Skip to content

PTQ-Level Efficiency, QAT-Level Accuracy | 训练后量化的效率,量化感知训练的精度

Create model_support_request.yml (#1738)

Xin He71f7946cf651
diff --git a/.github/ISSUE_TEMPLATE/model_support_request.yml b/.github/ISSUE_TEMPLATE/model_support_request.yml
new file mode 100644
index 000000000..9cb39e09d
--- /dev/null
+++ b/.github/ISSUE_TEMPLATE/model_support_request.yml
@@ -0,0 +1,105 @@
+name: Model Support Request
+description: Request support for a new model (similar to a feature request)
+title: "[Model]: "
+labels: ["Model Support"]
+body:
+ - type: markdown
+ attributes:
+ value: |
+ Thank you for the model support request! Please provide the details below so we can evaluate and prioritize support in AutoRound.
+
+ - type: input
+ id: model-name
+ attributes:
+ label: Model name
+ description: Model family/name (e.g., Qwen2.5, Llama-3.x, Mistral, Gemma).
+ placeholder: "e.g., Qwen2.5"
+ validations:
+ required: true
+
+ - type: input
+ id: model-variant
+ attributes:
+ label: Model variant / size
+ description: Parameter size and variant (base/instruct), and any specific checkpoint tag.
+ placeholder: "e.g., 7B-Instruct / 14B-Base / 70B-Instruct"
+ validations:
+ required: true
+
+ - type: input
+ id: model-link
+ attributes:
+ label: Model link
+ description: Link to the model page (Hugging Face / GitHub / official download).
+ placeholder: "https://huggingface.co/..."
+ validations:
+ required: true
+
+ - type: dropdown
+ id: request-type
+ attributes:
+ label: What do you want AutoRound to support?
+ multiple: true
+ options:
+ - Quantization / rounding workflow support for this model
+ - Example scripts / notebooks for this model
+ - Accuracy evaluation / benchmark recipe
+ - Export / integration with an inference backend (please specify below)
+ - Other (please specify below)
+ validations:
+ required: true
+
+ - type: textarea
+ id: motivation
+ attributes:
+ label: Motivation & use case
+ description: Why is this model important for you? What are you trying to do?
+ placeholder: "Describe your use case, constraints, and expected outcomes."
+ validations:
+ required: true
+
+ - type: textarea
+ id: environment
+ attributes:
+ label: Environment (optional but helpful)
+ description: Provide relevant environment details if you already tried something.
+ placeholder: |
+ - OS:
+ - Python:
+ - PyTorch:
+ - Transformers:
+ - AutoRound version/commit:
+ - Hardware (CPU/GPU):
+ validations:
+ required: false
+
+ - type: textarea
+ id: blockers
+ attributes:
+ label: Current blockers / errors (if any)
+ description: Paste logs, error messages, or describe what failed.
+ placeholder: "What doesn't work today? Any stack traces or minimal repro?"
+ validations:
+ required: false
+
+ - type: textarea
+ id: definition-of-done
+ attributes:
+ label: Definition of Done (acceptance criteria)
+ description: What would “done” look like for this model request?
+ placeholder: |
+ - [ ] Quantization pipeline runs end-to-end on the model
+ - [ ] Documented usage in README/docs
+ - [ ] Example command/script provided
+ - [ ] Basic accuracy/perf validation recorded
+ validations:
+ required: false
+
+ - type: textarea
+ id: additional-context
+ attributes:
+ label: Additional context
+ description: Add any extra context, references, or related issues/PRs.
+ placeholder: "Links, screenshots, related repos/issues, etc."
+ validations:
+ required: false