diff options
author | Gustaf Rydholm <gustaf.rydholm@gmail.com> | 2021-10-28 21:21:00 +0200 |
---|---|---|
committer | Gustaf Rydholm <gustaf.rydholm@gmail.com> | 2021-10-28 21:21:00 +0200 |
commit | 2203a1ba52ab2f72682fcee738844ee9ec584bda (patch) | |
tree | 8a1bcc6ceb3f1c91d47d0f35ca64554bcce2fcbd /text_recognizer/networks/transformer/embeddings | |
parent | 863c42ea67823ad616ac7485fc8a1e5018cb4233 (diff) |
Remove 2D positional embedding
Diffstat (limited to 'text_recognizer/networks/transformer/embeddings')
-rw-r--r-- | text_recognizer/networks/transformer/embeddings/fourier.py | 50 |
1 files changed, 2 insertions, 48 deletions
diff --git a/text_recognizer/networks/transformer/embeddings/fourier.py b/text_recognizer/networks/transformer/embeddings/fourier.py index c50afc3..ade589c 100644 --- a/text_recognizer/networks/transformer/embeddings/fourier.py +++ b/text_recognizer/networks/transformer/embeddings/fourier.py @@ -1,5 +1,4 @@ -"""A positional encoding for the image features, as the transformer has no notation of the order of the sequence.""" -from einops import repeat +"""Fourier positional embedding.""" import numpy as np import torch from torch import nn @@ -35,51 +34,6 @@ class PositionalEncoding(nn.Module): """Encodes the tensor with a postional embedding.""" # [T, B, D] if x.shape[2] != self.pe.shape[2]: - raise ValueError(f"x shape does not match pe in the 3rd dim.") + raise ValueError("x shape does not match pe in the 3rd dim.") x = x + self.pe[: x.shape[0]] return self.dropout(x) - - -class PositionalEncoding2D(nn.Module): - """Positional encodings for feature maps.""" - - def __init__(self, hidden_dim: int, max_h: int = 2048, max_w: int = 2048) -> None: - super().__init__() - if hidden_dim % 2 != 0: - raise ValueError(f"Embedding depth {hidden_dim} is not even!") - self.hidden_dim = hidden_dim - pe = self.make_pe(hidden_dim, max_h, max_w) - self.register_buffer("pe", pe) - - @staticmethod - def make_pe(hidden_dim: int, max_h: int, max_w: int) -> Tensor: - """Returns 2d postional encoding.""" - pe_h = PositionalEncoding.make_pe( - hidden_dim // 2, max_len=max_h - ) # [H, 1, D // 2] - pe_h = repeat(pe_h, "h w d -> d h (w tile)", tile=max_w) - - pe_w = PositionalEncoding.make_pe( - hidden_dim // 2, max_len=max_w - ) # [W, 1, D // 2] - pe_w = repeat(pe_w, "w h d -> d (h tile) w", tile=max_h) - - pe = torch.cat([pe_h, pe_w], dim=0) # [D, H, W] - return pe - - def forward(self, x: Tensor) -> Tensor: - """Adds 2D postional encoding to input tensor.""" - # Assumes x hase shape [B, D, H, W] - if x.shape[1] != self.pe.shape[0]: - raise ValueError("Hidden dimensions does not match.") - x += self.pe[:, : x.shape[2], : x.shape[3]] - return x - - -def target_padding_mask(trg: Tensor, pad_index: int) -> Tensor: - """Returns causal target mask.""" - trg_pad_mask = (trg != pad_index)[:, None, None] - trg_len = trg.shape[1] - trg_sub_mask = torch.tril(torch.ones((trg_len, trg_len), device=trg.device)).bool() - trg_mask = trg_pad_mask & trg_sub_mask - return trg_mask |