@ThomasTheMaker if you make an issue on the repo, I'll look into it!
Jordan Legg PRO
takarajordan
AI & ML interests
Chief AI Officer @takara.ai. Diffusion, Inference optimisation and all things MultiModal.
Recent Activity
replied to
their
post
about 3 hours ago
🎌 Two months in, https://github.com/takara-ai/go-attention has passed 429 stars on GitHub.
We built this library at takara.ai to bring attention mechanisms and transformer layers to Go — in a form that's lightweight, clean, and dependency-free.
We’re proud to say that every part of this project reflects what we set out to do.
- Pure Go — no external dependencies, built entirely on the Go standard library
- Core support for DotProductAttention and MultiHeadAttention
- Full transformer layers with LayerNorm, feed-forward networks, and residual connections
- Designed for edge, embedded, and real-time environments where simplicity and performance matter
Thank you to everyone who has supported this so far — the stars, forks, and feedback mean a lot.
replied to
their
post
1 day ago
🎌 Two months in, https://github.com/takara-ai/go-attention has passed 429 stars on GitHub.
We built this library at takara.ai to bring attention mechanisms and transformer layers to Go — in a form that's lightweight, clean, and dependency-free.
We’re proud to say that every part of this project reflects what we set out to do.
- Pure Go — no external dependencies, built entirely on the Go standard library
- Core support for DotProductAttention and MultiHeadAttention
- Full transformer layers with LayerNorm, feed-forward networks, and residual connections
- Designed for edge, embedded, and real-time environments where simplicity and performance matter
Thank you to everyone who has supported this so far — the stars, forks, and feedback mean a lot.
posted
an
update
1 day ago
🎌 Two months in, https://github.com/takara-ai/go-attention has passed 429 stars on GitHub.
We built this library at takara.ai to bring attention mechanisms and transformer layers to Go — in a form that's lightweight, clean, and dependency-free.
We’re proud to say that every part of this project reflects what we set out to do.
- Pure Go — no external dependencies, built entirely on the Go standard library
- Core support for DotProductAttention and MultiHeadAttention
- Full transformer layers with LayerNorm, feed-forward networks, and residual connections
- Designed for edge, embedded, and real-time environments where simplicity and performance matter
Thank you to everyone who has supported this so far — the stars, forks, and feedback mean a lot.
Organizations
takarajordan's activity
replied to
their
post
about 3 hours ago
replied to
their
post
1 day ago
@ThomasTheMaker it's just the raw attention and transformer architecture in golang designed for serverless so performance will definitely be less than ggml and llama.cpp since it's not accelerated by GPU's but if you're into edge AI CPU only, this is the first, only and best way to compute attention.
Quantization can definitely be supported as it's just a math model!
posted
an
update
1 day ago
Post
417
🎌 Two months in, https://github.com/takara-ai/go-attention has passed 429 stars on GitHub.
We built this library at takara.ai to bring attention mechanisms and transformer layers to Go — in a form that's lightweight, clean, and dependency-free.
We’re proud to say that every part of this project reflects what we set out to do.
- Pure Go — no external dependencies, built entirely on the Go standard library
- Core support for DotProductAttention and MultiHeadAttention
- Full transformer layers with LayerNorm, feed-forward networks, and residual connections
- Designed for edge, embedded, and real-time environments where simplicity and performance matter
Thank you to everyone who has supported this so far — the stars, forks, and feedback mean a lot.
We built this library at takara.ai to bring attention mechanisms and transformer layers to Go — in a form that's lightweight, clean, and dependency-free.
We’re proud to say that every part of this project reflects what we set out to do.
- Pure Go — no external dependencies, built entirely on the Go standard library
- Core support for DotProductAttention and MultiHeadAttention
- Full transformer layers with LayerNorm, feed-forward networks, and residual connections
- Designed for edge, embedded, and real-time environments where simplicity and performance matter
Thank you to everyone who has supported this so far — the stars, forks, and feedback mean a lot.
-
- 4 replies
posted
an
update
7 days ago
Post
1530
AI research over coffee ☕️
No abstracts, just bullet points.
