원클릭으로
efficient-transformers
efficient-transformers에는 quic에서 수집한 skills 3개가 있으며, 저장소 수준 직업 범위와 사이트 내 skill 상세 페이지를 제공합니다.
이 저장소의 skills
Add, modify, or review QEfficient transform code. Use when the user wants a new PyTorch transform, module mapping, method mapper, module mutator, quantization transform, prefill/sampler/SPD/blocking transform, ONNX-adjacent transform wiring, or tests for transform registration/parity in `QEfficient/base/pytorch_transforms.py`, `QEfficient/transformers/models/pytorch_transforms.py`, `QEfficient/transformers/quantizers/quant_transforms.py`, `QEfficient/base/modeling_qeff.py`, or `QEfficient/transformers/models/modeling_auto.py`.
Add support for a new Hugging Face model in this QEfficient repo. Use when the user provides a Hub model id, pasted model card, config excerpt, or architecture name and wants the QEff equivalent implemented, mapped to the closest existing wrapper family, validated with a tiny or config-derived test model, and covered through `tests/test_model_quickcheck.py`.
Rebase downstream QEff wrappers onto Hugging Face Transformers mainline with minimal divergence, preserving runtime/export parity and validating against tests/test_model_quickcheck.py.