skip the gpu memory checks if the device is set to 'auto' (#609)
Browse files* skip the gpu memory checks if the device is set to 'auto'
* skip gpu mem logging if cpu too
* don't worry about log_gpu_memory_usage since it calls another annotated fn
* rename decorator internal
- src/axolotl/utils/bench.py +27 -3
src/axolotl/utils/bench.py
CHANGED
|
@@ -1,14 +1,40 @@
|
|
| 1 |
"""Benchmarking and measurement utilities"""
|
|
|
|
| 2 |
|
| 3 |
import pynvml
|
| 4 |
import torch
|
| 5 |
from pynvml.nvml import NVMLError
|
| 6 |
|
| 7 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 8 |
def gpu_memory_usage(device=0):
|
| 9 |
return torch.cuda.memory_allocated(device) / 1024.0**3
|
| 10 |
|
| 11 |
|
|
|
|
| 12 |
def gpu_memory_usage_all(device=0):
|
| 13 |
usage = torch.cuda.memory_allocated(device) / 1024.0**3
|
| 14 |
reserved = torch.cuda.memory_reserved(device) / 1024.0**3
|
|
@@ -16,6 +42,7 @@ def gpu_memory_usage_all(device=0):
|
|
| 16 |
return usage, reserved - usage, max(0, smi - reserved)
|
| 17 |
|
| 18 |
|
|
|
|
| 19 |
def gpu_memory_usage_smi(device=0):
|
| 20 |
if isinstance(device, torch.device):
|
| 21 |
device = device.index
|
|
@@ -31,9 +58,6 @@ def gpu_memory_usage_smi(device=0):
|
|
| 31 |
|
| 32 |
|
| 33 |
def log_gpu_memory_usage(log, msg, device):
|
| 34 |
-
if not torch.cuda.is_available() or device == "auto":
|
| 35 |
-
return (0, 0, 0)
|
| 36 |
-
|
| 37 |
usage, cache, misc = gpu_memory_usage_all(device)
|
| 38 |
extras = []
|
| 39 |
if cache > 0:
|
|
|
|
| 1 |
"""Benchmarking and measurement utilities"""
|
| 2 |
+
import functools
|
| 3 |
|
| 4 |
import pynvml
|
| 5 |
import torch
|
| 6 |
from pynvml.nvml import NVMLError
|
| 7 |
|
| 8 |
|
| 9 |
+
def check_cuda_device(default_value):
|
| 10 |
+
"""
|
| 11 |
+
wraps a function and returns the default value instead of running the
|
| 12 |
+
wrapped function if cuda isn't available or the device is auto
|
| 13 |
+
:param default_value:
|
| 14 |
+
:return:
|
| 15 |
+
"""
|
| 16 |
+
|
| 17 |
+
def deco(func):
|
| 18 |
+
@functools.wraps(func)
|
| 19 |
+
def wrapper(*args, **kwargs):
|
| 20 |
+
device = kwargs.get("device", args[0] if args else None)
|
| 21 |
+
|
| 22 |
+
if not torch.cuda.is_available() or device == "auto" or device == "cpu":
|
| 23 |
+
return default_value
|
| 24 |
+
|
| 25 |
+
return func(*args, **kwargs)
|
| 26 |
+
|
| 27 |
+
return wrapper
|
| 28 |
+
|
| 29 |
+
return deco
|
| 30 |
+
|
| 31 |
+
|
| 32 |
+
@check_cuda_device(0.0)
|
| 33 |
def gpu_memory_usage(device=0):
|
| 34 |
return torch.cuda.memory_allocated(device) / 1024.0**3
|
| 35 |
|
| 36 |
|
| 37 |
+
@check_cuda_device((0.0, 0.0, 0.0))
|
| 38 |
def gpu_memory_usage_all(device=0):
|
| 39 |
usage = torch.cuda.memory_allocated(device) / 1024.0**3
|
| 40 |
reserved = torch.cuda.memory_reserved(device) / 1024.0**3
|
|
|
|
| 42 |
return usage, reserved - usage, max(0, smi - reserved)
|
| 43 |
|
| 44 |
|
| 45 |
+
@check_cuda_device(0.0)
|
| 46 |
def gpu_memory_usage_smi(device=0):
|
| 47 |
if isinstance(device, torch.device):
|
| 48 |
device = device.index
|
|
|
|
| 58 |
|
| 59 |
|
| 60 |
def log_gpu_memory_usage(log, msg, device):
|
|
|
|
|
|
|
|
|
|
| 61 |
usage, cache, misc = gpu_memory_usage_all(device)
|
| 62 |
extras = []
|
| 63 |
if cache > 0:
|