전체 검색

AI 모델 2119

• Gestura-MiniMax-H3-VAE-LiteRT↗

  ZawShiShawn

  App-level measurements on Samsung Galaxy S26 (SM-S9420), **256 × 256, 5-second output option (124 frames), 24 fps**: Four screen recordings from Gestura on Samsung Galaxy S26, showing generated video with audio. Gestura is an Android app for on-device AI generation, including text-to-video, first-frame image-to-video and reference-conditioned video with audio. The models in these repositories were converted or packaged during its development.

• chronos-bolt-tiny↗

  amazon

  🚀 **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 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.

• StreamRender-H3↗

  Efficient-Large-Model

  StreamRender-H3 splits an interactive world into a **Code Game Engine** and a **Streaming Diffusion Render**. The engine keeps the world state in code (maps, physics, rules) and streams a semantic map for every frame. A two-step streaming video model built on MiniMax-H3 Ref2VA renders those maps into photorealistic video as the player drives, with the look set by a reference image and a prompt. - Two denoising steps per chunk in the distilled student. - Diffusion render of 217.96 ms per streaming step on 8 GB200 GPUs, a 3.67× speedup over the PyTorch baseline. - Fast Causal H3 VAE decodes 2 latent frames at 720p in 24 ms on one GB200, built by a self-improving agent loop. There are five distinct model components. Qwen's two storage formats are the same model. The causal decoder is constructed directly from its architecture configuration and student checkpoint; no native H3 video VAE checkpoint is needed. Streaming reference encoding uses TAE, and video decoding uses the trained causal decoder.

• GLM-5.3-Flash-Uncensored-AWQ↗

  hell0ks

  The abliterated (refusal-removed) build of Z.ai's GLM-5.3-Flash — baked directly into the official block-FP8 checkpoint, byte-for-byte drop-in for the original > The **abliterated** (refusal-removed) build of `zai-org/GLM-5.3-Flash` > — a **320B / 18B-active** Mixture-of-Experts model with hybrid linear + sparse attention, > 4-wide **Manifold-Constrained Hyper-Connections**, a native **vision + video** tower, an **MTP** > speculative head, and a **1M-token** context. > > The refusal direction is baked **directly into the official block-FP8 shards** — same format, same > shard layout, same `model.safetensors.index.json`. Every tensor matches the base checkpoint in > name, dtype and shape, so this is a **drop-in replacement** for `zai-org/GLM-5.3-Flash` in any > stack that already serves it. > > **On the `-FP8` name:** block-FP8 *is* the format Z.ai ships GLM-5.3-Flash in — the suffix names the > checkpoint's own precision, not a quantization step we applied. There is no upstream BF16 release to > derive from, so this is the full-precision source as published. This model has had its **safety alignment substantially removed** via *abliteration* (orthogonalizing the refusal direction out of the residual stream). As a direct consequence:

• InternVL-Chat-V1-2↗

  OpenGVLab

  We are excited to introduce 🤗 InternVL-Chat-V1-2. Inspired by LLaVA-NeXT-34B, we have also adopted Nous-Hermes-2-Yi-34B as the language model. Below is the pipeline. From the experimental results, we've observed that **a stronger language model (34B) can better leverage the powerful capabilities of our vision foundation model.** For better training reproducibility, we follow the minimalist design and data efficiency similar to LLaVA-NeXT. To reduce training costs, we provide a pre-trained MLP projector and only employ around 1.2 million visual instruction tuning samples for SFT. Our model has a total of 40 billion parameters and can be trained within 1.5 days using 32 A100 GPUs. The code, data, and model have been made publicly available.

