xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-10 04:26:00 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Multi Modal][Performance] Fused Q,K's apply_rope in more models (#25005)

Browse Source 
Signed-off-by: wwl2755 <wangwenlong2755@gmail.com>
Co-authored-by: Roger Wang <hey@rogerw.io>
This commit is contained in:
Wenlong Wang 2025-09-20 20:55:10 -07:00 committed by GitHub
parent 1cd885bd54
commit 035fd2bd2c
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
3 changed files with 29 additions and 19 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
vllm/model_executor/models/ernie45_vl.py
View File

@ -234,8 +234,9 @@ class Ernie4_5_VisionAttention(nn.Module):
q, k, v = (rearrange(x, "s b ... -> b s ...").contiguous()
for x in (q, k, v))
if rotary_pos_emb is not None:
q = apply_rotary_pos_emb_vision(q, rotary_pos_emb)
k = apply_rotary_pos_emb_vision(k, rotary_pos_emb)
qk_concat = torch.cat([q, k], dim=0)
qk_rotated = apply_rotary_pos_emb_vision(qk_concat, rotary_pos_emb)
q, k = torch.chunk(qk_rotated, 2, dim=0)
if self.is_flash_attn_backend:
# from vllm_flash_attn.flash_attn_interface import (
@ -261,8 +262,8 @@ class Ernie4_5_VisionAttention(nn.Module):
causal=False)
context_layer = rearrange(output,
"(b s) ... -> b s ...",
b=batch_size)
"(b s) h d -> s b (h d)",
b=batch_size).contiguous()
elif self.attn_backend == _Backend.TORCH_SDPA:
# Execute attention entry by entry for speed & less VRAM.
outputs = []
@ -281,6 +282,8 @@ class Ernie4_5_VisionAttention(nn.Module):
output_i = rearrange(output_i, "b h s d -> b s h d ")
outputs.append(output_i)
context_layer = torch.cat(outputs, dim=1)
context_layer = rearrange(context_layer,
"b s h d -> s b (h d)").contiguous()
elif self.attn_backend == _Backend.XFORMERS:
from xformers import ops as xops
from xformers.ops.fmha.attn_bias import BlockDiagonalMask
@ -291,8 +294,8 @@ class Ernie4_5_VisionAttention(nn.Module):
context_layer = xops.memory_efficient_attention_forward(
q, k, v, attn_bias=attn_bias, p=0, scale=None)
context_layer = rearrange(context_layer,
"b s h d -> s b (h d)").contiguous()
context_layer = rearrange(context_layer,
"b s h d -> s b (h d)").contiguous()
output, _ = self.proj(context_layer)
return output

17
vllm/model_executor/models/glm4_1v.py
View File

@ -315,8 +315,10 @@ class Glm4vVisionAttention(nn.Module):
q, k, v = (rearrange(x, "s b ... -> b s ...").contiguous()
for x in (q, k, v))
if rotary_pos_emb is not None:
q = apply_rotary_pos_emb_vision(q, rotary_pos_emb)
k = apply_rotary_pos_emb_vision(k, rotary_pos_emb)
# [2 * b, s, heads, head_dim]
qk_concat = torch.cat([q, k], dim=0)
qk_rotated = apply_rotary_pos_emb_vision(qk_concat, rotary_pos_emb)
q, k = torch.chunk(qk_rotated, 2, dim=0)
if self.attn_backend == _Backend.FLASH_ATTN:
# from vllm_flash_attn.flash_attn_interface import (
@ -341,8 +343,8 @@ class Glm4vVisionAttention(nn.Module):
)
context_layer = rearrange(output,
"(b s) ... -> b s ...",
b=batch_size)
"(b s) h d -> s b (h d)",
b=batch_size).contiguous()
elif self.attn_backend == _Backend.TORCH_SDPA:
# Execute attention entry by entry for speed & less VRAM.
outputs = []
@ -361,6 +363,8 @@ class Glm4vVisionAttention(nn.Module):
output_i = rearrange(output_i, "b h s d -> b s h d ")
outputs.append(output_i)
context_layer = torch.cat(outputs, dim=1)
context_layer = rearrange(context_layer,
"b s h d -> s b (h d)").contiguous()
elif self.attn_backend == _Backend.XFORMERS:
from xformers import ops as xops
from xformers.ops.fmha.attn_bias import BlockDiagonalMask
@ -371,9 +375,8 @@ class Glm4vVisionAttention(nn.Module):
context_layer = xops.memory_efficient_attention_forward(
q, k, v, attn_bias=attn_bias, p=0, scale=None)
context_layer = rearrange(context_layer,
"b s h d -> s b (h d)").contiguous()
context_layer = rearrange(context_layer,
"b s h d -> s b (h d)").contiguous()
output, _ = self.proj(context_layer)
return output

16
vllm/model_executor/models/qwen2_vl.py
View File

@ -377,8 +377,10 @@ class Qwen2VisionAttention(nn.Module):
q, k, v = (rearrange(x, "s b ... -> b s ...").contiguous()
for x in (q, k, v))
if rotary_pos_emb is not None:
q = apply_rotary_pos_emb_vision(q, rotary_pos_emb)
k = apply_rotary_pos_emb_vision(k, rotary_pos_emb)
# [2 * b, s, heads, head_dim]
qk_concat = torch.cat([q, k], dim=0)
qk_rotated = apply_rotary_pos_emb_vision(qk_concat, rotary_pos_emb)
q, k = torch.chunk(qk_rotated, 2, dim=0)
if self.is_flash_attn_backend:
if self.attn_backend == _Backend.ROCM_AITER_FA:
@ -402,8 +404,8 @@ class Qwen2VisionAttention(nn.Module):
causal=False)
context_layer = rearrange(output,
"(b s) ... -> b s ...",
b=batch_size)
"(b s) h d -> s b (h d)",
b=batch_size).contiguous()
elif self.attn_backend == _Backend.TORCH_SDPA:
# Execute attention entry by entry for speed & less VRAM.
outputs = []
@ -422,6 +424,8 @@ class Qwen2VisionAttention(nn.Module):
output_i = rearrange(output_i, "b h s d -> b s h d ")
outputs.append(output_i)
context_layer = torch.cat(outputs, dim=1)
context_layer = rearrange(context_layer,
"b s h d -> s b (h d)").contiguous()
elif self.attn_backend == _Backend.XFORMERS:
from xformers import ops as xops
from xformers.ops.fmha.attn_bias import BlockDiagonalMask
@ -432,8 +436,8 @@ class Qwen2VisionAttention(nn.Module):
context_layer = xops.memory_efficient_attention_forward(
q, k, v, attn_bias=attn_bias, p=0, scale=None)
context_layer = rearrange(context_layer,
"b s h d -> s b (h d)").contiguous()
context_layer = rearrange(context_layer,
"b s h d -> s b (h d)").contiguous()
output, _ = self.proj(context_layer)
return output