biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Attention][UX][1/N] Add AttentionConfig and change attention env vars to CLI arguments (#26315)

Browse Source 
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
Signed-off-by: Matthew Bonanni <mbonanni001@gmail.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Lucas Wilkinson <LucasWilkinson@users.noreply.github.com>
This commit is contained in:
Matthew Bonanni
2025-12-05 12:48:43 -05:00
committed by GitHub
parent dff0a2b394
commit 66e674cdd5
22 changed files with 367 additions and 325 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
tests/compile/test_fusion_attn.py
View File

@@ -12,13 +12,13 @@ from vllm._custom_ops import cutlass_scaled_fp4_mm, scaled_fp4_quant
from vllm.attention.backends.abstract import AttentionMetadata
from vllm.attention.backends.registry import AttentionBackendEnum
from vllm.attention.layer import Attention
from vllm.attention.selector import global_force_attn_backend_context_manager
from vllm.compilation.fusion_attn import ATTN_OP, AttnFusionPass
from vllm.compilation.fx_utils import find_op_nodes
from vllm.compilation.matcher_utils import QUANT_OPS
from vllm.compilation.noop_elimination import NoOpEliminationPass
from vllm.compilation.post_cleanup import PostCleanupPass
from vllm.config import (
AttentionConfig,
CacheConfig,
CompilationConfig,
CompilationMode,
@@ -335,6 +335,7 @@ def test_attention_quant_pattern(
custom_ops=custom_ops_list,
),
cache_config=CacheConfig(cache_dtype="fp8"),
attention_config=AttentionConfig(backend=backend),
)
# Create test inputs
@@ -352,7 +353,6 @@ def test_attention_quant_pattern(
with (
set_current_vllm_config(vllm_config_unfused),
set_forward_context(attn_metadata=None, vllm_config=vllm_config_unfused),
global_force_attn_backend_context_manager(backend),
):
model_unfused = model_class(
num_qo_heads=num_qo_heads,
@@ -378,7 +378,6 @@ def test_attention_quant_pattern(
with (
set_current_vllm_config(vllm_config),
set_forward_context(attn_metadata=None, vllm_config=vllm_config),
global_force_attn_backend_context_manager(backend),
):
model_fused = model_class(
num_qo_heads=num_qo_heads,