Refined the docs. Simply very minor changes

Author: jrzaurin

pytorch_widedeep/bayesian_models/tabular/bayesian_linear/bayesian_wide.py

@@ -21,9 +21,17 @@ class BayesianWide(BaseBayesianModel):
     pred_dim: int
         size of the ouput tensor containing the predictions
     prior_sigma_1: float, default = 1.0
-        Prior of the sigma parameter for the first of the two Gaussian
-        distributions that will be mixed to produce the prior weight
-        distribution
+        The prior weight distribution is a scaled mixture of two Gaussian
+        densities:
+
+        .. math::
+           \begin{aligned}
+           P(\mathbf{w}) = \prod_{i=j} \pi N (\mathbf{w}_j | 0, \sigma_{1}^{2}) + (1 - \pi) N (\mathbf{w}_j | 0, \sigma_{2}^{2})
+           \end{aligned}
+
+        This is the prior of the sigma parameter for the first of the two
+        Gaussians that will be mixed to produce the prior weight
+        distribution.
     prior_sigma_2: float, default = 0.002
         Prior of the sigma parameter for the second of the two Gaussian
         distributions that will be mixed to produce the prior weight

pytorch_widedeep/bayesian_models/tabular/bayesian_mlp/bayesian_tab_mlp.py

@@ -57,9 +57,17 @@ class BayesianTabMlp(BaseBayesianModel):
         Activation function for the dense layers of the MLP. Currently
         `'tanh'`, `'relu'`, `'leaky_relu'` and `'gelu'` are supported
     prior_sigma_1: float, default = 1.0
-        Prior of the sigma parameter for the first of the two Gaussian
-        distributions that will be mixed to produce the prior weight
-        distribution for each Bayesian linear and embedding layer
+        The prior weight distribution is a scaled mixture of two Gaussian
+        densities:
+
+        .. math::
+           \begin{aligned}
+           P(\mathbf{w}) = \prod_{i=j} \pi N (\mathbf{w}_j | 0, \sigma_{1}^{2}) + (1 - \pi) N (\mathbf{w}_j | 0, \sigma_{2}^{2})
+           \end{aligned}
+
+        This is the prior of the sigma parameter for the first of the two
+        Gaussians that will be mixed to produce the prior weight
+        distribution.
     prior_sigma_2: float, default = 0.002
         Prior of the sigma parameter for the second of the two Gaussian
         distributions that will be mixed to produce the prior weight

pytorch_widedeep/models/image/vision.py

@@ -45,7 +45,7 @@ class Vision(nn.Module):
         List of strings containing the names (or substring within the name) of
         the parameters that will be trained. For example, if we use a
         `'resnet18'` pretrainable model and we set ``trainable_params =
-        ['layer4']`` only the parameters of `'layer4'` of the network(and the
+        ['layer4']`` only the parameters of `'layer4'` of the network (and the
         head, as mentioned before) will be trained. Note that setting this or
         the previous parameter involves some knowledge of the architecture
         used.

pytorch_widedeep/models/tabular/mlp/context_attention_mlp.py

@@ -68,8 +68,8 @@ class ContextAttentionMLP(BaseTabularModelWithAttention):
         Activation function to be applied to the continuous embeddings, if
         any. `'tanh'`, `'relu'`, `'leaky_relu'` and `'gelu'` are supported.
     input_dim: int, default = 32
-        The so-called *dimension of the model*. In general is the number of
-        embeddings used to encode the categorical and/or continuous columns
+        The so-called *dimension of the model*. Is the number of embeddings
+        used to encode the categorical and/or continuous columns
     attn_dropout: float, default = 0.2
         Dropout for each attention block
     with_addnorm: bool = False,

pytorch_widedeep/models/tabular/mlp/self_attention_mlp.py

@@ -67,7 +67,7 @@ class SelfAttentionMLP(BaseTabularModelWithAttention):
         Activation function to be applied to the continuous embeddings, if
         any. `'tanh'`, `'relu'`, `'leaky_relu'` and `'gelu'` are supported.
     input_dim: int, default = 32
-        The so-called *dimension of the model*. In general is the number of
+        The so-called *dimension of the model*. Is the number of
         embeddings used to encode the categorical and/or continuous columns
     attn_dropout: float, default = 0.2
         Dropout for each attention block

pytorch_widedeep/models/tabular/transformers/ft_transformer.py

@@ -73,7 +73,7 @@ class FTTransformer(BaseTabularModelWithAttention):
         (See `Linformer: Self-Attention with Linear Complexity
         <https://arxiv.org/abs/2006.04768>`_ ) The compression factor that
         will be used to reduce the input sequence length. If we denote the
-        resulting sequence length as :math:`k`
+        resulting sequence length as
         :math:`k = int(kv_{compression \space factor} \times s)`
         where :math:`s` is the input sequence length.
     kv_sharing: bool, default = False

pytorch_widedeep/models/tabular/transformers/saint.py

@@ -64,7 +64,7 @@ class SAINT(BaseTabularModelWithAttention):
         Activation function to be applied to the continuous embeddings, if
         any. `'tanh'`, `'relu'`, `'leaky_relu'` and `'gelu'` are supported.
     input_dim: int, default = 32
-        The so-called *dimension of the model*. In general is the number of
+        The so-called *dimension of the model*. Is the number of
         embeddings used to encode the categorical and/or continuous columns
     n_heads: int, default = 8
         Number of attention heads per Transformer block

pytorch_widedeep/models/tabular/transformers/tab_fastformer.py

@@ -66,7 +66,7 @@ class TabFastFormer(BaseTabularModelWithAttention):
         continuous embeddings, if any. `'tanh'`, `'relu'`, `'leaky_relu'` and
         `'gelu'` are supported.
     input_dim: int, default = 32
-        The so-called *dimension of the model*. In general is the number of
+        The so-called *dimension of the model*. Is the number of
         embeddings used to encode the categorical and/or continuous columns
     n_heads: int, default = 8
         Number of attention heads per FastFormer block

pytorch_widedeep/models/tabular/transformers/tab_perceiver.py

@@ -68,8 +68,8 @@ class TabPerceiver(BaseTabularModelWithAttention):
         Activation function to be applied to the continuous embeddings, if
         any. `'tanh'`, `'relu'`, `'leaky_relu'` and `'gelu'` are supported.
     input_dim: int, default = 32
-        The so-called *dimension of the model*. In general, is the number of
-        embeddings used to encode the categorical and/or continuous columns.
+        The so-called *dimension of the model*. Is the number of embeddings
+        used to encode the categorical and/or continuous columns.
     n_cross_attns: int, default = 1
         Number of times each perceiver block will cross attend to the input
         data (i.e. number of cross attention components per perceiver block).

pytorch_widedeep/models/tabular/transformers/tab_transformer.py

@@ -72,7 +72,7 @@ class TabTransformer(BaseTabularModelWithAttention):
         Activation function to be applied to the continuous embeddings, if
         any. `'tanh'`, `'relu'`, `'leaky_relu'` and `'gelu'` are supported.
     input_dim: int, default = 32
-        The so-called *dimension of the model*. In general is the number of
+        The so-called *dimension of the model*. Is the number of
         embeddings used to encode the categorical and/or continuous columns
     n_heads: int, default = 8
         Number of attention heads per Transformer block