Skip to content
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

Jump to bottom

llama3.2 + cross attn test #220

Merged
merged 7 commits into from
Oct 4, 2024
Merged

llama3.2 + cross attn test #220

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
64cdd32
llama3.2 + cross attn test
Oct 3, 2024
caf095b
lint issues fix
Oct 3, 2024
4b54363
mypy errors
Oct 3, 2024
34d2658
making yapf happy
Oct 4, 2024
9ed31a8
cut off WA for tunned gemms
Oct 4, 2024
2b25a0c
Merge branch 'main' into maleksan_llama32_support
maleksan85 Oct 4, 2024
cfe23d8
try and catch for non continuous tensor
Oct 4, 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
4 changes: 2 additions & 2 deletions tests/kernels/test_encoder_decoder_attn.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,7 +21,8 @@
from vllm.utils import is_hip

# List of support backends for encoder/decoder models
LIST_ENC_DEC_SUPPORTED_BACKENDS = [_Backend.XFORMERS]
LIST_ENC_DEC_SUPPORTED_BACKENDS = [_Backend.XFORMERS] if not is_hip() \
else [_Backend.ROCM_FLASH]

HEAD_SIZES = [64, 256]

Expand Down Expand Up @@ -807,7 +808,6 @@ def test_encoder_only(
assert_actual_matches_ideal(enc_test_params, enc_pckd_act_out)


@pytest.mark.skipif(is_hip(), reason=STR_NOT_IMPL_ENC_DEC_ROCM_HIP)
@pytest.mark.parametrize("num_heads", NUM_HEADS)
@pytest.mark.parametrize("head_size", HEAD_SIZES)
@pytest.mark.parametrize("attn_backend", LIST_ENC_DEC_SUPPORTED_BACKENDS)
Expand Down
11 changes: 8 additions & 3 deletions tests/kernels/utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,8 +12,8 @@
from torch._prims_common import TensorLikeType

from vllm.attention import AttentionBackend, AttentionMetadata, AttentionType
from vllm.utils import (STR_BACKEND_ENV_VAR, STR_XFORMERS_ATTN_VAL,
make_tensor_with_pad)
from vllm.utils import (STR_BACKEND_ENV_VAR, STR_ROCM_FLASH_ATTN_VAL,
STR_XFORMERS_ATTN_VAL, make_tensor_with_pad)

# For now, disable "test_aot_dispatch_dynamic" since there are some
# bugs related to this test in PyTorch 2.4.
Expand Down Expand Up @@ -524,8 +524,13 @@ def make_backend(backend_name: str) -> AttentionBackend:
if backend_name == STR_XFORMERS_ATTN_VAL:
# NOTE: xFormers backend cannot be imported for CPU and AMD GPUs.
from vllm.attention.backends.xformers import XFormersBackend

return XFormersBackend()

if backend_name == STR_ROCM_FLASH_ATTN_VAL:
from vllm.attention.backends.rocm_flash_attn import ( # noqa: F401
maleksan85 marked this conversation as resolved.
Show resolved Hide resolved
ROCmFlashAttentionBackend)
return ROCmFlashAttentionBackend

raise AssertionError(
f"Unrecognized backend_name {backend_name} for unit test")

Expand Down
Loading
Loading