google/translategemma-12b-it
Image-Text-to-Text
โข
13B
โข
Updated
โข
60.2k
โข
235
UygarUsta
RivianG
AI & ML interests
Computer Vision
Recent Activity
liked
a model
9 days ago
google/translategemma-12b-it
liked
a model
10 days ago
cyankiwi/Qwen3-30B-A3B-Instruct-2507-AWQ-8bit
liked
a model
about 1 month ago
unsloth/Qwen-Image-Edit-2511-GGUF
Organizations
None yet
liked
a
model
9 days ago
liked
a
model
10 days ago
liked
a
model
about 1 month ago
reacted to
hesamation's
post with โค๏ธ
about 2 months ago
Post
2997
this is big... 50 AI researchers from Bytedance, Alibaba, Tencent, and other labs/universities just published a 300-page paper with surprising lessons about coding models and agents (data, pre and post-training, etc).
key highlights:
> small LLMs can beat proprietary giants
RL (RLVR specifically) gives small open-source models an edge over big models in reasoning. a 14B model trained with RLVR on high-quality verified problems can match the performance of OpenAI's o3.
> models have a hard time learning Python.
mixing language models during pre-training is good, but Python behaves different from statically typed languages. languages with similar syntax (Java and C#, or JavaScript and TypeScript) creates high positive synergy. mixing Python heavily into the training of statically typed languages can actually hurt because of Python's dynamic typing.
> not all languages are equal (coding scaling laws)
the amount of data required to specialize a model on a language drastically depends on the language. paper argues like C# and Java are easier to learn (less training data required). languages like Python and Javascript are actually more tricky to learn, ironically (you see AI most used for these languages :)
> MoE vs Dense (ability vs stability)
MoE models offer higher capacity, but are much more fragile during SFT than dense models. hyperparams in training have a more drastic effect in MoE models, while dense models are more stable. MoE models also require constant learning rate schedules to avoid routing instability.
> code models are "insecure" by default (duh)
training on public repos makes models learn years of accumulated insecure coding patterns. safety fine-tuning often fails to work much on code. a model might refuse to write a hate speech email but will happily generate a SQL-injection vulnerable function because it "works."
read the full paper:
From Code Foundation Models to Agents and Applications: A Practical Guide to Code Intelligence (2511.18538)
key highlights:
> small LLMs can beat proprietary giants
RL (RLVR specifically) gives small open-source models an edge over big models in reasoning. a 14B model trained with RLVR on high-quality verified problems can match the performance of OpenAI's o3.
> models have a hard time learning Python.
mixing language models during pre-training is good, but Python behaves different from statically typed languages. languages with similar syntax (Java and C#, or JavaScript and TypeScript) creates high positive synergy. mixing Python heavily into the training of statically typed languages can actually hurt because of Python's dynamic typing.
> not all languages are equal (coding scaling laws)
the amount of data required to specialize a model on a language drastically depends on the language. paper argues like C# and Java are easier to learn (less training data required). languages like Python and Javascript are actually more tricky to learn, ironically (you see AI most used for these languages :)
> MoE vs Dense (ability vs stability)
MoE models offer higher capacity, but are much more fragile during SFT than dense models. hyperparams in training have a more drastic effect in MoE models, while dense models are more stable. MoE models also require constant learning rate schedules to avoid routing instability.
> code models are "insecure" by default (duh)
training on public repos makes models learn years of accumulated insecure coding patterns. safety fine-tuning often fails to work much on code. a model might refuse to write a hate speech email but will happily generate a SQL-injection vulnerable function because it "works."
