From cbc1bfe710e905b7d5fa1b246690c93ca53564db Mon Sep 17 00:00:00 2001 From: Disty0 Date: Mon, 26 May 2025 01:24:06 +0300 Subject: [PATCH] Cleanup --- modules/model_quant_nncf.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/modules/model_quant_nncf.py b/modules/model_quant_nncf.py index 2e07eced8..2acdbcbbe 100644 --- a/modules/model_quant_nncf.py +++ b/modules/model_quant_nncf.py @@ -575,7 +575,7 @@ class INT8SymmetricWeightsDecompressor(torch.nn.Module): def forward(self, x, input=None, *args, return_decompressed_only=False, skip_int8_matmul=False): # pylint: disable=unused-argument,keyword-arg-before-vararg if skip_int8_matmul: - return decompress_int4_symmetric_compiled(x.weight, self.scale.unsqueeze(-1), self.compressed_weight_shape, self.result_dtype, self.result_shape) + return decompress_symmetric_compiled(x.weight.transpose(0,1), self.scale.unsqueeze(-1), self.result_dtype, self.result_shape) result = decompress_symmetric_compiled(x.weight, self.scale, self.result_dtype, self.result_shape) if return_decompressed_only: return result