anchors

labml_nn/conv_mixer/__init__.py

@@ -44,9 +44,8 @@ from labml_nn.utils import clone_module_list
 
 class ConvMixerLayer(Module):
     """
-    <a id="ConvMixerLayer">
+    <a id="ConvMixerLayer"></a>
     ## ConvMixer layer
-    </a>
 
     This is a single ConvMixer layer. The model will have a series of these.
     """
@@ -100,9 +99,8 @@ class ConvMixerLayer(Module):
 
 class PatchEmbeddings(Module):
     """
-    <a id="PatchEmbeddings">
+    <a id="PatchEmbeddings"></a>
     ## Get patch embeddings
-    </a>
 
     This splits the image into patches of size $p \times p$ and gives an embedding for each patch.
     """
@@ -140,9 +138,8 @@ class PatchEmbeddings(Module):
 
 class ClassificationHead(Module):
     """
-    <a id="ClassificationHead">
+    <a id="ClassificationHead"></a>
     ## Classification Head
-    </a>
 
     They do average pooling (taking the mean of all patch embeddings) and a final linear transformation
     to predict the log-probabilities of the image classes.

labml_nn/experiments/mnist.py

@@ -23,9 +23,8 @@ from labml_nn.optimizers.configs import OptimizerConfigs
 
 class MNISTConfigs(MNISTDatasetConfigs, TrainValidConfigs):
     """
-    <a id="MNISTConfigs">
+    <a id="MNISTConfigs"></a>
     ## Trainer configurations
-    </a>
     """
 
     # Optimizer

labml_nn/experiments/nlp_autoregression.py

@@ -40,9 +40,8 @@ class CrossEntropyLoss(Module):
 
 class NLPAutoRegressionConfigs(TrainValidConfigs):
     """
-    <a id="NLPAutoRegressionConfigs">
+    <a id="NLPAutoRegressionConfigs"></a>
     ## Trainer configurations
-    </a>
 
     This has the basic configurations for NLP auto-regressive task training.
     All the properties are configurable.

labml_nn/experiments/nlp_classification.py

@@ -28,9 +28,8 @@ from labml_nn.optimizers.configs import OptimizerConfigs
 
 class NLPClassificationConfigs(TrainValidConfigs):
     """
-    <a id="NLPClassificationConfigs">
+    <a id="NLPClassificationConfigs"></a>
     ## Trainer configurations
-    </a>
 
     This has the basic configurations for NLP classification task training.
     All the properties are configurable.

labml_nn/optimizers/adam_warmup_cosine_decay.py

@@ -17,9 +17,8 @@ from labml_nn.optimizers.amsgrad import AMSGrad
 
 class AdamWarmupCosineDecay(AMSGrad):
     """
-    <a id="EmbeddingsWithPositionalEncoding">
+    <a id="EmbeddingsWithPositionalEncoding"></a>
     ## Adam Optimizer with Warmup and Cosine Decay
-    </a>
 
     This class extends from AMSGrad optimizer defined in [`amsgrad.py`](amsgrad.html).
     """

labml_nn/optimizers/configs.py

@@ -17,9 +17,8 @@ from labml_nn.optimizers import WeightDecay
 
 class OptimizerConfigs(BaseConfigs):
     """
-    <a id="OptimizerConfigs">
+    <a id="OptimizerConfigs"></a>
     ## Optimizer Configurations
-    </a>
     """
 
     # Optimizer

labml_nn/rl/game.py

@@ -16,7 +16,10 @@ import numpy as np
 
 class Game:
     """
-    ## <a name="game-environment"></a>Game environment
+    <a name="game-environment"></a>
+
+    ## Game environment
+
     This is a wrapper for OpenAI gym game environment.
     We do a few things here:
 

labml_nn/transformers/configs.py

@@ -20,9 +20,8 @@ from .models import EmbeddingsWithPositionalEncoding, EmbeddingsWithLearnedPosit
 
 class FeedForwardConfigs(BaseConfigs):
     """
-    <a id="FFN">
+    <a id="FFN"></a>
     ## FFN Configurations
-    </a>
 
     Creates a Position-wise FeedForward Network defined in
     [`feed_forward.py`](feed_forward.html).
@@ -143,10 +142,9 @@ aggregate(FeedForwardConfigs.glu_variant, 'SwiGLU',
 
 class TransformerConfigs(BaseConfigs):
     """
-    <a id="TransformerConfigs">
+    <a id="TransformerConfigs"></a>
     ## Transformer Configurations
-    </a>
-    
+
     This defines configurations for a transformer.
     The configurations are calculate using option functions.
     These are lazy loaded and therefore only the necessary modules

labml_nn/transformers/feedback/__init__.py

@@ -311,9 +311,8 @@ class FeedbackTransformer(Module):
         return self.norm(res)
 