read the full paper:
From Code Foundation Models to Agents and Applications: A Practical Guide to Code Intelligence (2511.18538)
- 1 reply
liked
a
model
about 2 months ago
upvoted
a
collection
about 2 months ago
liked
a
Space
2 months ago
liked
a
model
2 months ago
updated
a
model
2 months ago
published
a
model
2 months ago
reacted to
nouamanetazi's
post with ๐ค
3 months ago
Post
4352
After training ๐๐ฆ๐จ๐ฅ๐๐๐ on ๐๐๐ ๐๐๐๐๐ฌ for nearly a month, I've come to realize something most people overlook: ๐ข๐ง๐๐ซ๐๐ฌ๐ญ๐ซ๐ฎ๐๐ญ๐ฎ๐ซ๐ ๐ข๐ฌ ๐ญ๐ก๐ ๐ฆ๐๐ค๐-๐จ๐ซ-๐๐ซ๐๐๐ค ๐๐๐๐ญ๐จ๐ซ ๐ข๐ง ๐๐๐ ๐ญ๐ซ๐๐ข๐ง๐ข๐ง๐ . ๐ฅ
Everyone talks about model architecture and data quality. And yes, those matter immensely. But here's what nobody tells you: when your training run fails at 2 AM because of mysterious ๐๐๐๐ ๐๐ซ๐ซ๐จ๐ซ๐ฌ, or when your expensive GPU cluster is running at ๐๐% ๐๐๐๐ข๐๐ข๐๐ง๐๐ฒ, the problem isn't your model. It's most probably a ๐ฆ๐ข๐ฌ๐ฎ๐ฌ๐ ๐จ๐ ๐ญ๐ก๐ ๐ก๐๐ซ๐๐ฐ๐๐ซ๐. ๐ ๏ธ
Questions that seemed simple but had no clear answers: Why is ๐๐จ๐ ๐ญ๐ซ๐๐ข๐ง๐ข๐ง๐ ๐ฌ๐ฅ๐จ๐ฐ๐๐ซ ๐ญ๐ก๐๐ง ๐๐๐ง๐ฌ๐ ๐ฆ๐จ๐๐๐ฅ๐ฌ? Which ๐๐๐๐ ๐๐ฅ๐๐ ๐ฌ should we actually set? How often should we checkpoint without killing throughput?
That's why we built ๐๐ก๐ ๐๐ฆ๐จ๐ฅ ๐๐ซ๐๐ข๐ง๐ข๐ง๐ ๐๐ฅ๐๐ฒ๐๐จ๐จ๐ค ๐: a complete guide covering everything from model architecture and data curation to the SmolLM3 training marathon, post-training techniques, and crucially, the ๐ข๐ง๐๐ซ๐๐ฌ๐ญ๐ซ๐ฎ๐๐ญ๐ฎ๐ซ๐ ๐ฅ๐๐ฒ๐๐ซ that most teams get wrong.
We validated real vs theoretical bandwidth across the entire stack: ๐๐๐๐ ๐ก๐ข๐ญ๐ญ๐ข๐ง๐ ๐ ๐๐/๐ฌ, ๐๐๐๐ข๐ง๐ค ๐.๐ ๐ซ๐๐๐๐ก๐ข๐ง๐ ๐๐๐ ๐๐/๐ฌ, ๐๐๐๐ ๐๐๐ง๐ ๐๐ญ ๐๐.๐ ๐๐/๐ฌ. Then we ran collective operations across ๐๐๐ ๐๐๐๐ฌ (16 nodes, 8xH100s each) and measured how performance degrades at scale: all-reduce drops from ๐๐๐ ๐๐/๐ฌ on a single node to ๐๐๐-๐๐๐ ๐๐/๐ฌ across 16 nodes.
If you've ever wondered why your training runs are slower than they should be, or you're planning to scale up and want to avoid expensive mistakes, this guide might save you weeks of debugging.
