File size: 2,527 Bytes
36c78b1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import io
import time
import contextlib
from pathlib import Path
import sys
import torch

ROOT = Path(__file__).resolve().parents[1]
if str(ROOT) not in sys.path:
    sys.path.insert(0, str(ROOT))

from progressive_scaleup import progressive_scale_up_text
from unified_workflow import run_workflow
from bit_transformer.bit_io import text_to_bits
from bit_transformer.safety import hil_safe_inference


def capture_run(func, *args, **kwargs):
    buf = io.StringIO()
    start = time.time()
    with contextlib.redirect_stdout(buf):
        result = func(*args, **kwargs)
    duration = time.time() - start
    return result, buf.getvalue(), duration


def main() -> None:
    summary: list[str] = []

    _, log, dur = capture_run(
        progressive_scale_up_text,
        improve_thresh=0.01,
        steps=10,
        width_mult=2.0,
        max_len=64,
        dataset_size=512,
        forward_kwargs={"causal": True},
    )
    summary.append("### Progressive Scale-Up (causal=True)\n")
    summary.append(log.strip())
    summary.append(f"Duration: {dur:.2f}s\n")

    _, log, dur = capture_run(
        progressive_scale_up_text,
        improve_thresh=0.01,
        steps=10,
        width_mult=2.0,
        max_len=64,
        dataset_size=512,
        forward_kwargs={"causal": False},
    )
    summary.append("### Progressive Scale-Up (causal=False)\n")
    summary.append(log.strip())
    summary.append(f"Duration: {dur:.2f}s\n")

    (model, _), log, dur = capture_run(
        run_workflow,
        steps=2,
        max_len=32,
        dataset_size=32,
        plateau_steps=1,
        epochs_per_step=1,
        extra_steps=1,
        diffusion=False,
    )
    bits = text_to_bits("hi")
    tensor = torch.tensor(bits, dtype=torch.long).unsqueeze(0)
    out_bits, _ = hil_safe_inference(model, tensor, c_floor=0.0, s_floor=0.0)
    summary.append("### Unified Workflow (causal=True)\n")
    summary.append(log.strip())
    summary.append(f"Inference on 'hi': {out_bits.squeeze(0).tolist()}\n")
    summary.append(f"Duration: {dur:.2f}s\n")

    (_, _), log, dur = capture_run(
        run_workflow,
        steps=2,
        max_len=32,
        dataset_size=32,
        plateau_steps=1,
        epochs_per_step=1,
        extra_steps=1,
        diffusion=True,
    )
    summary.append("### Unified Workflow (causal=False / Diffusion)\n")
    summary.append(log.strip())
    summary.append(f"Duration: {dur:.2f}s\n")

    report = "\n".join(summary)
    print(report)


if __name__ == "__main__":
    main()