search

tensorflow / models

Models and examples built with TensorFlow
Other
77.05k stars 45.77k forks source link

the model generated by the api is different from the original one #5659

Closed 1453042287 closed 5 years ago

1453042287 commented 5 years ago

System information

Describe the problem

i use the api and the ssdlite_mobilenet_v2 pre-trained model to fine-tune my own dataset, then i check the variables' name and size in the checkpoint, but i find the size of the MobilenetV2/Conv_1/weights is different, in the model provided by api is [1,1,320,1280] and my is [1,1,320,256], it's weird because i didn't change the code at all, is there some details i miss? any suggestions and hints would be appreciate!

netanel-s commented 5 years ago

See issue #5597

mcsuy commented 5 years ago

Omung l MG

On Nov 1, 2018 5:17 PM, "a copy" notifications@github.com wrote:

System information

  • What is the top-level directory of the model you are using: object_detection
  • Have I written custom code (as opposed to using a stock example script provided in TensorFlow): no
  • OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Linux Ubuntu 16.04
  • TensorFlow installed from (source or binary): pypi tensorflow-gpu
  • TensorFlow version (use command below): 1.11
  • Bazel version (if compiling from source): none
  • CUDA/cuDNN version: CUDA9.0 cuDNN7.3
  • GPU model and memory: TITAN V 11G
  • Exact command to reproduce: CUDA_VISIBLE_DEVICES=0 python object_detection/model_main.py --pipeline_config_path=project/ssdlite_mobilenet_v2_coco_2018_05_09/ssdlite.config --model_dir=project/ssdlite_result --num_train_steps=100000 --sample_1_of_n_eval_examples=1 --alsologtostderr

Describe the problem

i use the api and the ssdlite_mobilenet_v2 pre-trained model to fine-tune my own dataset, then i check the variables' name and size in the checkpoint, but i find the size of the MobilenetV2/Conv_1/weights is different, in the model provided by api is [1,1,320,1280] and my is [1,1,320,256], it's weird cause i didn't change the code at all, is there some details i miss? any suggestions and hints would be appreciate!

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/tensorflow/models/issues/5659, or mute the thread https://github.com/notifications/unsubscribe-auth/Af3QWnsYfLoceaaCuOj9ocAuyfgjfoxSks5uqrwygaJpZM4YGCtS .

k-w-w commented 5 years ago

Closing this because it appears to be a duplicate issue. Please refer to #5597