Memory requirements for Deep Faune New England model

[aweaver1fandm]

Search before asking

• I have searched the Pytorch-Wildlife issues and found no similar bug report.

Question

I am attempting to use the Deep Faune New England model (DFNE) for classification (not detection) and I'm just wondering if anyone has any guidance on the memory requirements for this model. I initially tried to run it on an L40S GPU with 48GB of memory but it blew through that. Then I tried to run it CPU based and it dies after processing roughly 55 images hitting memory usage 64GB of RAM (which is what the machine has on it). I'm trying to batch process roughly 2300 images. Detection works just fine and runs on the GPU no problem

I'm just wondering if anyone has run this model for classification and how much memory did the machine have that you ran it on. Thank you

Update 12/22/2025: I increased the machines RAM from 64GB to 96GB, and it gets further but still dies after processing about 14% of the images. I even tried to run this on our compute cluster and after processing about 40% of the images we were using 300GB of memory.

I'm trying to run the classifier as:

# Perform basic image detection
detection_model = pw_detection.MegaDetectorV6(device=DEVICE, pretrained=True, version="MDV6-yolov10-e")
det_results = detection_model.batch_image_detection(args.base_dir, batch_size=args.batch_size)

# Perform image classification
classification_model = pw_classification.DFNE() 
clf_results = classification_model.batch_image_classification(det_results=det_results,
                                                              id_strip=args.base_dir,
                                                              batch_size=8)

Additional

No response