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

Allow installation/use without flash attention dependency #3

Merged
merged 4 commits into from
Dec 3, 2024
Merged

Allow installation/use without flash attention dependency #3

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
9f4cb96
Make flash-attn optional
ANaka Dec 2, 2024
677a715
dont import from `faesm.rotary` if fa not installed
ANaka Dec 2, 2024
4231f4c
only import from `faesm.utils` if not using fa
ANaka Dec 2, 2024
9652dfc
update readme
ANaka Dec 2, 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
20 changes: 14 additions & 6 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,13 +16,15 @@

# Table of Contents

- [FAESM: A Drop-in Efficient Pytorch Implementation of ESM2](#faesm-a-drop-in-efficient-pytorch-implementation-of-esm)
- [Installation](#installation)
- [Usage](#usage)
- [FAESM: A Drop-in Efficient Pytorch Implementation of ESM2](#faesm-a-drop-in-efficient-pytorch-implementation-of-esm2)
- [Table of Contents](#table-of-contents)
- [Installation](#installation)
- [Usage](#usage)
- [Training \[WIP\]](#training-wip)
- [Benchmarking](#benchmarking)
- [Appreciation](#appreciation)
- [Citation](#citation)
- [Benchmarking](#benchmarking)
- [TODOs](#todos)
- [Appreciation](#appreciation)
- [Citation](#citation)

# Installation

Expand All @@ -39,6 +41,12 @@ Having trouble installing flash attention but still want to use it? A workaround
3. Install FAESM from github:

```bash
# if you want to use flash attention
pip install faesm[flash_attn]
```

```bash
# if you want to forego flash attention and just use SDPA
pip install faesm
```

Expand Down
4 changes: 2 additions & 2 deletions faesm/esm.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,6 +5,8 @@
flash_attn_installed = True
try:
from flash_attn import flash_attn_varlen_qkvpacked_func
from faesm.rotary import RotaryEmbedding as FAEsmRotaryEmbedding
from faesm.utils import unpad
except ImportError:
flash_attn_installed = False
print(
Expand Down Expand Up @@ -42,8 +44,6 @@
EsmSelfOutput,
)

from faesm.rotary import RotaryEmbedding as FAEsmRotaryEmbedding
from faesm.utils import unpad

logger = logging.getLogger(__name__)

Expand Down
6 changes: 5 additions & 1 deletion setup.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,10 +7,13 @@
"setuptools",
"torch",
"einops",
"flash_attn",
"transformers",
]

extras_require = {
'flash_attn': ['flash_attn'],
}


setup(
name="faesm",
Expand All @@ -23,5 +26,6 @@
include_package_data=True,
zip_safe=True,
install_requires=requirements,
extras_require=extras_require,
test_suite="tests",
)
Loading