[ci] CI supports use cached models (#7874)

python/sglang/test/test_utils.py

@@ -2,6 +2,7 @@
 
 import argparse
 import copy
+import json
 import logging
 import os
 import random
@@ -102,6 +103,15 @@ def is_in_amd_ci():
     return get_bool_env_var("SGLANG_AMD_CI")
 
 
+def _use_cached_default_models(model_repo: str):
+    cache_dir = os.getenv("DEFAULT_MODEL_CACHE_DIR")
+    if cache_dir and model_repo:
+        model_path = os.path.join(cache_dir, model_repo)
+        if os.path.isdir(model_path):
+            return os.path.abspath(model_path)
+    return ""
+
+
 if is_in_ci():
     DEFAULT_PORT_FOR_SRT_TEST_RUNNER = (
         5000 + int(os.environ.get("CUDA_VISIBLE_DEVICES", "0")[0]) * 100
@@ -419,6 +429,31 @@ def get_call_select(args: argparse.Namespace):
     return func
 
 
+def _get_default_models():
+    import inspect
+
+    current_module = inspect.getmodule(_get_default_models)
+    default_models = set()
+    for name, value in current_module.__dict__.items():
+        if (
+            isinstance(name, str)
+            and "DEFAULT_" in name
+            and "MODEL_" in name
+            and isinstance(value, str)
+        ):
+            if "," in value:
+                parts = [part.strip() for part in value.split(",")]
+                default_models.update(parts)
+            else:
+                default_models.add(value.strip())
+    return json.dumps(list(default_models))
+
+
+def try_cached_model(model_repo: str):
+    model_dir = _use_cached_default_models(model_repo)
+    return model_dir if model_dir else model_repo
+
+
 def popen_launch_server(
     model: str,
     base_url: str,

scripts/ci_cache_models.sh

@@ -0,0 +1,40 @@
+#!/bin/bash
+set -euxo pipefail
+
+mapfile -t models < <(python3 -c "from sglang.test.test_utils import _get_default_models; print(_get_default_models())" | jq -r '.[]')
+
+if [ ${#models[@]} -eq 0 ]; then
+    echo "Failed to get default models."
+    exit 1
+fi
+
+cache_dir="${DEFAULT_MODEL_CACHE_DIR:-}"
+
+if [ -z "$cache_dir" ]; then
+    echo "DEFAULT_MODEL_CACHE_DIR environment variable is not set."
+    exit 1
+fi
+
+failed_models=()
+for model in "${models[@]}"; do
+    local_model_dir="$cache_dir/$model"
+    echo "Caching model: $model to $local_model_dir"
+    mkdir -p "$local_model_dir"
+
+    if ! huggingface-cli download "$model" \
+        --local-dir "$local_model_dir" \
+        --local-dir-use-symlinks False 2>/dev/null; then
+        echo "WARNING: Failed to cache model: $model"
+        rm -rf "$local_model_dir"
+        failed_models+=("$model")
+        continue
+    fi
+    echo "Successfully cached model: $model"
+done
+
+if [ ${#failed_models[@]} -gt 0 ]; then
+    echo -e "\n[Summary] Failed to cache following models:"
+    printf ' - %s\n' "${failed_models[@]}"
+else
+    echo -e "\n[Summary] All models cached successfully"
+fi