This repository contains the finetuned mT5-base model for Thai sentence summarization. The architecture of the model is based on mT5 model and fine-tuned on text-summarization pairs in Thai. Also, this project is a Senior Project of Computer Engineering Student at King Mongkut’s University of Technology Thonburi. (See the example on Google Colab )
SAVRN Model Hub
Open-Weight Models
An open-weight model is an AI model whose trained weights are published for anyone to download. The weights are what the model learned in training. With a copy of them you can run the model on hardware you control and train it further on your own data.
Open weights are not the same as open source. Many publishers release the weights without the training data or code, and the license sets what you may do with the model. This library puts each model's full card, architecture, files, license and published evaluations on one page.
2,760Models
859Datasets
254Papers
1,692Publishers
5,040Sourced relationships
Updated 2026-09-18 · How the library is built
2,760 models, sorted by most downloaded.
My mathematical formulation to utilize space projections to "measure" the Jump between points of discontinuity found in Non-Differentialable Functions. This Model underwent an additional merge between Qemma-redux and Qwen3-0.6B, in addition to adding Rope Scaling. Fusion Logic was updated to aid per layer fusion and post fusion embedding alignment. Qemma is a HuggingFace-native hybrid model that merges Gemma-3 (1B) and Qwen-3 (0.6B) at the weight level (no adapters). This variant uses Yarn based Rope Scaling with 1:1 Ratio from maxpositionembeddings Use: research, instruction following, code/help, analysis, further SFT/RLHF. Limits: may hallucinate; not for safety-critical, medical, legal…
VideoMAEv2-Huge model pre-trained for 1200 epochs in a self-supervised way on UnlabeldHybrid-1M dataset. It was introduced in the paper [[CVPR23]VideoMAE V2: Scaling Video Masked Autoencoders with Dual Masking](https://arxiv.org/abs/2203.12602) by Wang et al. and first released in GitHub. You can use the raw model for video feature extraction. Here is how to use this model to extract a video feature
Wan2.2-TI2V-5B video DiT + 48-joint action head, trainingmode=joint. The base is suhyeok's finalized B-method recipe: a teacher-forced (sigma=0.25) self-EMA teacher plus an iBOT prototype loss at L18 L18, gamma=0.01, two-view. On top of it the 3 PAST cond latent frames are pooled into one motion frame before a chosen block. These are NOT the surrogate ctxpool runs. The surrogate line (older base, pd8 x GA1) lives in hmkang/wamctxpoolxattn and hmkang/wamctxpoolavg. Do not compare across the two sets. Geometry: 4-latin (numframesin=25, numframesout=41, fdf 2) = 4 cond + 2 future latent slots, 96 tokens per latent frame, 576 tokens per row. Effective batch 16 clips x GA 2 x 2 views = 64 rows…
This model is a fine-tuned version of allenai/led-large-16384 on the BookSum dataset (kmfoda/booksum). It aims to generalize well and be useful in summarizing lengthy text for both academic and everyday purposes. - See the Colab demo linked above or try the demo on Spaces To improve summary quality, use encodernorepeatngramsize=3 when calling the pipeline object. This setting encourages the model to utilize new vocabulary and construct an abstractive summary. Load the model into a pipeline object: Feed the text into the pipeline object: Important: For optimal summary quality, use the global attention mask when decoding, as demonstrated in this community notebook, see the definition of…
UPDATE, 15.10.2021: Check out our new zero-shot classifiers, much more lightweight and even outperforming this one: zero-shot SELECTRA small and zero-shot SELECTRA medium. This model is a fine-tuned version of the spanish BERT model with the Spanish portion of the XNLI dataset. You can have a look at the training script for details of the training. You can use this model with Hugging Face's zero-shot-classification pipeline
Opir-multitask-large is the English, highest-accuracy multi-task checkpoint in the Opir family: an encoder-based GLiClass guardrail model for real-time LLM safety filtering. It supports binary safe/unsafe classification, toxicity detection, jailbreak and prompt-injection detection, and zero-shot harmful-content categorization over a hierarchical safety taxonomy. This card is for knowledgator/opir-multitask-large. The model is used through GLiClass zero-shot classification: pass text plus the candidate labels you want scored. Use single-label mode for binary safe/unsafe decisions and multi-label mode for taxonomy, toxicity, jailbreak, or custom policy labels. Use multi-label mode when you…
https://huggingface.co/MoritzLaurer/mDeBERTa-v3-base-xnli-multilingual-nli-2mil7 with ONNX weights to be compatible with Transformers.js. If you haven't already, you can install the Transformers.js JavaScript library from NPM using: Note: Having a separate repo for ONNX weights is intended to be a temporary solution until WebML gains more traction. If you would like to make your models web-ready, we recommend converting to ONNX using Optimum and structuring your repo like this one (with ONNX weights located in a subfolder named onnx).
