File size: 6,801 Bytes
681909f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
201
202
from dataclasses import dataclass

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


@dataclass
class SovynConfig:
    name: str = "SOVYN-120M-Cortex"
    vocab_size: int = 32000
    max_seq_len: int = 1024
    n_layers: int = 12
    hidden_size: int = 768
    n_heads: int = 12
    n_kv_heads: int = 4
    ffn_size: int = 2688
    dropout: float = 0.0
    rope_theta: float = 10000.0
    tie_embeddings: bool = True


class RMSNorm(nn.Module):
    def __init__(self, dim: int, eps: float = 1e-6):
        super().__init__()
        self.weight = nn.Parameter(torch.ones(dim))
        self.eps = eps

    def forward(self, x):
        normed = x * torch.rsqrt(x.pow(2).mean(dim=-1, keepdim=True) + self.eps)
        return normed * self.weight


def precompute_rope(head_dim: int, max_seq_len: int, theta: float):
    inv_freq = 1.0 / (theta ** (torch.arange(0, head_dim, 2).float() / head_dim))
    t = torch.arange(max_seq_len).float()
    freqs = torch.outer(t, inv_freq)
    return torch.cos(freqs), torch.sin(freqs)


def apply_rope(x, cos, sin):
    cos = cos[None, :, None, :]
    sin = sin[None, :, None, :]
    x_even = x[..., 0::2]
    x_odd = x[..., 1::2]
    out = torch.empty_like(x)
    out[..., 0::2] = x_even * cos - x_odd * sin
    out[..., 1::2] = x_even * sin + x_odd * cos
    return out


class Attention(nn.Module):
    def __init__(self, cfg: SovynConfig):
        super().__init__()
        if cfg.n_heads % cfg.n_kv_heads != 0:
            raise ValueError("n_heads must be divisible by n_kv_heads")
        self.n_heads = cfg.n_heads
        self.n_kv_heads = cfg.n_kv_heads
        self.head_dim = cfg.hidden_size // cfg.n_heads
        self.repeat = cfg.n_heads // cfg.n_kv_heads

        kv_dim = cfg.n_kv_heads * self.head_dim
        self.q_proj = nn.Linear(cfg.hidden_size, cfg.hidden_size, bias=False)
        self.k_proj = nn.Linear(cfg.hidden_size, kv_dim, bias=False)
        self.v_proj = nn.Linear(cfg.hidden_size, kv_dim, bias=False)
        self.o_proj = nn.Linear(cfg.hidden_size, cfg.hidden_size, bias=False)
        self.dropout = cfg.dropout

    def forward(self, x, cos, sin):
        bsz, seq_len, hidden = x.shape
        q = self.q_proj(x).view(bsz, seq_len, self.n_heads, self.head_dim)
        k = self.k_proj(x).view(bsz, seq_len, self.n_kv_heads, self.head_dim)
        v = self.v_proj(x).view(bsz, seq_len, self.n_kv_heads, self.head_dim)

        q = apply_rope(q, cos[:seq_len], sin[:seq_len])
        k = apply_rope(k, cos[:seq_len], sin[:seq_len])

        k = k.repeat_interleave(self.repeat, dim=2)
        v = v.repeat_interleave(self.repeat, dim=2)

        q = q.transpose(1, 2)
        k = k.transpose(1, 2)
        v = v.transpose(1, 2)

        y = F.scaled_dot_product_attention(
            q,
            k,
            v,
            attn_mask=None,
            dropout_p=self.dropout if self.training else 0.0,
            is_causal=True,
        )
        y = y.transpose(1, 2).contiguous().view(bsz, seq_len, hidden)
        return self.o_proj(y)


class SwiGLU(nn.Module):
    def __init__(self, cfg: SovynConfig):
        super().__init__()
        self.gate = nn.Linear(cfg.hidden_size, cfg.ffn_size, bias=False)
        self.up = nn.Linear(cfg.hidden_size, cfg.ffn_size, bias=False)
        self.down = nn.Linear(cfg.ffn_size, cfg.hidden_size, bias=False)

    def forward(self, x):
        return self.down(F.silu(self.gate(x)) * self.up(x))


class Block(nn.Module):
    def __init__(self, cfg: SovynConfig):
        super().__init__()
        self.attn_norm = RMSNorm(cfg.hidden_size)
        self.attn = Attention(cfg)
        self.ffn_norm = RMSNorm(cfg.hidden_size)
        self.ffn = SwiGLU(cfg)

    def forward(self, x, cos, sin):
        x = x + self.attn(self.attn_norm(x), cos, sin)
        x = x + self.ffn(self.ffn_norm(x))
        return x


class SovynForCausalLM(nn.Module):
    def __init__(self, cfg: SovynConfig):
        super().__init__()
        self.cfg = cfg
        self.embed = nn.Embedding(cfg.vocab_size, cfg.hidden_size)
        self.blocks = nn.ModuleList([Block(cfg) for _ in range(cfg.n_layers)])
        self.norm = RMSNorm(cfg.hidden_size)
        self.lm_head = nn.Linear(cfg.hidden_size, cfg.vocab_size, bias=False)
        if cfg.tie_embeddings:
            self.lm_head.weight = self.embed.weight

        cos, sin = precompute_rope(
            cfg.hidden_size // cfg.n_heads,
            cfg.max_seq_len,
            cfg.rope_theta,
        )
        self.register_buffer("rope_cos", cos, persistent=False)
        self.register_buffer("rope_sin", sin, persistent=False)
        self.apply(self._init_weights)

    def _init_weights(self, module):
        if isinstance(module, nn.Linear):
            nn.init.normal_(module.weight, mean=0.0, std=0.02)
        elif isinstance(module, nn.Embedding):
            nn.init.normal_(module.weight, mean=0.0, std=0.02)

    def forward(self, input_ids, labels=None):
        if input_ids.size(1) > self.cfg.max_seq_len:
            raise ValueError("Sequence length exceeds max_seq_len")

        x = self.embed(input_ids)
        for block in self.blocks:
            x = block(x, self.rope_cos, self.rope_sin)
        x = self.norm(x)
        logits = self.lm_head(x)

        loss = None
        if labels is not None:
            loss = F.cross_entropy(
                logits.view(-1, logits.size(-1)),
                labels.view(-1),
                ignore_index=-100,
            )
        return {"loss": loss, "logits": logits}

    @torch.no_grad()
    def generate(
        self,
        input_ids,
        max_new_tokens=96,
        temperature=0.8,
        top_k=50,
        eos_id=None,
        stop_ids=None,
        suppress_ids=None,
    ):
        self.eval()
        stop_ids = set(stop_ids or [])
        suppress_ids = list(suppress_ids or [])
        for _ in range(max_new_tokens):
            x = input_ids[:, -self.cfg.max_seq_len :]
            logits = self(x)["logits"][:, -1, :]
            if suppress_ids:
                logits[:, suppress_ids] = -float("inf")
            if temperature <= 0:
                next_id = torch.argmax(logits, dim=-1, keepdim=True)
            else:
                logits = logits / temperature
                if top_k > 0:
                    values, _ = torch.topk(logits, min(top_k, logits.size(-1)))
                    logits[logits < values[:, [-1]]] = -float("inf")
                probs = F.softmax(logits, dim=-1)
                next_id = torch.multinomial(probs, num_samples=1)
            input_ids = torch.cat([input_ids, next_id], dim=1)
            token_id = next_id.item()
            if eos_id is not None and token_id == eos_id:
                break
            if token_id in stop_ids:
                break
        return input_ids