Giters

dnth / yolov5-deepsparse-blogpost

By the end of this post, you will learn how to: Train a SOTA YOLOv5 model on your own data. Sparsify the model using SparseML quantization aware training, sparse transfer learning, and one-shot quantization. Export the sparsified model and run it using the DeepSparse engine at insane speeds. P/S: The end result - YOLOv5 on CPU at 180+ FPS using on

Home Page:https://dicksonneoh.com/portfolio/supercharging_yolov5_180_fps_cpu/

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Error in change epoic count

VYRION-Ai opened this issue · comments

commented

If i change the epoic count it be 239

`Image sizes 320 train, 320 val
Using 2 dataloader workers
Logging results to runs/train/exp5
Starting training for 5 epochs...
Disabling LR scheduler, managing LR using SparseML recipe
Overriding number of epochs from SparseML manager to 5

     Epoch   gpu_mem       box       obj       cls    labels  img_size
     0/239     2.01G   0.06361   0.01971   0.04937        59       320: 100% 512/512 [13:16<00:00,  1.56s/it]
               Class     Images     Labels          P          R     mAP@.5 mAP@.5:.95: 100% 64/64 [01:06<00:00,  1.05s/it]
                 all       8090      11821      0.324        0.5      0.441      0.236

     Epoch   gpu_mem       box       obj       cls    labels  img_size
     1/239     2.35G   0.04284   0.01539   0.02459        71       320: 100% 512/512 [13:04<00:00,  1.53s/it]
               Class     Images     Labels          P          R     mAP@.5 mAP@.5:.95: 100% 64/64 [01:03<00:00,  1.00it/s]
                 all       8090      11821      0.837      0.758      0.838      0.504

     Epoch   gpu_mem       box       obj       cls    labels  img_size
     2/239     2.37G   0.03649   0.01465   0.01271        80       320: 100% 512/512 [12:55<00:00,  1.52s/it]
               Class     Images     Labels          P          R     mAP@.5 mAP@.5:.95: 100% 64/64 [01:04<00:00,  1.00s/it]
                 all       8090      11821      0.857      0.741      0.835      0.545

     Epoch   gpu_mem       box       obj       cls    labels  img_size
     3/239     2.37G   0.03274   0.01415   0.00986        89       320: 100% 512/512 [13:07<00:00,  1.54s/it]
               Class     Images     Labels          P          R     mAP@.5 mAP@.5:.95: 100% 64/64 [01:05<00:00,  1.02s/it]
                 all       8090      11821      0.899      0.832      0.903      0.615

     Epoch   gpu_mem       box       obj       cls    labels  img_size
     4/239     2.37G   0.03092   0.01384  0.008138        78       320: 100% 512/512 [13:05<00:00,  1.53s/it]
               Class     Images     Labels          P          R     mAP@.5 mAP@.5:.95: 100% 64/64 [01:04<00:00,  1.00s/it]
                 all       8090      11821       0.91      0.844      0.914       0.64

     Epoch   gpu_mem       box       obj       cls    labels  img_size
     5/239     2.37G   0.02988    0.0135  0.007226        76       320: 100% 512/512 [13:08<00:00,  1.54s/it]
               Class     Images     Labels          P          R     mAP@.5 mAP@.5:.95: 100% 64/64 [01:05<00:00,  1.02s/it]
                 all       8090      11821      0.914      0.861      0.917      0.662

     Epoch   gpu_mem       box       obj       cls    labels  img_size
     6/239     2.37G   0.02958   0.01332  0.007077       202       320:  10% 51/512 [01:18<11:46,  1.53s/it]`