consistent nn. and nn.functional: p2 templates (#12153)

templates/adding_a_new_model/ADD_NEW_MODEL_PROPOSAL_TEMPLATE.md

@@ -711,7 +711,7 @@ defined by the name of the class attribute you give the layer. Let's
 define a dummy model in PyTorch, called `SimpleModel` as follows:
 
 ```python
-import torch.nn as nn
+from torch import nn
 
 class SimpleModel(nn.Module):
     def __init__(self):

templates/adding_a_new_model/cookiecutter-template-{{cookiecutter.modelname}}/modeling_{{cookiecutter.lowercase_modelname}}.py

@@ -1542,7 +1542,6 @@ import random
 from typing import Optional, Tuple
 
 import torch
-import torch.nn.functional as F
 from torch import nn
 from torch.nn import CrossEntropyLoss
 
@@ -1743,7 +1742,7 @@ class {{cookiecutter.camelcase_modelname}}Attention(nn.Module):
             attn_weights = attn_weights.view(bsz, self.num_heads, tgt_len, src_len) + attention_mask
             attn_weights = attn_weights.view(bsz * self.num_heads, tgt_len, src_len)
 
-        attn_weights = F.softmax(attn_weights, dim=-1)
+        attn_weights = nn.functional.softmax(attn_weights, dim=-1)
 
         if layer_head_mask is not None:
             if layer_head_mask.size() != (self.num_heads,):
@@ -1763,7 +1762,7 @@ class {{cookiecutter.camelcase_modelname}}Attention(nn.Module):
         else:
             attn_weights_reshaped = None
 
-        attn_probs = F.dropout(attn_weights, p=self.dropout, training=self.training)
+        attn_probs = nn.functional.dropout(attn_weights, p=self.dropout, training=self.training)
 
         attn_output = torch.bmm(attn_probs, value_states)
 
@@ -1823,15 +1822,15 @@ class {{cookiecutter.camelcase_modelname}}EncoderLayer(nn.Module):
             layer_head_mask=layer_head_mask,
             output_attentions=output_attentions,
         )
-        hidden_states = F.dropout(hidden_states, p=self.dropout, training=self.training)
+        hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training)
         hidden_states = residual + hidden_states
         hidden_states = self.self_attn_layer_norm(hidden_states)
 
         residual = hidden_states
         hidden_states = self.activation_fn(self.fc1(hidden_states))
-        hidden_states = F.dropout(hidden_states, p=self.activation_dropout, training=self.training)
+        hidden_states = nn.functional.dropout(hidden_states, p=self.activation_dropout, training=self.training)
         hidden_states = self.fc2(hidden_states)
-        hidden_states = F.dropout(hidden_states, p=self.dropout, training=self.training)
+        hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training)
         hidden_states = residual + hidden_states
         hidden_states = self.final_layer_norm(hidden_states)
 
@@ -1916,7 +1915,7 @@ class {{cookiecutter.camelcase_modelname}}DecoderLayer(nn.Module):
             layer_head_mask=layer_head_mask,
             output_attentions=output_attentions,
         )
-        hidden_states = F.dropout(hidden_states, p=self.dropout, training=self.training)
+        hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training)
         hidden_states = residual + hidden_states
         hidden_states = self.self_attn_layer_norm(hidden_states)
 
@@ -1936,7 +1935,7 @@ class {{cookiecutter.camelcase_modelname}}DecoderLayer(nn.Module):
                 past_key_value=cross_attn_past_key_value,
                 output_attentions=output_attentions,
             )
-            hidden_states = F.dropout(hidden_states, p=self.dropout, training=self.training)
+            hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training)
             hidden_states = residual + hidden_states
             hidden_states = self.encoder_attn_layer_norm(hidden_states)
 
@@ -1946,9 +1945,9 @@ class {{cookiecutter.camelcase_modelname}}DecoderLayer(nn.Module):
         # Fully Connected
         residual = hidden_states
         hidden_states = self.activation_fn(self.fc1(hidden_states))
-        hidden_states = F.dropout(hidden_states, p=self.activation_dropout, training=self.training)
+        hidden_states = nn.functional.dropout(hidden_states, p=self.activation_dropout, training=self.training)
         hidden_states = self.fc2(hidden_states)
-        hidden_states = F.dropout(hidden_states, p=self.dropout, training=self.training)
+        hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training)
         hidden_states = residual + hidden_states
         hidden_states = self.final_layer_norm(hidden_states)
 
@@ -2171,7 +2170,7 @@ class {{cookiecutter.camelcase_modelname}}Encoder({{cookiecutter.camelcase_model
 
     Args:
         config: {{cookiecutter.camelcase_modelname}}Config
-        embed_tokens (torch.nn.Embedding): output embedding
+        embed_tokens (nn.Embedding): output embedding
     """
 
     def __init__(self, config: {{cookiecutter.camelcase_modelname}}Config, embed_tokens: Optional[nn.Embedding] = None):
@@ -2270,7 +2269,7 @@ class {{cookiecutter.camelcase_modelname}}Encoder({{cookiecutter.camelcase_model
 
         hidden_states = inputs_embeds + embed_pos
         hidden_states = self.layernorm_embedding(hidden_states)
-        hidden_states = F.dropout(hidden_states, p=self.dropout, training=self.training)
+        hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training)
 
         # expand attention_mask
         if attention_mask is not None:
@@ -2337,7 +2336,7 @@ class {{cookiecutter.camelcase_modelname}}Decoder({{cookiecutter.camelcase_model
 
     Args:
         config: {{cookiecutter.camelcase_modelname}}Config
-        embed_tokens (torch.nn.Embedding): output embedding
+        embed_tokens (nn.Embedding): output embedding
     """
 
     def __init__(self, config: {{cookiecutter.camelcase_modelname}}Config, embed_tokens: Optional[nn.Embedding] = None):
@@ -2506,7 +2505,7 @@ class {{cookiecutter.camelcase_modelname}}Decoder({{cookiecutter.camelcase_model
         hidden_states = inputs_embeds + positions
         hidden_states = self.layernorm_embedding(hidden_states)
 
-        hidden_states = F.dropout(hidden_states, p=self.dropout, training=self.training)
+        hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training)
 
         # decoder layers
         all_hidden_states = () if output_hidden_states else None

templates/adding_a_new_model/open_model_proposals/ADD_BIG_BIRD.md

@@ -725,7 +725,7 @@ defined by the name of the class attribute you give the layer. Let's
 define a dummy model in PyTorch, called `SimpleModel` as follows:
 
 ```python
-import torch.nn as nn
+from torch import nn
 
 class SimpleModel(nn.Module):
     def __init__(self):