Skip to content
Snippets Groups Projects

Repository graph

You can move around the graph by using the arrow keys.
Select Git revision
  • master default protected
  • remove-decoder-and-encoder-ff
  • remove-decoder-ff
  • remove-encoder-ff
  • v0.6.2
  • v0.6.1
  • v0.6.0
  • v0.5.0
  • v0.4.0
  • v0.3.0
10 results
Created with Raphaël 2.2.06May532130Apr292726252422211716151210975432128Mar26191514121142128Feb252322191512865130Jan292524161514975428Dec27262418108765430Nov282726251918171614139876131Oct2926252220191753230Sep2524177316Aug131Jul262519111087228Jun26252421201524May2221912Apr28Mar2928262575127Feb12831Jan2926221256Dec124Nov1312118124Oct191714131211228Sep262419181514Add script for preparing WMT18-Multi30k with BPEmastermasterRemove two-layer ff network from transformer encoderremove-decoder-…remove-decoder-and-encoder-ffRemove two-layer ff network from transformer decoderremove-decoder-ffremove-decoder-ffRemove two-layer ff network from transformer encoderremove-encoder-ffremove-encoder-ffBugfix in size of multi-corpus datasetan option to raise exception if oom happens during fairseq.trainer.train_step (#2)added bert large architecture (#698)Make learned positional embedding optionalMove distributed_init into DistributedFairseqModel (#687)Validate on all sets based on --save-interval-updatesFix inconsistent gradient checkMake CTC work with more encoder-only modelsMake MultiCorpusSampledDataset and IndexedCachedDataset Picklableadd ConcatDataset support for XLMSupport dataset upsampling / relative ratio in PytorchTranslateTask (#494)Better OOM recoveryAdd default noising argument in WordNoiser initialization (#664)addding polynomial lr scheduler (#683)Merge internal changesAdd rm_pt.py helper script for removing checkpoint filesMerge internal changes (#654)Add more details in error message when sentence length > max tokens (#672)Fix upgrade_state_dict for XLM Transformer sentence encoder (#680)Update README.md (#679)Update comments and citationsAdd args and sys.argv to tensorboard (#673)Add small comments for MonolingualDataset and TokenBlockDatasetPassing kwargs in setup_task in fairseq_task (#670)Fix fairseq unittest timeouts (#667)XLM for NMT: option to only load encoder or decoder (#666)Load a XLM model into transformer encoder / decoder for MT training (#629)Add gelu and gelu_fast as possible activation functions (#653)Added link to blog post (#662)added link to sample storiesDon't reload best validation loss when using --reset-optimizerFix generation with --no-early-stop (#627)reduce memory footprint for average_checkpoints (#647)Open BlockPairDataset for MaskedLMData to work (#641)Enable custom sampling strategy in MultiCorpusSampledDataset (#639)Black formatting for multi_corpus_sampled_dataset.py (#638)