[refactor] refactor memory constants usage (#31865)
Signed-off-by: Andy Xie <andy.xning@gmail.com>
This commit is contained in:
@@ -14,8 +14,7 @@ from vllm.config.compilation import CUDAGraphMode
|
||||
from vllm.forward_context import set_forward_context
|
||||
from vllm.logger import init_logger
|
||||
from vllm.model_executor.model_loader import get_model_loader
|
||||
from vllm.utils.mem_constants import GiB_bytes
|
||||
from vllm.utils.mem_utils import DeviceMemoryProfiler
|
||||
from vllm.utils.mem_utils import DeviceMemoryProfiler, format_gib
|
||||
from vllm.utils.platform_utils import is_pin_memory_available
|
||||
from vllm.utils.torch_utils import STR_DTYPE_TO_TORCH_DTYPE
|
||||
from vllm.v1.core.sched.output import GrammarOutput, SchedulerOutput
|
||||
@@ -165,8 +164,8 @@ class GPUModelRunner(LoRAModelRunnerMixin, KVConnectorModelRunnerMixin):
|
||||
|
||||
self.model_memory_usage = m.consumed_memory
|
||||
logger.info(
|
||||
"Model loading took %.4f GiB and %.6f seconds",
|
||||
m.consumed_memory / GiB_bytes,
|
||||
"Model loading took %s GiB and %.6f seconds",
|
||||
format_gib(m.consumed_memory),
|
||||
time_after_load - time_before_load,
|
||||
)
|
||||
|
||||
|
||||
@@ -93,8 +93,7 @@ from vllm.tasks import GenerationTask, PoolingTask, SupportedTask
|
||||
from vllm.utils import length_from_prompt_token_ids_or_embeds
|
||||
from vllm.utils.jsontree import json_map_leaves
|
||||
from vllm.utils.math_utils import cdiv, round_up
|
||||
from vllm.utils.mem_constants import GiB_bytes
|
||||
from vllm.utils.mem_utils import DeviceMemoryProfiler
|
||||
from vllm.utils.mem_utils import DeviceMemoryProfiler, format_gib
|
||||
from vllm.utils.nvtx_pytorch_hooks import PytHooks
|
||||
from vllm.utils.platform_utils import is_pin_memory_available
|
||||
from vllm.utils.torch_utils import (
|
||||
@@ -3899,8 +3898,8 @@ class GPUModelRunner(
|
||||
logger.error(combined_msg)
|
||||
raise e
|
||||
logger.info_once(
|
||||
"Model loading took %.4f GiB memory and %.6f seconds",
|
||||
self.model_memory_usage / GiB_bytes,
|
||||
"Model loading took %s GiB memory and %.6f seconds",
|
||||
format_gib(self.model_memory_usage),
|
||||
time_after_load - time_before_load,
|
||||
scope="local",
|
||||
)
|
||||
|
||||
@@ -125,7 +125,7 @@ class Worker(WorkerBase):
|
||||
used_bytes = total - free_bytes_after_sleep
|
||||
assert freed_bytes >= 0, "Memory usage increased after sleeping."
|
||||
logger.info(
|
||||
"Sleep mode freed %f GiB memory, %f GiB memory is still in use.",
|
||||
"Sleep mode freed %s GiB memory, %s GiB memory is still in use.",
|
||||
format_gib(freed_bytes),
|
||||
format_gib(used_bytes),
|
||||
)
|
||||
@@ -342,19 +342,19 @@ class Worker(WorkerBase):
|
||||
|
||||
unrequested_memory = self.init_snapshot.free_memory - self.requested_memory
|
||||
logger.debug(
|
||||
"Initial free memory: %f GiB; Requested memory: %f (util), %f GiB",
|
||||
"Initial free memory: %s GiB; Requested memory: %f (util), %s GiB",
|
||||
format_gib(self.init_snapshot.free_memory),
|
||||
self.cache_config.gpu_memory_utilization,
|
||||
format_gib(self.requested_memory),
|
||||
)
|
||||
logger.debug(
|
||||
"Free memory after profiling: %f GiB (total), %f GiB (within requested)",
|
||||
"Free memory after profiling: %s GiB (total), %s GiB (within requested)",
|
||||
format_gib(free_gpu_memory),
|
||||
format_gib(free_gpu_memory - unrequested_memory),
|
||||
)
|
||||
logger.debug(profile_result)
|
||||
logger.info_once(
|
||||
"Available KV cache memory: %f GiB",
|
||||
"Available KV cache memory: %s GiB",
|
||||
format_gib(self.available_kv_cache_memory_bytes),
|
||||
scope="local",
|
||||
)
|
||||
|
||||
Reference in New Issue
Block a user