SigLIP (fine-tuned) (fastforks)

SigLIP So400m/14 image tower (Zhai et al. 2023) fine-tuned on ImageNet-1k with an attention-pool head.

This repo is the model card only: it holds no weights. fastforks downloads the authors' weights from the links below and checks them against the SHA-256 in the model card. Code: https://github.com/kleinicke/fastforks. Kernels for the kernels backend: https://huggingface.co/fastforks/kernels. Original code: https://github.com/huggingface/pytorch-image-models at 0df212b369a5.

Author weight repos: https://huggingface.co/timm/vit_so400m_patch14_siglip_378.webli_ft_in1k.

Task: classification. Tier: stable. Outputs: logits. Backends: portable, kernels.

Input: one RGB image. Output: ImageNet-1k class logits.

Variants

Weights are the authors' files (not hosted here). SHA-256 shortened to 12 hex digits; the full hashes are in card.toml.

variant	weights (author link)	SHA-256	weights licence	commercial use
vit_so400m_patch14_siglip_378.webli_ft_in1k@1	vit_so400m_patch14_siglip_378.webli_ft_in1k.safetensors	1aa26b75cd67	Apache-2.0	yes

Licence

Original-code strings, recorded as they are (research-use package; check them before any other use). Code: Apache-2.0. Files: .

Training data: WebLI (Google web image-text data, not released), ImageNet-1k (ILSVRC 2012 classification).

Licence evidence: https://huggingface.co/timm/vit_so400m_patch14_siglip_378.webli_ft_in1k/blob/efb07b5711bda4e3eb24630db9879e0832615ad4/README.md

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_in1k 378x378 (0.9 GiB peak allocated, 1.0 reserved)

Speed and memory

Classification, synthetic, 224x224

model	time (ms): original fp32	original	portable	kernels	kernels speed-up: vs original	vs portable	peak memory (MiB, allocated / reserved): original	portable	kernels
vit_so400m_patch14_siglip_378.webli_ft_in1k	76.86	25.81	13.13	12.94	2.0x	1.0x	2495 / 2564	879 / 1708	1216 / 1812

Startup and first call

model	load (s): original	portable	kernels	first call (s): original	portable	kernels
vit_so400m_patch14_siglip_378.webli_ft_in1k	7.52	7.29	3.57†	0.35	0.07	0.01

Fixed-size speed and peak memory, all models

model	mode	224x224
vit_so400m_patch14_siglip_378.webli_ft_in1k	portable	13.1 ms / 901 / 1050
vit_so400m_patch14_siglip_378.webli_ft_in1k	kernels	12.9 ms / 1254 / 1446

Fixed-size cells: time (ms) / peak allocated / peak reserved memory (MiB).

Hardware, protocol and caveats: benchmarks (RTX 5060 Ti 16 GB, sm_120). Original = the authors' code at the precision of portable (fp16, or (fp32) where fp16 is unsupported or not finite); original fp32 = its default fp32 run. Peak memory is the torch allocator's.

Accuracy

Parity with the original code

Every variant is checked against golden outputs of the original code at its pinned commit. CPU, float32: max |diff| <= 0.002 vs the original. GPU, float16 (the default), portable and kernels: mean |diff| <= 0.02 px and p99 <= 0.08 px.

Usage

from fastforks import classification

model = classification.load("vit_so400m_patch14_siglip_378.webli_ft_in1k")  # portable PyTorch on the GPU if there is one, else the CPU
result = model("photo.jpg")
result.top(5)  # top-5 classes
result.backend  # what ran, e.g. Backend(mode='portable', device='cuda:0', dtype='float16')

Install with pip install fastforks (see https://github.com/kleinicke/fastforks). Other variants: vit_so400m_patch14_siglip_378.webli_ft_in1k.

Kernels backend

classification.load("vit_so400m_patch14_siglip_378.webli_ft_in1k", mode="kernels") runs our CUDA kernels (cudnn_hacc, fvm_conv, fvm_extra, fvm_mb, fvm_r3). It needs a prebuilt kernel for your GPU architecture (downloaded from https://huggingface.co/fastforks/kernels on first use) and a recent NVIDIA driver; if there is none it raises BackendUnavailableError with the reason and never falls back silently. mode="auto" picks kernels when it can run and logs the choice.