File size: 12,775 Bytes
388fd6e | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 | """Lightweight script benchmarks for Gamma SSM variants.
This replaces the notebook-only timing loop for quick local/remote feedback.
It focuses on the model kernels themselves: full-sequence forward, optional
forward+backward, and optional recurrent decode.
"""
from __future__ import annotations
import argparse
from contextlib import nullcontext
import csv
import json
import os
import platform
import subprocess
import sys
import time
from pathlib import Path
from typing import Any, Iterable
import torch
import torch.nn as nn
REPO_ROOT = Path(__file__).resolve().parents[1]
if str(REPO_ROOT) not in sys.path:
sys.path.insert(0, str(REPO_ROOT))
from gamma_space_model import S4TernaryDPLRSSM, SSMGamma, SSMGammaS4
DTYPES = {
"float32": torch.float32,
"fp32": torch.float32,
"float16": torch.float16,
"fp16": torch.float16,
"bfloat16": torch.bfloat16,
"bf16": torch.bfloat16,
}
def parse_int_list(value: str) -> list[int]:
return [int(item.strip()) for item in value.split(",") if item.strip()]
def synchronize(device: torch.device) -> None:
if device.type == "cuda":
torch.cuda.synchronize(device)
def cuda_memory(device: torch.device) -> dict[str, float | None]:
if device.type != "cuda":
return {
"peak_allocated_mb": None,
"peak_reserved_mb": None,
}
return {
"peak_allocated_mb": torch.cuda.max_memory_allocated(device) / (1024**2),
"peak_reserved_mb": torch.cuda.max_memory_reserved(device) / (1024**2),
}
def reset_cuda_memory(device: torch.device) -> None:
if device.type == "cuda":
torch.cuda.reset_peak_memory_stats(device)
def nvidia_smi_snapshot() -> str | None:
try:
completed = subprocess.run(
[
"nvidia-smi",
"--query-gpu=name,memory.used,memory.total,utilization.gpu,utilization.memory,power.draw,temperature.gpu",
"--format=csv,noheader,nounits",
],
check=False,
capture_output=True,
text=True,
timeout=5,
)
except (OSError, subprocess.TimeoutExpired):
return None
if completed.returncode != 0:
return None
return completed.stdout.strip()
def make_model(name: str, d_model: int, hidden_dim: int, rank: int, kernel_mode: str) -> nn.Module:
if name == "baseline":
return SSMGamma(state_dim=d_model, hidden_dim=hidden_dim)
if name == "gamma_s4":
return SSMGammaS4(
state_dim=d_model,
hidden_dim=hidden_dim,
discretization="bilinear",
kernel_mode=kernel_mode,
kernel_threshold=1,
)
if name == "dplr":
return S4TernaryDPLRSSM(
state_dim=d_model,
hidden_dim=hidden_dim,
rank=rank,
kernel_mode=kernel_mode,
kernel_threshold=1,
)
raise ValueError(f"Unknown model '{name}'.")
def model_forward(model: nn.Module, x: torch.Tensor, return_state: bool = False) -> torch.Tensor:
if isinstance(model, SSMGamma):
y, _ = model(x)
else:
y, _ = model(x, return_state=return_state)
return y
def init_state(model: nn.Module, batch_size: int, device: torch.device, dtype: torch.dtype) -> torch.Tensor:
return model.init_state(batch_size=batch_size, device=device, dtype=dtype)
def allocate_cache(
model: nn.Module,
batch_size: int,
seq_len: int,
device: torch.device,
dtype: torch.dtype,
) -> dict[str, torch.Tensor] | None:
if hasattr(model, "allocate_inference_cache"):
return model.allocate_inference_cache(
batch_size=batch_size,
seq_len=seq_len,
device=device,
dtype=dtype,
)
return None
def recurrent_forward(model: nn.Module, x: torch.Tensor) -> torch.Tensor:
batch_size, seq_len, _ = x.shape
state = init_state(model, batch_size=batch_size, device=x.device, dtype=x.dtype)
cache = allocate_cache(model, batch_size=batch_size, seq_len=seq_len, device=x.device, dtype=x.dtype)
outputs = []
for step in range(seq_len):
token = x[:, step, :]
if cache is None:
y, state = model.step(token, state)
else:
y, state = model.step(token, state, cache=cache)
outputs.append(y)
return torch.stack(outputs, dim=1)
def timed_repeats(
fn,
*,
device: torch.device,
warmup: int,
repeats: int,
) -> tuple[float, float]:
for _ in range(warmup):
fn()
synchronize(device)
latencies = []
for _ in range(repeats):
reset_cuda_memory(device)
synchronize(device)
start = time.perf_counter()
fn()
synchronize(device)
latencies.append(time.perf_counter() - start)
mean_s = sum(latencies) / len(latencies)
min_s = min(latencies)
return mean_s, min_s
def benchmark_case(
*,
model_name: str,
batch_size: int,
seq_len: int,
d_model: int,
hidden_dim: int,
rank: int,
kernel_mode: str,
dtype: torch.dtype,
device: torch.device,
warmup: int,
repeats: int,
run_backward: bool,
run_recurrent: bool,
) -> list[dict[str, Any]]:
model = make_model(model_name, d_model=d_model, hidden_dim=hidden_dim, rank=rank, kernel_mode=kernel_mode)
model = model.to(device=device)
model.train()
x = torch.randn(batch_size, seq_len, d_model, device=device, dtype=dtype)
tokens = batch_size * seq_len
autocast_enabled = device.type == "cuda" and dtype in {torch.float16, torch.bfloat16}
autocast_dtype = dtype if dtype in {torch.float16, torch.bfloat16} else torch.float32
rows: list[dict[str, Any]] = []
def autocast_context():
if not autocast_enabled:
return nullcontext()
