Transparency
•
Analytics loading...
NEW
Deploy Sweelol-gemma3( loRA, KD and Prune) in a few clicks onInference Endpoints
Hub for Distilled Open Source LLMs.
Library of Expert Models.
Harness the power of distilled open-source LLMs to build, train, and deploy cutting-edge AI all within this hub.
Library of Expert Models.
The future of AI is personal, and its foundation is open source. At Swee.LOL, we are on a mission to democratize state-of-the-art AI by taking it out of the cloud and putting it onto the devices you use every day. We are an open-source research hub dedicated to creating "Personal LLMs"—a library of compact, powerful, and specialized models, distilled and fine-tuned for maximum efficiency in Edge and resource-constrained environments.
• Core Advantages: Open source platforms offer several compelling advantages over proprietary solutions.
Tasks
Problems solvers
Thousands of creators, one community—building the future of Audio, Vision, and Language with AI.
Stayed tuned.jane@sweelol:~
from transformers import AutoTokenizer, AutoModelForMaskedLM
tokenizer= AutoTokenizer.from_pretrained(
"sweelol-kd-gemma3"
)
model= AutoModelForMaskedLM.from_pretrained(
"sweelol-kd-gemma3"
) On demand
Inference API
Woking on an API interface which would simplifies access to the power of Google Cloud TPU accelerators, allowing you to efficiently run large-scale AI inference, fine-tune foundational models, and manage AI training workloads of any size, from small-scale experiments to massive distributed runs.
Fill-Mask
Examples
Computation time on Intel Xeon 3rd Gen Scalable cpu: cached
happiness
0.036survival
0.031salvation
0.017freedom
0.017unity
0.015Token Classification
Examples
Computation time on Intel Xeon 3rd Gen Scalable cpu: cached
My name isClaraPERand I live inBerkeleyLOC,CaliforniaLOC. I work at this cool company calledSwee.LOLINC.
Science
Our Research contributions
We’re on a journey to advance and democratize NLP for everyone. Along the way, we contribute to the development of technology for the better.
🌸
T0
Multitask Prompted Training Enables Zero-Shot Task Generalization
Open source state-of-the-art zero-shot language model out ofBigScience.
..........🐎
Gemma3
DistilBERT, a distilled version of GEMMA3: smaller, faster, cheaper and lighter
A smaller, faster, lighter, cheaper version of GEMMA3 obtained via model distillation.
Read more📚
HMTL
Hierarchical Multi-Task Learning
Learning embeddings from semantic tasks for multi-task learning. We have open-sourced code and a demo.
Read more🐸
Dynamical Language Models
Meta-learning for language modeling
A meta learner is trained via gradient descent to continuously and dynamically update language model weights.
Read more🤖
State of the art
Neuralcoref
Our open source coreference resolution library for coreference. You can train it on your own dataset and language.
Read more🦄
Auto-complete your thoughts
Write with Transformers
This web app is the official demo of the Transformers repository 's text generation capabilities.
Start writing