search

dong-x16 / PortraitNet

Code for the paper "PortraitNet: Real-time portrait segmentation network for mobile device" @ CAD&Graphics2019
286 stars 61 forks source link

Model training #12

Open rose-jinyang opened 4 years ago

rose-jinyang commented 4 years ago

Hello How are you? Thanks for contributing this project. I am training a new model with config model_mobilenetv2_with_two_auxiliary_losses and AISegment dataset for 8 days. But it is strange that the best model is not updated for 7 days. What do u think the reason?

shivSD commented 3 years ago

@rose-jinyang did you figure out the issue. Even we are having the same issue, Loss is not decreasing after awhile.

dpcq commented 3 years ago

i also have this problem ,change the lr and train many epoch (1000 at least)

DAVID-Hown commented 3 years ago

Hello, could you please tell me that I made a mistake when retraining the model? Do you have any good solutions? load model init finish... ===========> training <=========== Epoch: [0][0/23] Lr-deconv: [0.0] Lr-other: [0.001] Loss 13.0424 (13.0424) Traceback (most recent call last): File "train.py", line 809, in main(args) File "train.py", line 774, in main train(dataLoader_train, netmodel, optimizer, epoch, logger_train, exp_args) File "train.py", line 519, in train logger.scalar_summary(tag, value, step=i) File "/data/data_hao/PFLD-pytorch/20210817/PortraitNet/util/logger.py", line 26, in scalar_summary tf.summary.scalar(tag, value, step=step) File "/root/.local/lib/python3.6/site-packages/tensorboard/plugins/scalar/summary_v2.py", line 49, in scalar with tf.summary.summary_scope( AttributeError: module 'tensorboard.summary._tf.summary' has no attribute 'summary_scope'