Skip to content

Fix smoothquant minmax observer init #1421

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
merged 6 commits into from
Nov 28, 2023
Merged

Fix smoothquant minmax observer init #1421

Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
02f7b0f
Fix smoothquant minmax observer
changwangss Nov 27, 2023
160d75a
Update pytorch.py
changwangss Nov 27, 2023
adf01c8
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] Nov 27, 2023
67dca1f
Update pytorch.py
changwangss Nov 27, 2023
86d9d06
Update pytorch.py
changwangss Nov 27, 2023
266f07f
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] Nov 27, 2023
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
30 changes: 24 additions & 6 deletions neural_compressor/adaptor/pytorch.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3114,14 +3114,23 @@ def _get_quantizable_ops_recursively(self, model, prefix, quantizable_ops):
smooth_quant_args = self.recipes.get("smooth_quant_args", {})
folding = smooth_quant_args.get("folding", False)
if not folding:
if self.sq_minmax_init or self.version.release >= Version("2.1.1").release:
from torch.ao.quantization.observer import MinMaxObserver
from torch.ao.quantization.observer import MinMaxObserver

if self.version.release >= Version("2.1.1").release:
static_qconfig = ipex.quantization.get_smooth_quant_qconfig_mapping(
alpha=0.5, act_observer=MinMaxObserver
)
else:
static_qconfig = ipex.quantization.get_smooth_quant_qconfig_mapping(alpha=0.5)
if self.sq_minmax_init:
static_qconfig = ipex.quantization.get_smooth_quant_qconfig_mapping(
alpha=0.5, act_observer=MinMaxObserver()
)
logger.warning(
"The int8 model accuracy will be close to 0 with MinMaxobserver, "
+ "the suggested IPEX version is higher or equal than 2.1.100."
)
else:
static_qconfig = ipex.quantization.get_smooth_quant_qconfig_mapping(alpha=0.5)
if self.example_inputs is None:
self.example_inputs = get_example_inputs(model, self.q_dataloader)
from neural_compressor.adaptor.torch_utils.util import move_input_device
Expand Down Expand Up @@ -3304,14 +3313,23 @@ def qdq_quantize(self, model, q_model, tune_cfg, dataloader, q_func):
# Check save_qconf_summary part is a workaround for IPEX bug.
# Sometimes the prepared model from get_op_capablitiy loss this attribute
if not hasattr(model._model, "save_qconf_summary") or not hasattr(model._model, "load_qconf_summary"):
if self.sq_minmax_init or self.version.release >= Version("2.1.1").release:
from torch.ao.quantization.observer import MinMaxObserver
from torch.ao.quantization.observer import MinMaxObserver

if self.version.release >= Version("2.1.1").release:
static_qconfig = ipex.quantization.get_smooth_quant_qconfig_mapping(
alpha=0.5, act_observer=MinMaxObserver
)
else:
static_qconfig = ipex.quantization.get_smooth_quant_qconfig_mapping(alpha=0.5)
if self.sq_minmax_init:
static_qconfig = ipex.quantization.get_smooth_quant_qconfig_mapping(
alpha=0.5, act_observer=MinMaxObserver()
)
logger.warning(
"The int8 model accuracy will be close to 0 with MinMaxobserver, "
+ "the suggested IPEX version is higher or equal than 2.1.100+cpu."
)
else:
static_qconfig = ipex.quantization.get_smooth_quant_qconfig_mapping(alpha=0.5)
if isinstance(self.example_inputs, dict):
model._model = ipex.quantization.prepare(
model._model, static_qconfig, example_kwarg_inputs=self.example_inputs, inplace=inplace
Expand Down