A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 14, 05:30 UTC
add the codes to check AMD Instinct GPU number (#22367)
Signed-off-by: Zhang Jason <ning.zhang2@amd.com>
Zhang Jason committed
Aug 6, 2025 at 23:58 UTC
b4b9813b5e2076d510ae518252f64064e6646a3e
1 file changed
+8
-2
examples/others/lmcache/disagg_prefill_lmcache_v1/disagg_example_nixl.sh
+8
-2
@@ -21,8 +21,14 @@ check_hf_token() {
21
}
22
23
check_num_gpus() {
24
- # can you check if the number of GPUs are >=2 via nvidia-smi?
25
- num_gpus=$(nvidia-smi --query-gpu=name --format=csv,noheader | wc -l)
24
+ # can you check if the number of GPUs are >=2 via nvidia-smi/rocm-smi?
25
+ which rocm-smi > /dev/null 2>&1
26
+ if [ $? -ne 0 ]; then
27
+ num_gpus=$(nvidia-smi --query-gpu=name --format=csv,noheader | wc -l)
28
+ else
29
+ num_gpus=$(rocm-smi --showid | grep Instinct | wc -l)
30
+ fi
31
+
32
if [ "$num_gpus" -lt 2 ]; then
33
echo "You need at least 2 GPUs to run disaggregated prefill."
34
exit 1