Skip to content

[misc] add forward context for attention #9029

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 11 commits into from
Oct 3, 2024
Merged

[misc] add forward context for attention #9029

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
766e76b
add forward context
youkaichao Oct 2, 2024
496c7ad
add model_runner
youkaichao Oct 2, 2024
092f3f3
add enc decode model runner
youkaichao Oct 2, 2024
bb18e56
add embedding
youkaichao Oct 2, 2024
01e5a7d
add draft model runner
youkaichao Oct 2, 2024
9ed2a62
change flash attn
youkaichao Oct 2, 2024
55a3863
remove custom op inside custom op
youkaichao Oct 2, 2024
c6e98e6
fix batch expansion
youkaichao Oct 3, 2024
fa2c8fd
refactor tests
youkaichao Oct 3, 2024
0d445b2
remove custom op inside custom op
youkaichao Oct 3, 2024
d19cc10
remove custom op inside custom op
youkaichao Oct 3, 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
56 changes: 7 additions & 49 deletions tests/kernels/test_flash_attn.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,9 +3,9 @@
import pytest
import torch

import vllm.attention.backends.flash_attn # noqa: F401
from tests.kernels.utils import opcheck
from vllm.utils import seed_everything
from vllm.vllm_flash_attn import (flash_attn_varlen_func,
flash_attn_with_kvcache)

NUM_HEADS = [(4, 4), (8, 2), (16, 2)]
HEAD_SIZES = [128, 256]
Expand Down Expand Up @@ -112,36 +112,17 @@ def test_flash_attn_with_paged_kv(
(num_seqs, max_num_blocks_per_seq),
dtype=torch.int32)

output = torch.ops.vllm.flash_attn_with_kvcache(
decode_query=query.unsqueeze(1),
key_cache=key_cache,
value_cache=value_cache,
output = flash_attn_with_kvcache(
q=query.unsqueeze(1),
k_cache=key_cache,
v_cache=value_cache,
softmax_scale=scale,
causal=True,
block_table=block_tables,
cache_seqlens=kv_lens_tensor,
softcap=soft_cap if soft_cap is not None else 0,
).squeeze(1)

if num_blocks <= 2048:
test_utils = ["test_faketensor", "test_schema"]
else:
test_utils = ["test_faketensor"]

opcheck(torch.ops.vllm.flash_attn_with_kvcache,
args=tuple(),
kwargs=dict(
decode_query=query.unsqueeze(1),
key_cache=key_cache,
value_cache=value_cache,
softmax_scale=scale,
causal=True,
block_table=block_tables,
cache_seqlens=kv_lens_tensor,
softcap=soft_cap if soft_cap is not None else 0,
),
test_utils=test_utils)

ref_output = ref_paged_attn(
query=query,
key_cache=key_cache,
Expand Down Expand Up @@ -213,7 +194,7 @@ def test_varlen_with_paged_kv(
(num_seqs, max_num_blocks_per_seq),
dtype=torch.int32)

output = torch.ops.vllm.flash_attn_varlen_func(
output = flash_attn_varlen_func(
q=query,
k=key_cache,
v=value_cache,
Expand All @@ -228,29 +209,6 @@ def test_varlen_with_paged_kv(
softcap=soft_cap if soft_cap is not None else 0,
)

if num_blocks <= 2048:
test_utils = ["test_faketensor", "test_schema"]
else:
test_utils = ["test_faketensor"]

opcheck(torch.ops.vllm.flash_attn_varlen_func,
args=tuple(),
kwargs=dict(
q=query,
k=key_cache,
v=value_cache,
cu_seqlens_q=cu_query_lens,
cu_seqlens_k=cu_kv_lens,
max_seqlen_q=max_query_len,
max_seqlen_k=max_kv_len,
softmax_scale=scale,
causal=True,
window_size=window_size,
block_table=block_tables,
softcap=soft_cap if soft_cap is not None else 0,
),
test_utils=test_utils)

ref_output = ref_paged_attn(
query=query,
key_cache=key_cache,
Expand Down
Loading
Loading