サイト内検索
AI モデル 84762
- HY-WU↗
腾讯
- Qwen3.5-4B-Kazakh↗
issai
- Huihui-Qwen3.8-27B-Coder390-EfficientThink-abliterated-VAL-APEX-I-NanoPlus-V3-GGUF↗
IsValorum
- qwen3.8-flash-coder-44gb-selective-int8↗
Jab1718
**`Qwen3.8-Flash-Coder-44GB-Selective-INT8`** is a high-performance, selective-quantized Mixture-of-Experts (MoE) coding model. Sliced down from the monolithic **`Qwen/Qwen3.8-Flash-Next` (335GB)** and quantized from the **`Qwen3.8-Flash-Coder-85GB-BF16`** parent checkpoint, this model reduces disk and VRAM footprint to exactly **44.29 GB** (a **44.2% VRAM reduction** and **86.8% reduction from base**), enabling full zero-offload deployment on **only 2x 32GB GPUs** (e.g. 2x NVIDIA RTX 5000 Ada, 2x RTX 4090/3090, or 1x A100/H100 80GB). Traditional MoE post-training quantization often quantizes all layers uniformly, which severely degrades the Router Gate and causes **Routing Collapse** (routing tokens to sub-optimal experts). This checkpoint introduces **Selective MoE Quantization**: 1. **Critical High-Precision Modules (Kept in 100% Native BF16):** - **Router Gates:** Retain 100% floating-point routing fidelity across all 48 layers. - **Multi-Head Self-Attention & Linear Attention:** `q_proj`, `k_proj`, `v_proj`, `o_proj`. - **Shared Expert, RMSNorms, Embeddings & LM Head:** Zero quantization loss in embedding projections. 2. **High-Capacity Sparse Experts (Quantized to Symmetric Per-Channel INT8):** - 160 MoE Experts across 48 layers (`gate_up_proj`, `down_proj`). - Symmetrically quantized per-channel with dynamic scaling vectors (`gate_up_proj_scale`, `down_proj_scale`).
- jina-bert-implementation↗
jinaai
- yolov8s-protective-equipment-detection↗
keremberke
- activation↗
kernels-community
> [!CAUTION] > Starting from September 13, 2026, we will be removing the "model" type repositories of kernels (e.g., kernels-community/flash-attn3). Make sure you're using a latest version of kernels. If you face any disruption, please report them here: https://github.com/huggingface/kernels/issues/new. This is the repository card of kernels-community/activation that has been pushed on the Hub. It was built to be used with the `kernels` library. This card was automatically generated. - `silu_and_mul` - `mul_and_silu` - `gelu_and_mul` - `gelu_tanh_and_mul` - `fatrelu_and_mul` - `gelu_fast` - `gelu_new` - `gelu_quick` - `gelu_tanh` - `silu` - `gelu`
- deberta-v3-base-japanese↗
ku-nlp
The tokenizer of this model is based on huggingface/tokenizers Unigram byte-fallback model. The vocabulary entries were converted from `llm-jp-tokenizer v2.2 (100k)`. Please refer to README.md of `llm-jp/llm-ja-tokenizer` for details on the vocabulary construction procedure. Note that, unlike ku-nlp/deberta-v2-base-japanese, pre-segmentation by a morphological analyzer (e.g., Juman++) is no longer required for this model. We shuffled the corpora, which has 270B tokens in total, and trained the model for 2 epochs. Thus, the total number of tokens fed to the model was 540B.
- moss-tts-local-transformer-4.55b-voice-acting-v2↗
laion
- qwen3.5_4B_AiO_LR_lora_kimi_k2.5_distilation↗
leeminwaan
