Models · Classification

MobileNetV1

MobileNet V1 (Howard et al. 2017): a classifier built on depthwise-separable convolutions.

Commercial use: yes.

variantinputbackendtime (ms)peak memory (MiB, allocated / reserved)
mobilenetv1_100.ra4_e3600_r224_in1ksynthetic (224x224)original (fp32)
original
portable
kernels
portable + torch.compile
TensorRT fp16		2.19
2.96
1.77
0.370
1.21
0.176		59 / 72
58 / 74
44 / 72
76 / 114
77 / 90
36 (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/mobilenetv1_100.ra4_e3600_r224_in1k/blob/6e88b523abded44469eafcd52f7832d5e0300b1b/README.md
  • Training data: ImageNet-1k (ILSVRC 2012 classification)
  • Kernel libraries: cudnn_hacc, fvm_conv, fvm_extra, fvm_mb, fvm_r3
  • Variants: mobilenetv1_100.ra4_e3600_r224_in1k (default of mobilenetv1_100)
  • Outputs: logits; label_set: imagenet-1k
  • Input size: fixed 224x224: the input is resized (short side 256) and centre-cropped to it
  • Tested up to (portable, fp16, RTX 5060 Ti 16 GB): mobilenetv1_100.ra4_e3600_r224_in1k 224x224 (0.0 GiB peak allocated, 0.1 reserved)
  • mobilenetv1_100.ra4_e3600_r224_in1k 224x224, first call: portable + torch.compile 5 s (cold Inductor cache), TensorRT fp16 14 s (ONNX + engine).
  • mobilenetv1_100.ra4_e3600_r224_in1k original 224x224: torch.autocast(fp16) around the forward
  • mobilenetv1_100.ra4_e3600_r224_in1k load / first call: original 1.34 s / 1.502 s, portable 0.06 s / 0.011 s, kernels 1.7 s / 0.049 s
  • Protocol, median of 9 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.

variantdatasetoriginal fp32original fp16portablekernelsnote
mobilenetv1_100.ra4_e3600_r224_in1kImageNet-1k val, 5000 random images, top-1 / top-5 %76.44 / 92.0276.46 / 92.0876.48 / 92.0476.50 / 92.12

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| <= 0.08 and p99 <= 0.3 (absolute error of the output logits).

variantkernels error vs original, mean / p99GPU, torchkernels used
mobilenetv1_100.ra4_e3600_r224_in1k0.03969 / 0.12289RTX 5060 Ti, torch 2.14.1+cu130cudnn_hacc@cudnn92400, fvm_conv@f1d064e539e1:sm_120, fvm_extra@3208ce525a91:sm_120, fvm_mb@720dd3f29bf7:sm_120, fvm_r3@f54eb890e23c:sm_120