fairseq2
fairseq2 copied to clipboard
facebookresearch
FAIR Sequence Modeling Toolkit 2
**Describe the solution you would like:** - Enable the training script to access outputs of intermediate layers - Modify loss function to incorprate outputs of earlier layers **Describe the alternatives...
**Describe the solution you would like:** Would like to enable configuration of a different layer dropout rate for each layer. **Describe the alternatives you have considered:** Currently, layer dropout is...
**What does this PR do? Please describe:** This PR adds a wrapper to HuggingFace's `evaluate.Metric` to make it compatible to fairseq2.metrics APIs. This enables evaluating fairseq2 model on many downstream...
**Describe the bug:** Attempted to bump torch version to 2.3.0 and got an error: ERROR: fairseq2n 0.2.1 has requirement torch==2.2.2, but you'll have torch 2.3.0 which is incompatible **Describe how...
When I compiled the fairseq2n native source code, an error occurred.(my c++ is gcc-8) when i run "cmake -GNinja -DCMAKE_CUDA_ARCHITECTURES="80-real;80-virtual" -DFAIRSEQ2N_USE_CUDA=ON -B build cmake --build build" error msg: "src/fairseq2n/CMakeFiles/fairseq2n.dir/data/text/text_data_source.cc.o:/usr/include/c++/8/bits/fs_path.h:310: more...
This PR: - Introduces a new `RowMaskFactory` protocol which `compute_row_mask` follows. - Converts `Wav2Vec2Masker` into an interface and moves its existing implementation to a new `StandardWav2Vec2Masker`. - Introduces a new...
**What does this PR do? Please describe:** Adds support for the experimental FSDP2 data parallelism. **Does your PR introduce any breaking changes? If yes, please list them:** No. **Check list:**...
If possible, could you help to provide a `Dockerfile` and/or docker image that we can pull to use the fairseq2 library? Thank you in advance!
… indices to be masked **What does this PR do? Please describe:** A summary of the change or the issue that is fixed. Use rand + topk in place of...