Hugging Face's logo

mkd-hossain
/
finetune-model

Model card Files
xet
 Community
finetune-model
File size: 4,773 Bytes
aaf22a7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
# ---------- mkd_core.py ----------
# (์ฃผ์‹ํšŒ์‚ฌ์— ์ผ€์ด๋”” | MKD)
# MKDCore Transformer โ€” Optimized for 10GB GPU
# Author: MKD AI Systems

import warnings
warnings.filterwarnings("ignore", category=FutureWarning)

import torch
from torch import nn
from tokenizers import Tokenizer

# ============================================================
# CONFIGURATION
# ============================================================
SEQ_LEN = 1024       # Context window (tokens)
D_MODEL = 1024       # Hidden dimension
N_HEADS = 16         # Attention heads
N_LAYERS = 12        # Transformer depth
FF_DIM = D_MODEL * 4 # Feed-forward expansion
DROPOUT = 0.1        # Regularization rate


# ============================================================
# ๐Ÿ”ค TOKENIZER WRAPPER
# ============================================================
class MyTokenizer:
    """
    Wrapper for your trained BPE tokenizer (prod_tokenizer.json)
    Adds <|im_start|> and <|im_end|> for chat/sequence training.
    """
    def __init__(self, json_path: str):
        self.tok = Tokenizer.from_file(json_path)

    def encode(self, text: str, add_special: bool = True) -> list[int]:
        if add_special:
            text = "<|im_start|>" + text + "<|im_end|>"
        return self.tok.encode(text).ids

    def decode(self, ids: list[int]) -> str:
        return self.tok.decode(ids, skip_special_tokens=True)

    @property
    def vocab_size(self) -> int:
        return self.tok.get_vocab_size()


# ============================================================
#  MODEL COMPONENTS
# ============================================================
class FeedForward(nn.Module):
    """Feed-Forward Network (FFN) with GELU activation"""
    def __init__(self, dim: int):
        super().__init__()
        self.net = nn.Sequential(
            nn.Linear(dim, FF_DIM),
            nn.GELU(),
            nn.Linear(FF_DIM, dim),
            nn.Dropout(DROPOUT)
        )

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        return self.net(x)


class MultiHeadAttention(nn.Module):
    """Scaled Dot-Product Multi-Head Attention"""
    def __init__(self, dim: int, heads: int):
        super().__init__()
        self.heads = heads
        self.scale = (dim // heads) ** -0.5
        self.qkv = nn.Linear(dim, dim * 3)
        self.proj = nn.Linear(dim, dim)
        self.dropout = nn.Dropout(DROPOUT)

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        B, T, C = x.shape
        qkv = self.qkv(x).chunk(3, dim=-1)
        q, k, v = [t.view(B, T, self.heads, C // self.heads).transpose(1, 2) for t in qkv]

        attn = (q @ k.transpose(-2, -1)) * self.scale
        attn = attn.softmax(dim=-1)
        attn = self.dropout(attn)

        out = (attn @ v).transpose(1, 2).contiguous().view(B, T, C)
        return self.proj(out)


class Block(nn.Module):
    """Transformer block: Attention + FeedForward"""
    def __init__(self, dim: int, heads: int):
        super().__init__()
        self.norm1 = nn.LayerNorm(dim)
        self.attn = MultiHeadAttention(dim, heads)
        self.norm2 = nn.LayerNorm(dim)
        self.ff = FeedForward(dim)

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        x = x + self.attn(self.norm1(x))
        x = x + self.ff(self.norm2(x))
        return x


# ============================================================
# ๐Ÿง  MKDCore TRANSFORMER MODEL
# ============================================================
class MKDCore(nn.Module):
    """MKDCore Transformer Language Model"""
    def __init__(self, vocab_size: int):
        super().__init__()
        self.token_emb = nn.Embedding(vocab_size, D_MODEL)
        self.pos_emb = nn.Embedding(SEQ_LEN, D_MODEL)
        self.drop = nn.Dropout(DROPOUT)

        self.layers = nn.ModuleList([Block(D_MODEL, N_HEADS) for _ in range(N_LAYERS)])
        self.norm = nn.LayerNorm(D_MODEL)
        self.head = nn.Linear(D_MODEL, vocab_size, bias=False)

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        B, T = x.shape
        pos = torch.arange(0, T, device=x.device).unsqueeze(0)
        x = self.token_emb(x) + self.pos_emb(pos)
        x = self.drop(x)

        for layer in self.layers:
            x = layer(x)

        x = self.norm(x)
        logits = self.head(x)
        return logits


# ============================================================
# โœ… TEST (Run only if executed directly)
# ============================================================
if __name__ == "__main__":
    print("๐Ÿ” Testing MKDCore model...")
    vocab_size = 32000
    model = MKDCore(vocab_size=vocab_size)

    dummy = torch.randint(0, vocab_size, (2, 64))  # batch=2, seq=64
    out = model(dummy)
    print(f"โœ… Output shape: {out.shape}")  # [2, 64, 32000]