Shababo/build run example timeout fix (#1320)

* set optimum version back to 1.26.1 since 1.27 has come out and breaks dep on transformers

* update bpy for blender example

* update timeout to handle longer running examples

* allow timeout parameter to run_example::main

.github/workflows/build-and-run-example.yml

@@ -26,4 +26,4 @@ jobs:
 
       - name: Run a random example with MODAL_IGNORE_CACHE set
         run: |
-          MODAL_IGNORE_CACHE=1 python3 -m internal.run_example
+          MODAL_IGNORE_CACHE=1 python3 -m internal.run_example --timeout 1800

06_gpu_and_ml/blender/blender_video.py

@@ -39,7 +39,7 @@ app = modal.App("example-blender-video")
 rendering_image = (
     modal.Image.debian_slim(python_version="3.11")
     .apt_install("xorg", "libxkbcommon0")  # X11 (Unix GUI) dependencies
-    .pip_install("bpy==4.1.0")  # Blender as a Python package
+    .pip_install("bpy==4.5.0")  # Blender as a Python package
 )
 
 # ## Rendering a single frame

06_gpu_and_ml/embeddings/image_embeddings_infinity.py

@@ -104,9 +104,10 @@ infinity_image = (
             "hf_transfer==0.1.9",  # for fast huggingface data download
             "huggingface_hub[hf_xet]==0.33.2",
             "tqdm==4.67.1",  # progress bar for dataset download
-            "infinity_emb[all]==0.0.76",  # for Infinity inference lib
             "sentencepiece==0.2.0",  # for this particular chosen model
             "torchvision==0.22.1",  # for fast image loading
+            "infinity_emb[all]==0.0.76",  # for Infinity inference lib
+            "optimum==1.26.1",  # need to pin this because newer version requires
         ]
     )
     .env(

internal/run_example.py

@@ -1,3 +1,4 @@
+import argparse
 import os
 import random
 import subprocess
@@ -7,18 +8,20 @@ import time
 from . import utils
 
 MINUTES = 60
-TIMEOUT = 12 * MINUTES
+DEFAULT_TIMEOUT = 12 * MINUTES
 
 
-def run_script(example):
+def run_script(example, timeout=DEFAULT_TIMEOUT):
     t0 = time.time()
 
+    print(f"Running example {example.stem} with timeout {timeout}s")
+
     try:
         print(f"cli args: {example.cli_args}")
         process = subprocess.run(
             [str(x) for x in example.cli_args],
             env=os.environ | example.env | {"MODAL_SERVE_TIMEOUT": "5.0"},
-            timeout=TIMEOUT,
+            timeout=timeout,
         )
         total_time = time.time() - t0
         if process.returncode == 0:
@@ -31,32 +34,37 @@ def run_script(example):
         returncode = process.returncode
 
     except subprocess.TimeoutExpired:
-        print(f"Past timeout of {TIMEOUT}s :(")
+        print(f"Past timeout of {timeout}s :(")
         returncode = 999
 
     return returncode
 
 
-def run_single_example(stem):
+def run_single_example(stem, timeout=DEFAULT_TIMEOUT):
     examples = utils.get_examples()
     for example in examples:
         if stem == example.stem and example.metadata.get("lambda-test", True):
-            return run_script(example)
+            return run_script(example, timeout=timeout)
     else:
         print(f"Could not find example name {stem}")
         return 0
 
 
-def run_random_example():
+def run_random_example(timeout=DEFAULT_TIMEOUT):
     examples = filter(
         lambda ex: ex.metadata and ex.metadata.get("lambda-test", True),
         utils.get_examples(),
     )
-    return run_script(random.choice(list(examples)))
+    return run_script(random.choice(list(examples)), timeout=timeout)
 
 
 if __name__ == "__main__":
-    if len(sys.argv) > 1:
-        sys.exit(run_single_example(sys.argv[1]))
+    parser = argparse.ArgumentParser()
+    parser.add_argument("example", nargs="?", default=None)
+    parser.add_argument("--timeout", type=int, default=DEFAULT_TIMEOUT)
+    args = parser.parse_args()
+    print(args)
+    if args.example:
+        sys.exit(run_single_example(args.example, timeout=args.timeout))
     else:
-        sys.exit(run_random_example())
+        sys.exit(run_random_example(timeout=args.timeout))