huggingface/text-generation-inference
5
0
Fork 0
mirror of https://github.com/huggingface/text-generation-inference.git synced 2025-09-12 04:44:52 +00:00
Code Issues Packages Projects Releases Wiki Activity

use marlin even on 89

Browse Source
This commit is contained in:
OlivierDehaene 2024-07-23 10:35:32 +02:00
parent 473f968a01
commit 025f80dfd4
No known key found for this signature in database
GPG Key ID: BB104D67809DA93C
1 changed files with 5 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
server/text_generation_server/layers/fp8.py
View File

@ -32,8 +32,8 @@ def get_fp8_linear() -> torch.nn.Module:
"""
if SYSTEM == "cuda":
major, minor = torch.cuda.get_device_capability()
if major == 8 and minor < 9:
major, _ = torch.cuda.get_device_capability()
if major == 8:
from text_generation_server.layers.marlin import GPTQMarlinFP8Linear
return GPTQMarlinFP8Linear
@ -188,6 +188,9 @@ class Fp8Linear(torch.nn.Module):
dtype,
) -> None:
super().__init__()
if FBGEMM_MM_AVAILABLE:
log_once(logger.info, "Using FBGEMM fp8 optimized kernels")
self.dtype = dtype
self.qweight = qweight
self.scale = scale