--- library_name: "fastforks" pipeline_tag: "image-classification" license: "other" license_name: "see-model-card" license_link: "https://github.com/pytorch/vision/blob/1f1a920d685d46dfbe187934b785b7ecb7a182f2/LICENSE" tags: - "classification" - "image-classification" - "cuda" - "pytorch" - "fastforks" - "fastforks-card" --- # DenseNet (fastforks) DenseNet (Huang et al. 2016): densely connected convolution 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: . Kernels for the `kernels` backend: . Original code: 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`](https://github.com/kleinicke/fastforks/blob/main/src/fastforks/classification/models/densenet/card.toml). | variant | weights (author link) | SHA-256 | weights licence | commercial use | |---|---|---|---|---| | `densenet121.tv_in1k@1` | [densenet121-a639ec97.pth](https://download.pytorch.org/models/densenet121-a639ec97.pth) | `a639ec97d7c3` | 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. Code licence: BSD-3-Clause; the authors do not state a separate weights licence. Training data: ImageNet-1k (ILSVRC 2012 classification). Licence evidence: ## 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): `densenet121.tv_in1k` 224x224 (0.1 GiB peak allocated, 0.1 reserved) ## Speed and memory ### Classification, synthetic | model | input | time (ms): original fp32 | original | portable | kernels | kernels speed-up: vs original | vs portable | peak memory (MiB, allocated / reserved): original | portable | kernels | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | `densenet121.tv_in1k` | 224x224 | 10.7 | 13.6 | 9.41 | 1.28 | 8.4x | 7.4x | 80 / 102 | 53 / 90 | 83 / 132 | ### Startup and first call | model | load (s): original | portable | kernels | first call (s): original | portable | kernels | | --- | --- | --- | --- | --- | --- | --- | | `densenet121.tv_in1k` | 1.22 | 0.41 | 1.89 | 0.56 | 0.09 | 0.04 | ### torch.compile and TensorRT | model | input | time: portable | compile | TensorRT fp16 | kernels | preparation: compile (cold first call) | ONNX + engine build | peak MiB (alloc / reserved): portable | compile | kernels | TensorRT (engine + torch) | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | `densenet121.tv_in1k` | 224x224 | 9.41 ms | 5.49 ms | 0.686 ms | 1.28 ms | 20.4 s | 33.6 s | 53 / 90 | 89 / 112 | 83 / 132 | 85 / 106 | Fixed-size cells: time (ms) / peak allocated / peak reserved memory (MiB). Hardware, protocol and caveats: [benchmarks](https://github.com/kleinicke/fastforks/blob/main/docs/benchmarks.md) (RTX 5060 Ti 16 GB, sm_120). Original = the original code named above 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 ### On public datasets 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 | |---|---|---|---|---|---|---| | `densenet121.tv_in1k` | ImageNet-1k val, 5000 random images, top-1 / top-5 % | 75.60 / 92.12 | 75.58 / 92.14 | 75.64 / 92.14 | 75.74 / 92.14 | | Measured on NVIDIA GeForce RTX 5060 Ti, torch 2.14.1+cu130, fastforks 939adfe8, 2026-11-06. ### Parity with the original code Every variant is checked against reference 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 and p99 <= 0.08 (absolute error of the output `logits`). ## Usage ```python from fastforks import classification model = classification.load("densenet121.tv_in1k") # mode="auto": the preferred backend from recorded measurements 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 ). Other variants: `densenet121.tv_in1k`. ## Kernels backend `classification.load("densenet121.tv_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 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.