File size: 7,113 Bytes
671ce97
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import sys
import traceback
import os

print("🔮 Deep-NanoGPT Inference Script")

try:
    import torch
    import torch.nn as nn
    from torch.nn import functional as F
    import requests

    # --- Config (must match training) ---
    block_size = 256
    n_embd = 128
    n_head = 4
    n_layer = 72
    dropout = 0.1
    device = 'cuda' if torch.cuda.is_available() else 'cpu'
    
    # --- Storage ---
    storage_dir = "/home/user/app/storage/deep_experiment_v2"
    ckpt_path_a = os.path.join(storage_dir, 'ckpt_a.pt')
    ckpt_path_b = os.path.join(storage_dir, 'ckpt_b.pt')

    # --- Vocab (rebuild from data) ---
    url = 'https://raw.githubusercontent.com/karpathy/char-rnn/master/data/tinyshakespeare/input.txt'
    data = requests.get(url).text
    chars = sorted(list(set(data)))
    vocab_size = len(chars)
    stoi = { ch:i for i,ch in enumerate(chars) }
    itos = { i:ch for i,ch in enumerate(chars) }
    encode = lambda s: [stoi.get(c, 0) for c in s]
    decode = lambda l: ''.join([itos[i] for i in l])

    # --- Model Classes ---
    class Head(nn.Module):
        def __init__(self, head_size):
            super().__init__()
            self.key = nn.Linear(n_embd, head_size, bias=False)
            self.query = nn.Linear(n_embd, head_size, bias=False)
            self.value = nn.Linear(n_embd, head_size, bias=False)
            self.register_buffer('tril', torch.tril(torch.ones(block_size, block_size)))
            self.dropout = nn.Dropout(dropout)

        def forward(self, x):
            B,T,C = x.shape
            k = self.key(x)   
            q = self.query(x) 
            wei = q @ k.transpose(-2, -1) * C**-0.5
            wei = wei.masked_fill(self.tril[:T, :T] == 0, float('-inf'))
            wei = F.softmax(wei, dim=-1)
            wei = self.dropout(wei)
            v = self.value(x)
            return wei @ v

    class MultiHeadAttention(nn.Module):
        def __init__(self, num_heads, head_size):
            super().__init__()
            self.heads = nn.ModuleList([Head(head_size) for _ in range(num_heads)])
            self.proj = nn.Linear(n_embd, n_embd)
            self.dropout = nn.Dropout(dropout)

        def forward(self, x):
            out = torch.cat([h(x) for h in self.heads], dim=-1)
            return self.dropout(self.proj(out))

    class FeedForward(nn.Module):
        def __init__(self, n_embd):
            super().__init__()
            self.net = nn.Sequential(
                nn.Linear(n_embd, 4 * n_embd),
                nn.ReLU(),
                nn.Linear(4 * n_embd, n_embd),
                nn.Dropout(dropout),
            )
        def forward(self, x):
            return self.net(x)

    class BlockStandard(nn.Module):
        def __init__(self, n_embd, n_head):
            super().__init__()
            head_size = n_embd // n_head
            self.sa = MultiHeadAttention(n_head, head_size)
            self.ffwd = FeedForward(n_embd)
            self.ln1 = nn.LayerNorm(n_embd)
            self.ln2 = nn.LayerNorm(n_embd)

        def forward(self, x):
            x = x + self.sa(self.ln1(x))
            x = x + self.ffwd(self.ln2(x))
            return x

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

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

        def forward(self, x):
            return self._norm(x.float()).type_as(x) * self.weight

    class BlockMHC(nn.Module):
        def __init__(self, n_embd, n_head):
            super().__init__()
            head_size = n_embd // n_head
            self.sa = MultiHeadAttention(n_head, head_size)
            self.ffwd = FeedForward(n_embd)
            self.alpha1 = nn.Parameter(torch.tensor(0.9))
            self.beta1 = nn.Parameter(torch.tensor(0.1))
            self.ln1 = RMSNorm(n_embd)
            self.alpha2 = nn.Parameter(torch.tensor(0.9))
            self.beta2 = nn.Parameter(torch.tensor(0.1))
            self.ln2 = RMSNorm(n_embd)

        def forward(self, x):
            mix1 = self.alpha1 * x + self.beta1 * self.sa(x)
            x = self.ln1(mix1)
            mix2 = self.alpha2 * x + self.beta2 * self.ffwd(x)
            x = self.ln2(mix2)
            return x

    class GPT(nn.Module):
        def __init__(self, arch_type='standard'):
            super().__init__()
            self.arch_type = arch_type
            self.token_embedding_table = nn.Embedding(vocab_size, n_embd)
            self.position_embedding_table = nn.Embedding(block_size, n_embd)
            if arch_type == 'standard':
                self.blocks = nn.Sequential(*[BlockStandard(n_embd, n_head) for _ in range(n_layer)])
                self.ln_f = nn.LayerNorm(n_embd)
            elif arch_type == 'mhc':
                self.blocks = nn.Sequential(*[BlockMHC(n_embd, n_head) for _ in range(n_layer)])
                self.ln_f = RMSNorm(n_embd)
            self.lm_head = nn.Linear(n_embd, vocab_size)

        def forward(self, idx, targets=None):
            B, T = idx.shape
            tok_emb = self.token_embedding_table(idx)
            pos_emb = self.position_embedding_table(torch.arange(T, device=device))
            x = tok_emb + pos_emb
            x = self.blocks(x)
            x = self.ln_f(x)
            logits = self.lm_head(x)
            return logits, None
            
        def generate(self, idx, max_new_tokens):
            for _ in range(max_new_tokens):
                idx_cond = idx[:, -block_size:]
                logits, _ = self(idx_cond)
                logits = logits[:, -1, :]
                probs = F.softmax(logits, dim=-1)
                idx_next = torch.multinomial(probs, num_samples=1)
                idx = torch.cat((idx, idx_next), dim=1)
            return idx

    # --- Load Models ---
    print(f"📦 Loading Model A (Standard) from {ckpt_path_a}...")
    model_a = GPT(arch_type='standard').to(device)
    model_a.load_state_dict(torch.load(ckpt_path_a, map_location=device))
    model_a.eval()

    print(f"📦 Loading Model B (mHC) from {ckpt_path_b}...")
    model_b = GPT(arch_type='mhc').to(device)
    model_b.load_state_dict(torch.load(ckpt_path_b, map_location=device))
    model_b.eval()

    # --- Inference ---
    PROMPT = "ROMEO:"  # Shakespearean prompt
    MAX_TOKENS = 300

    print(f"\n🎭 Prompt: '{PROMPT}'")
    print(f"🔢 Max Tokens: {MAX_TOKENS}")

    context = torch.tensor([encode(PROMPT)], dtype=torch.long, device=device)

    print("\n--- MODEL A (Standard GPT, 72 Layers) ---")
    with torch.no_grad():
        out_a = model_a.generate(context.clone(), max_new_tokens=MAX_TOKENS)
    print(decode(out_a[0].tolist()))

    print("\n--- MODEL B (mHC GPT, 72 Layers) ---")
    with torch.no_grad():
        out_b = model_b.generate(context.clone(), max_new_tokens=MAX_TOKENS)
    print(decode(out_b[0].tolist()))

    print("\n✅ Inference Complete.")

except Exception as e:
    print(f"\n❌ FATAL ERROR: {e}")
    traceback.print_exc()