This model was trained on the MultiNLI dataset, which consists of 392 702 NLI hypothesis-premise pairs. The base model is DeBERTa-v3-base from Microsoft. The v3 variant of DeBERTa substantially outperforms previous versions of the model by including a different pre-training objective, see annex 11 of the original DeBERTa paper. For a more powerful model, check out DeBERTa-v3-base-mnli-fever-anli which was trained on even more data. This model was trained on the MultiNLI dataset, which consists of 392 702 NLI hypothesis-premise pairs. DeBERTa-v3-base-mnli was trained using the Hugging Face trainer with the following hyperparameters. The model was evaluated using the matched test set and…
Model · Video classification
videomae-base-finetuned-ssv2
VideoMAE model pre-trained for 2400 epochs in a self-supervised way and fine-tuned in a supervised way on Something-Something-v2. It was introduced in the paper VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training by Tong et al. and first released in this repository. Disclaimer: The team releasing VideoMAE did not write a model card for this model so this model card has been written by the Hugging Face team. VideoMAE is an extension of Masked Autoencoders (MAE) to video. The architecture of the model is very similar to that of a standard Vision Transformer (ViT), with a decoder on top for predicting pixel values for masked patches. Videos are…
VideoMAEv2-Large model pre-trained for 800 epochs in a self-supervised way on UnlabeldHybrid-1M dataset. It was introduced in the paper [[CVPR23]VideoMAE V2: Scaling Video Masked Autoencoders with Dual Masking](https://arxiv.org/abs/2203.12602) by Wang et al. and first released in GitHub. You can use the raw model for video feature extraction. Here is how to use this model to extract a video feature
A compact-but-capable ≈400M parameter causal LM that replaces dot-product attention with metric-native attention and augments sequence geometry with BlackHoleRoPE (a learnable, stable RoPE variant). Designed to train and run on modest hardware (CPU-first friendly) while staying fully compatible with Transformers. Datasets: yzhuang/Agentic-Long-Context-Understanding-QA, HuggingFaceH4/MATH-500 • Distance scores, not dot products. Heads score with L2, cosine, or diag-Mahalanobis distances. This gives direct control over geometry, often stabilizes training, and can be more sample-efficient. • BlackHoleRoPE positional encoding. • Q/K: pure unit-modulus rotation (unitary → numerically stable). •…
This is the model card of a transformers model that has been pushed on the Hub. Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. Use the code below to get started with the model. Carbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019). This model is part of the Convergent Intelligence LLC: Research Division portfolio. All models in this portfolio are developed under the Discrepancy Calculus (DISC) framework — a measure-theoretic approach to understanding and controlling the gap between what a model should produce and what it actually produces. DISC treats training…
This model is a fine-tuned version of LiquidAI/LFM2.5-8B-A1B, adapted on the angrygiraffe/claude-opus-4.6-4.7-reasoning-8.7k dataset for English text-generation and reasoning-style responses. The fine-tuning run used a custom Convergent Intelligence optimizer stack, CIxOpt, designed for heterogeneous routing across parameter types. The goal of this checkpoint is to test whether a Liquid Foundation Model backbone can be adapted efficiently through targeted sparse participation rather than broad full-model modification. This is an experimental research checkpoint intended for continued evaluation, domain adaptation, and architecture/optimizer testing.…
DeBERTa-v3-large fine-tuned with multi-task learning on 600 tasks of the tasksource collection You can further fine-tune this model to use it for any classification or multiple-choice task. This checkpoint has strong zero-shot validation performance on many tasks (e.g. 77% on WNLI). The untuned model CLS embedding also has strong linear probing performance (90% on MNLI), due to the multitask training. This is the shared model with the MNLI classifier on top. Its encoder was trained on many datasets including bigbench, Anthropic rlhf, anli... alongside many NLI and classification tasks with a SequenceClassification heads while using only one shared encoder. Each task had a specific CLS…
