Support Matrix#
Models#
GenMol NIM supports a generic model profile operates with any NVIDIA GPU that has sufficient memory capacity.
Precision |
Disk Space |
Profile ID |
|---|---|---|
FP32 |
340MB |
a525212dd4373ace3be568a38b5d189a6fb866e007adf56e750ccd11e896b036 |
Supported Hardware#
GPU#
GPU |
GPU Memory (GB) |
Precision |
Profile |
# of GPUs |
|---|---|---|---|---|
H100 |
80 |
FP32 |
generic |
1 |
A100 |
40 or 80 |
FP32 |
generic |
1 |
L40S |
48 |
FP32 |
generic |
1 |
A10G |
24 |
FP32 |
generic |
1 |
A6000 |
48 |
FP32 |
generic |
1 |
Minimum GPU Memory#
16GB
Disk Space#
NVIDIA recommends at least 20GB disk space for the container and model.
CUDA Compute Capability#
This NIM requires CUDA Compute Capability of ≥ 7.0.
Supported Software#
NVIDIA Driver and Prerequisite#
Prerequisite |
Version |
Link |
|---|---|---|
NVIDIA Graphic Drivers for linux |
535.104.05 or above |
|
NVIDIA Container Toolkit |
latest |
https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html |
Docker |
latest |
Ubuntu, CentOS, Debian: https://docs.docker.com/engine/install/ |