quark-torch-llm-ptq-eval

SkillAI & models

End-to-end Torch LLM PTQ recipe for AMD Quark — for PyTorch / HuggingFace transformers models (safetensors input): quantize, then validate, then evaluate accuracy, in one flow. Delegates the PTQ path (intake, planning, manifest, execution) to the quark-torch-ptq workflow, runs mandatory structural validation, and runs opt-in accuracy evaluation (perplexity / lm_eval / vLLM-accelerated). Trigger for "quantize and validate", "quantize and evaluate", "run PTQ end to end with accuracy check", "full PTQ pipeline with validation and eval", or "quantize Llama/Qwen/Mistral with FP8/INT4 and measure accuracy". For PTQ only (stop at the quantized output, no validation/eval) use quark-torch-ptq. Not for .onnx input models — use quark-onnx-ptq instead.

Available today. Use it from your connected AI after setup.

No other account needed.

Then ask your AI: use the quark-torch-llm-ptq-eval skill