Skip to content

[Misc] Pass cutlass_fp8_supported correctly in fbgemm_fp8 #6871

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 7 commits into from
Jul 28, 2024
Merged

[Misc] Pass cutlass_fp8_supported correctly in fbgemm_fp8 #6871

Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
2db58da
Use marlin kernel when cutlass fp8 is not support
zeyugao Jul 28, 2024
3b63202
Cleanup
zeyugao Jul 28, 2024
d75ace9
Revert "Cleanup"
zeyugao Jul 28, 2024
f803f60
Revert "Use marlin kernel when cutlass fp8 is not support"
zeyugao Jul 28, 2024
cbd081f
Pass cutlass_fp8_supported correctly
zeyugao Jul 28, 2024
1386d1e
Move cutlass_fp8_supported into __init__
zeyugao Jul 28, 2024
db7a375
yapf format
zeyugao Jul 28, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
19 changes: 11 additions & 8 deletions vllm/model_executor/layers/quantization/fbgemm_fp8.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,6 +9,7 @@
UnquantizedLinearMethod)
from vllm.model_executor.layers.quantization.base_config import (
QuantizationConfig, QuantizeMethodBase)
from vllm.model_executor.layers.quantization.fp8 import cutlass_fp8_supported
from vllm.model_executor.layers.quantization.utils.marlin_utils_fp8 import (
apply_fp8_marlin_linear, prepare_fp8_layer_for_marlin)
from vllm.model_executor.layers.quantization.utils.quant_utils import (
Expand Down Expand Up @@ -72,6 +73,7 @@ class FBGEMMFp8LinearMethod(LinearMethodBase):

def __init__(self, quant_config: FBGEMMFp8Config):
self.quant_config = quant_config
self.cutlass_fp8_supported = cutlass_fp8_supported()

def create_weights(
self,
Expand Down Expand Up @@ -139,11 +141,12 @@ def apply(self,
size_k=layer.input_size_per_partition,
bias=bias)

return apply_fp8_linear(input=x,
weight=layer.weight,
weight_scale=layer.weight_scale,
input_scale=None,
input_scale_ub=layer.input_scale_ub,
bias=bias,
cutlass_fp8_supported=True,
use_per_token_if_dynamic=True)
return apply_fp8_linear(
input=x,
weight=layer.weight,
weight_scale=layer.weight_scale,
input_scale=None,
input_scale_ub=layer.input_scale_ub,
bias=bias,
cutlass_fp8_supported=self.cutlass_fp8_supported,
use_per_token_if_dynamic=True)