You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: TensorFlow/Classification/ConvNets/resnet50v1.5/README.md
+1-1Lines changed: 1 addition & 1 deletion
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -383,7 +383,7 @@ a <a href="https://www.tensorflow.org/versions/r1.14/api_docs/python/tf/contrib/
383
383
for training QAT networks is to train a model until convergence and then finetune with the quantization layers. It is recommended that QAT is performed on a single GPU.
384
384
385
385
* For 1 GPU
386
-
* Command: `sh resnet50v1.5/training/QAT/GPU1_RN50_QAT.sh <path to pre-trained ckpt dir> <path to dataset directory> <result_directory>`
386
+
* Command: `sh resnet50v1.5/training/GPU1_RN50_QAT.sh <path to pre-trained ckpt dir> <path to dataset directory> <result_directory>`
387
387
388
388
It is recommended to finetune a model with quantization nodes rather than train a QAT model from scratch. The latter can also be performed by setting `quant_delay` parameter.
389
389
`quant_delay` is the number of steps after which quantization nodes are added for QAT. If we are fine-tuning, `quant_delay` is set to 0.
0 commit comments