Stability-AI
OSError: No such device (os error 19)
When testing radio_app_sv4d.py, the problem appears when loading the model.The model is already downloaded and device is set to default 'cpu', but it raises the OSError.
Hi I also got the same problem. Here are the logs for your reference:
python scripts/sampling/simple_video_sample.py --input_path /net/.../hydrant.jpg --version sv3d_p --elevations_deg 10.0
VideoTransformerBlock is using checkpointing
VideoTransformerBlock is using checkpointing
VideoTransformerBlock is using checkpointing
VideoTransformerBlock is using checkpointing
VideoTransformerBlock is using checkpointing
VideoTransformerBlock is using checkpointing
VideoTransformerBlock is using checkpointing
VideoTransformerBlock is using checkpointing
VideoTransformerBlock is using checkpointing
VideoTransformerBlock is using checkpointing
VideoTransformerBlock is using checkpointing
VideoTransformerBlock is using checkpointing
VideoTransformerBlock is using checkpointing
VideoTransformerBlock is using checkpointing
VideoTransformerBlock is using checkpointing
VideoTransformerBlock is using checkpointing
Initialized embedder #0: FrozenOpenCLIPImagePredictionEmbedder with 683800065 params. Trainable: False
Initialized embedder # 1: VideoPredictionEmbedderWithEncoder with 83653863 params. Trainable: False
Initialized embedder # 2: ConcatTimestepEmbedderND with 0 params. Trainable: False
Initialized embedder # 3: ConcatTimestepEmbedderND with 0 params. Trainable: False
Initialized embedder # 4: ConcatTimestepEmbedderND with 0 params. Trainable: False
Traceback (most recent call last):
File "/net/work/lau/generative-models/scripts/sampling/simple_video_sample.py", line 349, in
I was getting the same error with safetensors - turns out I was using a network mounted SSD, and safetensors memorymap doesn't play nicely with that.
A simple fix is to read the file into memory, and then pass that to safetensors:
def load_model_from_network_storage(checkpoint_path):
"""
Load a safetensors model from network storage by first copying to memory
Args:
checkpoint_path: Path to the safetensors file
"""
print(f"Loading model from: {checkpoint_path}")
# Read the entire file into memory
print("Reading file into memory...")
with open(checkpoint_path, 'rb') as f:
file_content = f.read()
print(f"Read {len(file_content) / (1024*1024*1024):.2f}GB into memory")
# Load using safetensors.torch.load
try:
print("Loading tensors...")
tensors = safetensors.torch.load(file_content, device="cpu")
print(f"Successfully loaded {len(tensors)} tensors")
return tensors
except Exception as e:
print(f"Error loading tensors: {str(e)}")
raise
