chore: Clarified docs, added more informative error prints (#342)

Co-authored-by: Olga Andreeva <oandreeva@oandreeva-mlt.client.nvidia.com>

docs/guides/dynamo_run.md

@@ -86,15 +86,20 @@ apt install -y build-essential libhwloc-dev libudev-dev pkg-config libssl-dev li
 ```
 
 Libraries macOS:
+- [Homebrew](https://brew.sh/)
+```
+# if brew is not installed on your system, install it
+/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"
+```
+- [Xcode](https://developer.apple.com/xcode/)
+
 ```
 brew install cmake protobuf
 
-# install Xcode from App Store and check that Metal is accessible
+# Check that Metal is accessible
 xcrun -sdk macosx metal
-
-# may have to install Xcode Command Line Tools:
-xcode-select --install
 ```
+If Metal is accessible, you should see an error like `metal: error: no input files`, which confirms it is installed correctly.
 
 Install Rust:
 ```

launch/dynamo-run/src/lib.rs

@@ -129,20 +129,44 @@ pub async fn run(
     let (maybe_card_path, maybe_card) = match (&model_path, &flags.model_config) {
         // --model-config takes precedence
         (_, Some(model_config)) => {
-            let card = ModelDeploymentCard::from_local_path(model_config, model_name.as_deref())
+            let card =
+                match ModelDeploymentCard::from_local_path(model_config, model_name.as_deref())
                     .await
-                .ok();
+                {
+                    Ok(card) => Some(card),
+                    Err(e) => {
+                        tracing::error!(
+                            "Failed to load model card from config path {}: {}",
+                            model_config.display(),
+                            e
+                        );
+                        None
+                    }
+                };
             (Some(model_config.clone()), card)
         }
         // If --model-path is an HF repo use that
         (Some(model_path), _) if model_path.is_dir() => {
-            let card = ModelDeploymentCard::from_local_path(model_path, model_name.as_deref())
+            let card = match ModelDeploymentCard::from_local_path(model_path, model_name.as_deref())
                 .await
-                .ok();
+            {
+                Ok(card) => Some(card),
+                Err(e) => {
+                    tracing::error!(
+                        "Failed to load model card from model path {}: {}",
+                        model_path.display(),
+                        e
+                    );
+                    None
+                }
+            };
             (Some(model_path.clone()), card)
         }
         // Otherwise we don't have one, but we only need it if we're tokenizing
-        _ => (None, None),
+        _ => {
+            tracing::debug!("No model card path provided (neither --model-config nor a directory in --model-path)");
+            (None, None)
+        }
     };
 
     #[cfg(any(feature = "vllm", feature = "sglang"))]
@@ -261,7 +285,8 @@ pub async fn run(
             };
             let Some(card) = maybe_card.clone() else {
                 anyhow::bail!(
-                    "out=vllm requires --model-path to be an HF repo, or for GGUF add flag --model-config <hf-repo>"
+                    "Failed to load model card: either unsupported HuggingFace repo format \
+                    or for GGUF files --model-config is missing."
                 );
             };
             let Some(sock_prefix) = zmq_socket_prefix else {