File size: 4,988 Bytes
4f7e31d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200


import math
import torch
import torch.nn as nn
import torch.nn.functional as F

from transformers import PreTrainedModel, PretrainedConfig
from transformers.modeling_outputs import CausalLMOutput


# =========================
# Config
# =========================

class TinyWayConfig(PretrainedConfig):
    model_type = "tinyway"

    def __init__(
        self,
        vocab_size=50257,
        n_positions=256,
        n_embd=512,
        n_layer=10,
        n_head=8,
        dropout=0.1,
        **kwargs
    ):
        super().__init__(**kwargs)

        self.vocab_size = vocab_size
        self.n_positions = n_positions
        self.n_embd = n_embd
        self.n_layer = n_layer
        self.n_head = n_head
        self.dropout = dropout

        # 🔥 HuggingFace-required aliases
        self.hidden_size = n_embd
        self.num_hidden_layers = n_layer
        self.num_attention_heads = n_head
        self.max_position_embeddings = n_positions


# =========================
# Causal Self-Attention
# =========================

class CausalSelfAttention(nn.Module):
    def __init__(self, config):
        super().__init__()
        assert config.n_embd % config.n_head == 0

        self.n_head = config.n_head
        self.head_dim = config.n_embd // config.n_head

        self.qkv = nn.Linear(config.n_embd, 3 * config.n_embd)
        self.proj = nn.Linear(config.n_embd, config.n_embd)

        self.attn_dropout = nn.Dropout(config.dropout)
        self.proj_dropout = nn.Dropout(config.dropout)

        self.register_buffer(
            "mask",
            torch.tril(
                torch.ones(
                    config.n_positions,
                    config.n_positions,
                    dtype=torch.bool
                )
            )
        )

        self.last_attn = None

    def forward(self, x):
        B, T, C = x.shape

        qkv = self.qkv(x)
        q, k, v = qkv.chunk(3, dim=-1)

        q = q.view(B, T, self.n_head, self.head_dim).transpose(1, 2)
        k = k.view(B, T, self.n_head, self.head_dim).transpose(1, 2)
        v = v.view(B, T, self.n_head, self.head_dim).transpose(1, 2)

        att = (q @ k.transpose(-2, -1)) / math.sqrt(self.head_dim)
        att = att.masked_fill(
            ~self.mask[:T, :T],
            torch.finfo(att.dtype).min
        )

        att = F.softmax(att, dim=-1)
        self.last_attn = att.detach()

        att = self.attn_dropout(att)

        out = att @ v
        out = out.transpose(1, 2).contiguous().view(B, T, C)

        out = self.proj(out)
        out = self.proj_dropout(out)

        return out


# =========================
# Transformer Block
# =========================

class Block(nn.Module):
    def __init__(self, config):
        super().__init__()

        self.ln1 = nn.LayerNorm(config.n_embd)
        self.attn = CausalSelfAttention(config)

        self.ln2 = nn.LayerNorm(config.n_embd)

        # 🔥 FFN EXACTLY MATCHES TRAINING
        self.ffn = nn.Sequential(
            nn.Linear(config.n_embd, 4 * config.n_embd),
            nn.GELU(),
            nn.Linear(4 * config.n_embd, config.n_embd),
            nn.Dropout(config.dropout),
        )

    def forward(self, x):
        x = x + self.attn(self.ln1(x))
        x = x + self.ffn(self.ln2(x))
        return x


# =========================
# TinyWay Language Model
# =========================

class TinyWayForCausalLM(PreTrainedModel):
    config_class = TinyWayConfig

    def __init__(self, config):
        super().__init__(config)

        self.token_emb = nn.Embedding(config.vocab_size, config.n_embd)
        self.pos_emb = nn.Embedding(config.n_positions, config.n_embd)

        self.blocks = nn.ModuleList([
            Block(config) for _ in range(config.n_layer)
        ])

        self.ln = nn.LayerNorm(config.n_embd)

        self.head = nn.Linear(
            config.n_embd,
            config.vocab_size,
            bias=False
        )

        # weight tying
        self.head.weight = self.token_emb.weight

        self.dropout = nn.Dropout(config.dropout)

        self.post_init()

    def forward(
        self,
        input_ids,
        labels=None,
        attention_mask=None,  # intentionally unused (causal LM)
        **kwargs              # 🔥 accept return_dict, use_cache, etc.
    ):
        B, T = input_ids.shape
        pos = torch.arange(T, device=input_ids.device)

        x = self.token_emb(input_ids) + self.pos_emb(pos)
        x = self.dropout(x)

        for block in self.blocks:
            x = block(x)

        x = self.ln(x)
        logits = self.head(x)

        loss = None
        if labels is not None:
            loss = F.cross_entropy(
                logits.view(-1, logits.size(-1)),
                labels.view(-1)
            )

        return CausalLMOutput(
            loss=loss,
            logits=logits
        )

    

    def prepare_inputs_for_generation(self, input_ids, **kwargs):
        return {"input_ids": input_ids}