stephenroller Torch Generator Agent (#1260)
* Dab.

* Simple code movement. Nothing generalizes.

* More code cut-paste.

* Docstring formatting.

* Moderate reorganization; minor docstring formatting.

* A little bit closer.

* Simplfiy slightly.

* Opinionated formatting.

* Beam search seems works for transformer, broken for seq2seq.

* Finally move the time iteration into decode for seq2seq.

* Fix some bugs around loading saved models.

* Support skipping generation for faster ppl-only validation.

* Move --input-dropout to seq2seq

* Support incremental decoding.

* Name be changin'

* Support multigpu. Integrate in Abi's fixes.

* Fix ranked candidates

* Dead code society.

* warn_once

* Lint.

* Forgot to reorder and select incremental states.

* More multigpu workarounds.

* lint.

* Be consistent in initialization.

* Make sure outputlayer has backwards compatibility.

* Bring #1267 into this branch.

* Add fast termination in greedy_decode. Bring back _init_cuda_buffer

* Fix bugs in hogwild

* Address review comments.

* Whoops forgot a cuda call.

* Simplify output layer since binary compatibility is no longer needed.

* Seq2seq version bump. (#1282)

* Seq2seq version bump.
Latest commit 548c67b Nov 15, 2018

README.md

Agents

This directory contains a variety of different agents which use ParlAI's interface.

Utility

  • local_human: receives human input from the terminal. used for interactive mode, e.g. parlai/scripts/interactive.py.
  • legacy_agents: contains deprecated agent code for posterity
  • random_candidate: returns a random candidate, if candidates are available. simple baseline agent.
  • remote_agent: uses ZMQ to communicate with a different process, either on the local machine or on a remote host.
  • repeat_label: sends back the label if available. good for sanity checks such as checking statistics of the base dataset.
  • repeat_query: repeats whatever is said to it. simple baseline agent.

Non-neural agents

  • ir_baseline: chooses response based on simple word overlap
  • retriever_reader: used primarily for OpenSquad evaluation. retrieves documents from database and reads them back
  • tfidf_retriever: returns candidate responses based on tfidf overlap
  • unigram: returns top unigrams

Text-based neural networks

  • drqa: context-based question answering system
  • fairseq: provides access to models from FAIR's FairSeq library (github.com/facebookresearch/fairseq)
  • language_model: simple RNN-based language model
  • memnn: memory network
  • ibm_seq2seq: IBM's RNN-based sequence to sequence model
  • seq2seq: our RNN-based sequence to sequence model
  • starspace: embedding model

Visual neural networks

  • mlb_vqa: visual question answering model
  • vsepp_caption: image captioning model