[Misc] Use logger instead of print() in utils/common.py (#29004)
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
co-authored by
Claude Fable 5
parent
10e0bcd622
commit
7430c56b20
@@ -566,9 +566,11 @@ def get_available_gpu_memory(
|
||||
assert gpu_id < num_gpus
|
||||
|
||||
if torch.cuda.current_device() != gpu_id:
|
||||
print(
|
||||
f"WARNING: current device is not {gpu_id}, but {torch.cuda.current_device()}, ",
|
||||
"which may cause useless memory allocation for torch CUDA context.",
|
||||
logger.warning(
|
||||
"current device is not %s, but %s, which may cause useless "
|
||||
"memory allocation for torch CUDA context.",
|
||||
gpu_id,
|
||||
torch.cuda.current_device(),
|
||||
)
|
||||
|
||||
if empty_cache:
|
||||
@@ -588,9 +590,11 @@ def get_available_gpu_memory(
|
||||
assert gpu_id < num_gpus
|
||||
|
||||
if torch.xpu.current_device() != gpu_id:
|
||||
print(
|
||||
f"WARNING: current device is not {gpu_id}, but {torch.xpu.current_device()}, ",
|
||||
"which may cause useless memory allocation for torch XPU context.",
|
||||
logger.warning(
|
||||
"current device is not %s, but %s, which may cause useless "
|
||||
"memory allocation for torch XPU context.",
|
||||
gpu_id,
|
||||
torch.xpu.current_device(),
|
||||
)
|
||||
|
||||
if empty_cache:
|
||||
@@ -604,9 +608,11 @@ def get_available_gpu_memory(
|
||||
assert gpu_id < num_gpus
|
||||
|
||||
if torch.hpu.current_device() != gpu_id:
|
||||
print(
|
||||
f"WARNING: current device is not {gpu_id}, but {torch.hpu.current_device()}, ",
|
||||
"which may cause useless memory allocation for torch HPU context.",
|
||||
logger.warning(
|
||||
"current device is not %s, but %s, which may cause useless "
|
||||
"memory allocation for torch HPU context.",
|
||||
gpu_id,
|
||||
torch.hpu.current_device(),
|
||||
)
|
||||
|
||||
free_gpu_memory, total_gpu_memory = torch.hpu.mem_get_info()
|
||||
@@ -621,9 +627,11 @@ def get_available_gpu_memory(
|
||||
assert gpu_id < num_gpus
|
||||
|
||||
if torch.npu.current_device() != gpu_id:
|
||||
print(
|
||||
f"WARNING: current device is not {gpu_id}, but {torch.npu.current_device()}, ",
|
||||
"which may cause useless memory allocation for torch NPU context.",
|
||||
logger.warning(
|
||||
"current device is not %s, but %s, which may cause useless "
|
||||
"memory allocation for torch NPU context.",
|
||||
gpu_id,
|
||||
torch.npu.current_device(),
|
||||
)
|
||||
if empty_cache:
|
||||
empty_device_cache(torch.npu)
|
||||
@@ -642,9 +650,11 @@ def get_available_gpu_memory(
|
||||
assert gpu_id < num_gpus
|
||||
|
||||
if torch.musa.current_device() != gpu_id:
|
||||
print(
|
||||
f"WARNING: current device is not {gpu_id}, but {torch.musa.current_device()}, ",
|
||||
"which may cause useless memory allocation for torch MUSA context.",
|
||||
logger.warning(
|
||||
"current device is not %s, but %s, which may cause useless "
|
||||
"memory allocation for torch MUSA context.",
|
||||
gpu_id,
|
||||
torch.musa.current_device(),
|
||||
)
|
||||
if empty_cache:
|
||||
empty_device_cache(torch.musa)
|
||||
@@ -1539,7 +1549,7 @@ def delete_directory(dirpath):
|
||||
# This will remove the directory and all its contents
|
||||
shutil.rmtree(dirpath)
|
||||
except OSError as e:
|
||||
print(f"Warning: {dirpath} : {e.strerror}")
|
||||
logger.warning("Failed to delete directory %s: %s", dirpath, e.strerror)
|
||||
|
||||
|
||||
# Temporary directory for prometheus multiprocess mode
|
||||
@@ -3848,7 +3858,7 @@ def get_nvidia_driver_version() -> tuple:
|
||||
|
||||
|
||||
@lru_cache(maxsize=1)
|
||||
def get_nvidia_driver_version_str() -> str:
|
||||
def get_nvidia_driver_version_str() -> str | None:
|
||||
"""Return the NVIDIA driver version string, e.g. '595.58.03'.
|
||||
Returns None on failure."""
|
||||
try:
|
||||
@@ -3930,7 +3940,7 @@ def get_device_sm_nvidia_smi():
|
||||
|
||||
except (subprocess.CalledProcessError, FileNotFoundError, ValueError) as e:
|
||||
# Handle cases where nvidia-smi isn't available or output is unexpected
|
||||
print(f"Error getting compute capability: {e}")
|
||||
logger.error("Error getting compute capability: %s", e)
|
||||
return (0, 0) # Default/fallback value
|
||||
|
||||
|
||||
|
||||
Reference in New Issue
Block a user