`neural_compressor.adaptor.torch_utils.pattern_detector`

Block detector for Transformer-based model.

Module Contents

Classes

TransformerBasedModelBlockPatternDetector

Detect the attention block and FFN block in transformer-based model.

class neural_compressor.adaptor.torch_utils.pattern_detector.TransformerBasedModelBlockPatternDetector(model: torch, pattern_lst: List[List[str | int]] = BLOCK_PATTERNS)[source]: Detect the attention block and FFN block in transformer-based model.