PyTorchModelHubMixin: Bridging the Gap for Custom AI Models on Hugging Face
β’
9
Lain
not-lain
AI & ML interests
custom AI models with HF integration, multimodal rag and open-source contributions
Articles
Organizations
not-lain's activity
reacted to
DavidGF's
post with π₯
8 days ago
Post
2915
π Celebrating One Year of #SauerkrautLM with Two Groundbreaking Releases!
We're thrilled to announce the release of SauerkrautLM-v2-14b in two specialized versions: VAGOsolutions/SauerkrautLM-v2-14b-SFT and VAGOsolutions/SauerkrautLM-v2-14b-DPO. Built on the robust Qwen2.5-14B foundation, these models represent a significant leap forward in multilingual AI capabilities.
π¬ Technical Breakthroughs:
π Innovative three-phase Fine-Tuning approach
π Two-step Spectrum SFT + one-step Spectrum DPO optimization phase for enhanced performance
π Balance of German and English language capabilities
π Advanced function calling - almost on par with Claude-3.5-Sonnet-20240620
π©πͺ German Language Excellence:
What sets this release apart is our unique achievement in simultaneously improving both German and English capabilities. Through our specialized training approach with over 1.2B tokens across two phases, we've managed to:
π Enhance German language understanding and generation (SFT Version > DPO Version)
π Maintain authentic German linguistic nuances
π Improve cross-lingual capabilities
π Preserve cultural context awareness
π Training Innovation:
Our three-phase approach targeted specific layer percentages (15%, 20% and 25%) with carefully curated datasets, including:
π Mathematics-focused content (proprietary classifier-selected)
π High-quality German training data
π Specialized function calling datasets
π Premium multilingual content
π Community Contribution:
We're also releasing two new datasets in a few days:
1οΈβ£ SauerkrautLM-Fermented-GER-DPO: 3,300 high-quality German training samples
2οΈβ£ SauerkrautLM-Fermented-Irrelevance-GER-DPO: 2,000 specialized samples for optimized function call irrelevance handling
Thank you to our incredible community and partners who have supported us throughout this journey. Here's to another year of AI innovation!Β π
We're thrilled to announce the release of SauerkrautLM-v2-14b in two specialized versions: VAGOsolutions/SauerkrautLM-v2-14b-SFT and VAGOsolutions/SauerkrautLM-v2-14b-DPO. Built on the robust Qwen2.5-14B foundation, these models represent a significant leap forward in multilingual AI capabilities.
π¬ Technical Breakthroughs:
π Innovative three-phase Fine-Tuning approach
π Two-step Spectrum SFT + one-step Spectrum DPO optimization phase for enhanced performance
π Balance of German and English language capabilities
π Advanced function calling - almost on par with Claude-3.5-Sonnet-20240620
π©πͺ German Language Excellence:
What sets this release apart is our unique achievement in simultaneously improving both German and English capabilities. Through our specialized training approach with over 1.2B tokens across two phases, we've managed to:
π Enhance German language understanding and generation (SFT Version > DPO Version)
π Maintain authentic German linguistic nuances
π Improve cross-lingual capabilities
π Preserve cultural context awareness
π Training Innovation:
Our three-phase approach targeted specific layer percentages (15%, 20% and 25%) with carefully curated datasets, including:
π Mathematics-focused content (proprietary classifier-selected)
π High-quality German training data
π Specialized function calling datasets
π Premium multilingual content
π Community Contribution:
We're also releasing two new datasets in a few days:
1οΈβ£ SauerkrautLM-Fermented-GER-DPO: 3,300 high-quality German training samples
2οΈβ£ SauerkrautLM-Fermented-Irrelevance-GER-DPO: 2,000 specialized samples for optimized function call irrelevance handling
Thank you to our incredible community and partners who have supported us throughout this journey. Here's to another year of AI innovation!Β π
reacted to
reach-vb's
post with π₯π
9 days ago
Post
2888