Start your day here: https://tldr.takara.ai
No abstracts, just bullet points.
Start your day here: https://tldr.takara.ai
- 1 reply
This is a pretty big update for sure. The models have improved significantly which is great for everyone involved, especially the end user. Those datasets look very promising as well!
Sounds interesting, I’ll check it out!
This is a really interesting post. I’ve been looking at the DeepSeek models for sure. This shows a pretty nice improvement, would love to see some example changes!
Very cool
posted
an
update
16 days ago
Post
1855
Takara takes 3rd place in the {tech:munich} AI hackathon with Fudeno!
A little over 2 weeks ago @aldigobbler and I set out to create the largest MultiModal SVG dataset ever created, we succeeded in this and when I was in Munich, Germany I took it one step further and made an entire app with it!
We fine-tuned Mistral Small, made a Next.JS application and blew some minds, taking 3rd place out of over 100 hackers. So cool!
If you want to see the dataset, please see below.
takara-ai/fudeno-instruct-4M
A little over 2 weeks ago @aldigobbler and I set out to create the largest MultiModal SVG dataset ever created, we succeeded in this and when I was in Munich, Germany I took it one step further and made an entire app with it!
We fine-tuned Mistral Small, made a Next.JS application and blew some minds, taking 3rd place out of over 100 hackers. So cool!
If you want to see the dataset, please see below.
takara-ai/fudeno-instruct-4M
replied to
their
post
3 months ago
Sir, basically I want to create a generative AI university helpdesk chatbot, and for this, I have created datasets myself and also fine-tuned models, but I am not getting satisfactory results. Sir, if you have time, could you please check my datasets in my profile and help me understand how I can improve my dataset and work on it so that my task gets completed? I would be very grateful to you.
I would enhance your dataset to use multi turn conversations if you can at all for llama2 you could do something like this:
<s>[INST] Is the BS Physics program a part-time or full-time course? [/INST] The BS Physics program is a full-time undergraduate program that requires regular on-campus attendance. </s><s>[INST] How many units per semester? [/INST] A typical semester load consists of 15-18 units. </s>
hope this helps! Again, please reach out to me on discord here: takarajordan_82155
gimme an invite! :D
reacted to
s3nh's
post with ❤️
4 months ago
Post
2045
Welcome back,
Small Language Models Enthusiasts and GPU Poor oss enjoyers lets connect.
Just created an organization which main target is to have fun with smaller models tuneable on consumer range GPUs, feel free to join and lets have some fun, much love ;3
SmolTuners
Small Language Models Enthusiasts and GPU Poor oss enjoyers lets connect.
Just created an organization which main target is to have fun with smaller models tuneable on consumer range GPUs, feel free to join and lets have some fun, much love ;3
SmolTuners
-
- 3 replies
Amazing work
reacted to
merve's
post with 🚀
4 months ago
Post
2840
Aya by Cohere For AI can now see! 👀
C4AI community has built Maya 8B, a new open-source multilingual VLM built on SigLIP and Aya 8B 🌱 works on 8 languages! 🗣️
The authors extend Llava dataset using Aya's translation capabilities with 558k examples!
ry it here kkr5155/maya_demo
Dataset maya-multimodal/pretrain
Model maya-multimodal/maya 👏
kudos @nahidalam and team
C4AI community has built Maya 8B, a new open-source multilingual VLM built on SigLIP and Aya 8B 🌱 works on 8 languages! 🗣️
The authors extend Llava dataset using Aya's translation capabilities with 558k examples!
ry it here kkr5155/maya_demo
Dataset maya-multimodal/pretrain
Model maya-multimodal/maya 👏
kudos @nahidalam and team
-
- 1 reply
reacted to
merve's
post with 🚀
4 months ago
Post
3581
Apollo is a new family of open-source video language models by Meta, where 3B model outperforms most 7B models and 7B outperforms most 30B models 🧶
✨ the models come in 1.5B https://huggingface.co/Apollo-LMMs/Apollo-1_5B-t32, 3B https://huggingface.co/Apollo-LMMs/Apollo-3B-t32 and 7B https://huggingface.co/Apollo-LMMs/Apollo-7B-t32 with A2.0 license, based on Qwen1.5 & Qwen2
✨ the authors also release a benchmark dataset https://huggingface.co/spaces/Apollo-LMMs/ApolloBench
The paper has a lot of experiments (they trained 84 models!) about what makes the video LMs work ⏯️
Try the demo for best setup here https://huggingface.co/spaces/Apollo-LMMs/Apollo-3B
they evaluate sampling strategies, scaling laws for models and datasets, video representation and more!
