Skip to content

[Bugfix] Remove constant pruning from 2of4 examples #1267

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

Merged
rahul-tuli merged 4 commits into from
Mar 19, 2025
Merged

[Bugfix] Remove constant pruning from 2of4 examples #1267

Show file tree
Hide file tree
Changes from 3 commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
4b5fe76
update example, do not start on initialize
kylesayrs Mar 18, 2025
60a536a
update tests
kylesayrs Mar 18, 2025
acca004
start on initialize
kylesayrs Mar 18, 2025
93a12f1
remove CMP from all 2of4 examples
kylesayrs Mar 19, 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
35 changes: 14 additions & 21 deletions examples/sparse_2of4_quantization_fp8/llama3_8b_2of4.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,7 +5,6 @@

from llmcompressor import oneshot
from llmcompressor.modifiers.obcq import SparseGPTModifier
from llmcompressor.modifiers.pruning import ConstantPruningModifier
from llmcompressor.modifiers.quantization import QuantizationModifier

# Configuration
Expand Down Expand Up @@ -52,29 +51,23 @@ def get_recipe(fp8_enabled):
save_dir = MODEL_ID.split("/")[1] + "2of4-sparse"

if fp8_enabled:
base_recipe.extend(
[
QuantizationModifier(
targets=["Linear"],
ignore=["lm_head"],
scheme="FP8_DYNAMIC",
),
ConstantPruningModifier(
targets=[
r"re:.*q_proj.weight",
r"re:.*k_proj.weight",
r"re:.*v_proj.weight",
r"re:.*o_proj.weight",
r"re:.*gate_proj.weight",
r"re:.*up_proj.weight",
r"re:.*down_proj.weight",
],
start=0,
),
]
base_recipe.append(
QuantizationModifier(
targets=["Linear"],
ignore=["lm_head"],
scheme="FP8_DYNAMIC",
)
)
save_dir = MODEL_ID.split("/")[1] + "2of4-W8A8-FP8-Dynamic-Per-Token"

# check that asymmetric quantization is not being used
q_scheme = base_recipe[1].scheme
if not isinstance(q_scheme, str) and not q_scheme["weights"].symmetric:
raise ValueError(
"Asymmetric quantization with 2of4 sparsity is not supported by vLLM. "
"Please use symmetric quantization"
)

return base_recipe, save_dir


Expand Down
1 change: 0 additions & 1 deletion tests/llmcompressor/pytorch/modifiers/pruning/constant/test_pytorch.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -74,7 +74,6 @@ def test_constant_pruning_modifier_e2e(model, optimizer):
state.update(
model=model,
optimizer=optimizer(model.parameters(), lr=0.1),
start=0,
)
modifier = ConstantPruningModifier(
targets="__ALL_PRUNABLE__",
Expand Down