huggingface/text-generation-inference
5
0
Fork 0
mirror of https://github.com/huggingface/text-generation-inference.git synced 2025-04-22 15:32:08 +00:00
Code Issues Packages Projects Releases Wiki Activity

Wait 2sec once shard is ready to improve stability (#92) (#94)

Browse Source 
Co-authored-by: mswiniarsk <156412439+mswiniarsk@users.noreply.github.com>
This commit is contained in:
Karol Damaszke 2024-03-04 12:17:24 +01:00 committed by GitHub
parent 80ae9ead28
commit 8e14780bf4
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 5 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
launcher/src/main.rs
View File

@ -463,7 +463,7 @@ fn shard_manager(
envs.push(("MAX_TOTAL_TOKENS".into(), max_total_tokens.to_string().into()));
// Torch Distributed Env vars
if world_size == 1 {
if world_size == 1 {
envs.push(("RANK".into(), rank.to_string().into()));
}
envs.push(("WORLD_SIZE".into(), world_size.to_string().into()));
@ -603,6 +603,7 @@ fn shard_manager(
// Shard is ready
if uds.exists() && !ready {
tracing::info!("Shard ready in {:?}", start_time.elapsed());
sleep(Duration::from_millis(2000));
status_sender.send(ShardStatus::Ready).unwrap();
ready = true;
} else if !ready && wait_time.elapsed() > Duration::from_secs(10) {

4
router/src/main.rs
View File

@ -142,7 +142,9 @@ fn main() -> Result<(), RouterError> {
// This will only be used to validate payloads
let local_path = Path::new(&tokenizer_name);
let local_model = local_path.exists() && local_path.is_dir();
let skip_tokenizer_in_tgi = env::var("SKIP_TOKENIZER_IN_TGI").ok().map_or(false, |value| value.to_lowercase() == "true");
let skip_tokenizer_in_tgi = env::var("SKIP_TOKENIZER_IN_TGI")
.ok()
.map_or(false, |value| value.to_lowercase() == "true");
let tokenizer = if skip_tokenizer_in_tgi {
None
} else if local_model {