File size: 7,927 Bytes
b88c26d | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 | #!/usr/bin/env python3
"""
Benchmark SmoothLeaky activation against standard activations.
License: BSL-1.1
Contact: jessica@collectivekitty.com
SmoothLeaky activation:
f(x) = alpha * x + (1 - alpha) * x * sigmoid(x)
where alpha = 0.01 (default)
This combines the non-dying gradient property of LeakyReLU with the
smooth, differentiable nature of SiLU/Swish. Benchmarks forward pass,
backward pass, and peak memory usage against standard activations.
"""
import argparse
import sys
import time
from dataclasses import dataclass
from typing import Callable, List
import torch
import torch.nn as nn
import torch.nn.functional as F
class SmoothLeaky(nn.Module):
"""SmoothLeaky activation: alpha * x + (1-alpha) * x * sigmoid(x).
Combines LeakyReLU's non-dying gradients with SiLU's smoothness.
At alpha=0, reduces to SiLU/Swish.
At alpha=1, reduces to identity.
"""
def __init__(self, alpha: float = 0.01):
super().__init__()
self.alpha = alpha
def forward(self, x: torch.Tensor) -> torch.Tensor:
return self.alpha * x + (1.0 - self.alpha) * x * torch.sigmoid(x)
def extra_repr(self) -> str:
return f"alpha={self.alpha}"
class Mish(nn.Module):
"""Mish activation: x * tanh(softplus(x))."""
def forward(self, x: torch.Tensor) -> torch.Tensor:
return x * torch.tanh(F.softplus(x))
@dataclass
class BenchmarkResult:
"""Result of benchmarking a single activation function."""
name: str
forward_time_ms: float
backward_time_ms: float
peak_memory_mb: float
forward_std_ms: float
backward_std_ms: float
def benchmark_activation(
activation: nn.Module,
name: str,
batch_size: int,
dim: int,
num_iters: int,
device: torch.device,
warmup_iters: int = 50,
) -> BenchmarkResult:
"""Benchmark a single activation function."""
# Warmup
for _ in range(warmup_iters):
x = torch.randn(batch_size, dim, device=device, requires_grad=True)
y = activation(x)
loss = y.sum()
loss.backward()
if device.type == "cuda":
torch.cuda.synchronize()
torch.cuda.reset_peak_memory_stats(device)
# Benchmark forward pass
forward_times = []
for _ in range(num_iters):
x = torch.randn(batch_size, dim, device=device, requires_grad=True)
if device.type == "cuda":
torch.cuda.synchronize()
start = time.perf_counter()
y = activation(x)
if device.type == "cuda":
torch.cuda.synchronize()
end = time.perf_counter()
forward_times.append((end - start) * 1000.0) # ms
# Clean up
del y
del x
# Benchmark backward pass
backward_times = []
for _ in range(num_iters):
x = torch.randn(batch_size, dim, device=device, requires_grad=True)
y = activation(x)
loss = y.sum()
if device.type == "cuda":
torch.cuda.synchronize()
start = time.perf_counter()
loss.backward()
if device.type == "cuda":
torch.cuda.synchronize()
end = time.perf_counter()
backward_times.append((end - start) * 1000.0) # ms
del x, y, loss
# Measure peak memory
if device.type == "cuda":
torch.cuda.reset_peak_memory_stats(device)
x = torch.randn(batch_size, dim, device=device, requires_grad=True)
y = activation(x)
loss = y.sum()
loss.backward()
torch.cuda.synchronize()
peak_memory_mb = torch.cuda.max_memory_allocated(device) / (1024 * 1024)
del x, y, loss
else:
# On CPU, we estimate memory based on tensor sizes
peak_memory_mb = (batch_size * dim * 4 * 3) / (1024 * 1024) # input + output + grad (float32)
forward_times_t = torch.tensor(forward_times)
backward_times_t = torch.tensor(backward_times)
return BenchmarkResult(
name=name,
forward_time_ms=forward_times_t.mean().item(),
backward_time_ms=backward_times_t.mean().item(),
peak_memory_mb=peak_memory_mb,
forward_std_ms=forward_times_t.std().item(),
backward_std_ms=backward_times_t.std().item(),
)
def main():
parser = argparse.ArgumentParser(
description="Benchmark SmoothLeaky activation against standard activations."
