fairseq2
fairseq2 copied to clipboard
facebookresearch
FAIR Sequence Modeling Toolkit 2
Until now, I don't seem to have found any details about the specific example of dataset organizing and training from scratch , Would fairseq2 intend to provide a separate trainer...
**What does this PR do? Please describe:** This PR adds integration for whisper model with the `hg` evaluation CLI interface. In the process, I have refactored some functions to be...
**What does this PR do? Please describe:** This PR extends the CLI with support for tab completions. Initially, I tried using argcomplete, but it was slow and ran a Fairseq2...
**What does this PR do? Please describe:** Add a new optional argument to dynamic_bucket---bucket_creation_fn---to allow the user to customize how buckets are created. bucket_creation_fn allows the user to customize what...
**What does this PR do? Please describe:** Adds a utility function, read_and_pickle_wrap_iterator, to which any iterator/generator can be passed. This will wrap the iterator/generator in a class whose pickle behavior...
**What does this PR do? Please describe:** To run pretraining: ``` fairseq2 wav2vec2 train /checkpoint/$USER/wav2vec2_train ``` Run this code snippet for testing the dataloader: ``` import torch from fairseq2.datasets.speech import...
**Describe the solution you would like:** Create a fairseq2 wrapper class/script to enable LM Evaluation Harness, https://github.com/EleutherAI/lm-evaluation-harness. This We need to create a [wrapper class](https://github.com/EleutherAI/lm-evaluation-harness/blob/main/docs/interface.md#external-library-usage) that implements the following functions:...
Thank you so much for the great work. I wonder if there are documentation and training examples for quick use.
I have numerous audio files along with their corresponding subtitles. I need to load specific segments of these audio files, defined by particular offsets and durations. However, the current python...
**Describe the solution you would like:** Implement self-speculative decoding as described in this [paper](https://arxiv.org/abs/2404.16710) where the earlier layers act as the draft stage and remaining layers act as the verification...