fairseq
fairseq copied to clipboard
Facebook AI Research Sequence-to-Sequence Toolkit written in Python.
## ❓ Questions and Help ### Before asking: 1. search the issues. 2. search the docs. #### What is your question? How to use monolingual data ? Do I understand...
## 📚 Documentation Hey there, I am trying to play with nllb but there isn't a basic code sample to try it. I can download the `checkpoint.pt` but I am...
## ❓ Questions and Help #### What is your question? I tried to fine tune bart.base using CNN-DM. However, I cannot learn because I got the following error. ”Cannot load...
# Before submitting - [ ] Was this discussed/approved via a Github issue? (no need for typos, doc improvements) - [ ] Did you read the [contributor guideline](https://github.com/pytorch/fairseq/blob/main/CONTRIBUTING.md)? - [...
# Before submitting - [ ] Was this discussed/approved via a Github issue? (no need for typos, doc improvements) - [ ] Did you read the [contributor guideline](https://github.com/pytorch/fairseq/blob/main/CONTRIBUTING.md)? - [...
## 🐛 Bug `torch.jit.frontend.NotSupportedError` while JIT tracing/scripting `FairseqLanguageModel` on CPU. ### To Reproduce Reopening #2741 #### Code sample Please refer to #2741 ### Expected behavior @myleott, this behavior is [expected...
Hi, I use the released NLLB checkpoint to decode flroes Chinese testset, overall the results looks good. However, I found that a lot of very common Chinese characters/tokens are missing...
Looking at "MultilingualMaskedLMTask" code, dictionaries seems to be required to setup this task. To build the dictionaries, we require to preprocess the sentence pieces upfront. Preprocessing raw text upfront doesn't...
不知道怎么形容,但是NLLB 200 demo所展示的,翻译为简体中文的效果真的太差劲了。 方向很高大上,但是实际体验的效果真的不如人意。 加油吧,小伙伴们,期待作出更精彩绝伦的作品。
According to [this link](https://pytorch.org/blog/a-better-transformer-for-fast-transformer-encoder-inference/), torch 1.12.0 improve inferece speed of [TransformerEncoder](https://pytorch.org/docs/stable/generated/torch.nn.TransformerEncoder.html), [TransformerEncoderLayer](https://pytorch.org/docs/stable/generated/torch.nn.TransformerEncoderLayer.html) or [MultiheadAttention](https://pytorch.org/docs/stable/generated/torch.nn.MultiheadAttention.html) (MHA) in specific conditions (when we use lots of padding tokens) by fusing cuda kernel so...