mirror of
https://github.com/huggingface/text-generation-inference.git
synced 2025-04-21 14:52:20 +00:00
26b3916612
This was ignored up to now, even with `--cuda-graphs 0`. With this fix, `--cuda-graphs` is obeyed to.
18 lines
506 B
Python
18 lines
506 B
Python
import torch
|
|
import os
|
|
|
|
MEM_POOL = torch.cuda.graph_pool_handle()
|
|
# This is overridden by the cli
|
|
cuda_graphs = os.getenv("CUDA_GRAPHS")
|
|
if cuda_graphs is not None and cuda_graphs != "0":
|
|
try:
|
|
cuda_graphs = [int(item) for item in cuda_graphs.split(",")]
|
|
except Exception as e:
|
|
raise RuntimeError(
|
|
f"Could not parse cuda graphs {cuda_graphs}, expected comma separated list for batch sizes to run on: {e}"
|
|
)
|
|
else:
|
|
cuda_graphs = None
|
|
|
|
CUDA_GRAPHS = cuda_graphs
|