microsoft / microsoft/foundry-local
[Bug] Model Qwen3.5 9b does not run with WebGPU
- Dominant language
- C++
- Stars
- 2.6k
- Forks
- 369
- Avg merge
- 2d 17h
- Merged PRs (30d)
- 39
Description
### Describe the issue
```
natke@timmy % foundry run qwen3.5-9b
● error: IPC error 'op_handler_failed': Error loading model qwen3.5-9b-generic-gpu:3: Error:
Microsoft.ML.OnnxRuntimeGenAI.OnnxRuntimeGenAIException: This is an invalid model. Error:
Duplicate definition of name (image_grid_thw_CUDAExecutionProvider).
at Microsoft.ML.OnnxRuntimeGenAI.Result.VerifySuccess(IntPtr) + 0x48
at Microsoft.ML.OnnxRuntimeGenAI.Model..ctor(Config) + 0x20
at Microsoft.Neutron.OpenAI.Provider.OnnxLoadedModel..ctor(String, Config, GenAIConfig,
InferenceModel, OnnxEP) + 0x64
at Microsoft.AI.Foundry.Local.ModelManager.d__9.MoveNext() + 0xb78
--- End of stack trace from previous location ---
at
Microsoft.AI.Foundry.Local.NativeInterop.<>c__DisplayClass10_0.<b__5>d.Move
Next() + 0x84
--- End of stack trace from previous location ---
at Microsoft.AI.Foundry.Local.NativeInterop.d__9.MoveNext() + 0xa8
Hint: Try 'foundry server stop' then 'foundry server start'.
```
Foundry Local version: `0.10.3`
O/S: macOS
### To reproduce
```
brew install foundrylocal
foundry run qwen3.5-9b
```
### Urgency
_No response_
### Platform and architecture
Windows X64
### OS Version
macOS 15.7.9
### Installation type
Released package/binary
### Foundry Local version
0.10.3
### API or surface area
CLI
### Hardware acceleration/backend
WebGPU/CUDA
### Backend/runtime version
_No response_
Contributor guide
Research direction
Start by reproducing `foundry run qwen3.5-9b` on macOS 15.7.9 with Foundry Local 0.10.3, then trace the model-loading path shown in the error. Investigate the duplicate `image_grid_thw_CUDAExecutionProvider` definition and verify that the command loads the model without the invalid-model error.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- cpp, macos
- Domain
- ai, cli
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Active
- Clarity
- Mostly clear
- Newbie friendliness
- 45/100