you provide some help?

[zhongzhenluo]

This is mainly caused by the breaking API change diffusers library, we are working on the support and will keep you updated

Hi @guoyaol , do you have estimated time for a fix or which version of diffuser work ? Thanks

[0xWOLAND]

@guoyaol I have the same issue. Bumping this.
OS: Windows 11 WSL
Python: 3.10.12

[mei-ye]

@guoyaol: If you have a last-known-good, could you please give versions of torch, diffusers and transformers? Thanks!

[sunzj]

@mei-ye @0xWOLAND @zhongzhenluo @JChunX try torch 2.0.0.

[sunzj]

should fixed by: apache/tvm#16690

[sidsingla]

Hi,
Still same issue.