Skip to content

Training Models using Built-in Keras Function #118

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 4 commits into from
Jan 23, 2021
Merged

Training Models using Built-in Keras Function #118

merged 4 commits into from
Jan 23, 2021

Conversation

@nglehuy
Copy link
Collaborator

@nglehuy nglehuy commented Jan 18, 2021 •
edited
Loading

  • Implement Slice Dataset that yield data in keras format (x, y)
  • Implement TFRecord Dataset that yield data in keras format (x, y)
  • Add new RnntLoss subclass of tf.keras.losses.Loss to use in compile function of tf.keras.Model
  • Add Transducer with custom train_step and test_step that uses RnntLoss
  • Add new CtcLoss subclass of tf.keras.losses.Loss to use in compile function of tf.keras.Model
  • Add CtcModel with custom train_step and test_step that uses CtcLoss

@nglehuy nglehuy self-assigned this Jan 18, 2021
@dathudeptrai dathudeptrai self-requested a review January 22, 2021 04:30
@dathudeptrai dathudeptrai added the enhancement New feature or request label Jan 22, 2021
dathudeptrai
dathudeptrai reviewed Jan 22, 2021
View reviewed changes
tensorflow_asr/models/keras/transducer.py Outdated

def test_step(self, batch):
x, y_true = batch
y_pred = self(x, training=True)
Copy link

@dathudeptrai dathudeptrai Jan 22, 2021

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

shoud be y_pred = self(x, training=False) ?
`

Copy link
Collaborator Author

@nglehuy nglehuy Jan 23, 2021

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

fixed

dathudeptrai
dathudeptrai reviewed Jan 22, 2021
View reviewed changes
tensorflow_asr/models/keras/transducer.py
with tf.GradientTape() as tape:
y_pred = self(x, training=True)
loss = self.loss(y, y_pred)
loss = self.loss(y_true, y_pred)
Copy link

@dathudeptrai dathudeptrai Jan 22, 2021

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pls adding Mixed precision support here

Copy link
Collaborator Author

@nglehuy nglehuy Jan 23, 2021

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

loss scaling added, in fact, it doesn't need loss scaling because the loss is already too big at the beginning but I added anyway

@nglehuy nglehuy changed the title Training and Testing Models using Built-in Keras Function Training Models using Built-in Keras Function Jan 23, 2021
@nglehuy nglehuy merged commit 3a4d4ee into main Jan 23, 2021
This was referenced Jan 23, 2021
Closed
Closed
@nglehuy nglehuy deleted the dev/keras branch February 15, 2021 17:30
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@dathudeptrai dathudeptrai dathudeptrai left review comments

Assignees

@nglehuy nglehuy

Labels

enhancement New feature or request

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@nglehuy @dathudeptrai