Unsupported operator: com.microsoft.QAttention
Open
Nobody has claimed this yet.
- Dominant language
- C++
- Stars
- 333
- Forks
- 150
- Avg merge
- 4d 19h
- Merged PRs (30d)
- 54
Description
- MIGraphX commit: 431e714
- GPU: MI210 (gfx90a)
- ROCm version: 6.0.1
Model(s) to reproduce:
- https://huggingface.co/Intel/distilbert-base-uncased-finetuned-sst-2-english-int8-static/blob/main/model.onnx
- https://huggingface.co/Intel/distilbert-base-uncased-distilled-squad-int8-static/blob/main/model.onnx
- https://huggingface.co/focia/yt-title-grader-quantized/blob/main/model_optimized_quantized.onnx
- https://huggingface.co/focia/yt-desire-quantized/blob/main/model_optimized_quantized.onnx
- https://huggingface.co/focia/yt-list-quantized/blob/main/model_optimized_quantized.onnx
- https://huggingface.co/focia/yt-negativity-quantized/blob/main/model_optimized_quantized.onnx
- https://huggingface.co/focia/yt-curiosity-quantized/blob/main/model_optimized_quantized.onnx
- https://huggingface.co/focia/yt-openloop-quantized/blob/main/model_optimized_quantized.onnx
Fails at: /code/AMDMIGraphX/src/onnx/onnx_parser.cpp:419: parse_graph: Unknown operator: QAttention
Note: More info at ContribOperators.md/com.microsoft.QAttention
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start at src/onnx/onnx_parser.cpp:419, where parsing stops with Unknown operator: QAttention. Run the listed quantized ONNX models on the stated MI210/ROCm setup to reproduce the failure and read the linked com.microsoft.QAttention operator documentation. Done means these models parse successfully without the unsupported-operator error.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- cpp
- Domain
- machine-learning
- Issue type
- Feature
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 35/100