Function dequantize_state_dict

src/quant_utils.py:286–297 · view source on GitHub ↗

(state_dict: dict[str, torch.Tensor], dtype: torch.dtype = torch.float16)

Source from the content-addressed store, hash-verified

source not stored for this graph (policy: none)

nothing calls this directly

dequantize_weight_from_fp8Function · 0.85

no test coverage detected