OpenNMT-py v0.7.1

Many fixes and code refactoring thanks @bpopeters, @flauted, @guillaumekln

New features
Random sampling thanks @daphnei
Enable sharding for huge files at translation

OpenNMT-py v0.7.0

Many fixes and code refactoring thanks @benopeters
Migrated to Pytorch 1.0

OpenNMT-py v0.6.0

Mostly fixes and code improvements.

New: yml config files. See the config folder

OpenNMT-py v0.5.0

Ability to reset the optimizer when using -train_from

-reset_optim = ['none', 'all', 'states', 'keep_states']
none: default behavior as before
all: reset the optimizer !! steps start at zero again.
states: reset only states, keep all other parameters from checkpoint
keep_states: keep current states from checkpoint, but allow to change parameters (learning_rate for instance)

Bug fixes.
Tested with Pytorch 1.0RC works fine.

OpenNMT-py v0.4.1

• fix preprocess filenames introduced by new sharding.

OpenNMT-py v0.4

Fixed Speech2Text training (thanks Yuntian)

Removed -max_shard_size, replaced by -shard_size = number of examples in a shard.

Default value = 1M which works fine in most Text dataset cases. (will avoid Ram OOM in most cases)

OpenNMT-py v0.3

Now requires Pytorch 0.4.1

Multi-node Multi-GPU with Torch Distributed

New options are:
-master_ip: ip address of the master node
-master_port: port number of th emaster node
-world_size = total number of processes to be run (total GPUs accross all nodes)
-gpu_ranks = list of indices of processes accross all nodes

-gpuid is deprecated

See examples in https://github.com/OpenNMT/OpenNMT-py/blob/master/docs/source/FAQ.md

Fixes to img2text now working

New sharding based on number of examples

Fixes to avoid 0.4.1 deprecated functions.

OpenNMT-py v0.2.1

Fixes and improvements

• First compatibility steps with Pytorch 0.4.1 (non breaking)
• Fix TranslationServer (when various request try to load the same model at the same time)
• Fix StopIteration error (python 3.7)

New features

• Ensemble at inference (thanks @Waino) see FAQ

Last Pytorch 0.4.0 version

New in this release:

Multi-GPU based on torch distributed (acknowledgement to Fairseq)
Change from Epoch to Step (see opts.py)
Average Attention Network (AAN) for the Transformer (thanks @francoishernandez )
New fast beam search (see -fast in translate.py) (thanks @guillaumekln)
Sparse attention / sparsemax (thanks to @bpopeters)

and many fixes.

This is the last version with pytorch 0.4.0
Next 0.4.1 pytorch version includes breakings changes.

Pytorch 0.3 Last Release

Merge pull request #680 from OpenNMT/torch0.4

Fix softmaxes