Skip to content

Add asym for XPU backend. #1316

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

Draft
luoyu-intel wants to merge 1 commit into
base: main
Choose a base branch
from
Draft

Add asym for XPU backend. #1316

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
21 changes: 3 additions & 18 deletions auto_round/inference/backend.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -456,7 +456,7 @@ def fp8_static_scheme_checker(

BackendInfos["auto_round_kernel_xpu"] = BackendInfo(
device=["xpu"],
sym=[True],
sym=[True, False],
packing_format=GPTQ_FORMAT_NO_ZP,
bits=[4, 8],
group_size=None,
Expand Down Expand Up @@ -486,7 +486,7 @@ def fp8_static_scheme_checker(

BackendInfos["auto_round_kernel_zp_xpu"] = BackendInfo(
device=["xpu"],
sym=[True],
sym=[True, False],
packing_format=GPTQ_FORMAT,
bits=[4, 8],
group_size=None,
Expand All @@ -500,7 +500,7 @@ def fp8_static_scheme_checker(
)

BackendInfos["auto_round_kernel_awq"] = BackendInfo(
device=["cpu"],
device=["cpu", "xpu"],
sym=[True, False],
packing_format=AWQ_FORMAT,
bits=[4],
Expand All @@ -514,21 +514,6 @@ def fp8_static_scheme_checker(
requirements=["torch>=2.8.0", "auto_round_kernel"],
)

BackendInfos["auto_round_kernel_awq_xpu"] = BackendInfo(
device=["xpu"],
sym=[True],
packing_format=AWQ_FORMAT,
bits=[4],
group_size=None,
priority=6,
checkers=[ark_feature_checker],
alias=["ark"],
compute_dtype=["float32", "float16"],
data_type=["int"],
act_bits=WOQ_DEFAULT_ACT_BITS,
requirements=["torch>=2.8.0", "auto_round_kernel"],
)

BackendInfos["ipex_gptq_cpu"] = BackendInfo(
device=["cpu"],
sym=[True, False],
Expand Down
3 changes: 1 addition & 2 deletions auto_round_extension/ark/qlinear.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -116,8 +116,7 @@ def post_init(self):
raise NotImplementedError(
f"Device type {self.qweight.device.type} is not supported. Only CPU and XPU devices are supported."
)
if self.qweight.device.type != "cpu" and self.asym:
raise NotImplementedError("Asymmetric quantization is only supported on CPU devices")

if "awq" in self.QUANT_TYPE:
intweight, zeros = unpack_awq(
self.qweight, self.qzeros, self.bits
Expand Down
6 changes: 2 additions & 4 deletions test/test_ark/test_model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -29,8 +29,6 @@ def main_op(self, format, bits, group_size, sym, dtype, device, fast_cfg=True, t
limit = 1000
if not torch.xpu.is_available():
pytest.skip("No XPU device")
if sym is False:
pytest.skip("No asym support for XPU")
model = AutoModelForCausalLM.from_pretrained(self.model_name, dtype="auto")
tokenizer = AutoTokenizer.from_pretrained(self.model_name)
if fast_cfg:
Expand All @@ -56,9 +54,9 @@ def main_op(self, format, bits, group_size, sym, dtype, device, fast_cfg=True, t
shutil.rmtree(self.save_folder, ignore_errors=True)

@pytest.mark.parametrize("format", ["auto_round", "auto_round:gptqmodel"])
@pytest.mark.parametrize("bits, group_size, sym", [(4, 128, True), (8, 128, True)])
@pytest.mark.parametrize("bits, group_size, sym", [(4, 128, False), (8, 128, True)])
@pytest.mark.parametrize("dtype", [torch.bfloat16])
@pytest.mark.parametrize("device", ["cpu", "xpu"])
@pytest.mark.parametrize("device", ["xpu", "cpu"])
def test_formats(self, format, bits, group_size, sym, dtype, device):
self.main_op(format, bits, group_size, sym, dtype, device)

Expand Down