전체 검색

AI 모델 2119

• gpt-oss-120b-DFlash↗

  z-lab

  **DFlash** is a novel speculative decoding method that utilizes a lightweight **block diffusion** model for drafting. It enables efficient, high-quality parallel drafting that pushes the limits of inference speed. This model serves as the **drafter** component and contains **0.8B parameters**. It must be used in conjunction with the target model `openai/gpt-oss-120b`. **gpt-oss-120b-DFlash** is trained on **800K samples**, drawn from: - `nvidia/Nemotron-Post-Training-Dataset-v2` - `theblackcat102/evol-codealpaca-v1`

• vakyansh-wav2vec2-sanskrit-sam-60↗

  Harveenchadha

• Nemotron-Labs-Diffusion-8B-Base↗

  NVIDIA

  Nemotron-Labs-Diffusion is a tri-mode language model that supports both AR decoding and diffusion-based parallel decoding by simply switching the attention pattern of the same model during inference. The synergy between these two modes enables a third mode, called self-speculation: the same model performs diffusion-based parallel drafting and AR verification with shared KV cache, achieving high acceptance lengths and decoding efficiency. The seamless mode switching by simply changing attention patterns enables high efficiency at different concurrency levels in varying deployment scenarios with one single model. - SOTA 3B, 8B, 14B dense LM family (base, instruct, and vision-language variants) supporting AR, diffusion, and self-speculation with the focus on decode efficiency. - Generation moved from a memory-bound regime toward a compute-bound regime. Model weights are loaded once and reused to compute multiple tokens during generation. - Self-speculation uses diffusion for drafting and AR for verification, providing a stronger alternative to MTP approaches: * 3x higher acceptance length and 2.2x speed-up vs. Qwen3-8B-Eagle3 in SGLang. * 5.9× tokens per forward over Qwen3-8B (no MTP) with the same accuracy. - Real-device speed-up across platforms: * DGX Spark (8B, concurrency 1): 2.7x faster with 112 tok/sec vs. 41.8 tok/sec AR using w4a16. * GB200 (8B, concurrency 1): 3.3x faster with 850 tok/sec vs. 253 tok/sec AR and 360 tok/sec Eagle3. Custom CUDA kernels boost to 1015 tok/sec (4x). - Diffusion speedup-of-light analysis shows that throughput can be further doubled (vs. current best) for a single user with better sampling - future research. Use of this model is governed by the NVIDIA Nemotron Open Model License.

• InternVL2_5-38B↗

  OpenGVLab

  We are excited to introduce **InternVL 2.5**, an advanced multimodal large language model (MLLM) series that builds upon InternVL 2.0, maintaining its core model architecture while introducing significant enhancements in training and testing strategies as well as data quality. As shown in the following figure, InternVL 2.5 retains the same model architecture as its predecessors, InternVL 1.5 and 2.0, following the "ViT-MLP-LLM" paradigm. In this new version, we integrate a newly incrementally pre-trained InternViT with various pre-trained LLMs, including InternLM 2.5 and Qwen 2.5, using a randomly initialized MLP projector. As in the previous version, we applied a pixel unshuffle operation, reducing the number of visual tokens to one-quarter of the original. Besides, we adopted a similar dynamic resolution strategy as InternVL 1.5, dividing images into tiles of 448×448 pixels. The key difference, starting from InternVL 2.0, is that we additionally introduced support for multi-image and video data.

• TieFighter-Holodeck-Holomax-Mythomax-F1-V1-COMPOS-20B-gguf↗

  DavidAU

  Creative, Brainstorming, Roleplay, Long Form, Story, Scene, And Fiction Generation: Model Version 1, Formula 1 This model is a mixture of 4 highly creative models using a very different form of merging: composite. This model EXCELS in long form [often exceeding 1500 tokens - without being told to do so], highly creative output with a minimum of input. (see samples below)

• phikon-v2↗

  owkin

  Phikon-v2 is a Vision Transformer Large pre-trained with Dinov2 self-supervised method on PANCAN-XL, a dataset of 450M 20x magnification histology images sampled from 60K whole slide images. PANCAN-XL only incorporates publicly available datasets: CPTAC (6,193 WSI) and TCGA (29,502 WSI) for malignant tissue, and GTEx for normal tissue (13,302 WSI). Phikon-v2 improves upon Phikon, our previous foundation model pre-trained with iBOT on 40M histology images from TCGA (6k WSI), on a large variety of weakly-supervised tasks tailored for biomarker discovery. Phikon-v2 is evaluated on external cohorts to avoid any data contamination with PANCAN-XL pre-training dataset, and benchmarked against an exhaustive panel of representation learning and foundation models. - **Developed by:** Owkin, Inc - **Model type:** Pretrained vision backbone (ViT-L/16 via DINOv2) - **Pretraining dataset:** PANCAN-XL, sourced from public histology collections (TCGA, CPTAC, GTEx, TCIA and others). - **Paper:** Arxiv - **License:** Owkin non-commercical licence

• sentence-croissant-llm-base↗

  Wissam42

• DeepSeek-V4.1-Flash-1017-Uncensored-DwarfStar-GGUF↗

  AtomicChat

  The DwarfStar (ds4) Q2 build of DeepSeek-V4.1-Flash-1017 with DeepSeek's refusal direction projected out of the weights. Built by the team behind Atomic Chat , the app that runs open-weight models locally and offline. - This is antirez/deepseek-v4.1-flash-gguf's `DeepSeek-V4.1-Flash-1017-IQ2_XXS-Q2_K-imatrix.gguf` (revision `9dcee51692e6`) with **only the 122 tensors that write into the residual stream replaced**: attention output (`attn_output_b`), shared-expert down projection, the down projection of all 384 routed experts of every block, and Engram's value projection. Every other byte, the header included, is the original's, so it runs wherever that file runs, in the same memory. - The replaced tensors carry the edit of AtomicChat/DeepSeek-V4.1-Flash-1017-Uncensored-GGUF (one direction, strength 2.0), baked into DeepSeek's weights at full precision and quantized to the original file's types (Q8_0, Q2_K with an importance matrix, F16). After the bake, 99.98% of the edit is in every writer.

• bi-encoder_msmarco_bert-base_german↗

  PM-AI

  This model can be used for **semantic search** and **documents retrieval** to find relevant passages based on a query. It was trained on a machine translated **MSMARCO dataset** for _german_ with **hard negatives** and **Margin MSE loss**. Combining these elements results in a SOTA transformer for asymmetric search. Details are presented below. The model is based on training with samples from **MSMARCO Passage Ranking** dataset. It contains about 500.000 questions and 8.8 million passages. The training objective is to identify the relevant passages or answers for an input question. In terms of content, the texts deal with diverse domains. Questions are available as sentences but also keyword-based variants can be found. Consequently, models trained on MSMARCO can be used in a variety of domains. The dataset was originally published in English, but has been translated into other languages by researchers with the help of machine translation. To be more specific, **"mMARCO: A Multilingual Version of the MS MARCO Passage Ranking Dataset"** is used, which contains 13 Google based translations, German is one of them.

• gpt2-mini↗

  erwanf

  A smaller GPT-2 model with (only) 39M parameters. It was pretrained on a subset of OpenWebText, the open-source version of the pretraining dataset used by OpenAI for the original GPT-2 models. The purpose of this model is mainly for research and education. Its small size allows for fast experiments in resource-limited settings, while still being able of generating complex and coherent text. The architecture relies on the GPT-2 model, with smaller dimensions and less layers. It uses the same tokenizer as GPT-2. We used the first 2M rows from the OpenWebText dataset, out of which we use 1k for test and validation sets.