xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-20 05:55:01 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Misc] Add assertion and helpful message for marlin24 compressed models (#11388)

Browse Source
This commit is contained in:
Dipika Sikka 2024-12-23 13:23:38 -05:00 committed by GitHub
parent 2e726680b3
commit b866cdbd05
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 4 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
vllm/model_executor/layers/quantization/compressed_tensors/schemes/compressed_tensors_w4a16_24.py
View File

@ -61,6 +61,10 @@ class CompressedTensorsW4A16Sparse24(CompressedTensorsScheme):
params_dtype: torch.dtype, weight_loader: Callable, params_dtype: torch.dtype, weight_loader: Callable,
**kwargs): **kwargs):
assert params_dtype == torch.float16, (
"float16 is required for marlin24 compressd models. Set dtype=torch.float16" # noqa: E501
)
pack_factor = 32 // self.quant_type.size_bits pack_factor = 32 // self.quant_type.size_bits
output_size_per_partition = sum(output_partition_sizes) output_size_per_partition = sum(output_partition_sizes)