File size: 6,235 Bytes
bb7acef
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
import math
import torch
import torch.nn as nn
from transformers import BertConfig
from torch.utils.checkpoint import checkpoint

class ConvBlock(nn.Module):
    def __init__(self, hidden_size, kernel_size=3, padding=1):
        super().__init__()
        self.conv_dw = nn.Conv1d(
            in_channels=hidden_size,
            out_channels=hidden_size,
            kernel_size=kernel_size,
            padding=padding,
            groups=hidden_size
        )
        self.conv_pw = nn.Conv1d(
            in_channels=hidden_size,
            out_channels=hidden_size,
            kernel_size=1
        )
        self.norm1 = nn.LayerNorm(hidden_size)
        self.ffn = nn.Sequential(
            nn.Linear(hidden_size, hidden_size * 4),
            nn.GELU(),
            nn.Linear(hidden_size * 4, hidden_size)
        )
        self.norm2 = nn.LayerNorm(hidden_size)
        self.dropout = nn.Dropout(0.1)

    def forward(self, x):
        residual = x
        x_conv = x.transpose(1, 2)
        x_conv = self.conv_dw(x_conv)
        x_conv = self.conv_pw(x_conv)
        x_conv = x_conv.transpose(1, 2)
        x = self.norm1(residual + self.dropout(x_conv))
        residual = x
        x_ffn = self.ffn(x)
        x = self.norm2(residual + self.dropout(x_ffn))
        return x


class AttentionBlock(nn.Module):
    def __init__(self, hidden_size, num_heads):
        super().__init__()
        self.self_attn = nn.MultiheadAttention(
            embed_dim=hidden_size,
            num_heads=num_heads,
            dropout=0.1,
            batch_first=True
        )
        self.norm1 = nn.LayerNorm(hidden_size)
        self.ffn = nn.Sequential(
            nn.Linear(hidden_size, hidden_size * 4),
            nn.GELU(),
            nn.Linear(hidden_size * 4, hidden_size)
        )
        self.norm2 = nn.LayerNorm(hidden_size)
        self.dropout = nn.Dropout(0.1)

    def forward(self, x, attention_mask=None):
        residual = x
        if attention_mask is not None:
            key_padding_mask = (attention_mask == 0)
        else:
            key_padding_mask = None
            
        attn_output, _ = self.self_attn(
            query=x,
            key=x,
            value=x,
            key_padding_mask=key_padding_mask,
            need_weights=False
        )
        x = self.norm1(residual + self.dropout(attn_output))
        residual = x
        x_ffn = self.ffn(x)
        x = self.norm2(residual + self.dropout(x_ffn))
        return x


class HCAEModel(nn.Module):
    def __init__(self, vocab_size=30522, hidden_size=384, max_seq_len=512, 
                 conv_layers=5, attn_layers=3, num_heads=12):
        super().__init__()
        self.vocab_size = vocab_size
        self.hidden_size = hidden_size
        self.word_embeddings = nn.Embedding(vocab_size, hidden_size, padding_idx=0)
        self.position_embeddings = nn.Embedding(max_seq_len, hidden_size)
        self.LayerNorm = nn.LayerNorm(hidden_size)
        self.dropout = nn.Dropout(0.1)
        self.conv_blocks = nn.ModuleList([
            ConvBlock(hidden_size) for _ in range(conv_layers)
        ])
        self.attn_blocks = nn.ModuleList([
            AttentionBlock(hidden_size, num_heads) for _ in range(attn_layers)
        ])
        self.use_gradient_checkpointing = False

    def forward(self, input_ids, attention_mask=None):
        seq_length = input_ids.size(1)
        position_ids = torch.arange(seq_length, dtype=torch.long, device=input_ids.device)
        position_ids = position_ids.unsqueeze(0).expand_as(input_ids)
        
        words_embeddings = self.word_embeddings(input_ids)
        position_embeddings = self.position_embeddings(position_ids)
        x = words_embeddings + position_embeddings
        x = self.LayerNorm(x)
        x = self.dropout(x)
        
        for i, block in enumerate(self.conv_blocks):
            if self.use_gradient_checkpointing and self.training:
                def create_custom_forward(module):
                    def custom_forward(*args):
                        return module(*args)
                    return custom_forward
                x = checkpoint(create_custom_forward(block), x, use_reentrant=False)
            else:
                x = block(x)
                
        for i, block in enumerate(self.attn_blocks):
            if self.use_gradient_checkpointing and self.training:
                def create_custom_forward(module):
                    def custom_forward(hidden_states, mask):
                        return module(hidden_states, attention_mask=mask)
                    return custom_forward
                x = checkpoint(create_custom_forward(block), x, attention_mask, use_reentrant=False)
            else:
                x = block(x, attention_mask=attention_mask)
        
        if attention_mask is not None:
            input_mask_expanded = attention_mask.unsqueeze(-1).expand(x.size()).float()
            sum_embeddings = torch.sum(x * input_mask_expanded, 1)
            sum_mask = torch.clamp(input_mask_expanded.sum(1), min=1e-9)
            sentence_embeddings = sum_embeddings / sum_mask
        else:
            sentence_embeddings = x.mean(dim=1)
            
        return sentence_embeddings

if __name__ == "__main__":
    model = HCAEModel()
    total_params = sum(p.numel() for p in model.parameters())
    print(f"Total parameters: {total_params / 1e6:.2f} M")
    
    batch_size = 32
    seq_len = 128
    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
    model.to(device)
    
    dummy_input = torch.randint(0, 30522, (batch_size, seq_len)).to(device)
    dummy_mask = torch.ones((batch_size, seq_len)).to(device)
    
    model.use_gradient_checkpointing = True
    
    with torch.cuda.amp.autocast(dtype=torch.float16):
        output = model(dummy_input, attention_mask=dummy_mask)
        
    print(f"Output shape: {output.shape}")
    
    if torch.cuda.is_available():
        memory_allocated = torch.cuda.memory_allocated(device) / (1024 ** 2)
        memory_reserved = torch.cuda.memory_reserved(device) / (1024 ** 2)
        print(f"CUDA memory allocated: {memory_allocated:.2f} MB")
        print(f"CUDA memory reserved: {memory_reserved:.2f} MB")