PocketFlow
PocketFlow copied to clipboard
Tencent
An Automatic Model Compression (AutoMC) framework for developing smaller and faster AI applications.
After having a glance on channel pruning gpu version, i found that gpu version may not strictly implement lasso regression. Neither Coordinate descent method nor LARS optimization algorithm are used....
Hello! I use `ChannelPrunedGpuLearner` to compress a regression model which predicts a monocular depth image. The loss become `nan` and I want to know is it normal or is it...
``` INFO:tensorflow:learner model name: resnet_v1_18, run_mode: TRAIN, num_classes: 5, weight_decay: 0, data_format: NHWC, image_size: 224 Quantized: model/resnet_v1_18/block1/unit_1/residual_unit_v1/conv1/Conv2D/quantize Quantized: model/resnet_v1_18/block1/unit_1/residual_unit_v1/conv2/Conv2D/quantize Quantized: model/resnet_v1_18/block1/unit_2/residual_unit_v1/conv1/Conv2D/quantize Quantized: model/resnet_v1_18/block1/unit_2/residual_unit_v1/conv2/Conv2D/quantize Quantized: model/resnet_v1_18/block2/unit_1/residual_unit_v1/shortcut/Conv2D/quantize Quantized: model/resnet_v1_18/block2/unit_1/residual_unit_v1/conv1/Conv2D/quantize Quantized: model/resnet_v1_18/block2/unit_1/residual_unit_v1/conv2/Conv2D/quantize Quantized:...
Hanging during the pruning. extra_args: --learner channel --cp_preserve_ratio 0.5 --cp_prune_option auto --ddpg_noise_std_init=0.5 --ddpg_actor_width 300 --ddpg_critic_width 300 --cp_nb_rlouts_min 100 --cp_nb_rlouts 400 --cp_finetune=False --cp_retrain=True --enbl_dst=True --batch_size_eval 64 --data_dir_local /data/NewTFRecord/ [mobilev1_img_cp_auto_0.5_noise0.5_width300_roll100-400_save2_4GPU.log](https://github.com/Tencent/PocketFlow/files/2708333/mobilev1_img_cp_auto_0.5_noise0.5_width300_roll100-400_save2_4GPU.log)
**Describe the bug** A clear and concise description of what the bug is. I ran two distinct experiments, one on uniform quantization, and one on channel pruning with the same...
The following code snippet makes me puzzled. I know it is used to select `c_new` channels from c_in. WHY the lbound, rbound for channels and left, right for `alpha` are...
### When I used horovod docker and trainning a network in the docker container, I met the error below: ### horovod docker tag: 0.14.1-tf1.10.0-torch0.4.0-py3.5 ### GPU: Titan Xp ### GPU_NUM:...
I am trying to compress a lot classic models using pocketflow with different compressed ratios(0.4,0.6,0.8). But the GPU resources are limited and time is also limited, the process time last...
error info: ``` Traceback (most recent call last): File "/home//git/pocket_flow/compress_model_run.py", line 100, in compress_model learner = create_learner(sm_writer, model_helper) File "/home/git/pocket_flow/learners/learner_utils.py", line 49, in create_learner learner = ChannelPrunedLearner(sm_writer, model_helper) File "/home/git/pocket_flow/learners/channel_pruning/learner.py",...
**Describe the bug** A clear and concise description of what the bug is. **To Reproduce** Steps to reproduce the behavior: 1. Go to '...' 2. Click on '....' 3. Scroll...
