quantization - fix for ConvTranspose2d and BatchNorm merge. Do not merge weights...
authorManu Mathew <a0393608@ti.com>
Tue, 12 May 2020 16:03:18 +0000 (21:33 +0530)
committerManu Mathew <a0393608@ti.com>
Wed, 13 May 2020 03:00:58 +0000 (08:30 +0530)
commitd12caf8a972cc2df8153015274dd78c56d5033c1
tree0458a5757d9d663e445f34a73a767f7153f00b54
parentcae916dd80067ab6a37bd122917894032248f98e
quantization - fix for ConvTranspose2d and BatchNorm merge. Do not merge weights upfront in QuantCalibrateModule, but will be done in PAct2.
modules/pytorch_jacinto_ai/xnn/quantize/quant_calib_module.py
modules/pytorch_jacinto_ai/xnn/quantize/quant_graph_module.py
modules/pytorch_jacinto_ai/xnn/quantize/quant_test_module.py
modules/pytorch_jacinto_ai/xnn/quantize/quant_train_module.py