Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Add model_load_time metric #2311

Open
wants to merge 3 commits into
base: main
Choose a base branch
from
Open

Add model_load_time metric #2311

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
c27075d
added implementation that requires new cli argument
Edwinhr716 Jul 25, 2024
9697d16
enviroment variable approach
Edwinhr716 Jul 25, 2024
592ea3f
removed tracing logs
Edwinhr716 Jul 26, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 6 additions & 0 deletions launcher/src/main.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1162,6 +1162,7 @@ fn spawn_webserver(
max_input_tokens: usize,
max_total_tokens: usize,
max_batch_prefill_tokens: u32,
download_time: u64,
shutdown: Arc<AtomicBool>,
shutdown_receiver: &mpsc::Receiver<()>,
) -> Result<Child, LauncherError> {
Expand Down Expand Up @@ -1275,6 +1276,8 @@ fn spawn_webserver(
envs.push(("COMPUTE_TYPE".into(), compute_type.into()))
}

envs.push(("DOWNLOAD_TIME".into(), download_time.to_string().into()));

let mut webserver = match Command::new("text-generation-router")
.args(router_args)
.envs(envs)
Expand Down Expand Up @@ -1341,6 +1344,7 @@ fn terminate(process_name: &str, mut process: Child, timeout: Duration) -> io::R
fn main() -> Result<(), LauncherError> {
// Pattern match configuration
let args: Args = Args::parse();
let start_time = Instant::now();

// Filter events with LOG_LEVEL
let varname = "LOG_LEVEL";
Expand Down Expand Up @@ -1622,12 +1626,14 @@ fn main() -> Result<(), LauncherError> {
return Ok(());
}

let download_time = start_time.elapsed().as_secs();
let mut webserver = spawn_webserver(
num_shard,
args,
max_input_tokens,
max_total_tokens,
max_batch_prefill_tokens,
download_time,
shutdown.clone(),
&shutdown_receiver,
)
Expand Down
7 changes: 7 additions & 0 deletions router/src/server.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -54,6 +54,7 @@ use tower_http::cors::{AllowOrigin, CorsLayer};
use tracing::{info_span, instrument, Instrument};
use utoipa::OpenApi;
use utoipa_swagger_ui::SwaggerUi;
use tokio::time::Duration;

/// Generate tokens if `stream == false` or a stream of token if `stream == true`
#[utoipa::path(
Expand Down Expand Up @@ -1512,6 +1513,8 @@ pub async fn run(
)
)]
struct ApiDoc;
let download_time = std::env::var("DOWNLOAD_TIME").unwrap_or("30".to_string()).parse::<u64>().unwrap_or(30);
let length_time = Instant::now();

// Create state
if print_schema_command {
Expand Down Expand Up @@ -1892,6 +1895,10 @@ pub async fn run(
.layer(cors_layer);

tracing::info!("Connected");
let total_time = length_time.elapsed() + Duration::from_secs(download_time);
metrics::gauge!("tgi_model_load_time").set(total_time.as_secs_f64());



if ngrok {
#[cfg(feature = "ngrok")]
Expand Down
Loading