huggingface/text-generation-inference
5
0
Fork 0
mirror of https://github.com/huggingface/text-generation-inference.git synced 2025-09-12 04:44:52 +00:00
Code Issues Packages Projects Releases Wiki Activity

refine get xpu free memory

Browse Source 
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>
This commit is contained in:
Wang, Yi A 2024-06-27 05:18:57 -07:00
parent b53b21c63a
commit 886bfab23d
1 changed files with 8 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
server/text_generation_server/utils/import_utils.py
View File

@ -1,6 +1,7 @@
import torch import torch
from loguru import logger from loguru import logger
import subprocess import subprocess
import os
def is_ipex_available(): def is_ipex_available():
@ -21,10 +22,13 @@ def get_cuda_free_memory(device, memory_fraction):
def get_xpu_free_memory(device, memory_fraction): def get_xpu_free_memory(device, memory_fraction):
total_memory = torch.xpu.get_device_properties(device).total_memory total_memory = torch.xpu.get_device_properties(device).total_memory
device_id = device.index device_id = device.index
query = f"xpu-smi dump -d {device_id} -m 18 -n 1" memory_fraction = float(os.getenv("XPU_MEMORY_FRACTION", "1.0"))
output = subprocess.check_output(query.split()).decode("utf-8").split("\n") free_memory = max(
used_memory = float(output[1].split(",")[-1]) * 1024 * 1024 0,
free_memory = int(total_memory * 0.95 - used_memory) int(
total_memory * 0.9 * memory_fraction - torch.xpu.memory_reserved(device_id)
),
)
return free_memory return free_memory