Post
229
At takara I'm constantly reading papers, I wonder if anyone can train a model to predict popular papers on our dataset?
takara-ai/daily-papers-popularity
takara-ai/daily-papers-popularity
- 1 reply
Jordan Legg PRO
takarajordan
AI & ML interests
Chief AI Officer @takara.ai. Diffusion, Inference optimisation and all things MultiModal.
Organizations
posted an update 6 months ago
reacted to danielhanchen's post with 🔥 6 months ago
Post
2936
You can now do reinforcement learning training with 7× longer context and no accuracy loss, via our new batching algorithms.
Long reasoning chains in RL are costly, but now we enable you to train gpt-oss with GRPO & reach 380K context on a 192GB GPU.
Blog: https://unsloth.ai/docs/new/grpo-long-context
Long reasoning chains in RL are costly, but now we enable you to train gpt-oss with GRPO & reach 380K context on a 192GB GPU.
Blog: https://unsloth.ai/docs/new/grpo-long-context
posted an update 8 months ago
Post
3209
Two weeks ago I had an engaging discussion with locals in Cockermouth about AI and the broader industry, a reminder that hearing candid perspectives beyond our professional circles is invaluable and something anyone working full-time in this field should make time for.
Thank you!
Thank you!
posted an update 8 months ago
Post
270
🌞 LOVABLE IS CRACKED
Built a golden hour tracker in under 15 minutes with Lovable: uses your phone’s Geolocation API, the SunCalc library, and runs fully client-side with no servers. https://goldenhour.404missing.link
Built a golden hour tracker in under 15 minutes with Lovable: uses your phone’s Geolocation API, the SunCalc library, and runs fully client-side with no servers. https://goldenhour.404missing.link
posted an update 10 months ago
Post
470
Yay I made an in memory vector DB in pure golang, check it out here https://github.com/takara-ai/serverlessVector
posted an update 10 months ago
Post
2644
Are we really back to storing access tokens in plain text again?
{
"mcpServers": {
"hf-mcp-server": {
"url": "https://huggingface.co/mcp",
"headers": {
"Authorization": "Bearer <YOUR_HF_TOKEN>"
}
}
}
}
- 3 replies
posted an update 10 months ago
Post
3041
I'm currently looking into what makes a scientific paper more popular than others on a platform like Hugging Face. I conducted a huge array of tests, content length, time based information even semantic feature extraction to get to some sort of answer around...
What actually drives popularity of these papers, why do some papers get zero upvotes and why do some get thousands?
The answer is absolutely nothing. Yes that's right. Nothing about the actual paper itself drives popularity, the paper's popularity is driven by external factors like it's authors, external marketing and others.
So next time you see a research paper with a lot of upvotes, just remember it's not because of the efforts of the authors. Remain objective.
What actually drives popularity of these papers, why do some papers get zero upvotes and why do some get thousands?
The answer is absolutely nothing. Yes that's right. Nothing about the actual paper itself drives popularity, the paper's popularity is driven by external factors like it's authors, external marketing and others.
So next time you see a research paper with a lot of upvotes, just remember it's not because of the efforts of the authors. Remain objective.
reacted to tomaarsen's post with ❤️ 11 months ago
Post
4569
😎 I just published Sentence Transformers v5.1.0, and it's a big one. 2x-3x speedups of SparseEncoder models via ONNX and/or OpenVINO backends, easier distillation data preparation with hard negatives mining, and more:
1️⃣ Faster ONNX and OpenVINO backends for SparseEncoder models
Usage is as simple as
2️⃣ New
This new output format is immediately compatible with the MarginMSELoss and SparseMarginMSELoss for training SentenceTransformer, CrossEncoder, and SparseEncoder losses.
3️⃣ Gathering across devices
When doing multi-GPU training using a loss that has in-batch negatives (e.g. MultipleNegativesRankingLoss), you can now use
4️⃣ Trackio support
If you also upgrade
5️⃣ MTEB Documentation
We've added some documentation on evaluating SentenceTransformer models properly with MTEB. It's rudimentary as the documentation on the MTEB side is already great, but it should get you started.
Plus many more smaller features & fixes (crash fixes, compatibility with datasets v4, FIPS compatibility, etc.).
See the full release notes here: https://github.com/UKPLab/sentence-transformers/releases/tag/v5.1.0
Big thanks to all of the contributors for helping with the release, many of the features from this release were proposed by others. I have a big list of future potential features that I'd love to add, but I'm
1️⃣ Faster ONNX and OpenVINO backends for SparseEncoder models
Usage is as simple as
backend="onnx" or backend="openvino" when initializing a SparseEncoder to get started, but I also included utility functions for optimization, dynamic quantization, and static quantization, plus benchmarks.2️⃣ New
n-tuple-scores output format from mine_hard_negativesThis new output format is immediately compatible with the MarginMSELoss and SparseMarginMSELoss for training SentenceTransformer, CrossEncoder, and SparseEncoder losses.
3️⃣ Gathering across devices
When doing multi-GPU training using a loss that has in-batch negatives (e.g. MultipleNegativesRankingLoss), you can now use
gather_across_devices=True to load in-batch negatives from the other devices too! Essentially a free lunch, pretty big impact potential in my evals.4️⃣ Trackio support
If you also upgrade
transformers, and you install trackio with pip install trackio, then your experiments will also automatically be tracked locally with trackio. Just open up localhost and have a look at your losses/evals, no logins, no metric uploading.5️⃣ MTEB Documentation
We've added some documentation on evaluating SentenceTransformer models properly with MTEB. It's rudimentary as the documentation on the MTEB side is already great, but it should get you started.
Plus many more smaller features & fixes (crash fixes, compatibility with datasets v4, FIPS compatibility, etc.).
See the full release notes here: https://github.com/UKPLab/sentence-transformers/releases/tag/v5.1.0
Big thanks to all of the contributors for helping with the release, many of the features from this release were proposed by others. I have a big list of future potential features that I'd love to add, but I'm
posted an update 11 months ago
Post
293
What do you all actually think about the open source OpenAI models? Are they legitimately any good or are they hype?
- 3 replies
posted an update about 1 year ago
Post
389
Cool to see the new model lightonai/Reason-ModernColBERT
Made with late interaction I'd love to recreate the dataset to see a proper apache 2.0 version!
Made with late interaction I'd love to recreate the dataset to see a proper apache 2.0 version!
reacted to clem's post with ❤️ about 1 year ago
Post
4136
What are you using to evaluate models or AI systems? So far we're building lighteval & leaderboards on the hub but still feels early & a lot more to build. What would be useful to you?
- 6 replies
I'm using https://artificialanalysis.ai/ just because it puts everything in one place! It's not the best resource but these days I'm all about saving time.
replied to their post about 1 year ago
@ThomasTheMaker if you make an issue on the repo, I'll look into it!
replied to their post about 1 year 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 about 1 year ago
Post
647
🎌 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 over 1 year ago
Post
1599
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!