test_vine.py

"""
Test script for VINE model loaded from HuggingFace Hub
"""
import os
import sys
from pathlib import Path
import torch

os.environ['OPENAI_API_KEY'] = "dummy-key"

# Add src to path
sys.path.insert(0, str(Path(__file__).parent / "src"))

# Determine device
device = "cuda" if torch.cuda.is_available() else "cpu"
print(f"Using device: {device}")

print("=" * 80)
print("Testing VINE Model from video-fm/vine")
print("=" * 80)

# Load VINE from HuggingFace
print("\n1. Loading VINE model from HuggingFace Hub...")
from transformers import AutoModel, AutoConfig

# Load config and set device properly
config = AutoConfig.from_pretrained('video-fm/vine', trust_remote_code=True)
config._device = device  # Override the device setting

# Load model with config
model = AutoModel.from_pretrained('video-fm/vine', config=config, trust_remote_code=True)
print("✓ Model loaded successfully")

# Verify checkpoint files
print("\n2. Verifying checkpoint files...")
checkpoint_dir = Path(__file__).parent / "checkpoints"
checkpoints = {
    "SAM2 config": checkpoint_dir / "sam2_hiera_t.yaml",
    "SAM2 checkpoint": checkpoint_dir / "sam2_hiera_tiny.pt",
    "GroundingDINO config": checkpoint_dir / "GroundingDINO_SwinT_OGC.py",
    "GroundingDINO checkpoint": checkpoint_dir / "groundingdino_swint_ogc.pth",
}

all_exist = True
for name, path in checkpoints.items():
    if path.exists():
        size_mb = path.stat().st_size / (1024 * 1024)
        print(f"✓ {name}: {path.name} ({size_mb:.1f} MB)")
    else:
        print(f"✗ {name}: NOT FOUND at {path}")
        all_exist = False

# Create pipeline
print("\n3. Creating VINE pipeline...")
from vine_hf import VinePipeline

pipeline = VinePipeline(
    model=model,
    tokenizer=None,
    sam_config_path=str(checkpoints["SAM2 config"]),
    sam_checkpoint_path=str(checkpoints["SAM2 checkpoint"]),
    gd_config_path=str(checkpoints["GroundingDINO config"]),
    gd_checkpoint_path=str(checkpoints["GroundingDINO checkpoint"]),
    device=device,
    trust_remote_code=True
)
print("✓ Pipeline created successfully")

print("\n" + "=" * 80)
print("✅ VINE Setup Complete and Working!")
print("=" * 80)
print("\nYou can now use the model for video understanding:")
print("""
from transformers import AutoModel
from vine_hf import VinePipeline

model = AutoModel.from_pretrained('video-fm/vine', trust_remote_code=True)
pipeline = VinePipeline(model=model, ...)
results = pipeline('video.mp4', categorical_keywords=['person', 'dog'], ...)
""")