huggingface/text-generation-inference
5
0
Fork 0
mirror of https://github.com/huggingface/text-generation-inference.git synced 2025-09-11 20:34:54 +00:00
Code Issues Packages Projects Releases Wiki Activity

Adding L40.

Browse Source
This commit is contained in:
Nicolas Patry 2024-12-09 10:54:14 +01:00
parent d701f9e866
commit 908dec63d4
No known key found for this signature in database
GPG Key ID: D2920555C90F704C
1 changed files with 5 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
launcher/src/main.rs
View File

@ -1628,6 +1628,7 @@ enum Gpu {
RTX4090, RTX4090,
T4, T4,
L4, L4,
L40,
A10G, A10G,
H100, H100,
A100, A100,
@ -1646,6 +1647,7 @@ impl From<&str> for Gpu {
"nvidia-4090" => Gpu::RTX4090, "nvidia-4090" => Gpu::RTX4090,
"nvidia-t4" => Gpu::T4, "nvidia-t4" => Gpu::T4,
"nvidia-l4" => Gpu::L4, "nvidia-l4" => Gpu::L4,
"nvidia-l40" => Gpu::L40,
"nvidia-a10g" => Gpu::A10G, "nvidia-a10g" => Gpu::A10G,
"nvidia-h100-80gb-hbm3" => Gpu::H100, "nvidia-h100-80gb-hbm3" => Gpu::H100,
"nvidia-a100-sxm4-80gb" => Gpu::A100, "nvidia-a100-sxm4-80gb" => Gpu::A100,
@ -1661,6 +1663,7 @@ impl std::fmt::Display for Gpu {
Gpu::RTX4090 => write!(f, "nvida-4090"), Gpu::RTX4090 => write!(f, "nvida-4090"),
Gpu::T4 => write!(f, "nvida-t4"), Gpu::T4 => write!(f, "nvida-t4"),
Gpu::L4 => write!(f, "nvida-l4"), Gpu::L4 => write!(f, "nvida-l4"),
Gpu::L40 => write!(f, "nvida-l40"),
Gpu::A10G => write!(f, "nvidia-a10g"), Gpu::A10G => write!(f, "nvidia-a10g"),
Gpu::H100 => write!(f, "nvidia-h100-80fb-hbm3"), Gpu::H100 => write!(f, "nvidia-h100-80fb-hbm3"),
Gpu::A100 => write!(f, "nvida-a100-sxm4-80gb"), Gpu::A100 => write!(f, "nvida-a100-sxm4-80gb"),
@ -1679,6 +1682,8 @@ impl ComputeType {
Gpu::T4 => Some(65 * 10u64.pow(12)), Gpu::T4 => Some(65 * 10u64.pow(12)),
// https://www.nvidia.com/en-us/data-center/l4/ // https://www.nvidia.com/en-us/data-center/l4/
Gpu::L4 => Some(121 * 10u64.pow(12)), Gpu::L4 => Some(121 * 10u64.pow(12)),
// https://www.nvidia.com/en-us/data-center/l40/
Gpu::L40 => Some(181 * 10u64.pow(12)),
// https://www.nvidia.com/en-us/data-center/products/a10-gpu/ // https://www.nvidia.com/en-us/data-center/products/a10-gpu/
Gpu::A10G => Some(125 * 10u64.pow(12)), Gpu::A10G => Some(125 * 10u64.pow(12)),
// https://www.nvidia.com/en-us/data-center/h100/ // https://www.nvidia.com/en-us/data-center/h100/