TorchStudio 0.9.19

• fix an issue when installing python packages

TorchStudio 0.9.18

• add AdamWScheduleFree optimizer, which become the new default optimizer
• add NoSchedule scheduler, which become the new default scheduler
• 10 times faster project saving and 1.5 faster project loading
• 2 times faster tensors transfer
• new Start or Resume All Trainings/Stop All Trainings button in the Dashboard tab
• early stop steps raised to 30 (previously 20)
• update deprecated functions with newer functions in pytorch and matplotlib

TorchStudio 0.9.17

• add a new command line option "--instance" to launch separate, independent instances of TorchStudio
• fix an issue that prevented TorchStudio from properly installing PyTorch with CUDA support
• fix an issue that prevented using remote servers
• fix an issue when parsing modules

TorchStudio 0.9.16

This release fixes a couple issues and uses the new recommendations for PyTorch installation, resulting in a more reliable TorchStudio environment initialization.

• replace conda by pip as the main package installer
• keep batch dimension in graph
• fix a crash with the graph process
• fix an issue with the generic loader and images not having frame numbers

TorchStudio 0.9.15

TorchStudio is now fully compatible with PyTorch 2.x and conda 23.x.

• update ONNX export to version 17
• ONNX export can now be customized
• fix a crash when rescaling tensor displays
• dataset shuffle state is now properly saved
• fix a crash when python install failed
• spectrogram renderer improved to handle more tensor types
• fix issues when installing pytorch
• fix issues when installing python
• fix an issue where models settings could be changed when switching back from the dashboard

TorchStudio 0.9.14

This release brings lots of performance, stability and UI improvements. It also adds a new GenericLoader dataset, which can handle most common image, audio and numpy tensors datasets formats, for either classification, segmentation or regression.

• new GenericLoader dataset, which can load several kinds of image, audio and numpy tensors datasets for either classification, segmentation or regression
• TorchStudio projects now use much less RAM
• transfers with local and remote servers is now much faster
• new weights and state transfers from local and remote servers is now asynchronous
• only best weights and state are transferred to speed up training and optimize memory use
• dataset can now be cached on local and remote trainign server to speed up the start of new trainings
• environment installer now compatible with the newest conda installers
• Patience value can now be defined (in number of epochs) when Early Stopping is checked
• channel colors can now be defined for Bitmap, Spectrogram and Volume renderers
• default threshold value for accuracy metric set to 0.01
• new best weights indicator in the loss and metric plots (both in Model tabs and Dashboard panel)
• loss indicator now plotted on a square root scale for easier readibility
• model color indicator added to the parameters plot (in the Dashboard panel)
• tool tips improved
• fix an issue when loading remote dataset
• fix an issue where inference could stop working
• fix an issue where tensor display and analysis could stop working
• fix an issue where tcp decoding could break
• fix project opening when a dataset is loaded from a remote server
• fix metrics not properly displaying epochs beyond 100

TorchStudio 0.9.13

This release brings compatibility with PyTorch 1.13 and fix server connection issues.

• compatibility with PyTorch 1.13
• fix connection lost issues with some servers and larger models
• progress bar indicator when loading and saving projects
• more log infos

TorchStudio 0.9.12

• fix module parameters not properly restored when loading a project
• fix an issue introduced with mixed precision training
• fix pin on top state when saving and loading a project
• minor improvements to RandomGenerator dataset

TorchStudio 0.9.11

This release adds support for FP32, TF32, FP16 and BF16 precision modes both for training and inference. If you had previously added a remote server, refresh it to make those modes visible. They will appear next to the inference or training device selector.

• Add support for FP32, TF32, FP16 and BF16 precision modes
• Improve SSH connectivity and sync
• Fix python clients staying alive on local and remote servers when tasks were complete

TorchStudio 0.9.10

• IDEs extensions for PyCharm, VS Code, Spyder and Sublime Text
• the main window can be pinned at the top to ease interactions with IDEs
• layout optimized to ease interactions with IDEs
• customizable text size and UI brightness
• simplified remote server configuration
• associate .tsz projects with TorchStudio
• fix an issue when trying to connect to SSH servers using only a password
• fix several issues when configuring or using Windows-based remote servers
• various stability improvements
• various UI and tooltips improvements