> The authors find out that whatever design decision was applied to small models also scale properly when the model and dataset are scaled 📈 scaling dataset has diminishing returns for smaller models
> They evaluate frame sampling strategies, and find that FPS sampling is better than uniform sampling, and they find 8-32 tokens per frame optimal
> They also compare image encoders, they try a variation of models from shape optimized SigLIP to DINOv2
they find google/siglip-so400m-patch14-384 to be most powerful 🔥
> they also compare freezing different parts of models, training all stages with some frozen parts give the best yield
They eventually release three models, where Apollo-3B outperforms most 7B models and Apollo 7B outperforms 30B models 🔥
✨ the models come in 1.5B https://huggingface.co/Apollo-LMMs/Apollo-1_5B-t32, 3B https://huggingface.co/Apollo-LMMs/Apollo-3B-t32 and 7B https://huggingface.co/Apollo-LMMs/Apollo-7B-t32 with A2.0 license, based on Qwen1.5 & Qwen2
✨ the authors also release a benchmark dataset https://huggingface.co/spaces/Apollo-LMMs/ApolloBench
The paper has a lot of experiments (they trained 84 models!) about what makes the video LMs work ⏯️
Try the demo for best setup here https://huggingface.co/spaces/Apollo-LMMs/Apollo-3B
they evaluate sampling strategies, scaling laws for models and datasets, video representation and more!
> The authors find out that whatever design decision was applied to small models also scale properly when the model and dataset are scaled 📈 scaling dataset has diminishing returns for smaller models
> They evaluate frame sampling strategies, and find that FPS sampling is better than uniform sampling, and they find 8-32 tokens per frame optimal
> They also compare image encoders, they try a variation of models from shape optimized SigLIP to DINOv2
they find google/siglip-so400m-patch14-384 to be most powerful 🔥
> they also compare freezing different parts of models, training all stages with some frozen parts give the best yield
They eventually release three models, where Apollo-3B outperforms most 7B models and Apollo 7B outperforms 30B models 🔥
- 8 replies
you guys are amazing!
reacted to
sayakpaul's
post with 🚀
4 months ago
Post
2235
In the past seven days, the Diffusers team has shipped:
1. Two new video models
2. One new image model
3. Two new quantization backends
4. Three new fine-tuning scripts
5. Multiple fixes and library QoL improvements
Coffee on me if someone can guess 1 - 4 correctly.
1. Two new video models
2. One new image model
3. Two new quantization backends
4. Three new fine-tuning scripts
5. Multiple fixes and library QoL improvements
Coffee on me if someone can guess 1 - 4 correctly.
-
- 1 reply
reacted to
lorraine2's
post with 🚀
4 months ago
Post
2008
🦙New NVIDIA paper: LLaMA-Mesh 🦙
We enable large language models to generate and understand 3D meshes by representing them as text and fine-tuning. This unifies the 3D and text modalities in a single model and preserves language abilities, unlocking conversational 3D creation with mesh understanding.
🔎 Project Page: https://research.nvidia.com/labs/toronto-ai/LLaMA-Mesh/
🕹️ Interactive Demo: Zhengyi/LLaMA-Mesh (courtesy of HuggingFace and Gradio)
📖 Full Paper: https://arxiv.org/abs/2411.09595
👨💻Code: https://github.com/nv-tlabs/LLaMa-Mesh
💾 Model Checkpoint: Zhengyi/LLaMA-Mesh
🧩 Blender Addon: https://github.com/huggingface/meshgen (courtesy of Dylan Ebert)
🎥 5-min Overview Video: https://youtu.be/eZNazN-1lPo?si=-idQa5aaceVw0Bbj (courtesy of AI Papers Academy)
We enable large language models to generate and understand 3D meshes by representing them as text and fine-tuning. This unifies the 3D and text modalities in a single model and preserves language abilities, unlocking conversational 3D creation with mesh understanding.
