How to use these flags

Archives DeepSpeech
#1

Hello, I’m training a germen model with Mozilla DeepSpeech. The Training, Validation and Test go well. Following is my command:
python DeepSpeech.py
–train_files /workspace/de/clips/train.csv
–test_files /workspace/de/clips/test.csv
–dev_files /workspace/de/clips/dev.csv
–export_dir /workspace/DeepSpeech/data/model
–train_batch_size=100
–dev_batch_size=100
–test_batch_size=100
–epochs=30
–n_hidden=512
–learning_rate=0.0001
–dropout_rate=0.2
–export_file_name output_615
–export_author_id sun
–export_model_name 615
–export_model_version 1
–summary_dir /workspace/DeepSpeech/data/summary

Problem 1. After testing the summaries were still saved in /root/.local/share/deepspeech/summaries and I can’t touch it.
Problem 2. After testing the model was exported at /workspace/DeepSpeech/data/model and metadata was saved to /workspace/DeepSpeech/data/model/sun_615_1. But the folder model in this path is empty.

Did I write the command right? Did I use the flag right?

How can I get the model, metadata and the summaries after training?

Thx a lot Guys

#2

Without comprehensive infos, nobody can help:

  • logs,
  • environments used,
  • etc.

You also share a command line as multiline text: please share as code as discourse allows, and make sure what you share is accurate.

e.g.,

python DeepSpeech.py \
 --train_files /workspace/de/clips/train.csv \
[...]
 --summary_dir /workspace/DeepSpeech/data/summary
#3

sry for that

env:
Ubuntu 20.04 RTX3090 run in Nvidia container which is built by Nvidia image 20.11-tf1-py3.

the latest command line:

python3 -u DeepSpeech.py
–train_files /workspace/de/clips/train.csv
–test_files /workspace/de/clips/test.csv
–dev_files /workspace/de/clips/dev.csv
–export_dir /workspace/DeepSpeech/data/model
–train_batch_size=100
–dev_batch_size=100
–test_batch_size=100
–epochs=1
–n_hidden=512
–learning_rate=0.0001
–dropout_rate=0.2
–export_file_name output_615
–export_author_id sun
–export_model_name 615
–export_model_version 1
–summary_dir /workspace/DeepSpeech/data/model

the latest logs:

I Could not find best validating checkpoint.
I Could not find most recent checkpoint.
I Initializing all variables.
I STARTING Optimization
Epoch 0 | Training | Elapsed Time: 0:03:53 | Steps: 2371 | Loss: 118.941052
Epoch 0 | Validation | Elapsed Time: 0:00:11 | Steps: 151 | Loss: 94.769755 | Dataset: /workspace/de/clips/dev.csv
I Saved new best validating model with loss 94.769755 to: /root/.local/share/deepspeech/checkpoints/best_dev-2371

I FINISHED optimization in 0:04:04.418811
I Loading best validating checkpoint from /root/.local/share/deepspeech/checkpoints/best_dev-2371
I Loading variable from checkpoint: cudnn_lstm/rnn/multi_rnn_cell/cell_0/cudnn_compatible_lstm_cell/bias
I Loading variable from checkpoint: cudnn_lstm/rnn/multi_rnn_cell/cell_0/cudnn_compatible_lstm_cell/kernel
I Loading variable from checkpoint: global_step
I Loading variable from checkpoint: layer_1/bias
I Loading variable from checkpoint: layer_1/weights
I Loading variable from checkpoint: layer_2/bias
I Loading variable from checkpoint: layer_2/weights
I Loading variable from checkpoint: layer_3/bias
I Loading variable from checkpoint: layer_3/weights
I Loading variable from checkpoint: layer_5/bias
I Loading variable from checkpoint: layer_5/weights
I Loading variable from checkpoint: layer_6/bias
I Loading variable from checkpoint: layer_6/weights
Testing model on /workspace/de/clips/test.csv
Test epoch | Steps: 151 | Elapsed Time: 1:21:49
Test on /workspace/de/clips/test.csv - WER: 0.949567, CER: 0.449052, loss: 94.426300
…
I Models exported at /workspace/DeepSpeech/data/model
I Model metadata file saved to /workspace/DeepSpeech/data/model/sun_615_1.md. Before submitting the exported model for publishing make sure all information in the metadata file is correct, and complete the URL fields.

the logs shows no error or warning.

Problems:

  1. I can’t get the pb and md file in the export_dir.
  2. The summaries are not saved in the new path as the flag says.
  3. In the default path there are no new summary saved. Each time I use tensorboard, I can just get the same graphic which is generated by the oldest summaries.

which I want:

  1. new summaries after Training in the path which is provided in flag. At least in the default path.
  2. new model after Training in the path which is provided in flag.
#4

You don’t share informations as I requested, I can’t help.

#5

I share the env and logs. There is no Error in the logs.

Could you please tell me the specific information you need?

Thx for answering

#6

@Xuetong_Sun This, it’s really important.