update

Signed-off-by: Roger Wang <hey@rogerw.io>

.buildkite/pipeline.yml

@@ -1,7 +1,6 @@
 steps:
-  - label: "Dummy pipeline"
-    if: "false"
-    agents:
-      queue: cpu_queue_premerge_us_east_1
+  - label: ":toolbox: Qwen2.5 omni offline example"
     commands:
-      - "echo 'This pipeline is intentionally a no-op.'"
+      - ".buildkite/scripts/run_qwen_offline_example.sh"
+    agents:
+      queue: "gpu_1_queue" # g6.4xlarge instance on AWS, has 1 L4 GPU

.buildkite/scripts/run_qwen_offline_example.sh

@@ -0,0 +1,33 @@
+#!/usr/bin/env bash
+
+set -euo pipefail
+
+# Move to repo root (script lives in .buildkite/scripts/)
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/../.." && pwd)"
+cd "${ROOT_DIR}"
+
+UV_INSTALL_DIR="${ROOT_DIR}/.local/bin"
+case ":${PATH}:" in
+  *":${UV_INSTALL_DIR}:"*) ;;
+  *) export PATH="${UV_INSTALL_DIR}:${PATH}" ;;
+esac
+
+if ! command -v uv >/dev/null 2>&1; then
+  mkdir -p "${UV_INSTALL_DIR}"
+  curl -LsSf https://astral.sh/uv/install.sh | env UV_INSTALL_DIR="${UV_INSTALL_DIR}" sh
+fi
+
+UV_BIN=(uv)
+
+"${UV_BIN[@]}" python install 3.12
+"${UV_BIN[@]}" venv --python 3.12 .venv
+source .venv/bin/activate
+
+"${UV_BIN[@]}" pip install -e .
+"${UV_BIN[@]}" pip install vllm==0.11.0 --torch-backend=auto
+
+EXAMPLE_DIR="examples/offline_inference/qwen2_5_omni"
+cd "${EXAMPLE_DIR}"
+
+python end2end.py --query-type text \
+                  --no-save-results

examples/offline_inference/qwen2_5_omni/end2end.py

@@ -5,6 +5,7 @@ This example shows how to use vLLM-omni for running offline inference
 with the correct prompt format on Qwen2.5-Omni
 """
 
+import argparse
 import os
 from typing import NamedTuple
 
@@ -195,35 +196,43 @@ def main(args):
 
     omni_outputs = omni_llm.generate(prompts, sampling_params_list)
 
-    # Determine output directory: prefer --output-dir; fallback to --output-wav
-    output_dir = args.output_dir if getattr(args, "output_dir", None) else args.output_wav
-    os.makedirs(output_dir, exist_ok=True)
+    output_dir = None
+    if args.save_results:
+        # Determine output directory: prefer --output-dir; fallback to --output-wav
+        output_dir = args.output_dir if getattr(args, "output_dir", None) else args.output_wav
+        os.makedirs(output_dir, exist_ok=True)
     for stage_outputs in omni_outputs:
         if stage_outputs.final_output_type == "text":
             for output in stage_outputs.request_output:
                 request_id = int(output.request_id)
                 text_output = output.outputs[0].text
-                # Save aligned text file per request
-                prompt_text = prompts[request_id]["prompt"]
-                out_txt = os.path.join(output_dir, f"{request_id:05d}.txt")
-                lines = []
-                lines.append("Prompt:\n")
-                lines.append(str(prompt_text) + "\n")
-                lines.append("vllm_text_output:\n")
-                lines.append(str(text_output).strip() + "\n")
-                try:
-                    with open(out_txt, "w", encoding="utf-8") as f:
-                        f.writelines(lines)
-                except Exception as e:
-                    print(f"[Warn] Failed writing text file {out_txt}: {e}")
-                print(f"Request ID: {request_id}, Text saved to {out_txt}")
+                if args.save_results:
+                    # Save aligned text file per request
+                    prompt_text = prompts[request_id]["prompt"]
+                    out_txt = os.path.join(output_dir, f"{request_id:05d}.txt")
+                    lines = []
+                    lines.append("Prompt:\n")
+                    lines.append(str(prompt_text) + "\n")
+                    lines.append("vllm_text_output:\n")
+                    lines.append(str(text_output).strip() + "\n")
+                    try:
+                        with open(out_txt, "w", encoding="utf-8") as f:
+                            f.writelines(lines)
+                    except Exception as e:
+                        print(f"[Warn] Failed writing text file {out_txt}: {e}")
+                    print(f"Request ID: {request_id}, Text saved to {out_txt}")
+                else:
+                    print(f"Request ID: {request_id}, Text output:\n{text_output}")
         elif stage_outputs.final_output_type == "audio":
             for output in stage_outputs.request_output:
                 request_id = int(output.request_id)
                 audio_tensor = output.multimodal_output["audio"]
-                output_wav = os.path.join(output_dir, f"output_{output.request_id}.wav")
-                sf.write(output_wav, audio_tensor.detach().cpu().numpy(), samplerate=24000)
-                print(f"Request ID: {request_id}, Saved audio to {output_wav}")
+                if args.save_results:
+                    output_wav = os.path.join(output_dir, f"output_{output.request_id}.wav")
+                    sf.write(output_wav, audio_tensor.detach().cpu().numpy(), samplerate=24000)
+                    print(f"Request ID: {request_id}, Saved audio to {output_wav}")
+                else:
+                    print(f"Request ID: {request_id}, Received audio output (not saved).")
 
 
 def parse_args():
@@ -266,6 +275,12 @@ def parse_args():
         default=65536,
         help="Threshold for using shared memory in bytes (default: 65536)",
     )
+    parser.add_argument(
+        "--save-results",
+        action=argparse.BooleanOptionalAction,
+        default=True,
+        help="Save text/audio outputs to files (default: enabled). Use --no-save-results to disable.",
+    )
     parser.add_argument(
         "--output-wav",
         default="output_audio",