ztensor 1.2.3

Unified, zero-copy, and safe I/O for deep learning formats
Documentation 
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

import os

import config
from datagen import generate_tensor_dict
from runners import run_single_benchmark
from plot import draw_plot


def run_sweep():
    import pandas as pd

    sizes = [128, 512, 1024, 2048]  # MB
    distributions = ["mixed", "small", "large"]
    formats = config.NATIVE_FORMATS  # Only native formats in main sweep

    results = []

    os.makedirs("bench_out", exist_ok=True)
    os.makedirs("bench_out/plots", exist_ok=True)

    # Define sweep matrix: (data_style, scenario, distributions, sizes, formats)
    sweep_configs = [
        # Original sweep: random data, full load, all distributions and sizes
        {
            "data_style": "random",
            "scenario": "full_load",
            "distributions": distributions,
            "sizes": sizes,
            "formats": formats,
        },
        # Structured data: shows realistic compression ratios
        {
            "data_style": "structured",
            "scenario": "full_load",
            "distributions": ["mixed"],
            "sizes": [512, 1024, 2048],
            "formats": formats,
        },
        # Zero-copy: ztensor's killer feature for inference
        {
            "data_style": "random",
            "scenario": "zero_copy",
            "distributions": ["mixed"],
            "sizes": [512, 1024, 2048],
            "formats": ["ztensor"],  # only ztensor supports zero-copy
        },
        # Selective loading: load 10% of tensors
        {
            "data_style": "random",
            "scenario": "selective",
            "distributions": ["mixed"],
            "sizes": [512, 1024, 2048],
            "formats": formats,
        },
        # Cross-format: ztensor reading other formats
        {
            "data_style": "random",
            "scenario": "full_load",
            "distributions": ["mixed"],
            "sizes": sizes,
            "formats": config.CROSS_READ_FORMATS,
        },
        # Realistic workload: Llama 1B shapes (fixed size, ignores sizes param)
        {
            "data_style": "random",
            "scenario": "full_load",
            "distributions": ["llama-1b"],
            "sizes": [0],  # ignored, llama-1b generates its own fixed size
            "formats": formats,
        },
        # Zstd compression level sweep
        {
            "data_style": "random",
            "scenario": "full_load",
            "distributions": ["mixed"],
            "sizes": [512],
            "formats": ["ztensor"] + config.ZSTD_FORMATS,
        },
        {
            "data_style": "structured",
            "scenario": "full_load",
            "distributions": ["mixed"],
            "sizes": [512],
            "formats": ["ztensor"] + config.ZSTD_FORMATS,
        },
    ]

    print(f"=== Benchmark Sweep ===")
    print(f"Runs per data point: {config.BENCH_RUNS} (+ {config.WARMUP_RUNS} warmup)")
    print()

    for cfg in sweep_configs:
        data_style = cfg["data_style"]
        scenario = cfg["scenario"]

        print(f"--- {scenario} / {data_style} ---")
        print(
            f"  Sizes: {cfg['sizes']}, Dists: {cfg['distributions']}, "
            f"Formats: {cfg['formats']}"
        )
        print()

        for dist in cfg["distributions"]:
            for size_mb in cfg["sizes"]:
                tensors = generate_tensor_dict(size_mb, dist, data_style)
                tensor_count = len(tensors)

                for fmt in cfg["formats"]:
                    filepath = config.filepath_for_format(fmt, "bench_out", "sweep")

                    result = run_single_benchmark(
                        fmt,
                        tensors,
                        filepath,
                        scenario=scenario,
                    )

                    if result is None:
                        continue

                    # Clean up
                    if os.path.exists(filepath):
                        os.remove(filepath)

                    w_speed = result["write_median"]
                    r_lat = result["read_median"]
                    size_gb = result["file_size_gb"]

                    print(
                        f"  {fmt:<16} | {size_mb:>4}MB ({dist:<5}) | "
                        f"W: {w_speed:.2f} GB/s | R: {r_lat:.4f}s | "
                        f"File: {size_gb*1024:.1f}MB | Tensors: {tensor_count} "
                        f"| {scenario}/{data_style}"
                    )

                    results.append(
                        {
                            "Format": fmt,
                            "SizeMB": size_mb,
                            "Distribution": dist,
                            "DataStyle": data_style,
                            "Scenario": scenario,
                            "WriteGBs": w_speed,
                            "ReadSeconds": r_lat,
                            "FileSizeGB": size_gb,
                            "DataSizeGB": result["data_size_gb"],
                            "TensorCount": tensor_count,
                        }
                    )

                print()

    # Save CSV
    df = pd.DataFrame(results)
    df.to_csv("bench_out/sweep_results.csv", index=False)
    print("Sweep complete. Results saved to bench_out/sweep_results.csv")
    draw_plot("bench_out/sweep_results.csv")