Open
Conversation
* Fixed topk decoder.
* Use torchtext from pipe. * Fixed torch text sorting order.
…BM#90) * attention is not required when only using teacher forcing in decoder
* 0.1.5 (IBM#91) * Modified parameter order of DecoderRNN.forward (IBM#85) * Updated TopKDecoder (IBM#86) * Fixed topk decoder. * Use torchtext from pipy (IBM#87) * Use torchtext from pipe. * Fixed torch text sorting order. * attention is not required when only using teacher forcing in decoder (IBM#90) * attention is not required when only using teacher forcing in decoder * Updated docs and version. * Fixed code style. * shuffle the training data
* fix example of inflate function in TopKDecoer.py
* Fix hidden_layer size for one-directional decoder Hidden layer size of the decoder was given `hidden_size * 2 if bidirectional else 1`, resulting in a dimensionality error for non-bidirectional decoders. Changed `1` to `hidden_size`.
* Adapt load to allow CPU loading of GPU models Add storage parameter to torch.load to allow loading models on a CPU that are trained on the GPU, depending on availability of cuda.
* Fix wrong parameter use on DecoderRNN
# Conflicts: # seq2seq/models/TopKDecoder.py # seq2seq/trainer/supervised_trainer.py
* Modified parameter order of DecoderRNN.forward (IBM#85) * Updated TopKDecoder (IBM#86) * Fixed topk decoder. * Use torchtext from pipy (IBM#87) * Use torchtext from pipe. * Fixed torch text sorting order. * attention is not required when only using teacher forcing in decoder (IBM#90) * attention is not required when only using teacher forcing in decoder * Updated docs and version. * Fixed code style. * bugfix (IBM#92) Fixed field arguments validation. * Removed `initial_lr` when resuming optimizer with scheduler. (IBM#95) * shuffle the training data (IBM#97) * 0.1.5 (IBM#91) * Modified parameter order of DecoderRNN.forward (IBM#85) * Updated TopKDecoder (IBM#86) * Fixed topk decoder. * Use torchtext from pipy (IBM#87) * Use torchtext from pipe. * Fixed torch text sorting order. * attention is not required when only using teacher forcing in decoder (IBM#90) * attention is not required when only using teacher forcing in decoder * Updated docs and version. * Fixed code style. * shuffle the training data * fix example of inflate function in TopKDecoer.py (IBM#98) * fix example of inflate function in TopKDecoer.py * Fix hidden_layer size for one-directional decoder (IBM#99) * Fix hidden_layer size for one-directional decoder Hidden layer size of the decoder was given `hidden_size * 2 if bidirectional else 1`, resulting in a dimensionality error for non-bidirectional decoders. Changed `1` to `hidden_size`. * Adapt load to allow CPU loading of GPU models (IBM#100) * Adapt load to allow CPU loading of GPU models Add storage parameter to torch.load to allow loading models on a CPU that are trained on the GPU, depending on availability of cuda. * Fix wrong parameter use on DecoderRNN (IBM#103) * Fix wrong parameter use on DecoderRNN
* Upgrade to pytorch-0.3.0 * Use pytorch 3.0 in travis env.
…eturns several seqs for a given seq (IBM#116) * Adding a predictor method to return n predicted seqs for a src_seq input (intended to be used along to Beam Search using TopKDecoder)
when attention is turned off, pytorch (well, 0.4 at least) gets angry about calling view on a non-contiguous tensor
* Modified parameter order of DecoderRNN.forward (IBM#85) * Updated TopKDecoder (IBM#86) * Fixed topk decoder. * Use torchtext from pipy (IBM#87) * Use torchtext from pipe. * Fixed torch text sorting order. * attention is not required when only using teacher forcing in decoder (IBM#90) * attention is not required when only using teacher forcing in decoder * Updated docs and version. * Fixed code style. * bugfix (IBM#92) Fixed field arguments validation. * Removed `initial_lr` when resuming optimizer with scheduler. (IBM#95) * shuffle the training data (IBM#97) * 0.1.5 (IBM#91) * Modified parameter order of DecoderRNN.forward (IBM#85) * Updated TopKDecoder (IBM#86) * Fixed topk decoder. * Use torchtext from pipy (IBM#87) * Use torchtext from pipe. * Fixed torch text sorting order. * attention is not required when only using teacher forcing in decoder (IBM#90) * attention is not required when only using teacher forcing in decoder * Updated docs and version. * Fixed code style. * shuffle the training data * fix example of inflate function in TopKDecoer.py (IBM#98) * fix example of inflate function in TopKDecoer.py * Fix hidden_layer size for one-directional decoder (IBM#99) * Fix hidden_layer size for one-directional decoder Hidden layer size of the decoder was given `hidden_size * 2 if bidirectional else 1`, resulting in a dimensionality error for non-bidirectional decoders. Changed `1` to `hidden_size`. * Adapt load to allow CPU loading of GPU models (IBM#100) * Adapt load to allow CPU loading of GPU models Add storage parameter to torch.load to allow loading models on a CPU that are trained on the GPU, depending on availability of cuda. * Fix wrong parameter use on DecoderRNN (IBM#103) * Fix wrong parameter use on DecoderRNN * Upgrade to pytorch-0.3.0 (IBM#111) * Upgrade to pytorch-0.3.0 * Use pytorch 3.0 in travis env. * Make sure tensor contiguous when attention's not used. (IBM#112) * Implementing the predict_n method. Using the beam search outputs it returns several seqs for a given seq (IBM#116) * Adding a predictor method to return n predicted seqs for a src_seq input (intended to be used along to Beam Search using TopKDecoder) * Checkpoint after batches not epochs (IBM#119) * Pytorch 0.4 (IBM#134) * add contiguous call to tensor (IBM#127) when attention is turned off, pytorch (well, 0.4 at least) gets angry about calling view on a non-contiguous tensor * Fixed shape documentation (IBM#131) * Update to pytorch-0.4 * Remove pytorch manual install in travis. * Allow using pre-trained embedding (IBM#135) * updated docs
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
11 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
copy_decoderdef and minor changes.TODO: fix dimensions for compatibility with
top_k_decoder.