SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1911.08717 · 2019

Fine-Tuning by Curriculum Learning for Non-Autoregressive Neural Machine Translation

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 12 functions out of this paper's own repositories and ran 11 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
lemmonation/fcl-nat canonical 11 of 12
FunctionStatusWhere it lives
add_standard_attention_hparams Ran lemmonation/fcl-nat/tensor2tensor/layers/common_attention.py
code served (permissive licence) · get_code("44e847faaba1328c")
attention_bias_without_self Ran lemmonation/fcl-nat/tensor2tensor/models/transformer_nat_cl_word.py
code served (permissive licence) · get_code("1cca0cf169d3512e")
copy_batches Ran lemmonation/fcl-nat/tensor2tensor/models/transformer_nat_cl_word.py
code served (permissive licence) · get_code("ff01c7e565eb434e")
hard_sigmoid Ran lemmonation/fcl-nat/tensor2tensor/layers/common_layers.py
code served (permissive licence) · get_code("62078081af34b8ee")
hard_tanh Ran lemmonation/fcl-nat/tensor2tensor/layers/common_layers.py
code served (permissive licence) · get_code("dfdf8307a155fa8b")
is_in_expr Ran lemmonation/fcl-nat/tensor2tensor/data_generators/algorithmic_math.py
code served (permissive licence) · get_code("7e2bc85a4d4189a9")
lower_endian_to_number Ran lemmonation/fcl-nat/tensor2tensor/data_generators/algorithmic.py
code served (permissive licence) · get_code("b1896a2536602d28")
random_expr Ran lemmonation/fcl-nat/tensor2tensor/data_generators/algorithmic_math.py
code served (permissive licence) · get_code("916cc023d1d1ddbd")
random_expr_with_required_var Ran lemmonation/fcl-nat/tensor2tensor/data_generators/algorithmic_math.py
code served (permissive licence) · get_code("0a76872f8ddc4b1e")
zipf_distribution Ran lemmonation/fcl-nat/tensor2tensor/data_generators/algorithmic.py
code served (permissive licence) · get_code("68fe9a15fc61f6bd")
zipf_random_sample Ran lemmonation/fcl-nat/tensor2tensor/data_generators/algorithmic.py
code served (permissive licence) · get_code("d0d6f2cc06f67920")
saturating_sigmoid Not yet run lemmonation/fcl-nat/tensor2tensor/layers/common_layers.py
code served (permissive licence) · get_code("c53fad116e23fd5d")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Non-autoregressive translation (NAT) models remove the dependence on previous target tokens and generate all target tokens in parallel, resulting in significant inference speedup but at the cost of inferior translation accuracy compared to autoregressive translation (AT) models. Considering that AT models have higher accuracy and are easier to train than NAT models, and both of them share the same model configurations, a natural idea to improve the accuracy of NAT models is to transfer a well-trained AT model to an NAT model through fine-tuning. However, since AT and NAT models differ greatly in training strategy, straightforward fine-tuning does not work well. In this work, we introduce curriculum learning into fine-tuning for NAT. Specifically, we design a curriculum in the fine-tuning process to progressively switch the training from autoregressive generation to non-autoregressive generation. Experiments on four benchmark translation datasets show that the proposed method achieves good improvement (more than $1$ BLEU score) over previous NAT baselines in terms of translation accuracy, and greatly speed up (more than $10$ times) the inference process over AT baselines.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("1911.08717")
get_code_for_paper("1911.08717")
have("1911.08717")

Connect an agent — have() is free.