Search FoxAI
AI Models 2119
- distilroberta-finetuned-financial-news-sentiment-analysis↗
mrm8488
This model is a fine-tuned version of distilroberta-base on the financial_phrasebank dataset. It achieves the following results on the evaluation set: - Loss: 0.1116 - Accuracy: **0.98**23 This model is a distilled version of the RoBERTa-base model. It follows the same training procedure as DistilBERT. The code for the distillation process can be found here. This model is case-sensitive: it makes a difference between English and English. The model has 6 layers, 768 dimension and 12 heads, totalizing 82M parameters (compared to 125M parameters for RoBERTa-base). On average DistilRoBERTa is twice as fast as Roberta-base.
- whisper-large-v3-turbo-ar-quran↗
naazimsnh02
Fine-tuned Whisper Large V3 Turbo model for high-accuracy Arabic Quranic speech recognition. This model is a fine-tuned version of openai/whisper-large-v3-turbo trained on the tarteel-ai/everyayah dataset. The model was evaluated on the **full test split** of the EveryAyah dataset, comprising **23,473 samples**.
- ThinkingCap-Qwen3.8-27B-GGUF↗
ngquocvinh
☕ If this GGUF made your day easier, a coffee would make mine. Send a coffee ☕ I build and test these releases myself. Your coffee helps keep me going. Thank you for supporting this work. This is a community GGUF package of BottleCap AI's 27B dense vision-language model, based on Qwen3.8-27B. The upstream model card describes hybrid linear and full attention, image and video understanding, an MTP head, and a native context length of 262,144 tokens. See the upstream model card for its capabilities, evaluation results, and recommended sampling settings. Those upstream task measurements apply to the upstream checkpoint and its exact quantized files; this package does not claim equivalent scores for these files. The package includes the upstream F16 multimodal projector. The ten quantizations from the initial release passed text generation smoke tests. Nine additional formats have also passed the same check; all 19 quantizations are published here. A held-out fidelity comparison to the locked F16 source has not been completed, so this card makes no fidelity ranking or quality percentage claim. Image input did not pass smoke validation with the pinned llama.cpp build: image encoding completed, but token decoding failed on both CUDA and CPU profiles. Treat image input as unverified with this runtime/package combination.
- StartLux-Decision-v1-4B-Q4_K_M-GGUF↗
startlux-models
StartLux-Decision-v1-4B in Q4_K_M (4-bit) as a GGUF file for llama.cpp. The decision server in this repository turns it into the same typed decisions, with a probability for every option, as the original model. "Same decision" is the share of the 231 public JevBench items on which the file picks the same answer as the original weights run through the `startlux_decision` package, with the same prompts, readout and temperatures. Q8_0 and Q4_K_M are llama.cpp's standard quantizations of BF16, without an importance matrix. `-ngl 99` puts every layer on the GPU (CUDA or Metal); leave it out on a CPU-only machine. llama.cpp has to be recent enough for this model (build b10454 or newer).
