huggingface/text-generation-inference
5
0
Fork 0
mirror of https://github.com/huggingface/text-generation-inference.git synced 2025-04-24 00:12:08 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix comment

Browse Source 
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>
This commit is contained in:
Wang, Yi A 2025-03-28 00:09:26 -07:00
parent 7914e980e2
commit 787dbe98a8
2 changed files with 4 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
launcher/src/env_runtime.rs
View File

@ -28,8 +28,8 @@ impl Env {
} }
} }
pub fn is_hpu_device(&self) -> bool { pub fn should_start_a_single_hpu_shard(&self) -> bool {
self.hpu_env != "N/A" self.hpu_env != "N/A" && std::env::var("ATTENTION").as_deref() != Ok("paged")
} }
} }

7
launcher/src/main.rs
View File

@ -1559,10 +1559,7 @@ fn spawn_shards(
) -> Result<(), LauncherError> { ) -> Result<(), LauncherError> {
// Start shard processes // Start shard processes
for rank in 0..num_shard { for rank in 0..num_shard {
if rank != 0 if rank != 0 && env_runtime::Env::new().should_start_a_single_hpu_shard() {
&& env_runtime::Env::new().is_hpu_device()
&& std::env::var("ATTENTION").as_deref() != Ok("paged")
{
tracing::info!("Running on HPU, the launcher will not do any sharding as actual sharding is done in the server"); tracing::info!("Running on HPU, the launcher will not do any sharding as actual sharding is done in the server");
break; break;
} }
@ -1642,7 +1639,7 @@ fn spawn_shards(
if shard_ready == num_shard { if shard_ready == num_shard {
break; break;
} }
if env_runtime::Env::new().is_hpu_device() { if env_runtime::Env::new().should_start_a_single_hpu_shard() {
tracing::info!("HPU detected, shard is ready"); tracing::info!("HPU detected, shard is ready");
break; break;
} }