DeiT · fastforks

DeiT (Touvron et al. 2021): a vision transformer trained on ImageNet-1k only.

Commercial use: yes.

variant	input	backend	time (ms)	peak memory (MiB, allocated / reserved)
`deit_base_patch16_224.fb_in1k`	synthetic (224x224)	original (fp32) original portable kernels portable + torch.compile TensorRT fp16	6.39 5.46 2.98 1.64 2.09 1.12	371 / 386 531 / 548 203 / 386 300 / 370 238 / 266 196 (engine)

Details

Original code: https://github.com/huggingface/pytorch-image-models at 0df212b369a5
Code licence: Apache-2.0 (permissive); files:
Weights licence: Apache-2.0
Licence evidence: https://huggingface.co/timm/deit_base_patch16_224.fb_in1k/blob/b78cc5532a69df6bcad9c3a8d76653fd20b31ac6/README.md
Training data: ImageNet-1k (ILSVRC 2012 classification)
Kernel libraries: cudnn_hacc, fvm_conv, fvm_extra, fvm_mb, fvm_r3
Variants: deit_base_patch16_224.fb_in1k (default of deit_base_patch16_224)
Outputs: logits; label_set: imagenet-1k
Input size: fixed 224x224: the input is resized (short side 248) and centre-cropped to it
Tested up to (portable, fp16, RTX 5060 Ti 16 GB): deit_base_patch16_224.fb_in1k 224x224 (0.2 GiB peak allocated, 0.3 reserved)
deit_base_patch16_224.fb_in1k 224x224, first call: portable + torch.compile 7 s (cold Inductor cache), TensorRT fp16 28 s (ONNX + engine).
deit_base_patch16_224.fb_in1k original 224x224: torch.autocast(fp16) around the forward
deit_base_patch16_224.fb_in1k load / first call: original 2.52 s / 0.309 s, portable 0.72 s / 0.004 s, kernels 1.96 s / 0.067 s
Protocol, median of 5 after 3 warm-up, uint8 CUDA input (synthetic for bench_inputs sizes), card defaults; gc before each load; peak = max over the first call and the timed calls

Measured

fastforks: the model's own resize / crop on the GPU; original: the library's pretrained loader and its PIL eval transform. Original = the original code named under Details, same checkpoint and protocol.

variant	dataset	original fp32	original fp16	portable	kernels	note
`deit_base_patch16_224.fb_in1k`	ImageNet-1k val, 5000 random images, top-1 / top-5 %	83.12 / 95.92	83.14 / 95.88	83.22 / 95.84	83.16 / 95.84

Measured on NVIDIA GeForce RTX 5060 Ti, torch 2.14.1+cu130, fastforks 939adfe8, 2026-11-06.

Parity

CPU, float32: max |diff| <= 0.002 vs the original. GPU, float16 (the default), portable and kernels: mean |diff| <= 0.02 and p99 <= 0.08 (absolute error of the output logits).

variant	kernels error vs original, mean / p99	GPU, torch	kernels used
`deit_base_patch16_224.fb_in1k`	0.00223 / 0.0071	RTX 5060 Ti, torch 2.14.1+cu130	`cudnn_hacc@cudnn92400`, `fvm_conv@f1d064e539e1:sm_120`, `fvm_extra@3208ce525a91:sm_120`, `fvm_mb@720dd3f29bf7:sm_120`, `fvm_r3@f54eb890e23c:sm_120`