File size: 6,601 Bytes
d757a4a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
# use.py - Chat with VDrontMoE model (text continuation)
import torch
import argparse
import torch.nn.functional as F
import os
from transformers import AutoTokenizer

from create import VDrontMoEConfig, VDrontMoEModel

# --- Configuration ---
MODEL_DIR = "./VDrontMoE-2m-5e"
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
TEMPERATURE = 0.4
MAX_NEW_TOKENS = 100
TOP_P = 0.9
TOP_K = 50
REPETITION_PENALTY = 1.1

# --- Load model ---
print(f"Loading model from {MODEL_DIR}...")
tokenizer = AutoTokenizer.from_pretrained(MODEL_DIR)
if tokenizer.pad_token is None:
    tokenizer.pad_token = tokenizer.eos_token

model = VDrontMoEModel.from_pretrained(MODEL_DIR).to(DEVICE)
model.eval()
print(f"Model loaded! Temperature: {TEMPERATURE}")


def generate(prompt: str, max_tokens: int = MAX_NEW_TOKENS) -> str:
    """Continue text from prompt."""

    # Tokenize prompt
    input_ids = tokenizer.encode(prompt, return_tensors="pt").to(DEVICE)

    generated = input_ids.clone()

    with torch.no_grad():
        for _ in range(max_tokens):
            # Truncate if too long
            if generated.shape[1] > 512:
                generated = generated[:, -512:]

            # Forward pass
            outputs = model(generated)
            logits = outputs["logits"][:, -1, :]  # Take last token

            # Temperature
            logits = logits / TEMPERATURE

            # Top-K filter
            if TOP_K > 0:
                top_k_values, _ = torch.topk(logits, min(TOP_K, logits.size(-1)))
                logits[logits < top_k_values[:, -1:]] = float('-inf')

            # Top-P (nucleus) filter
            if TOP_P < 1.0:
                sorted_logits, sorted_indices = torch.sort(logits, descending=True)
                cumulative_probs = torch.cumsum(F.softmax(sorted_logits, dim=-1), dim=-1)

                # Remove tokens with cumulative prob > TOP_P
                sorted_indices_to_remove = cumulative_probs > TOP_P
                sorted_indices_to_remove[:, 1:] = sorted_indices_to_remove[:, :-1].clone()
                sorted_indices_to_remove[:, 0] = False

                indices_to_remove = sorted_indices_to_remove.scatter(1, sorted_indices, sorted_indices_to_remove)
                logits[indices_to_remove] = float('-inf')

            # Repetition penalty
            if REPETITION_PENALTY != 1.0:
                for token_id in set(generated[0].tolist()[-10:]):  # Last 10 tokens
                    if logits[0, token_id] > 0:
                        logits[0, token_id] /= REPETITION_PENALTY
                    else:
                        logits[0, token_id] *= REPETITION_PENALTY

            # Sample next token
            probs = F.softmax(logits, dim=-1)
            next_token = torch.multinomial(probs, num_samples=1)

            # Add to generated
            generated = torch.cat([generated, next_token], dim=-1)

            # Check for end of text
            if next_token.item() == tokenizer.eos_token_id:
                break

    # Decode
    full_text = tokenizer.decode(generated[0], skip_special_tokens=True)
    return full_text


def chat():
    """Interactive text continuation mode."""
    print("\n" + "=" * 60)
    print("VDrontMoE-2m-5e - Text Continuation Mode")
    print("=" * 60)
    print("Commands:")
    print("  /temp <value>  - Set temperature (0.1 - 2.0)")
    print("  /max <value>   - Set max new tokens (10 - 500)")
    print("  /clear         - Clear screen")
    print("  /exit          - Exit")
    print("=" * 60)

    global TEMPERATURE, MAX_NEW_TOKENS

    while True:
        try:
            # Get prompt
            prompt = input("\nPrompt: ").strip()

            # Commands
            if prompt.startswith("/"):
                parts = prompt.split()
                cmd = parts[0]

                if cmd == "/exit":
                    print("Goodbye!")
                    break
                elif cmd == "/clear":
                    os.system('cls' if os.name == 'nt' else 'clear')
                    continue
                elif cmd == "/temp" and len(parts) > 1:
                    TEMPERATURE = float(parts[1])
                    print(f"Temperature set to {TEMPERATURE}")
                    continue
                elif cmd == "/max" and len(parts) > 1:
                    MAX_NEW_TOKENS = int(parts[1])
                    print(f"Max tokens set to {MAX_NEW_TOKENS}")
                    continue
                else:
                    print("Unknown command")
                    continue

            if not prompt:
                print("Please enter a prompt!")
                continue

            # Generate
            print("\nGenerating...")
            result = generate(prompt)

            # Display
            print("\n" + "=" * 60)
            print("GENERATED TEXT:")
            print("=" * 60)
            print(result)
            print("=" * 60)

            # Statistics
            generated_part = result[len(prompt):]
            new_tokens = len(tokenizer.encode(generated_part))
            print(f"Generated {new_tokens} new tokens")

        except KeyboardInterrupt:
            print("\n\nInterrupted. Type /exit to quit.")
        except Exception as e:
            print(f"\nError: {e}")


if __name__ == "__main__":
    parser = argparse.ArgumentParser(description="VDrontMoE Text Generation")
    parser.add_argument("--prompt", type=str, help="Single prompt mode (no chat)")
    parser.add_argument("--temperature", type=float, default=0.4, help="Temperature (default: 0.4)")
    parser.add_argument("--max_tokens", type=int, default=100, help="Max new tokens")
    parser.add_argument("--top_p", type=float, default=0.9, help="Nucleus sampling threshold")
    parser.add_argument("--top_k", type=int, default=50, help="Top-K filtering")
    parser.add_argument("--model_dir", type=str, default="./VDrontMoE-2m-5e")

    args = parser.parse_args()

    # Update parameters
    MODEL_DIR = args.model_dir
    TEMPERATURE = args.temperature
    MAX_NEW_TOKENS = args.max_tokens
    TOP_P = args.top_p
    TOP_K = args.top_k

    if args.prompt:
        # Single mode
        print(f"\nPrompt: {args.prompt}")
        print(f"Temperature: {TEMPERATURE} | Max tokens: {MAX_NEW_TOKENS}")
        print("\n" + "=" * 60)
        result = generate(args.prompt)
        print(result)
        print("=" * 60)
    else:
        # Interactive mode
        chat()