• sam2_hiera_small.fb_r896↗

  timm

• gemma-3-270m↗

  unsloth

  Gemma is a family of lightweight, state-of-the-art open models from Google, built from the same research and technology used to create the Gemini models. Gemma 3 models are multimodal, handling text and image input and generating text output, with open weights for both pre-trained variants and instruction-tuned variants. Gemma 3 has a large, 128K context window, multilingual support in over 140 languages, and is available in more sizes than previous versions. Gemma 3 models are well-suited for a variety of text generation and image understanding tasks, including question answering, summarization, and reasoning. Their relatively small size makes it possible to deploy them in environments with limited resources such as laptops, desktops or your own cloud infrastructure, democratizing access to state of the art AI models and helping foster innovation for everyone. These models were trained on a dataset of text data that includes a wide variety of sources. The 27B model was trained with 14 trillion tokens, the 12B model was trained with 12 trillion tokens, 4B model was trained with 4 trillion tokens, the 1B with 2 trillion tokens, and the 270M with 6 trillion tokens. The knowledge cutoff date for the training data was August 2024. Here are the key components: The combination of these diverse data sources is crucial for training a powerful multimodal model that can handle a wide variety of different tasks and data formats.

• prompt-injection-guard-base↗

  Horizon-Labs

  A fast, multilingual classifier that flags **prompt injection and jailbreak attempts**, both in **user messages** (direct) and in **untrusted content an AI agent reads**: emails, web pages, documents, RAG chunks, and tool/API outputs (indirect). Labels: `SAFE` (0) and `INJECTION` (1). This is the same convention as `protectai/deberta-v3-base-prompt-injection-v2`, so the model is a drop-in replacement in code and tools built for that one. `INJECTION` means the text tries to **change what the AI reading it does**, against the instructions of its developer or user:

• Vela-1.0-Encoder-307M-Domain↗

  vllm-sr

  Vela Domain identifies the subject of a request across 14 domains to help select relevant models and tools. Macro F1 (×100) on the same development sets, compared with the original mmBERT32K Domain model. Higher is better. Short requests cover English, Chinese, Spanish, French, German and Japanese. Long-context evaluation uses 28 English/Chinese examples extended to each length. Both models process the complete input in FP32. These development sets informed Vela checkpoint selection; they are not an independent test set.

• Gemma-4-26B-A4B-it-Uncensored-NVFP4↗

  AEON-7

  Complete copy-paste recipe — pull the container, pull the model (fresh), pull the DFlash drafter (fresh), then serve with the vetted DGX Spark recipe (DFlash n=10, drafter `flash_attn`, body `triton_attn`). The recipe notes below cover tag-specific drafter backends and the production GPU-util / context-tier profile; see also Container Image Details. **Recipe notes — the drafter backend is tied to the image tag:** - **On `:latest` / `:2026-07-01-v0.24.0` (current; same behavior as the `:2026-06-18-v0.23.0-dflashfix` fix lineage):** the drafter **must** use `attention_backend: flash_attn` (enabled for multimodal Gemma targets by PR #41703's `use_mm_prefix=False`). `flex_attention` **crashes at the first request** on this image (non-contiguous KV view in the new KV-sharing path). `--enable-prefix-caching` is safe and recommended — soak-validated with DFlash (see the fixes section below). - **On the rollback tag `:2026-06-04-pr44389` (pre-fix):** the reverse — only `flex_attention` loads (`flash_attn` fails with `partial multimodal token full attention not supported`), the drafter silently runs without sliding-window support (~0% acceptance beyond 2k-token contexts), and `--enable-prefix-caching` triggers a slow acceptance collapse (see below). Not recommended. - Either way the body stays on `triton_attn` (Gemma's heterogeneous head dims), and do **not** set `--kv-cache-dtype fp8` — DFlash's non-causal drafter requires BF16 KV. `--reasoning-parser gemma4` cleanly splits `enable_thinking` output into `reasoning_content`. `num_speculative_tokens=10` is the value our production fleet runs; the drafter's trained block size is 16 and z-lab ships 15 — an n=10-vs-15 re-bench on the fixed drafter is pending (the earlier "10 beats 15" result was measured on the pre-fix drafter and shouldn't be treated as settled). The model's `config.json` already ships the vision ignore-list fix (`re:.*embed_vision.*`, `re:.*vision_tower.*`) so vLLM keeps the vision tower BF16. This default profile suits agentic gateways — one large full-context chat plus many short-lived subagents under the `--max-num-seqs 64` cap. For short-context throughput benchmarking, use `--max-model-len 32768 --max