ResNet (fastforks)

ResNet (He et al. 2015): residual basic / bottleneck blocks.

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/pytorch/vision at 1f1a920d685d.

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
resnet152.tv2_in1k@1	resnet152-f82ba261.pth	f82ba2614b9f	BSD-3-Clause (assumed from code)	yes
resnet18.tv_in1k@1	resnet18-f37072fd.pth	f37072fd47e8	BSD-3-Clause (assumed from code)	yes
resnet50.tv2_in1k@1	resnet50-11ad3fa6.pth	11ad3fa62ca7	BSD-3-Clause (assumed from code)	yes

Licence

Original-code strings, recorded as they are (research-use package; check them before any other use). Code: BSD-3-Clause. Files: .

Code licence: BSD-3-Clause; the authors do not state a separate weights licence.

Training data: ImageNet-1k (ILSVRC 2012 classification).

Licence evidence: https://github.com/pytorch/vision/blob/1f1a920d685d46dfbe187934b785b7ecb7a182f2/LICENSE

Input size

• resnet152.tv2_in1k, resnet50.tv2_in1k: fixed 224x224: the input is resized (short side 232) and centre-cropped to it
• resnet18.tv_in1k: fixed 224x224: the input is resized (short side 256) and centre-cropped to it Tested up to (portable, fp16, RTX 5060 Ti 16 GB): resnet152.tv2_in1k 224x224 (0.2 GiB peak allocated, 0.2 reserved); resnet18.tv_in1k 224x224 (0.1 GiB peak allocated, 0.1 reserved); resnet50.tv2_in1k 224x224 (0.1 GiB peak allocated, 0.1 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
resnet152.tv2_in1k	12.79	16.27	8.36	2.17	5.9x	3.9x	378 / 390	155 / 186	184 / 248
resnet18.tv_in1k	1.91	2.44	1.48	0.55	3.5x	2.7x	105 / 130	61 / 90	91 / 154
resnet50.tv2_in1k	4.74	5.84	3.15	0.99	4.8x	3.2x	180 / 192	89 / 142	118 / 204

Startup and first call

model	load (s): original	portable	kernels	first call (s): original	portable	kernels
resnet152.tv2_in1k	2.34	0.43	0.49†	0.24	0.02	0.00
resnet18.tv_in1k	0.27	0.07	0.08†	0.00	0.00	0.00
resnet50.tv2_in1k	0.51	0.16	0.18†	0.01	0.00	0.00

Fixed-size speed and peak memory, all models

model	mode	224x224
resnet152.tv2_in1k	portable	8.4 ms / 157 / 192
resnet152.tv2_in1k	kernels	2.2 ms / 188 / 270
resnet18.tv_in1k	portable	1.4 ms / 60 / 68
resnet18.tv_in1k	kernels	0.6 ms / 91 / 144
resnet50.tv2_in1k	portable	3.1 ms / 91 / 110
resnet50.tv2_in1k	kernels	1.0 ms / 121 / 190

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("resnet152.tv2_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: resnet152.tv2_in1k, resnet18.tv_in1k, resnet50.tv2_in1k.

Kernels backend

classification.load("resnet152.tv2_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.