전체 검색
AI 모델 2119
- parakeet-ultra-coreml↗
FluidInference
Core ML build of moondream/parakeet-ultra, a full-precision post-training of `nvidia/parakeet-tdt-0.6b-v3`. Same architecture, 25 languages, tokenizer and output conventions as v3; drop-in for the v3 component contract. Encoder is int8 linear per-channel (595 MB); decoder and joint are re-exported from the Ultra checkpoint; preprocessor and vocabulary are the v3 ones. Ultra wins all 24 FLEURS languages. The int8 encoder scores the same WER as an fp16 export (2.13 % / 3.79 %). Speed is within a few percent of v3 run back to back on the same machine (test-clean 126.7× vs 128.6× on the ANE). Conversion recipe: FluidInference/mobius `models/stt/parakeet-ultra/coreml`. License CC-BY-4.0, as the upstream checkpoint.
- C-RADIOv3-B↗
NVIDIA
This model performs visual feature extraction. For instance, RADIO generates image embeddings that can be used by a downstream model to classify images. C-RADIOv3 models are available in multiple sizes: * Base (90M parameters). * Large (320M parameters). * Huge (653M parameters). (In training) * Gigantic (1.1B parameters). C-RADIOv3 was trained for 1M steps (400k more steps than v1), using inverse frequency sampling for data balancing, and PHI Standardization for teacher distribution balancing. As well as new techniques for summary distribution matching, and domain generalization.
- NV-Reason-CXR-3B-GGUF↗
samwell
- TinyLlama-1.1B-intermediate-step-1431k-3T↗
TinyLlama
The TinyLlama project aims to **pretrain** a **1.1B Llama model on 3 trillion tokens**. With some proper optimization, we can achieve this within a span of "just" 90 days using 16 A100-40G GPUs 🚀🚀. The training has started on 2023-09-01. We adopted exactly the same architecture and tokenizer as Llama 2. This means TinyLlama can be plugged and played in many open-source projects built upon Llama. Besides, TinyLlama is compact with only 1.1B parameters. This compactness allows it to cater to a multitude of applications demanding a restricted computation and memory footprint. This collection contains all checkpoints after the 1T fix. Branch name indicates the step and number of tokens seen.
- MiMo-V2.6-Pro-MOPD↗
XiaomiMiMo
> [!IMPORTANT] > **This is the MOPD upgrade of the MiMo-V2.6-Pro-RL checkpoint.** > > - **MOPD2** (👉 **Technical Report §5.6**) > Fuses several domain-specialized teachers into one model, extending to domains where reliable training-time verification is hard, such as long-horizon game development, scientific research and embodied intelligence. > - **Diagnosing and Mitigating Tool-Call Repetition in MiMo-V2.6** (👉 **Technical Blog**) > An easy-to-overlook failure mode in which the model keeps issuing the same or highly similar tool calls, appearing busy while making no progress. Nothing fails outright, so it tends to go unnoticed. The MOPD stage handles it efficiently, with a short specialized-teacher run that converges quickly. MOPD2 distills several domain-specialized teachers into the student on-policy. The teachers fall into two families: **mixRL teachers**, trained on verifiable tasks, and **SFT teachers**, trained on synthetic demonstrations for open-domain tasks where a reliable reward is hard to design. Three streams contribute to a single update: Following the release of MiMo-V2.6, tool-call repetition emerged as one of the most noticeable issues in agentic settings: the model would sometimes issue the same or highly similar tool calls repeatedly, consuming time and context without making progress. This checkpoint mitigates it.
- mask2former-swin-base-coco-panoptic↗
facebook
Mask2Former model trained on COCO panoptic segmentation (base-sized version, Swin backbone). It was introduced in the paper Masked-attention Mask Transformer for Universal Image Segmentation and first released in this repository. Disclaimer: The team releasing Mask2Former did not write a model card for this model so this model card has been written by the Hugging Face team. Mask2Former addresses instance, semantic and panoptic segmentation with the same paradigm: by predicting a set of masks and corresponding labels. Hence, all 3 tasks are treated as if they were instance segmentation. Mask2Former outperforms the previous SOTA, MaskFormer both in terms of performance an efficiency by (i) replacing the pixel decoder with a more advanced multi-scale deformable attention Transformer, (ii) adopting a Transformer decoder with masked attention to boost performance without without introducing additional computation and (iii) improving training efficiency by calculating the loss on subsampled points instead of whole masks.
- Wald-4B-GGUF↗
org2ai
GGUF files of Wald-Q4B v2.1 (tag `v2.1`) for llama.cpp: an open 4B decision model that returns a calibrated probability for every option. - One pass, text only: the bundled server reads a probability for every option from the model's option-letter logits (Qwen chat template), calibrated with the same frozen temperature table as the BF16 release. - `--effort auto` (Auto 0.7) also runs here but has not been parity-checked on GGUF; for images, use the BF16 release. - On a 1,000-request Decision Index sample (5,922 questions), Q8_0 gives the BF16 answer on 99.5 % of questions and Q4_K_M on 97.3 %. All numbers: evaluation/v2.1-gguf-parity.json. - v2 (`Wald-4B-v2-*.gguf`, `v2/`): Wald-Q4B v2, unchanged in this repository. - v1.2 (`Wald-4B-v1.2-*.gguf`, `server/`): unchanged in this repository.
- Qwen-Image-2.1-viggle-turbo-v0.3-6step-GGUF↗
Abiray
This repository provides standalone **GGUF quantizations** of Alibaba's **Qwen-Image-2.1**, natively fused with **Viggle's v0.3 6-Step Turbo LoRA (Rank 256 / Alpha 256)**. By baking the distilled velocity field permanently into the base DiT in high-precision **FP32** prior to quantization, these checkpoints eliminate runtime LoRA overhead, avoid stock runtime merge degradation (~30% update precision loss in stock BF16 loaders), and run at full speed inside ComfyUI's native GGUF loader in **just 6 steps with zero CFG**. > **Note on Single-File GGUF:** Because the v0.3 LoRA is already fused into the model, you do not need to load or chain any LoRA nodes in ComfyUI. Load the GGUF directly into your sampler pipeline.
- Moonlight-16B-A3B-Instruct-GGUF↗
gabriellarson
Recently, the Muon optimizer has demonstrated strong results in training small-scale language models, but the scalability to larger models has not been proven. We identify two crucial techniques for scaling up Muon: These techniques allow Muon to work out-of-the-box on large-scale training without the need of hyper-parameter tuning. Scaling law experiments indicate that Muon is $\sim2\times$ more sample efficient than Adam with compute optimal training. Based on these improvements, we introduce **Moonlight**, a 3B/16B-parameter Mixture-of-Expert (MoE) model trained with 5.7T tokens using Muon. Our model improves the current Pareto frontier, achieving better performance with much fewer training FLOPs compared to prior models.
- Samee-ur_-_NeuralPipe-7B-slerp-DPO-gguf↗
RichardErkhov