๐๐ก๐ ๐๐ฆ๐จ๐ฅ ๐๐ซ๐๐ข๐ง๐ข๐ง๐ ๐๐ฅ๐๐ฒ๐๐จ๐จ๐ค: https://lnkd.in/e5MKXUHS
Shared with โค๏ธ by the HuggingFace team
Everyone talks about model architecture and data quality. And yes, those matter immensely. But here's what nobody tells you: when your training run fails at 2 AM because of mysterious ๐๐๐๐ ๐๐ซ๐ซ๐จ๐ซ๐ฌ, or when your expensive GPU cluster is running at ๐๐% ๐๐๐๐ข๐๐ข๐๐ง๐๐ฒ, the problem isn't your model. It's most probably a ๐ฆ๐ข๐ฌ๐ฎ๐ฌ๐ ๐จ๐ ๐ญ๐ก๐ ๐ก๐๐ซ๐๐ฐ๐๐ซ๐. ๐ ๏ธ
Questions that seemed simple but had no clear answers: Why is ๐๐จ๐ ๐ญ๐ซ๐๐ข๐ง๐ข๐ง๐ ๐ฌ๐ฅ๐จ๐ฐ๐๐ซ ๐ญ๐ก๐๐ง ๐๐๐ง๐ฌ๐ ๐ฆ๐จ๐๐๐ฅ๐ฌ? Which ๐๐๐๐ ๐๐ฅ๐๐ ๐ฌ should we actually set? How often should we checkpoint without killing throughput?
That's why we built ๐๐ก๐ ๐๐ฆ๐จ๐ฅ ๐๐ซ๐๐ข๐ง๐ข๐ง๐ ๐๐ฅ๐๐ฒ๐๐จ๐จ๐ค ๐: a complete guide covering everything from model architecture and data curation to the SmolLM3 training marathon, post-training techniques, and crucially, the ๐ข๐ง๐๐ซ๐๐ฌ๐ญ๐ซ๐ฎ๐๐ญ๐ฎ๐ซ๐ ๐ฅ๐๐ฒ๐๐ซ that most teams get wrong.
We validated real vs theoretical bandwidth across the entire stack: ๐๐๐๐ ๐ก๐ข๐ญ๐ญ๐ข๐ง๐ ๐ ๐๐/๐ฌ, ๐๐๐๐ข๐ง๐ค ๐.๐ ๐ซ๐๐๐๐ก๐ข๐ง๐ ๐๐๐ ๐๐/๐ฌ, ๐๐๐๐ ๐๐๐ง๐ ๐๐ญ ๐๐.๐ ๐๐/๐ฌ. Then we ran collective operations across ๐๐๐ ๐๐๐๐ฌ (16 nodes, 8xH100s each) and measured how performance degrades at scale: all-reduce drops from ๐๐๐ ๐๐/๐ฌ on a single node to ๐๐๐-๐๐๐ ๐๐/๐ฌ across 16 nodes.
If you've ever wondered why your training runs are slower than they should be, or you're planning to scale up and want to avoid expensive mistakes, this guide might save you weeks of debugging.
๐๐ก๐ ๐๐ฆ๐จ๐ฅ ๐๐ซ๐๐ข๐ง๐ข๐ง๐ ๐๐ฅ๐๐ฒ๐๐จ๐จ๐ค: https://lnkd.in/e5MKXUHS
Shared with โค๏ธ by the HuggingFace team
liked
a
model
3 months ago
liked
2
models
4 months ago
reacted to
sergiopaniego's
post with ๐ฅ
4 months ago
Post
1496
Super nice intro to fine-tuning with TRL, just dropped by
@google
(runs free on Colab)!
