SigLIP (fine-tuned)
SigLIP So400m/14 (Zhai et al. 2023) fine-tuned on ImageNet-1k.
Commercial use: yes.
| variant | input | backend | time (ms) | peak memory (MiB, allocated / reserved) |
|---|---|---|---|---|
vit_so400m_patch14_siglip_378.webli_ft_in1k | synthetic (224x224) | original (fp32) original portable kernels portable + torch.compile TensorRT fp16 | 76.9 25.8 13.1 12.8 13.3 11.4 | 1710 / 1744 2495 / 2564 879 / 1708 1216 / 1346 909 / 944 848 (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/vit_so400m_patch14_siglip_378.webli_ft_in1k/blob/efb07b5711bda4e3eb24630db9879e0832615ad4/README.md
- Training data: WebLI (Google web image-text data, not released), ImageNet-1k (ILSVRC 2012 classification)
- Kernel libraries:
cudnn_hacc,fvm_conv,fvm_extra,fvm_mb,fvm_r3 - Variants:
vit_so400m_patch14_siglip_378.webli_ft_in1k(default ofvit_so400m_patch14_siglip_378) - Outputs: logits; label_set:
imagenet-1k - Input size: fixed 378x378: the input is resized (short side 378) and centre-cropped to it
- Tested up to (portable, fp16, RTX 5060 Ti 16 GB):
vit_so400m_patch14_siglip_378.webli_ft_in1k378x378 (0.9 GiB peak allocated, 1.0 reserved) vit_so400m_patch14_siglip_378.webli_ft_in1k224x224, first call: portable + torch.compile 14 s (cold Inductor cache), TensorRT fp16 61 s (ONNX + engine).vit_so400m_patch14_siglip_378.webli_ft_in1koriginal 224x224: torch.autocast(fp16) around the forwardvit_so400m_patch14_siglip_378.webli_ft_in1kload / first call: original 7.52 s / 0.346 s, portable 7.29 s / 0.069 s, kernels 5.82 s / 0.093 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 |
|---|---|---|---|---|---|---|
vit_so400m_patch14_siglip_378.webli_ft_in1k | ImageNet-1k val, 5000 random images, top-1 / top-5 % | 90.04 / 98.90 | 90.02 / 98.88 | 89.98 / 98.94 | 90.00 / 98.92 |
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 |
|---|---|---|---|
vit_so400m_patch14_siglip_378.webli_ft_in1k | 0.00232 / 0.00745 | 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 |