Employ timm
for ImageNet evaluation.
Different from standard timm
scripts, we separate the root directory of train and eval data, as the images are reconstructed in the quantization process.
Besides, you can change the select_indices
parameter to specify the sample-level quantized sample indices. Multiple indices can be specified here.
We use the ResNet50
model as the template here. For the other models, you can refer to the timm documentation and conduct the above modifications.
sh distributed_train.sh 9 [TRAIN_ROOT] [EVAL_ROOT] --select-indices [INDICES1] [INDICES2] --output [OUTPUT_DIR] --model resnet50 --sched cosine --epochs 260 --lr 0.6 --reprob 0.6 --remode pixel --batch-size 128 --amp --aug-splits 3 -aa rand-m9-mstd0.5-inc1 --resplit --split-bn --jsd --dist-bn reduce
Current documentation for timm
covers the basics.
Getting Started with PyTorch Image Models (timm): A Practitioner’s Guide by Chris Hughes is an extensive blog post covering many aspects of timm
in detail.
timmdocs is quickly becoming a much more comprehensive set of documentation for timm
. A big thanks to Aman Arora for his efforts creating timmdocs.
paperswithcode is a good resource for browsing the models within timm
.