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

Jump to bottom

Enhance GPTQ act_order UT #1804

Merged
merged 5 commits into from
May 21, 2024
Merged

Enhance GPTQ act_order UT #1804

Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
4ea57b4
add act_order ut
Kaihui-intel May 20, 2024
ce4fe59
add act_order ut
Kaihui-intel May 20, 2024
6bf3c1e
minor fix
Kaihui-intel May 20, 2024
60516a7
minor fix
Kaihui-intel May 20, 2024
de509e1
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] May 20, 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
24 changes: 24 additions & 0 deletions test/3x/torch/quantization/weight_only/test_gptq.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -142,6 +142,30 @@ def test_mse_search(self):
atol_false > atol_true
), "use_mse_search=True doesn't help accuracy, maybe is reasonable, please double check."

def test_act_order(self):
# act_order=False
model = copy.deepcopy(self.tiny_gptj)
quant_config = GPTQConfig(
act_order=False,
)
model = prepare(model, quant_config)
run_fn(model)
model = convert(model)
out = model(self.example_inputs)[0]
atol_false = (out - self.label).amax()
# act_order=True
model = copy.deepcopy(self.tiny_gptj)
quant_config = GPTQConfig(
act_order=True,
)
model = prepare(model, quant_config)
run_fn(model)
model = convert(model)
out = model(self.example_inputs)[0]
atol_true = (out - self.label).amax()
# compare atol, this case is an ideal case.
assert atol_false > atol_true, "act_order=True doesn't help accuracy, maybe is reasonable, please double check."

# def test_layer_wise(self):
# model = copy.deepcopy(self.tiny_gptj)
# quant_config = GPTQConfig(
Expand Down
Loading