mlverse / luz

Higher Level API for torch

https://mlverse.github.io/luz/

mlverse/luz Issues

Apparent issue with macOS Sonoma 14.4.1
Updated 2 months ago1
How to use luz with custom optimizer?
Updated 3 months ago
cpp_tensor_from_buffer error
Updated 3 months ago
Error in setup() with nn_sequential()
Closed 3 years ago3
rerun the same 'setup' codes a second time results in error
Closed 4 months ago5
CRAN and GitHub versions failed to install WIN11
Updated 4 months ago1
luz_callback_auto_resume doesn't resume ctx$epoch
Closed 10 months ago4
Weights
Updated a year ago1
R crashes when attempting to save a model or make checkpoints before moving tensors to CPU
Closed a year ago1
`predict`: How to get embedding layer?
Updated a year ago3
luz metrics do not work with optim_lbfgs.
Closed a year ago2
How to just continue training
Closed a year ago2
Release luz 0.4.0
Closed a year ago
Rstudio crashes when fitting using luz
Closed 2 years ago8
Consider only computing metrics for the validation set.
Closed 2 years ago
Debug errors in callbacks/metrics
Closed 2 years ago
MixUp callback should be configurable
Closed 2 years ago
Avoid computing metrics on train data
Closed 2 years ago
MNIST example error: missing value where TRUE/FALSE needed
Closed 2 years ago1
progressr callback
Closed 2 years ago
Examples dont run on their own
Closed 2 years ago1
lr_finder changes minibatch size
Closed 2 years ago1
Update luz examples
Closed 2 years ago
Stop/resume model fitting
Closed 2 years ago
predict.luz_module_fitted callbacks help
Closed 2 years ago
Release luz 0.3.1
Closed 2 years ago
Release luz 0.3.0
Closed 2 years ago
Learning rates tested by lr_finder() run into saturation
Closed 2 years ago1
Leak the `fit` function environment
Closed 2 years ago8
FIgure out how to modify how `backward()` is called.
Closed 2 years ago1
Move `master` branch to `main`
Closed 3 years ago1
Release luz 0.2.0
Closed 3 years ago
Error with last batch that is smaller than the others
Closed 3 years ago4
Early stopping with patience=1 shouldn't stop in the first epoch
Closed 3 years ago
Combining early stopping with csv logger callback, metrics for last epoch don't get logged
Closed 3 years ago1
Implement `evaluate` so we can quickly compute metrics in a specific dataset.
Closed 3 years ago1
Training data should be shuffled by default
Closed 3 years ago
Support for `valid_data = p`
Closed 3 years ago
Minimum and maximum number of epochs
Closed 3 years ago2
Issue with length of progress bar for validation loop
Closed 3 years ago
`luz_callback_csv_logger()` saves a space-delimited file, not a comma-separated one. I think `sep = ","` needs to be added here:
Closed 3 years ago
Support for mixed precision
Updated 3 years ago
Release luz 0.1.0
Closed 3 years ago
plot method for fitted models
Closed 3 years ago
add URL to the description
Closed 3 years ago1
Profile callback
Closed 3 years ago
luz_callback_model_checkpoint(save_best_only = TRUE) does not save a file
Closed 3 years ago
Error if module has default initialization parameters
Closed 3 years ago2
Allow interrupting the training loop
Closed 3 years ago
luz_metric_auroc does not work on GPU
Closed 3 years ago