tocmo0nlord/axolotl
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

upgrade flash-attn to 2.8.3 for gpt-oss attn sink support (#3082)

Browse Source
This commit is contained in:
Wing Lian
2025-08-21 15:04:10 -04:00
committed by GitHub
parent 08e517ea48
commit 0fa752e58b
6 changed files with 7 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
examples/gpt-oss/gpt-oss-120b-fft-fsdp2-offload.yaml
View File

@@ -44,7 +44,7 @@ bf16: true
tf32: true tf32: true
flash_attention: true flash_attention: true
attn_implementation: kernels-community/vllm-flash-attn3 attn_implementation: kernels-community/vllm-flash-attn3 # this is not needed if using flash_attn >= 2.8.3
gradient_checkpointing: true gradient_checkpointing: true
activation_offloading: true activation_offloading: true

2
examples/gpt-oss/gpt-oss-20b-fft-deepspeed-zero3.yaml
View File

@@ -40,7 +40,7 @@ bf16: true
tf32: true tf32: true
flash_attention: true flash_attention: true
attn_implementation: kernels-community/vllm-flash-attn3 attn_implementation: kernels-community/vllm-flash-attn3 # this is not needed if using flash_attn >= 2.8.3
gradient_checkpointing: true gradient_checkpointing: true
activation_offloading: true activation_offloading: true

2
examples/gpt-oss/gpt-oss-20b-fft-fsdp2-offload.yaml
View File

@@ -41,7 +41,7 @@ bf16: true
tf32: true tf32: true
flash_attention: true flash_attention: true
attn_implementation: kernels-community/vllm-flash-attn3 attn_implementation: kernels-community/vllm-flash-attn3 # this is not needed if using flash_attn >= 2.8.3
gradient_checkpointing: true gradient_checkpointing: true
activation_offloading: true activation_offloading: true

2
examples/gpt-oss/gpt-oss-20b-fft-fsdp2.yaml
View File

@@ -40,7 +40,7 @@ bf16: true
tf32: true tf32: true
flash_attention: true flash_attention: true
attn_implementation: kernels-community/vllm-flash-attn3 attn_implementation: kernels-community/vllm-flash-attn3 # this is not needed if using flash_attn >= 2.8.3
gradient_checkpointing: true gradient_checkpointing: true
activation_offloading: true activation_offloading: true

2
examples/gpt-oss/gpt-oss-20b-sft-lora-singlegpu.yaml
View File

@@ -53,7 +53,7 @@ bf16: true
tf32: true tf32: true
flash_attention: true flash_attention: true
attn_implementation: kernels-community/vllm-flash-attn3 attn_implementation: kernels-community/vllm-flash-attn3 # this is not needed if using flash_attn >= 2.8.3
gradient_checkpointing: true gradient_checkpointing: true
activation_offloading: true activation_offloading: true

4
setup.py
View File

@@ -118,9 +118,9 @@ def get_package_version():
extras_require = { extras_require = {
"flash-attn": ["flash-attn==2.8.2"], "flash-attn": ["flash-attn==2.8.3"],
"ring-flash-attn": [ "ring-flash-attn": [
"flash-attn==2.8.2", "flash-attn==2.8.3",
"ring-flash-attn>=0.1.7", "ring-flash-attn>=0.1.7",
"yunchang==0.6.0", "yunchang==0.6.0",
], ],