This model was fine-tuned on a novel financial news dataset, which consists of 2K articles from Bloomberg, on topics such as stock, markets, currencies, rate and cryptocurrencies. It is based on the PEGASUS model and in particular PEGASUS fine-tuned on the Extreme Summarization (XSum) dataset: google/pegasus-xsum model. PEGASUS was originally proposed by Jingqing Zhang, Yao Zhao, Mohammad Saleh and Peter J. Liu in PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization. Note: This model serves as a base version. For an even more advanced model with significantly enhanced performance, please check out our advanced version on Rapid API. The advanced model offers more…
A compact-but-capable ≈150M parameter causal LM that replaces dot-product attention with metric-native attention and augments sequence geometry with BlackHoleRoPE (a learnable, stable RoPE variant). Designed to train and run on modest hardware (CPU-first friendly) while staying fully compatible with • Distance scores, not dot products. Heads score with L2, cosine, or diag-Mahalanobis distances. This gives direct control over geometry, often stabilizes training, and can be more sample-efficient. • BlackHoleRoPE positional encoding. • Q/K: pure unit-modulus rotation (unitary → numerically stable). • V: bounded-energy gating (Penrose-inspired), optionally modulated by a discrepancy signal. •…
This is the model card of a transformers model that has been pushed on the Hub. Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. Use the code below to get started with the model. Carbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019). This model is part of the Convergent Intelligence LLC: Research Division portfolio. All models in this portfolio are developed under the Discrepancy Calculus (DISC) framework — a measure-theoretic approach to understanding and controlling the gap between what a model should produce and what it actually produces. DISC treats training…
A geometry‑aware Transformer that mixes several attention mechanisms and routes them with a metric‑based router. MoA replaces the classic dot‑product attention with metric‑based attention and blends four distinct heads per Transformer block: A token‑wise router decides, for each token, which head(s) to use and applies feature‑gates (FiLM‑style) and router‑bias gates for up/down‑scaling. The FFN is a HyperFFN – three parallel branches (SwiGLU MLP, separable‑conv, low‑rank) combined by a branch router. LayerScale and optional DropPath keep training stable. Triangle‑inequality (TI) penalty on sampled triples to encourage true‑metric behaviour. Ball pruning – each head learns an origin \(oh\)…
This model is a fine-tuned derivative of google/gemma-3-270m, adapted using the Convergent Intelligence sparse fine-tuning setup originally tested on Liquid Foundation Models. The checkpoint was trained on reasoning-style English examples from angrygiraffe/claude-opus-4.6-4.7-reasoning-8.7k using a targeted adaptation strategy and the custom CIxOpt optimizer framework. The goal of this model is to test whether a compact Gemma 3 270M backbone can be shaped toward reasoning-style text generation through selective parameter participation rather than broad full-model modification. This is an experimental research checkpoint intended for evaluation, local testing, optimizer research, and…
Evaluated on the SQuAD 2.0 dev set with the official eval script. Haystack is an AI orchestration framework to build customizable, production-ready LLM applications. You can use this model in Haystack to do extractive question answering on documents. To load and run the model with Haystack: For a complete example with an extractive question answering pipeline that scales over many documents, check out the corresponding Haystack tutorial. Vaishali Pal vaishali.pal [at] deepset.ai Timo Möller: timo.moeller [at] deepset.ai deepset is the company behind the production-ready open-source AI framework Haystack. We also have a Discord community open to everyone!