- Huihui-Qwen3.6-35B-A3B-abliterated-NVFP4↗
YuYu1015
> [!TIP] > **Quantized on 2026-04-21** using llm-compressor with mixed-domain calibration and sensitive-layer protection for maximum accuracy recovery. > [!IMPORTANT] > **Native W4A4 on DGX Spark (SM121) — confirmed working** > > Unlike earlier NVFP4 models on SM121, this checkpoint runs **true W4A4** via FlashInfer CUTLASS NVFP4 MoE kernel (verified in vLLM logs: `FlashInferCutlassNvFp4LinearKernel` + `FLASHINFER_CUTLASS NvFp4 MoE backend`). Requires: > > - vLLM `0.19.1rc1.dev374+g1174723eb` or later (includes PR #37725 arch-suffix fix) > - FlashInfer ≥ 0.6.8 with SM120f compilation (PR #2650) > - CUDA ≥ 12.9 > > **MTP speculative decoding supported** — MTP layers preserved in BF16 from the original checkpoint via `save_mtp_tensors_to_checkpoint`. > [!WARNING] > **Abliteration changes the optimal speculative decoding setup — this is a known trade-off, not a defect.** > > This release's distinguishing feature is **mixed-domain calibration** (`ultrachat_200k` chat + `Nemotron-Post-Training-Dataset-v2` reasoning, 256 samples total). The calibration recovers quantization accuracy, but it **cannot undo the distribution shift introduced upstream by abliteration itself** — the DFlash drafter was trained on the original Qwen3.6-35B-A3B weights, and the abliterated residual distribution no longer matches the drafter's prior, so acceptance rate drops. > > Measured throughput on DGX Spark: > > - **DFlash (`num_speculative_tokens: 15`)** — ~50 t/s sustained, occasional bursts up to ~100 t/s > - **MTP (`num_speculative_tokens: 1`)** — ~40 t/s sustained, occasional bursts up to ~70 t/s > > Counter-intuitively, **MTP with a single speculative token outperforms DFlash** on this abliterated variant — MTP reuses the model's own hidden state, so it stays aligned with the abliterated distribution that the mixed-domain calibration was tuned against. **Prefer `--speculative-config '{"method":"mtp","num_speculative_tokens":1}'` as the default**; only fall back to DFlash if you specifically need it.
- chronos-bolt-small↗
autogluon
🚀 **Update Feb 14, 2025**: Chronos-Bolt models are now available on Amazon SageMaker JumpStart! Check out the tutorial notebook to learn how to deploy Chronos endpoints for production use in a few lines of code. Chronos-Bolt is a family of pretrained time series forecasting models which can be used for zero-shot forecasting. It is based on the T5 encoder-decoder architecture and has been trained on nearly 100 billion time series observations. It chunks the historical time series context into patches of multiple observations, which are then input into the encoder. The decoder then uses these representations to directly generate quantile forecasts across multiple future steps—a method known as direct multi-step forecasting. Chronos-Bolt models are **more accurate**, up to **250 times faster** and **20 times more memory-efficient** than the original Chronos models of the same size. The following plot compares the inference time of Chronos-Bolt against the original Chronos models for forecasting 1024 time series with a context length of 512 observations and a prediction horizon of 64 steps.
- lab1-cifar10↗
Springman
Submission branch: `s_109197_103216`. Fine-tuned ImageNet ResNet-18 in Google Colab; only layer4 and fc are updated. The learning rate and epoch are selected by validation accuracy. The test split is used only after model selection. Seed: 42. Samples: train=10000, validation=2000, test=2000. Validation accuracy: 0.8905; test accuracy: 0.8815. Classes: airplane, automobile, bird, cat, deer, dog, frog, horse, ship, truck. This model recognizes these ten classes and does not detect out-of-distribution images. CIFAR-10 consists of small 32x32 images, so predictions on arbitrary photographs may be unreliable.
- SaProt_650M_AF2↗
westlake-repl
Note: SaProt requires structural (SA token) input for optimal performance. AA-sequence-only mode works but must be finetuned - frozen embeddings work only for SA, not AA sequences! With structural input, SaProt surpasses ESM2 in most tasks. --> We provide two ways to use SaProt, including through huggingface class and through the same way as in esm github. Users can choose either one to use. The esm version is also stored in the same folder, named `SaProt_650M_AF2.pt`. We provide a function to load the model.
- parakeet-tdt-0.6b-v3-GGUF↗
cstr
GGUF / ggml conversions of `nvidia/parakeet-tdt-0.6b-v3` for use with the `parakeet-main` CLI from **CrispStrobe/CrispASR**. This repo provides four quantisations, all converted from the same `.nemo` checkpoint via the streaming `convert-parakeet-to-gguf.py` script and quantised with `cohere-quantize`. All quantisations produce identical text on `samples/jfk.wav`: > And so my fellow Americans. Ask not what your country can do for you. Ask what you can do for your country.
- medsam-vit-base↗
flaviagiammarino