fix group quant bug (#203)

neuralmagic · Nov 8, 2024 · db6ccb2 · db6ccb2
1 parent a43dad2
commit db6ccb2
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/src/compressed_tensors/compressors/sparse_quantized_compressors/marlin_24.py b/src/compressed_tensors/compressors/sparse_quantized_compressors/marlin_24.py
@@ -238,7 +238,7 @@ def pack_scales_24(scales, quantization_args, w_shape):
     _, scale_perm_2_4, scale_perm_single_2_4 = get_permutations_24(num_bits)
 
     if (
-        quantization_args.strategy is QuantizationStrategy.GROUP
+        quantization_args.strategy == QuantizationStrategy.GROUP
         and quantization_args.group_size < size_k
     ):
         scales = scales.reshape((-1, len(scale_perm_2_4)))[:, scale_perm_2_4]