Spaces:
Running
on
Zero
Silent Crash During Hunyuan3D Model Loading Due to Probable VRAM Exhaustion
The application is consistently crashing silently (exiting without a Python traceback) immediately after successfully downloading and attempting to load ≈18.7G of model checkpoints (including hunyuan3d-dit-v2-1/model.fp16.ckpt and model.safetensors) within the Docker container. Despite successfully implementing volume mapping (-v /home/bisgin/.cache/huggingface...) to ensure persistent model caching, the failure point appears to be a hardware limitation, specifically a CUDA Out of Memory (OOM) error during the final model loading phase, as the combined model size far exceeds the typical VRAM capacity of consumer GPUs. The goal is blocked by this probable resource constraint, and we lack a documented low-VRAM mode argument to mitigate the issue.
Help and inform me if any eco-usage options that im missing. 16gb vram and 16gb ram should work something like that