🔎 Project Page: https://research.nvidia.com/labs/toronto-ai/LLaMA-Mesh/
🕹️ Interactive Demo: Zhengyi/LLaMA-Mesh (courtesy of HuggingFace and Gradio)
📖 Full Paper: https://arxiv.org/abs/2411.09595
👨💻Code: https://github.com/nv-tlabs/LLaMa-Mesh
💾 Model Checkpoint: Zhengyi/LLaMA-Mesh
🧩 Blender Addon: https://github.com/huggingface/meshgen (courtesy of Dylan Ebert)
🎥 5-min Overview Video: https://youtu.be/eZNazN-1lPo?si=-idQa5aaceVw0Bbj (courtesy of AI Papers Academy)
reacted to
DualityAI-RebekahBogdanoff's
post with ❤️🔥
4 months ago
Post
1978
Training YOLO with Synthetic Data from Duality AI's Falcon Simulation Software 🎮📊
Hello again! 👋 Duality.ai has released a second Google Colab and tutorial for training a YOLOv8 model using synthetic data from our Falcon simulation software!
https://falcon.duality.ai/secure/documentation/see-synth-work-no-specs?sidebarMode=learn#download-the-colab-notebook
Train using synthetic images of a soup can twin this time, and see it work on real-world images. 🥫🍜
The tutorial also walks you through how to add your own twin from our FalconCloud library, and our goal is to equip people like you to be able to create your own data for your own projects.
You'll have to create a free account to access the files, but once you do, you'll have access to not only this colab file, but also all of our lessons and our digital twin library. 🎓
Instructions for creating the synthetic data accessed by the colab notebook can be found here: https://falcon.duality.ai/secure/documentation/ex2-objdetection-newtwin?sidebarMode=learn
This method is a game-changer for cost-effective, scalable, and customizable datasets in computer vision.
Why Synthetic Data?🤔
- Precise Annotations: Get bounding boxes, segmentation masks, and more without manual effort.
- Customizable Scenarios: Get comprehensive data and cover all corner cases by simulating diverse conditions like lighting, weather, visual occlusions, and more.
What’s in the Notebook?📓
- Training & Evaluation: Train YOLOv8 with synthetic data and test its performance on real-world samples.
Let’s Discuss!💬
Check out our discord to see how people are using the Falcon simulation software to develop strong datasets and train robust models. https://discord.com/invite/dualityfalconcommunity
Hello again! 👋 Duality.ai has released a second Google Colab and tutorial for training a YOLOv8 model using synthetic data from our Falcon simulation software!
https://falcon.duality.ai/secure/documentation/see-synth-work-no-specs?sidebarMode=learn#download-the-colab-notebook
Train using synthetic images of a soup can twin this time, and see it work on real-world images. 🥫🍜
The tutorial also walks you through how to add your own twin from our FalconCloud library, and our goal is to equip people like you to be able to create your own data for your own projects.
You'll have to create a free account to access the files, but once you do, you'll have access to not only this colab file, but also all of our lessons and our digital twin library. 🎓
Instructions for creating the synthetic data accessed by the colab notebook can be found here: https://falcon.duality.ai/secure/documentation/ex2-objdetection-newtwin?sidebarMode=learn
This method is a game-changer for cost-effective, scalable, and customizable datasets in computer vision.
Why Synthetic Data?🤔
- Precise Annotations: Get bounding boxes, segmentation masks, and more without manual effort.
- Customizable Scenarios: Get comprehensive data and cover all corner cases by simulating diverse conditions like lighting, weather, visual occlusions, and more.
What’s in the Notebook?📓
- Training & Evaluation: Train YOLOv8 with synthetic data and test its performance on real-world samples.
Let’s Discuss!💬
Check out our discord to see how people are using the Falcon simulation software to develop strong datasets and train robust models. https://discord.com/invite/dualityfalconcommunity
- 2 replies