File size: 6,272 Bytes
a9bd396 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 | # 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
# This file was automatically generated from examples/modular-transformers/modular_from_uppercase_model.py.
# Do NOT edit this file manually as any edits will be overwritten by the generation of
# the file from the modular. If any change should be done, please apply the change to the
# modular_from_uppercase_model.py file directly. One of our CI enforces this.
# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
from collections.abc import Callable
import torch
from torch import nn
from ...activations import ACT2FN
from ...modeling_layers import GradientCheckpointingLayer
from ...modeling_utils import ALL_ATTENTION_FUNCTIONS
from ...processing_utils import Unpack
from ...utils import TransformersKwargs
from .configuration_from_uppercase_model import FromUppercaseModelTextConfig, FromUppercaseModelVisionConfig
def eager_attention_forward(
module: nn.Module,
query: torch.Tensor,
key: torch.Tensor,
value: torch.Tensor,
attention_mask: torch.Tensor | None,
scaling: float,
dropout: float = 0.0,
**kwargs: Unpack[TransformersKwargs],
):
attn_weights = torch.matmul(query, key.transpose(-1, -2)) * scaling
if attention_mask is not None:
attn_weights = attn_weights + attention_mask
attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)
attn_weights = nn.functional.dropout(attn_weights, p=dropout, training=module.training)
attn_output = torch.matmul(attn_weights, value)
attn_output = attn_output.transpose(1, 2).contiguous()
return attn_output, attn_weights
class FromUppercaseModelAttention(nn.Module):
"""Multi-headed attention from 'Attention Is All You Need' paper"""
def __init__(self, config: FromUppercaseModelVisionConfig | FromUppercaseModelTextConfig):
super().__init__()
self.config = config
self.embed_dim = config.hidden_size
self.num_heads = config.num_attention_heads
self.head_dim = self.embed_dim // self.num_heads
if self.head_dim * self.num_heads != self.embed_dim:
raise ValueError(
f"embed_dim must be divisible by num_heads (got `embed_dim`: {self.embed_dim} and `num_heads`:"
f" {self.num_heads})."
)
self.scale = self.head_dim**-0.5
self.dropout = config.attention_dropout
self.is_causal = False
self.k_proj = nn.Linear(self.embed_dim, self.embed_dim)
self.v_proj = nn.Linear(self.embed_dim, self.embed_dim)
self.q_proj = nn.Linear(self.embed_dim, self.embed_dim)
self.out_proj = nn.Linear(self.embed_dim, self.embed_dim)
def forward(
self,
hidden_states: torch.Tensor,
attention_mask: torch.Tensor | None = None,
**kwargs: Unpack[TransformersKwargs],
) -> tuple[torch.Tensor, torch.Tensor | None]:
"""Input shape: Batch x Time x Channel"""
batch_size, seq_length, embed_dim = hidden_states.shape
queries = self.q_proj(hidden_states)
keys = self.k_proj(hidden_states)
values = self.v_proj(hidden_states)
queries = queries.view(batch_size, seq_length, -1, self.head_dim).transpose(1, 2)
keys = keys.view(batch_size, seq_length, -1, self.head_dim).transpose(1, 2)
values = values.view(batch_size, seq_length, -1, self.head_dim).transpose(1, 2)
attention_interface: Callable = eager_attention_forward
if self.config._attn_implementation != "eager":
attention_interface = ALL_ATTENTION_FUNCTIONS[self.config._attn_implementation]
attn_output, attn_weights = attention_interface(
self,
queries,
keys,
values,
attention_mask,
scaling=self.scale,
dropout=0.0 if not self.training else self.dropout,
**kwargs,
)
attn_output = attn_output.reshape(batch_size, seq_length, -1).contiguous()
attn_output = self.out_proj(attn_output)
return attn_output, attn_weights
class FromUppercaseModelMLP(nn.Module):
def __init__(self, config):
super().__init__()
self.config = config
self.activation_fn = ACT2FN[config.hidden_act]
self.fc1 = nn.Linear(config.hidden_size, config.intermediate_size)
self.fc2 = nn.Linear(config.intermediate_size, config.hidden_size)
def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
hidden_states = self.fc1(hidden_states)
hidden_states = self.activation_fn(hidden_states)
hidden_states = self.fc2(hidden_states)
return hidden_states
class FromUppercaseModelEncoderLayer(GradientCheckpointingLayer):
def __init__(self, config: FromUppercaseModelVisionConfig | FromUppercaseModelTextConfig):
super().__init__()
self.embed_dim = config.hidden_size
self.self_attn = FromUppercaseModelAttention(config)
self.layer_norm1 = nn.LayerNorm(self.embed_dim, eps=config.layer_norm_eps)
self.mlp = FromUppercaseModelMLP(config)
self.layer_norm2 = nn.LayerNorm(self.embed_dim, eps=config.layer_norm_eps)
def forward(
self,
hidden_states: torch.Tensor,
attention_mask: torch.Tensor,
**kwargs: Unpack[TransformersKwargs],
) -> torch.FloatTensor:
residual = hidden_states
hidden_states = self.layer_norm1(hidden_states)
hidden_states, _ = self.self_attn(
hidden_states=hidden_states,
attention_mask=attention_mask,
**kwargs,
)
hidden_states = residual + hidden_states
residual = hidden_states
hidden_states = self.layer_norm2(hidden_states)
hidden_states = self.mlp(hidden_states)
hidden_states = residual + hidden_states
return hidden_states
|