Skip to content

fix: migrate FP8 support from float8_experimental to torchao #618

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

Open
aryasaatvik wants to merge 3 commits into
base: main
Choose a base branch
from
+5 −5,153
Open

fix: migrate FP8 support from float8_experimental to torchao #618

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
2512b58
refactor quantization interface to use torchao for FP8 support and up…
aryasaatvik Jul 10, 2025
c2a202f
Update pyproject.toml
michaelfeil Aug 28, 2025
9d13111
Delete libs/infinity_emb/poetry.lock
michaelfeil Aug 28, 2025
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
15 changes: 5 additions & 10 deletions libs/infinity_emb/infinity_emb/transformer/quantization/interface.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,20 +51,15 @@ def quant_interface(model: Any, dtype: Union[Dtype] = Dtype.int8, device: Device
model.to(device_orig)
elif device == Device.cuda and dtype in [Dtype.fp8, torch.float8_e5m2]:
try:
from float8_experimental.float8_dynamic_linear import ( # type: ignore
Float8DynamicLinear,
)
from float8_experimental.float8_linear_utils import ( # type: ignore
swap_linear_with_float8_linear,
)
from torchao.quantization import quantize_, float8_dynamic_activation_float8_weight
except ImportError:
raise ImportError(
"float8_experimental is not installed."
"https://github.com/pytorch-labs/float8_experimental "
"with commit `88e9e507c56e59c5f17edf513ecbf621b46fc67d`"
"FP8 quantization requires torchao. "
"Install with: pip install torchao\n"
)
logger.info("using dtype=fp8")
swap_linear_with_float8_linear(model, Float8DynamicLinear)
# Use quantize_ with the float8 configuration
quantize_(model, float8_dynamic_activation_float8_weight())
else:
raise ValueError(f"Quantization is not supported on {device} with dtype {dtype}.")
return model
Expand Down
21 changes: 20 additions & 1 deletion libs/infinity_emb/poetry.lock
View file
Open in desktop

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

4 changes: 4 additions & 0 deletions libs/infinity_emb/pyproject.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -100,6 +100,10 @@ black = "^24.10.0"
types-chardet = "^5.0.4.6"
mypy-protobuf = "^3.0.0"


[tool.poetry.group.dev.dependencies]
torchao = "^0.11.0"

@greptile-apps greptile-apps Bot Jul 10, 2025

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

style: Consider moving torchao to main dependencies section with optional=true if it's needed for core FP8 functionality


[tool.poetry.extras]
ct2=["ctranslate2","sentence-transformers","torch","transformers"]
optimum=["optimum"]
Expand Down