They use SFT + QLoRA to fine-tune the tiny Gemma 3 270M model for emoji generation
Hereโs what the fine-tuned model generates for the prompt: โI'm learning to tweetโ โ ๐ฆ๐ฃ๐ป
Colab: https://colab.research.google.com/github/google-gemini/gemma-cookbook/blob/main/Demos/Emoji-Gemma-on-Web/resources/Fine_tune_Gemma_3_270M_for_emoji_generation.ipynb
Try it out: google/emoji-gemma
Learn more: https://developers.googleblog.com/en/own-your-ai-fine-tune-gemma-3-270m-for-on-device/
They use SFT + QLoRA to fine-tune the tiny Gemma 3 270M model for emoji generation
Hereโs what the fine-tuned model generates for the prompt: โI'm learning to tweetโ โ ๐ฆ๐ฃ๐ป
Colab: https://colab.research.google.com/github/google-gemini/gemma-cookbook/blob/main/Demos/Emoji-Gemma-on-Web/resources/Fine_tune_Gemma_3_270M_for_emoji_generation.ipynb
Try it out: google/emoji-gemma
Learn more: https://developers.googleblog.com/en/own-your-ai-fine-tune-gemma-3-270m-for-on-device/
reacted to
hba123's
post with ๐ฅ
4 months ago
Post
4062
๐ค What if building your own robot arm costs less than ยฃ220?
For years, robotics has been locked behind high prices and complex systems.
So we decided to change that.
Today, weโre open-sourcing Ark-Bot โ a fully 3D-printed, 6-DOF robot arm that works seamlessly with our Python robotics library, Ark.
And yesโฆ Itโs only ยฃ215.86 to build.
๐ง ArkBot Specs ๐ง
1๏ธโฃ Reach: 1 meter
2๏ธโฃ Weight: 2.6 kg
3๏ธโฃ Payload: 1.8 kg ๐ช
4๏ธโฃ DOF: 6
5๏ธโฃ Input Voltage: DC 12V
๐คFully 3D-printable & open-source
๐คIntegrated with Ark โ no ROS required
๐น Weโve also released a video showing the full assembly process โ because robotics should be something everyone can learn, build, and improve on.
๐ฉโ๐ With Ark-Bot, anyone โ from students to AI researchers โ can experiment with embodied AI, robot learning, and control algorithms on real hardware, affordably.
If you could control a 1-meter robot arm from your laptop for under ยฃ220โฆ
๐ What would you build first?
๐https://github.com/Robotics-Ark/ark_bot
๐ฅ https://www.youtube.com/watch?v=Kuk4pC0EaEw&feature=youtu.be
For years, robotics has been locked behind high prices and complex systems.
So we decided to change that.
Today, weโre open-sourcing Ark-Bot โ a fully 3D-printed, 6-DOF robot arm that works seamlessly with our Python robotics library, Ark.
And yesโฆ Itโs only ยฃ215.86 to build.
๐ง ArkBot Specs ๐ง
1๏ธโฃ Reach: 1 meter
2๏ธโฃ Weight: 2.6 kg
3๏ธโฃ Payload: 1.8 kg ๐ช
4๏ธโฃ DOF: 6
5๏ธโฃ Input Voltage: DC 12V
๐คFully 3D-printable & open-source
๐คIntegrated with Ark โ no ROS required
๐น Weโve also released a video showing the full assembly process โ because robotics should be something everyone can learn, build, and improve on.
๐ฉโ๐ With Ark-Bot, anyone โ from students to AI researchers โ can experiment with embodied AI, robot learning, and control algorithms on real hardware, affordably.
If you could control a 1-meter robot arm from your laptop for under ยฃ220โฆ
๐ What would you build first?
๐https://github.com/Robotics-Ark/ark_bot
๐ฅ https://www.youtube.com/watch?v=Kuk4pC0EaEw&feature=youtu.be
- 2 replies
reacted to
sergiopaniego's
post with ๐ฅ
4 months ago
Post
3007
A few days ago, Thinking Machines Lab released โLoRA Without Regretโ, showing that LoRA can match full fine-tuning performance when configured right.
Naturally, we decided to reproduce the results with TRL and release a guide!
https://huggingface.co/docs/trl/main/en/lora_without_regret
Naturally, we decided to reproduce the results with TRL and release a guide!
https://huggingface.co/docs/trl/main/en/lora_without_regret
liked
a
model
4 months ago
upvoted
an
article
4 months ago
Article
Smol2Operator: Post-Training GUI Agents for Computer Use
- +3
โข
135
liked
a
dataset
4 months ago