 
-# <a id="shared_kv">
+# <a id="shared_kv"></a>
 # # Shared keys and values among layers
-# </a>
 
 class StackFunction(torch.autograd.Function):
     """

labml_nn/transformers/mha.py

@@ -29,9 +29,8 @@ from labml_helpers.module import Module
 
 class PrepareForMultiHeadAttention(Module):
     """
-    <a id="PrepareMHA">
+    <a id="PrepareMHA"></a>
     ## Prepare for multi-head attention
-    </a>
 
     This module does a linear transformation and splits the vector into given
     number of heads for multi-head attention.
@@ -65,9 +64,8 @@ class PrepareForMultiHeadAttention(Module):
 
 class MultiHeadAttention(Module):
     r"""
-    <a id="MHA">
+    <a id="MHA"></a>
     ## Multi-Head Attention Module
-    </a>
 
     This computes scaled multi-headed attention for given `query`, `key` and `value` vectors.
 

labml_nn/transformers/models.py

@@ -22,9 +22,8 @@ from .positional_encoding import get_positional_encoding
 
 class EmbeddingsWithPositionalEncoding(Module):
     """
-    <a id="EmbeddingsWithPositionalEncoding">
+    <a id="EmbeddingsWithPositionalEncoding"></a>
     ## Embed tokens and add [fixed positional encoding](positional_encoding.html)
-    </a>
     """
 
     def __init__(self, d_model: int, n_vocab: int, max_len: int = 5000):
@@ -40,9 +39,8 @@ class EmbeddingsWithPositionalEncoding(Module):
 
 class EmbeddingsWithLearnedPositionalEncoding(Module):
     """
-    <a id="EmbeddingsWithLearnedPositionalEncoding">
+    <a id="EmbeddingsWithLearnedPositionalEncoding"></a>
     ## Embed tokens and add parameterized positional encodings
-    </a>
     """
 
     def __init__(self, d_model: int, n_vocab: int, max_len: int = 5000):
@@ -58,9 +56,8 @@ class EmbeddingsWithLearnedPositionalEncoding(Module):
 
 class TransformerLayer(Module):
     """
-    <a id="TransformerLayer">
+    <a id="TransformerLayer"></a>
     ## Transformer Layer
-    </a>
 
     This can act as an encoder layer or a decoder layer.
 
@@ -138,9 +135,8 @@ class TransformerLayer(Module):
 
 class Encoder(Module):
     """
-    <a id="Encoder">
+    <a id="Encoder"></a>
     ## Transformer Encoder
-    </a>
     """
 
     def __init__(self, layer: TransformerLayer, n_layers: int):
@@ -160,9 +156,8 @@ class Encoder(Module):
 
 class Decoder(Module):
     """
-    <a id="Decoder">
+    <a id="Decoder"></a>
     ## Transformer Decoder
-    </a>
     """
 
     def __init__(self, layer: TransformerLayer, n_layers: int):
@@ -182,9 +177,8 @@ class Decoder(Module):
 
 class Generator(Module):
     """
-    <a id="Generator">
+    <a id="Generator"></a>
     ## Generator
-    </a>
 
     This predicts the tokens and gives the lof softmax of those.
     You don't need this if you are using `nn.CrossEntropyLoss`.
@@ -200,9 +194,8 @@ class Generator(Module):
 
 class EncoderDecoder(Module):
     """
-    <a id="EncoderDecoder">
+    <a id="EncoderDecoder"></a>
     ## Combined Encoder-Decoder
-    </a>
     """
 
     def __init__(self, encoder: Encoder, decoder: Decoder, src_embed: Module, tgt_embed: Module, generator: Module):

labml_nn/transformers/vit/__init__.py

@@ -52,9 +52,8 @@ from labml_nn.utils import clone_module_list
 
 class PatchEmbeddings(Module):
     """
-    <a id="PatchEmbeddings">
+    <a id="PatchEmbeddings"></a>
     ## Get patch embeddings
-    </a>
 
     The paper splits the image into patches of equal size and do a linear transformation
     on the flattened pixels for each patch.
@@ -93,9 +92,8 @@ class PatchEmbeddings(Module):
 
 class LearnedPositionalEmbeddings(Module):
     """
-    <a id="LearnedPositionalEmbeddings">
+    <a id="LearnedPositionalEmbeddings"></a>
     ## Add parameterized positional encodings
-    </a>
 
     This adds learned positional embeddings to patch embeddings.
     """
@@ -121,9 +119,8 @@ class LearnedPositionalEmbeddings(Module):
 
 class ClassificationHead(Module):
     """
-    <a id="ClassificationHead">
+    <a id="ClassificationHead"></a>
     ## MLP Classification Head
-    </a>
 
     This is the two layer MLP head to classify the image based on `[CLS]` token embedding.
     """

labml_nn/utils/tokenizer.py

@@ -5,9 +5,8 @@ from labml.configs import BaseConfigs, option
 
 class TokenizerConfigs(BaseConfigs):
     """
-    <a id="OptimizerConfigs">
-    ## Optimizer Configurations
-    </a>
+    <a id="TokenizerConfigs"></a>
+    ## Tokenizer Configurations
     """
 
     tokenizer: Callable = 'character'