huggingface/text-generation-inference
5
0
Fork 0
mirror of https://github.com/huggingface/text-generation-inference.git synced 2025-09-11 12:24:53 +00:00
Code Issues Packages Projects Releases Wiki Activity

Upgrading axum=0.7

Browse Source
This commit is contained in:
Nicolas Patry 2024-02-02 12:15:32 +01:00
parent 0e97af456a
commit 4c6c39e491
3 changed files with 472 additions and 201 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

586
Cargo.lock generated
View File

File diff suppressed because it is too large Load Diff

10
router/Cargo.toml
View File

@ -16,8 +16,8 @@ path = "src/main.rs"
[dependencies]
async-stream = "0.3.5"
axum = { version = "0.6.20", features = ["json"] }
axum-tracing-opentelemetry = "0.14.1"
axum = { version = "0.7.4", features = ["json"] }
axum-tracing-opentelemetry = "0.16"
text-generation-client = { path = "client" }
clap = { version = "4.4.5", features = ["derive", "env"] }
futures = "0.3.28"
@ -35,12 +35,12 @@ thiserror = "1.0.48"
tokenizers = { version = "0.15.1", features = ["http"] }
tokio = { version = "1.32.0", features = ["rt", "rt-multi-thread", "parking_lot", "signal", "sync"] }
tokio-stream = "0.1.14"
tower-http = { version = "0.4.4", features = ["cors"] }
tower-http = { version = "0.5.1", features = ["cors"] }
tracing = "0.1.37"
tracing-opentelemetry = "0.21.0"
tracing-subscriber = { version = "0.3.17", features = ["json", "env-filter"] }
utoipa = { version = "3.5.0", features = ["axum_extras"] }
utoipa-swagger-ui = { version = "3.1.5", features = ["axum"] }
utoipa = { version = "4.2.0", features = ["axum_extras"] }
utoipa-swagger-ui = { version = "6.0.0", features = ["axum"] }
ngrok = { version = "0.13.1", features = ["axum"], optional = true }
init-tracing-opentelemetry = { version = "0.14.1", features = ["opentelemetry-otlp"] }
minijinja = "1.0.10"

77
router/src/server.rs
View File

@ -973,46 +973,47 @@ pub async fn run(
if ngrok {
#[cfg(feature = "ngrok")]
{
use ngrok::config::TunnelBuilder;
// use ngrok::config::TunnelBuilder;
let _ = addr;
// let _ = addr;
let authtoken =
ngrok_authtoken.expect("`ngrok-authtoken` must be set when using ngrok tunneling");
// let authtoken =
// ngrok_authtoken.expect("`ngrok-authtoken` must be set when using ngrok tunneling");
let edge = ngrok_edge.expect("`ngrok-edge` must be set when using ngrok tunneling");
// let edge = ngrok_edge.expect("`ngrok-edge` must be set when using ngrok tunneling");
let tunnel = ngrok::Session::builder()
.authtoken(authtoken)
.connect()
.await
.unwrap()
.labeled_tunnel()
.label("edge", edge);
// let tunnel = ngrok::Session::builder()
// .authtoken(authtoken)
// .connect()
// .await
// .unwrap()
// .labeled_tunnel()
// .label("edge", edge);
let listener = tunnel.listen().await.unwrap();
// let listener = tunnel.listen().await.unwrap();
// let router = Router::new()
// .route("/health", get(health))
// .route("/metrics", get(metrics))
// .layer(Extension(health_ext))
// .layer(Extension(prom_handle))
// .into_make_service();
// Run prom metrics and health locally too
tokio::spawn(
axum::Server::bind(&addr)
.serve(
Router::new()
.route("/health", get(health))
.route("/metrics", get(metrics))
.layer(Extension(health_ext))
.layer(Extension(prom_handle))
.into_make_service(),
)
//Wait until all requests are finished to shut down
.with_graceful_shutdown(shutdown_signal()),
);
// // Run prom metrics and health locally too
// let tokio_listener = tokio::net::TcpListener::bind(&addr).await.unwrap();
// axum::serve(tokio_listener, router)
// .with_graceful_shutdown(shutdown_signal())
// .await?;
// axum::serve(listener, app)
// .with_graceful_shutdown(shutdown_signal())
// .await?;
// Run server
axum::Server::builder(listener)
.serve(app.into_make_service())
//Wait until all requests are finished to shut down
.with_graceful_shutdown(shutdown_signal())
.await?;
// axum::Server::builder(listener)
// .serve(app.into_make_service())
// //Wait until all requests are finished to shut down
// .with_graceful_shutdown(shutdown_signal())
// .await?;
}
#[cfg(not(feature = "ngrok"))]
{
@ -1025,11 +1026,17 @@ pub async fn run(
}
} else {
// Run server
axum::Server::bind(&addr)
.serve(app.into_make_service())
// Wait until all requests are finished to shut down
let listener = tokio::net::TcpListener::bind(&addr).await.unwrap();
axum::serve(listener, app)
.with_graceful_shutdown(shutdown_signal())
.await?;
// axum::Server::bind(&addr)
// .serve(app.into_make_service())
// // Wait until all requests are finished to shut down
// .with_graceful_shutdown(shutdown_signal())
// .await?;
}
Ok(())
}