frozenleaves/pytorch
1
0
Fork 0
mirror of https://github.com/pytorch/pytorch.git synced 2025-10-21 13:44:15 +08:00
Code Issues Packages Projects Releases Wiki Activity

Graph-mode quantization for convolution from traced model (#30245)

Browse Source 
Summary:
In the PR, we enhance the graph-mode quantization for aten::_convolution, which could be generated from tracing path.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/30245

Differential Revision: D18671597

Pulled By: lly-zero-one

fbshipit-source-id: 78a2470fbb0fe0def55d63c6bda7cbb5c89f7848
This commit is contained in:
Lingyi Liu
2019-11-23 00:47:09 -08:00
committed by Facebook Github Bot
parent 2a7a39c1af
commit 59ca9b7430
2 changed files with 62 additions and 16 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
test/test_quantization.py
View File

@ -842,15 +842,17 @@ class GraphModePostTrainingQuantTest(QuantizationTestCase):
qconfig_dict = {
'': default_qconfig
}
model_script = quantize_script(
torch.jit.script(conv_model_to_script),
qconfig_dict,
default_eval_fn,
[self.img_data],
inplace=False)
model_traced = torch.jit.trace(conv_model_to_script, self.img_data[0][0])
model_script = torch.jit.script(conv_model_to_script)
result_eager = model_eager(self.img_data[0][0])
result_script = model_script(self.img_data[0][0])
self.assertEqual(result_eager, result_script)
for model_under_test in [model_traced, model_script]:
model_quantized = quantize_script(
model_under_test,
qconfig_dict,
default_eval_fn,
[self.img_data],
inplace=False)
self.assertEqual(model_quantized(self.img_data[0][0]), result_eager)
@unittest.skip("This doesn't work right now, re-enable after fold_convbn is fixed")
def test_conv_bn(self):