Releases | Open Source Science

hezar - 0.42.0

- Python
Published by github-actions[bot] over 1 year ago

hezar - 0.41.1

- Python
Published by github-actions[bot] over 1 year ago

hezar - 0.41.0

- Python
Published by github-actions[bot] over 1 year ago

hezar - 0.40.3

- Python
Published by github-actions[bot] over 1 year ago

hezar - 0.40.2

- Python
Published by github-actions[bot] over 1 year ago

hezar - 0.40.1

- Python
Published by github-actions[bot] over 1 year ago

Main Changes

Add text_detection to models
Add CRAFT model for text detection
Suppress some dependencies and make them optional
Fix minor bugs and issues

- Python
Published by github-actions[bot] over 1 year ago

Changes: - Now all build functions (previously only accessible from hezar.builders) are also importable from their base modules. (from hezar.models import build_model) - Fixed the bug of requiring librosa for the base installation. - Some updates to the docs.

- Python
Published by github-actions[bot] over 1 year ago

hezar - 0.39.0

Main Changes

All dataset classes now have a preprocessor parameter which can be a path or object. Meaning that all manual preprocessor configs or paths are removed from the dataset configs.
Dataset configs now have a hf_load_kwargs to be passed to the datasets.load_dataset() .
Dataset config names are now supported either in hf_load_kwargs (hf_load_kwargs={"name": "fa"}) or by passing to path like Dataset.load("<path>:<config_name>")`.
Datasets now have the _load() method which is responsible for loading data files either from the hub or custom data reading.
Datasets now have a max_size in their config to overwrite the length of the dataset (all __len__ implementations are moved to the base class). This value can be a fraction too e.g, 0.3 means 30% of the original length.
Better LR scheduling implemented in the Trainer.
The docs have been rewritten or improved a lot!
Make log_steps and save_steps accept float values between 0 and 1 representing a fraction.
Other bug fixes and improvements.

- Python
Published by github-actions[bot] over 1 year ago

hezar - 0.38.5

- Python
Published by github-actions[bot] over 1 year ago

hezar - 0.38.4

- Python
Published by github-actions[bot] almost 2 years ago

hezar - 0.38.3

- Python
Published by github-actions[bot] almost 2 years ago

hezar - 0.38.2

- Python
Published by github-actions[bot] almost 2 years ago

hezar - 0.38.1

- Python
Published by github-actions[bot] almost 2 years ago

hezar - 0.38.0

Main changes

Resuming from checkpoint moved to the TrainerConfig (setting resume_from_checkpoint in Trainer.train() is now deprecated and raises error)
Resuming from checkpoints now supports inner-loop steps instead of only epochs
Add data sampler for slicing data loaders (mainly used for training resumption)
Re-order objects initialization in the Trainer's init function
Add support for optimizer checkpointing in Trainer
Add option to disable preprocess and post processing in Model.predict()
Seperate generation config in Whisper's model config with a separate data class
Drop support for Python 3.9

- Python
Published by github-actions[bot] almost 2 years ago

hezar - 0.37.0

- Python
Published by github-actions[bot] almost 2 years ago

hezar - 0.36.1

- Python
Published by github-actions[bot] almost 2 years ago

hezar - 0.36.0

v0.36.0

1. Add gradient accumulation support to the `Trainer`

Now, you can set gradient_accumulation_steps (defaults to 1 which is the same as regular training) in the TrainerConfig to enable this feature. This technique can mimic having larger batch sizes without changing the actual batch size! For example, having batch size of 16 and gradient accumulation steps of 4 equals to having batch size of 64! This can lead to faster convergance.

2. Implement tools for training speech recognition models

In this release we added SpeechRecognitionDataset, SpeechRecognitionDataCollator and SpeechRecognitionMetricsHandler so that you can easily train or finetune a Whisper model. Take a look at this example.

3. Split and refactor in `Trainer` for better subclassing

We split the training_step function of the Trainer in a way that now it only takes care of forward/backward pass and the optimization step is now moved to its own method called optimization_step. Also added lr_scheduler_step for customizing LR scheduling step.

4. Add support for more LR schedulers

5. Other bug fixes and improvements

- Python
Published by github-actions[bot] about 2 years ago

hezar - 0.35.1

- Python
Published by github-actions[bot] about 2 years ago

hezar - 0.35.0

This is a big one! We made a lot of changes and improvements in Hezar.

Improvements

Add support for accelerate for distributed training
Add resume from checkpoint feature to Trainer
Improve saving/logging capabilities in Trainer
Improve print_info()
Add ImageCaptioningDataset and ImageCaptioningDataCollator
Enhance padding in tokenizers
Rewrite contribution docs
Add tests workflow to actions
Add cache_dir parameter to all load() methods
Improve OCRDataset and bug fixes
Add training scripts for image captioning
Add training script for CRNN training
Clean registry.py
Change license from MIT to Apache 2.0
Some improvements and bug fixes in ViTRobertaImage2Text
Bug fixes in tests
Safe class var handling in configs
Add return_scores to CRNNImage2Text
Add get_state_dict_from_hub to support loading from any (non-Hezar) model on the Hub
Set default LR scheduler (reduce on plateau) to Trainer

Bug fixes

Fix image captioning decoding bug
Fix mixed precision bug on CPU
Fix embedding config bug