claf.modules package

Subpackages

• claf.modules.attention package
  • Submodules
  • Module contents
• claf.modules.conv package
  • Submodules
  • Module contents
• claf.modules.encoder package
  • Submodules
  • Module contents
• claf.modules.layer package
  • Submodules
  • Module contents

Submodules

claf.modules.activation.get_activation_fn(name)

PyTorch built-in activation functions

some functional codes from allennlp: https://github.com/allenai/allennlp

• add_masked_value : replace_masked_values (allennlp)

• get_mask_from_tokens : get_mask_from_tokens (allennlp)

• last_dim_masked_softmax : last_dim_masked_softmax (allennlp)

• masked_softmax : masked_softmax (allennlp)

• weighted_sum : weighted_sum (allennlp)

claf.modules.functional.add_masked_value(tensor, mask, value=-10000000.0)

claf.modules.functional.forward_rnn_with_pack(rnn_module, tensor, seq_config)

claf.modules.functional.get_mask_from_tokens(tokens)

claf.modules.functional.get_sorted_seq_config(features, pad_index=0)

claf.modules.functional.last_dim_masked_softmax(x, mask)

claf.modules.functional.masked_log_softmax(vector, mask)

claf.modules.functional.masked_softmax(x, mask)

claf.modules.functional.masked_zero(tensor, mask)

Tensor masking operation

claf.modules.functional.weighted_sum(attention, matrix)

claf.modules.initializer.weight(module)

weight initialization (according to module type)

• Args:

  module: torch.nn.Module

Module contents