search

TemryL / ComfyUI-IDM-VTON

ComfyUI adaptation of IDM-VTON for virtual try-on.
GNU General Public License v3.0
218 stars 26 forks source link

Allocation on device 0 would exceed allowed memory. (out of memory) #25

Open mehbayat opened 1 month ago

mehbayat commented 1 month ago

Error occurred when executing IDM-VTON:

Allocation on device 0 would exceed allowed memory. (out of memory) Currently allocated : 18.33 GiB Requested : 3.43 GiB Device limit : 22.02 GiB Free (according to CUDA): 31.62 MiB PyTorch limit (set by user-supplied memory fraction) : 17179869184.00 GiB

TemryL commented 1 month ago

Hi, try loading in half precision (fp16 or bfp16)

mehbayat commented 1 month ago

On the pipeline node? This is what I assume you meant, so I updated it, and now I get this error - "When loading the graph, the following node types were not found: IDM-VTON PipelineLoader Nodes that have failed to load will show as red on the graph.". It is installed on Manager, but now I get this.

Maik-96 commented 1 month ago

Almost same issue

Error occurred when executing PipelineLoader:

Allocation on device 0 would exceed allowed memory. (out of memory) Currently allocated : 24.78 GiB Requested : 5.00 MiB Device limit : 44.56 GiB Free (according to CUDA): 13.00 MiB PyTorch limit (set by user-supplied memory fraction) : 17179869184.00 GiB

Previous release worked well instead