Skip to content

Reduce FLUX int8 test peak memory with sequential offload #13776

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
jiqing-feng wants to merge 4 commits into
base: main
Choose a base branch
from
+6 −2
Open

Reduce FLUX int8 test peak memory with sequential offload #13776

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
6d36ba9
fix oom
jiqing-feng May 20, 2026
53e0a7c
revert
jiqing-feng May 20, 2026
862eb67
adjust tol
jiqing-feng May 21, 2026
1ee339d
Merge branch 'main' into test_xpu
jiqing-feng May 21, 2026
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
8 changes: 6 additions & 2 deletions tests/quantization/bnb/test_mixed_int8.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -663,7 +663,11 @@ def setUp(self) -> None:
transformer=transformer_8bit,
torch_dtype=torch.float16,
)
self.pipeline_8bit.enable_model_cpu_offload()
# Use sequential CPU offload to keep peak GPU memory minimal (one layer at a time).
# enable_model_cpu_offload moves an entire sub-model to GPU at once, which OOMs on
# <=24 GB cards for FLUX.1-dev even with int8 quantization.
# This requires the bitsandbytes fix that preserves Int8Params.SCB across .to() calls.
self.pipeline_8bit.enable_sequential_cpu_offload()

def tearDown(self):
del self.pipeline_8bit
Expand Down Expand Up @@ -709,7 +713,7 @@ def test_lora_loading(self):
expected_slice = np.array([0.3916, 0.3916, 0.3887, 0.4243, 0.4155, 0.4233, 0.4570, 0.4531, 0.4248])

max_diff = numpy_cosine_similarity_distance(expected_slice, out_slice)
self.assertTrue(max_diff < 1e-3)
self.assertTrue(max_diff < 2e-3)


@require_transformers_version_greater("4.44.0")
Expand Down
Loading