return torch.autocast(device_type=device.type, dtype=autocast_dtype, enabled=True)
def add_row(mode: str, mean_s: float, min_s: float) -> None:
mem = cuda_memory(device)
rows.append(
{
"model": model_name,
"mode": mode,
"batch_size": batch_size,
"seq_len": seq_len,
"d_model": d_model,
"hidden_dim": hidden_dim,
"rank": rank if model_name == "dplr" else None,
"kernel_mode": kernel_mode,
"dtype": str(dtype).replace("torch.", ""),
"device": str(device),
"mean_ms": mean_s * 1000.0,
"min_ms": min_s * 1000.0,
"tokens_per_s_mean": tokens / max(mean_s, 1e-12),
"tokens_per_s_best": tokens / max(min_s, 1e-12),
**mem,
}
)
def forward_only() -> None:
with torch.no_grad():
with autocast_context():
y = model_forward(model, x, return_state=False)
_ = y.sum()
mean_s, min_s = timed_repeats(forward_only, device=device, warmup=warmup, repeats=repeats)
add_row("forward", mean_s, min_s)
if run_backward:
def forward_backward() -> None:
model.zero_grad(set_to_none=True)
with autocast_context():
y = model_forward(model, x, return_state=False)
loss = y.square().mean()
loss.backward()
mean_s, min_s = timed_repeats(forward_backward, device=device, warmup=warmup, repeats=repeats)
add_row("forward_backward", mean_s, min_s)
if run_recurrent:
model.eval()
def recurrent() -> None:
with torch.no_grad():
y = recurrent_forward(model, x)
_ = y.sum()
mean_s, min_s = timed_repeats(recurrent, device=device, warmup=max(1, warmup // 2), repeats=repeats)
add_row("recurrent", mean_s, min_s)
return rows
def write_outputs(rows: list[dict[str, Any]], output_dir: Path, metadata: dict[str, Any]) -> None:
output_dir.mkdir(parents=True, exist_ok=True)
json_path = output_dir / "ssm_variant_benchmark.json"
csv_path = output_dir / "ssm_variant_benchmark.csv"
json_path.write_text(
json.dumps(
{
"metadata": metadata,
"results": rows,
},
indent=2,
),
encoding="utf-8",
)
fieldnames = list(rows[0].keys()) if rows else []
with csv_path.open("w", newline="", encoding="utf-8") as handle:
writer = csv.DictWriter(handle, fieldnames=fieldnames)
writer.writeheader()
writer.writerows(rows)
print(f"Wrote {json_path}")
print(f"Wrote {csv_path}")
def print_table(rows: Iterable[dict[str, Any]]) -> None:
columns = [
"model",
"mode",
"batch_size",
"seq_len",
"mean_ms",
"tokens_per_s_mean",
"peak_allocated_mb",
]
print("\t".join(columns))
for row in rows:
values = []
for column in columns:
value = row[column]
if isinstance(value, float):
values.append(f"{value:.3f}")
else:
values.append(str(value))
print("\t".join(values))
def main() -> None:
parser = argparse.ArgumentParser(description="Benchmark Gamma SSM variants.")
parser.add_argument("--models", default="dplr,gamma_s4,baseline")
parser.add_argument("--batch-sizes", default="1,4")
parser.add_argument("--seq-lens", default="128,512")
parser.add_argument("--d-model", type=int, default=128)
parser.add_argument("--hidden-dim", type=int, default=256)
parser.add_argument("--rank", type=int, default=1)
parser.add_argument("--kernel-mode", choices=["auto", "conv", "recurrent"], default="conv")
parser.add_argument("--dtype", choices=sorted(DTYPES), default="bf16")
parser.add_argument("--device", default="auto")
parser.add_argument("--warmup", type=int, default=2)
parser.add_argument("--repeats", type=int, default=5)
parser.add_argument("--backward", action="store_true")
parser.add_argument("--recurrent", action="store_true")
parser.add_argument("--output-dir", default=os.environ.get("REPOBRIDGE_OUTPUT_DIR", "output/benchmarks"))
args = parser.parse_args()
if args.device == "auto":
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
else:
device = torch.device(args.device)
dtype = DTYPES[args.dtype]
if device.type != "cuda" and dtype == torch.float16:
raise ValueError("float16 benchmark requires CUDA.")
if device.type == "cuda":
torch.backends.cuda.matmul.allow_tf32 = True
torch.backends.cudnn.allow_tf32 = True
models = [item.strip() for item in args.models.split(",") if item.strip()]
batch_sizes = parse_int_list(args.batch_sizes)
seq_lens = parse_int_list(args.seq_lens)
metadata = {
"python": platform.python_version(),
"platform": platform.platform(),
"torch": torch.__version__,
"cuda_available": torch.cuda.is_available(),
"cuda_device": torch.cuda.get_device_name(device) if device.type == "cuda" else None,
"nvidia_smi_before": nvidia_smi_snapshot(),
"args": vars(args),
}
rows: list[dict[str, Any]] = []
for model_name in models:
for batch_size in batch_sizes:
for seq_len in seq_lens:
print(f"Benchmarking model={model_name} batch={batch_size} seq={seq_len}")
rows.extend(
benchmark_case(
model_name=model_name,
batch_size=batch_size,
seq_len=seq_len,
d_model=args.d_model,
hidden_dim=args.hidden_dim,
rank=args.rank,
kernel_mode=args.kernel_mode,
dtype=dtype,
device=device,
warmup=args.warmup,
repeats=args.repeats,
run_backward=args.backward,
run_recurrent=args.recurrent,
)
)
metadata["nvidia_smi_after"] = nvidia_smi_snapshot()
print_table(rows)
write_outputs(rows, Path(args.output_dir), metadata)
if __name__ == "__main__":
main()
|