Local AI with Docker's Testcontainers
•
4
Joseph Pollack
Tonic
AI & ML interests
🤖Making robots to help people learn things quicker 👩🏻🚀🚀
Articles
Organizations
Posts
22
Post
595
🙋🏻♂️Hey there folks ,
I found this cool (new?) thing by Docker called Testcontainers , and there's an @ollama object that you can use to programmatically serve ephemeral containers and LLMs.
I made a post about it here : https://huggingface.co/blog/Tonic/localai-testcontainers
It's really useful, powerful and fun !
Demo coming soon 🤗
I found this cool (new?) thing by Docker called Testcontainers , and there's an @ollama object that you can use to programmatically serve ephemeral containers and LLMs.
I made a post about it here : https://huggingface.co/blog/Tonic/localai-testcontainers
It's really useful, powerful and fun !
Demo coming soon 🤗
Post
1628
🙋🏻♂️ Hey there folks
made a demo for Nvidia Minitron on an A100.
Minitron is a family of small language models (SLMs) obtained by pruning NVIDIA's Nemotron-4 15B model. We prune model embedding size, attention heads, and MLP intermediate dimension, following which, we perform continued training with distillation to arrive at the final models.
Deriving the Minitron 8B and 4B models from the base 15B model using our approach requires up to 40x fewer training tokens per model compared to training from scratch; this results in compute cost savings of 1.8x for training the full model family (15B, 8B, and 4B). Minitron models exhibit up to a 16% improvement in MMLU scores compared to training from scratch, perform comparably to other community models such as Mistral 7B, Gemma 7B and Llama-3 8B, and outperform state-of-the-art compression techniques from the literature. Please refer to our arXiv paper for more details.
Minitron models are for research and development only.
source : nvidia/Minitron-8B-Base
demo : Tonic/Minitron
made a demo for Nvidia Minitron on an A100.
Minitron is a family of small language models (SLMs) obtained by pruning NVIDIA's Nemotron-4 15B model. We prune model embedding size, attention heads, and MLP intermediate dimension, following which, we perform continued training with distillation to arrive at the final models.
Deriving the Minitron 8B and 4B models from the base 15B model using our approach requires up to 40x fewer training tokens per model compared to training from scratch; this results in compute cost savings of 1.8x for training the full model family (15B, 8B, and 4B). Minitron models exhibit up to a 16% improvement in MMLU scores compared to training from scratch, perform comparably to other community models such as Mistral 7B, Gemma 7B and Llama-3 8B, and outperform state-of-the-art compression techniques from the literature. Please refer to our arXiv paper for more details.
Minitron models are for research and development only.
source : nvidia/Minitron-8B-Base
demo : Tonic/Minitron
spaces
108
pinned
Running
on
Zero
6
🚐🤖💽
Minitron
State-of-the-Art Performance With a SOTA Compression
pinned
Runtime error
2
💬
01aiYi NvidiaEmbed
pinned
Running
on
Zero
8
🦅👁️
Falcon Vision
Falcon2-11B-vlm is a vision model by TIL
pinned
Sleeping
9
💻
Granite Code
IBM-Granite for Code Generation
pinned
Running
on
Zero
8
☯️9️⃣🅱️
Yi 9B
Yi-9B-BaseModel Serve while i try to fine tune it further 🤗
pinned
Running
17
🐙🐙
Octopus
🐙🐙Octopus2 is an On-Device Function Calling Model
models
10
Tonic/voyage-2-large
Text Generation
•
Updated
•
14
Tonic/mirnet
Updated
•
3
Tonic/video-swin-transformer
Updated
•
8
Tonic/mobile-vit
Updated
•
10
Tonic/paligemma-3b-pt-896
Updated
•
1
Tonic/stablemed
Updated
•
7
•
5
Tonic/GaiaMiniMed
Question Answering
•
Updated
•
8
Tonic/experiments
Updated
•
1
Tonic/mistralmed
Updated
•
4
•
10
Tonic/LunarLander
Reinforcement Learning
•
Updated
•
1
datasets
6
Tonic/medquad
Viewer
•
Updated
•
15.5k
•
5
•
2
Tonic/Tonic-AI-Transcript-1-10-2024
Viewer
•
Updated
•
101
•
1
Tonic/Tonic-AI-Transcript-1-09-2024
Updated
•
3
•
1
Tonic/EasyReddit
Viewer
•
Updated
•
2.02M
•
3
•
3
Tonic/K23-Huggingface-Certificate-of-Excellence
Viewer
•
Updated
•
1
•
1
Tonic/image-classification-mistakes
Updated
•
1
•
1