This is the model card of a transformers model that has been pushed on the Hub. Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. Use the code below to get started with the model. Carbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019). Part of the DistilQwen3 Series by Convergent Intelligence LLC: Research Division This model is part of a distillation chain built on Discrepancy Calculus — a measure-theoretic framework where the teacher's output distribution is decomposed via the Mesh Fundamental Identity into smooth (AC), jump, and Cantor components. The discrepancy operator…
DualMind TKD Agentic 1.7B is a two-stage derivative of Qwen/Qwen3-1.7B. It combines topology-guided mathematical knowledge distillation with assistant-masked agentic and function-calling specialization. teacher distillation topology, gap-energy diagnostics, and phase-weighted Explore/Examine/Response supervision Stage 1 was designed to transfer mathematical reasoning behavior while placing additional learning pressure on derivation, verification, and high-discrepancy reasoning transitions. - Tool schemas, user messages, and tool-result messages were visible as context but excluded from direct loss - Mathematical replay was mixed into Stage 2 to reduce catastrophic forgetting The files in…
Claude Opus 4.6 Reasoning Traces → 1.7B via DualMind SFT A 1.7B model trained on 2.5M+ tokens of Claude Opus 4.6 reasoning traces using the DualMind SFT methodology. The training data comes from Opus-4.6-Reasoning-3000x-filtered — a curated dataset of extended reasoning chains from Anthropic's most capable model, with refusals removed. This is the Opus variant of the DualMind family. Where the base DualMind model was trained on LogicInference data, this model absorbs the reasoning patterns of Claude Opus 4.6 — longer chains, more nuanced self-correction, and richer deliberative structure. The Opus teacher produces qualitatively different reasoning than synthetic logic datasets: it…
Model Collections
Hand-picked starting points, each with the reason it exists.
Collection · 4 entries
Embedding models for retrieval
Sentence and document embedding models used to build retrieval systems. Dimension and sequence length matter more than size here, and both come from the publisher.
Collection · 4 entries
Models that fit on one accelerator
Models whose publisher-reported parameter count puts them within reach of a single accelerator at common precisions. Memory needed depends on precision and serving configuration, so treat the parameter count as the starting point, not the answer.
Collection · 6 entries
Open-weight text models worth knowing
Widely used open-weight language models, chosen because each one is a distinct family rather than a variant of the one above it. Selection, not a ranking.
Collection · 3 entries
Speech and audio models
Recognition and synthesis models, grouped so the two directions are easy to compare.
Open-Weight Models Explained
What is an open-weight model?
An AI model whose trained weights are published for anyone to download, so it can be run, tested and fine-tuned on hardware the user controls.
Is an open-weight model the same as open source?
Not always. Open weights means the trained model can be downloaded. Open source usually also means the training code and data are available and the license allows broad reuse. Many open-weight models release the weights only.
Can I use an open-weight model commercially?
It depends on the license. Apache 2.0 and MIT allow commercial use. Other licenses limit it, for example to non-commercial use or below a set number of users. Every model page here shows its license.
How much memory does an open-weight model need?
About two bytes per parameter at 16-bit precision, so a 7-billion-parameter model needs roughly 14 GB for its weights, plus memory for the context it processes. Each model page lists its parameter count and the size of its files.
Related SAVRN Research
The hub sits beside SAVRN's market data and infrastructure research: what models cost to run, and what it takes to run them.
SAVRN Index
What open models cost to run
The same open-weight model priced by every host that serves it, per million tokens.
Research Hub
Data center trackers and maps
Moratoriums, permits, power, water and capital behind the facilities that run these models.
Method
How the Model Hub is built
Sources, evidence labels, refresh behaviour, and the limits of every comparison here.