Spaces:

zero-gpu-explorers

/

README

like 82

Running

I got these errors running a Gradio space (v 4.44.0) with ZeroGPU:

2024-09-20T16:01:13.384319Z [info     ] Device 0 seems unavailable, Proceeding to check subsequent devices. [accelerate.utils.modeling] func_name=get_max_memory lineno=916
2024-09-20T16:01:13.443918Z [error    ] inference_exception:
                        Some modules are dispatched on the CPU or the disk. Make sure you have enough GPU RAM to fit
                        the quantized model. If you want to dispatch the model on the CPU or the disk while keeping
                        these modules in 32-bit, you need to set `load_in_8bit_fp32_cpu_offload=True` and pass a custom
                        `device_map` to `from_pretrained`. Check
                        https://huggingface.co/docs/transformers/main/en/main_classes/quantization#offload-between-cpu-and-gpu
                        for more details.

This is a complex project, then I moved almost all the code in the python package samgis-lisa that imports these two custom python package:

1. samgis-web
2. lisa-for-cuda

Any idea or suggestion about this? these are the build and the container logs:
https://drive.google.com/file/d/1omtdKa38xBEoelEF4y9N4w6LFszuuD4t/view?usp=sharing
https://drive.google.com/file/d/1zKF8t8i0_ubIdVY7HOMh-7WSXIo9XKyI/view?usp=sharing
The link for the space: https://huggingface.co/spaces/aletrn/samgis-lisa-on-zero