)
parser.add_argument("--batch-size", type=int, default=1024, help="Batch size (default: 1024)")
parser.add_argument("--dim", type=int, default=256, help="Hidden dimension (default: 256)")
parser.add_argument("--num-iters", type=int, default=1000, help="Number of iterations (default: 1000)")
parser.add_argument("--device", type=str, default="cuda:0", help="Device (default: cuda:0)")
parser.add_argument("--alpha", type=float, default=0.01, help="SmoothLeaky alpha (default: 0.01)")
args = parser.parse_args()
# Check device availability
if "cuda" in args.device and not torch.cuda.is_available():
print(f"CUDA not available, falling back to CPU.")
args.device = "cpu"
device = torch.device(args.device)
print("BURT-IMMA Activation Benchmark")
print("=" * 80)
print(f" Batch size: {args.batch_size}")
print(f" Dimension: {args.dim}")
print(f" Iterations: {args.num_iters}")
print(f" Device: {device}")
print(f" Alpha: {args.alpha}")
if device.type == "cuda":
print(f" GPU: {torch.cuda.get_device_name(device)}")
print("=" * 80)
print()
# Define activations to benchmark
activations = [
(SmoothLeaky(alpha=args.alpha), "SmoothLeaky (ours)"),
(nn.ReLU(), "ReLU"),
(nn.LeakyReLU(negative_slope=0.01), "LeakyReLU"),
(nn.GELU(), "GELU"),
(nn.SiLU(), "SiLU/Swish"),
(Mish(), "Mish"),
]
results: List[BenchmarkResult] = []
for activation, name in activations:
activation = activation.to(device)
print(f" Benchmarking {name}...", end="", flush=True)
result = benchmark_activation(
activation=activation,
name=name,
batch_size=args.batch_size,
dim=args.dim,
num_iters=args.num_iters,
device=device,
)
results.append(result)
print(f" done ({result.forward_time_ms:.4f} ms fwd)")
print()
print("=" * 80)
print(f"{'Activation':<20} {'Fwd (ms)':<12} {'Bwd (ms)':<12} {'Mem (MB)':<12} {'Fwd Speedup':<12}")
print("-" * 80)
# Use SmoothLeaky as baseline for speedup comparison
baseline_fwd = results[0].forward_time_ms
for r in results:
speedup = baseline_fwd / r.forward_time_ms if r.forward_time_ms > 0 else float("inf")
print(
f"{r.name:<20} "
f"{r.forward_time_ms:<12.4f} "
f"{r.backward_time_ms:<12.4f} "
f"{r.peak_memory_mb:<12.2f} "
f"{speedup:<12.2f}x"
)
print("-" * 80)
print()
# Detailed statistics
print("Detailed Statistics (mean +/- std):")
print(f"{'Activation':<20} {'Forward':<24} {'Backward':<24}")
print("-" * 68)
for r in results:
fwd_str = f"{r.forward_time_ms:.4f} +/- {r.forward_std_ms:.4f} ms"
bwd_str = f"{r.backward_time_ms:.4f} +/- {r.backward_std_ms:.4f} ms"
print(f"{r.name:<20} {fwd_str:<24} {bwd_str:<24}")
print()
print("Notes:")
print(" - Speedup > 1.0 means faster than SmoothLeaky")
print(" - Speedup < 1.0 means slower than SmoothLeaky")
print(f" - SmoothLeaky(alpha={args.alpha}): f(x) = {args.alpha}*x + {1-args.alpha:.2f}*x*sigmoid(x)")
if __name__ == "__main__":
main()
|