Compare commits

..

5 Commits

Author SHA1 Message Date
sayakpaul
67f4691cab resolve conflicts. 2026-02-19 18:22:49 +05:30
sayakpaul
e10fe61303 fix version and force updated kernels. 2026-02-19 18:00:01 +05:30
Sayak Paul
348350cf24 Merge branch 'main' into update-kernel-hub-repos 2026-02-19 17:53:46 +05:30
Sayak Paul
af35e3806c Merge branch 'main' into update-kernel-hub-repos 2026-02-19 09:35:15 +05:30
sayakpaul
d6bc647932 change to updated repo and version. 2026-02-18 23:46:06 +05:30
6 changed files with 37 additions and 10 deletions

View File

@@ -117,7 +117,7 @@ jobs:
- name: Install dependencies
run: |
uv pip install -e ".[quality]"
uv pip install -e ".[quality,test]"
#uv pip uninstall transformers huggingface_hub && uv pip install --prerelease allow -U transformers@git+https://github.com/huggingface/transformers.git
uv pip uninstall transformers huggingface_hub && uv pip install transformers==4.57.1
uv pip uninstall accelerate && uv pip install -U accelerate@git+https://github.com/huggingface/accelerate.git --no-deps

View File

@@ -114,7 +114,7 @@ jobs:
- name: Install dependencies
run: |
uv pip install -e ".[quality]"
uv pip install -e ".[quality,test]"
#uv pip uninstall transformers huggingface_hub && uv pip install --prerelease allow -U transformers@git+https://github.com/huggingface/transformers.git
uv pip uninstall transformers huggingface_hub && uv pip install transformers==4.57.1
uv pip uninstall accelerate && uv pip install -U accelerate@git+https://github.com/huggingface/accelerate.git --no-deps
@@ -191,7 +191,7 @@ jobs:
- name: Install dependencies
run: |
uv pip install -e ".[quality]"
uv pip install -e ".[quality,test]"
- name: Environment
run: |
@@ -242,7 +242,7 @@ jobs:
- name: Install dependencies
run: |
uv pip install -e ".[quality]"
uv pip install -e ".[quality,test]"
# TODO (sayakpaul, DN6): revisit `--no-deps`
uv pip install -U peft@git+https://github.com/huggingface/peft.git --no-deps
uv pip install -U tokenizers

View File

@@ -41,7 +41,7 @@ jobs:
shell: arch -arch arm64 bash {0}
run: |
${CONDA_RUN} python -m pip install --upgrade pip uv
${CONDA_RUN} python -m uv pip install -e ".[quality]"
${CONDA_RUN} python -m uv pip install -e ".[quality,test]"
${CONDA_RUN} python -m uv pip install torch torchvision torchaudio
${CONDA_RUN} python -m uv pip install accelerate@git+https://github.com/huggingface/accelerate.git
${CONDA_RUN} python -m uv pip install transformers --upgrade

View File

@@ -38,6 +38,7 @@ from ..utils import (
is_flash_attn_available,
is_flash_attn_version,
is_kernels_available,
is_kernels_version,
is_sageattention_available,
is_sageattention_version,
is_torch_npu_available,
@@ -265,6 +266,7 @@ class _HubKernelConfig:
repo_id: str
function_attr: str
revision: str | None = None
version: int | None = None
kernel_fn: Callable | None = None
wrapped_forward_attr: str | None = None
wrapped_backward_attr: str | None = None
@@ -274,27 +276,31 @@ class _HubKernelConfig:
# Registry for hub-based attention kernels
_HUB_KERNELS_REGISTRY: dict["AttentionBackendName", _HubKernelConfig] = {
# TODO: temporary revision for now. Remove when merged upstream into `main`.
AttentionBackendName._FLASH_3_HUB: _HubKernelConfig(
repo_id="kernels-community/flash-attn3", function_attr="flash_attn_func", revision="fake-ops-return-probs"
repo_id="kernels-community/flash-attn3", function_attr="flash_attn_func", version=1
),
AttentionBackendName._FLASH_3_VARLEN_HUB: _HubKernelConfig(
repo_id="kernels-community/flash-attn3",
function_attr="flash_attn_varlen_func",
# revision="fake-ops-return-probs",
version=1,
),
AttentionBackendName.FLASH_HUB: _HubKernelConfig(
repo_id="kernels-community/flash-attn2",
function_attr="flash_attn_func",
version=1,
revision=None,
wrapped_forward_attr="flash_attn_interface._wrapped_flash_attn_forward",
wrapped_backward_attr="flash_attn_interface._wrapped_flash_attn_backward",
),
AttentionBackendName.FLASH_VARLEN_HUB: _HubKernelConfig(
repo_id="kernels-community/flash-attn2", function_attr="flash_attn_varlen_func", revision=None
repo_id="kernels-community/flash-attn2",
function_attr="flash_attn_varlen_func",
version=1,
),
AttentionBackendName.SAGE_HUB: _HubKernelConfig(
repo_id="kernels-community/sage_attention", function_attr="sageattn", revision=None
repo_id="kernels-community/sage-attention",
function_attr="sageattn",
version=1,
),
}
@@ -464,6 +470,10 @@ def _check_attention_backend_requirements(backend: AttentionBackendName) -> None
raise RuntimeError(
f"Backend '{backend.value}' is not usable because the `kernels` package isn't available. Please install it with `pip install kernels`."
)
if not is_kernels_version(">=", "0.12"):
raise RuntimeError(
f"Backend '{backend.value}' needs to be used with a `kernels` version of at least 0.12. Please update with `pip install -U kernels`."
)
elif backend == AttentionBackendName.AITER:
if not _CAN_USE_AITER_ATTN:

View File

@@ -86,6 +86,7 @@ from .import_utils import (
is_inflect_available,
is_invisible_watermark_available,
is_kernels_available,
is_kernels_version,
is_kornia_available,
is_librosa_available,
is_matplotlib_available,

View File

@@ -724,6 +724,22 @@ def is_transformers_version(operation: str, version: str):
return compare_versions(parse(_transformers_version), operation, version)
@cache
def is_kernels_version(operation: str, version: str):
"""
Compares the current Kernels version to a given reference with an operation.
Args:
operation (`str`):
A string representation of an operator, such as `">"` or `"<="`
version (`str`):
A version string
"""
if not _kernels_available:
return False
return compare_versions(parse(_kernels_version), operation, version)
@cache
def is_hf_hub_version(operation: str, version: str):
"""