aprender-gpu 0.31.1

Pure Rust PTX generation for NVIDIA CUDA - no LLVM, no nvcc
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
158
159
160
161
162
163
164
165
166
167
168

//! PMAT-018: ops.rs elementwise coverage tests for 95%+ target

use super::*;

#[test]
fn test_ops_softmax_with_stream_warp_path() {
    let ctx = fresh_ctx!();
    let (seq_len, row_size) = (4u32, 16u32);
    let tensor = upload(&ctx, &ramp_f32((seq_len * row_size) as usize, 0.1, 0.0));
    let stream = CudaStream::new(&ctx).unwrap();

    let result = tensor.softmax_with_stream(&ctx, seq_len, &stream).unwrap();
    stream.synchronize().unwrap();
    assert_eq!(result.len(), (seq_len * row_size) as usize);
}

#[test]
fn test_ops_softmax_with_stream_long_row_path() {
    let ctx = fresh_ctx!();
    let (seq_len, row_size) = (4u32, 64u32);
    let tensor = upload(&ctx, &ramp_f32((seq_len * row_size) as usize, 0.1, 0.0));
    let stream = CudaStream::new(&ctx).unwrap();

    let result = tensor.softmax_with_stream(&ctx, seq_len, &stream).unwrap();
    stream.synchronize().unwrap();
    assert_eq!(result.len(), (seq_len * row_size) as usize);
}

#[test]
fn test_ops_softmax_with_stream_dimension_error() {
    let ctx = fresh_ctx!();
    let tensor = upload(&ctx, &vec![1.0; 10]);
    let stream = CudaStream::new(&ctx).unwrap();

    assert!(tensor.softmax_with_stream(&ctx, 3, &stream).is_err());
}

#[test]
fn test_ops_add_with_stream_dimension_error() {
    let ctx = fresh_ctx!();
    let a = upload(&ctx, &vec![1.0f32; 10]);
    let b = upload(&ctx, &vec![1.0f32; 5]);
    let stream = CudaStream::new(&ctx).unwrap();

    assert!(a.add_with_stream(&ctx, &b, &stream).is_err());
}

#[test]
fn test_ops_scale_larger_tensor() {
    let ctx = fresh_ctx!();
    let tensor = upload(&ctx, &ramp_f32(1024, 0.01, 0.0));

    let mut scaled = tensor.scale(&ctx, 0.5).unwrap();
    let result = scaled.to_host().unwrap();

    assert!((result[0] - 0.0).abs() < 1e-5);
    assert!((result[100] - 0.5).abs() < 1e-5);
    assert!((result[1000] - 5.0).abs() < 1e-5);
}

#[test]
fn test_ops_gelu_larger_tensor() {
    let ctx = fresh_ctx!();
    let tensor = upload(&ctx, &ramp_f32(1024, 0.01, -512.0 * 0.01));
    assert_eq!(tensor.gelu(&ctx).unwrap().len(), 1024);
}

#[test]
fn test_ops_layer_norm_larger_batch() {
    let ctx = fresh_ctx!();
    let (hidden, batch) = (32u32, 16u32);

    let input = upload(&ctx, &ramp_f32((hidden * batch) as usize, 0.01, 0.0));
    let gamma = upload(&ctx, &vec![1.0f32; hidden as usize]);
    let beta = upload(&ctx, &vec![0.0f32; hidden as usize]);

    let output = input
        .layer_norm(&ctx, &gamma, &beta, hidden, batch)
        .unwrap();
    assert_eq!(output.len(), (hidden * batch) as usize);
}

#[test]
fn test_ops_bias_add_larger_tensor() {
    let ctx = fresh_ctx!();
    let (n_rows, bias_size) = (256usize, 64usize);

    let input = upload(&ctx, &vec![1.0f32; n_rows * bias_size]);
    let bias = upload(&ctx, &ramp_f32(bias_size, 0.1, 0.0));

    let mut output = input.bias_add(&ctx, &bias).unwrap();
    let result = output.to_host().unwrap();

    assert!((result[0] - 1.0).abs() < 1e-5);
    assert!((result[1] - 1.1).abs() < 1e-5);
    assert!((result[64] - 1.0).abs() < 1e-5);
}

#[test]
fn test_ops_add_larger_tensor() {
    let ctx = fresh_ctx!();
    let size = 1024;

    let a = upload(&ctx, &ramp_f32(size, 1.0, 0.0));
    let b_data: Vec<f32> = (0..size).map(|i| (size - i) as f32).collect();
    let b = upload(&ctx, &b_data);

    let mut c = a.add(&ctx, &b).unwrap();
    let result = c.to_host().unwrap();
    assert!(result.iter().all(|&v| (v - 1024.0).abs() < 1e-3));
}

#[test]
fn test_ops_gelu_with_stream_larger() {
    let ctx = fresh_ctx!();
    let tensor = upload(&ctx, &ramp_f32(2048, 0.001, -1.024));
    let stream = CudaStream::new(&ctx).unwrap();

    let output = tensor.gelu_with_stream(&ctx, &stream).unwrap();
    stream.synchronize().unwrap();
    assert_eq!(output.len(), 2048);
}

#[test]
fn test_ops_layer_norm_with_stream_larger() {
    let ctx = fresh_ctx!();
    let (hidden, batch) = (64u32, 8u32);

    let input = upload(&ctx, &ramp_f32((hidden * batch) as usize, 0.01, 0.0));
    let gamma = upload(&ctx, &vec![1.0f32; hidden as usize]);
    let beta = upload(&ctx, &vec![0.5f32; hidden as usize]);

    let stream = CudaStream::new(&ctx).unwrap();
    let output = input
        .layer_norm_with_stream(&ctx, &gamma, &beta, hidden, batch, &stream)
        .unwrap();
    stream.synchronize().unwrap();
    assert_eq!(output.len(), (hidden * batch) as usize);
}

#[test]
fn test_ops_bias_add_with_stream_larger() {
    let ctx = fresh_ctx!();
    let (n_rows, bias_size) = (128usize, 32usize);

    let input = upload(&ctx, &vec![0.5f32; n_rows * bias_size]);
    let bias = upload(&ctx, &ramp_f32(bias_size, 0.05, 0.0));

    let stream = CudaStream::new(&ctx).unwrap();
    let output = input.bias_add_with_stream(&ctx, &bias, &stream).unwrap();
    stream.synchronize().unwrap();
    assert_eq!(output.len(), n_rows * bias_size);
}

#[test]
fn test_ops_add_with_stream_larger() {
    let ctx = fresh_ctx!();
    let size = 2048;

    let a = upload(&ctx, &ramp_f32(size, 0.001, 0.0));
    let b_data: Vec<f32> = (0..size).map(|i| (size - i) as f32 * 0.001).collect();
    let b = upload(&ctx, &b_data);

    let stream = CudaStream::new(&ctx).unwrap();
    let c = a.add_with_stream(&ctx, &b, &stream).unwrap();
    stream.synchronize().unwrap();
    assert_eq!(c.len(), size);
}