Smol models ftw! AMD released AMD OLMo 1B - beats OpenELM, tiny llama on MT Bench, Alpaca Eval - Apache 2.0 licensed π₯
> Trained with 1.3 trillion (dolma 1.7) tokens on 16 nodes, each with 4 MI250 GPUs
> Three checkpoints:
- AMD OLMo 1B: Pre-trained model
- AMD OLMo 1B SFT: Supervised fine-tuned on Tulu V2, OpenHermes-2.5, WebInstructSub, and Code-Feedback datasets
- AMD OLMo 1B SFT DPO: Aligned with human preferences using Direct Preference Optimization (DPO) on UltraFeedback dataset
Key Insights:
> Pre-trained with less than half the tokens of OLMo-1B
> Post-training steps include two-phase SFT and DPO alignment
> Data for SFT:
- Phase 1: Tulu V2
- Phase 2: OpenHermes-2.5, WebInstructSub, and Code-Feedback
> Model checkpoints on the Hub & Integrated with Transformers β‘οΈ
Congratulations & kudos to AMD on a brilliant smol model release! π€
amd/amd-olmo-6723e7d04a49116d8ec95070
> Trained with 1.3 trillion (dolma 1.7) tokens on 16 nodes, each with 4 MI250 GPUs
> Three checkpoints:
- AMD OLMo 1B: Pre-trained model
- AMD OLMo 1B SFT: Supervised fine-tuned on Tulu V2, OpenHermes-2.5, WebInstructSub, and Code-Feedback datasets
- AMD OLMo 1B SFT DPO: Aligned with human preferences using Direct Preference Optimization (DPO) on UltraFeedback dataset
Key Insights:
> Pre-trained with less than half the tokens of OLMo-1B
> Post-training steps include two-phase SFT and DPO alignment
> Data for SFT:
- Phase 1: Tulu V2
- Phase 2: OpenHermes-2.5, WebInstructSub, and Code-Feedback
> Model checkpoints on the Hub & Integrated with Transformers β‘οΈ
Congratulations & kudos to AMD on a brilliant smol model release! π€
amd/amd-olmo-6723e7d04a49116d8ec95070
reacted to
nroggendorff's
post with π€―
10 days ago
Post
1791
Did you guys know that if you try to link a prepaid card to huggingface it won't work, but then if you press the button again it links anyway? Then you can lock the card (deny any charges), and get resources for free? You're welcome :P
- 4 replies
reacted to
merve's
post with β€οΈπ₯
12 days ago
Post
5147
Another great week in open ML!
Here's a small recap π«°π»
Model releases
β―οΈ Video Language Models
AI at Meta released Vision-CAIR/LongVU_Qwen2_7B, a new state-of-the-art long video LM model based on DINOv2, SigLIP, Qwen2 and Llama 3.2
π¬ Small language models
Hugging Face released HuggingFaceTB/SmolLM2-1.7B, a family of new smol language models with Apache 2.0 license that come in sizes 135M, 360M and 1.7B, along with datasets.
Meta released facebook/MobileLLM-1B, a new family of on-device LLMs of sizes 125M, 350M and 600M
πΌοΈ Image Generation
Stability AI released stabilityai/stable-diffusion-3.5-medium, a 2B model with commercially permissive license
πΌοΈπ¬Any-to-Any
gpt-omni/mini-omni2 is closest reproduction to GPT-4o, a new LLM that can take image-text-audio input and output speech is released!
Dataset releases
πΌοΈ Spawning/PD12M, a new captioning dataset of 12.4 million examples generated using Florence-2
Here's a small recap π«°π»
Model releases
β―οΈ Video Language Models
AI at Meta released Vision-CAIR/LongVU_Qwen2_7B, a new state-of-the-art long video LM model based on DINOv2, SigLIP, Qwen2 and Llama 3.2
π¬ Small language models
Hugging Face released HuggingFaceTB/SmolLM2-1.7B, a family of new smol language models with Apache 2.0 license that come in sizes 135M, 360M and 1.7B, along with datasets.
Meta released facebook/MobileLLM-1B, a new family of on-device LLMs of sizes 125M, 350M and 600M
πΌοΈ Image Generation
Stability AI released stabilityai/stable-diffusion-3.5-medium, a 2B model with commercially permissive license
πΌοΈπ¬Any-to-Any
gpt-omni/mini-omni2 is closest reproduction to GPT-4o, a new LLM that can take image-text-audio input and output speech is released!
Dataset releases
πΌοΈ Spawning/PD12M, a new captioning dataset of 12.4 million examples generated using Florence-2
reacted to
vikhyatk's
post with π₯
16 days ago
Post
4055
Pushed a new update to
vikhyatk/moondream2 today. TextVQA up from 60.2 to 65.2, DocVQA up from 61.9 to 70.5.
Space has been updated to the new model if you want to try it out! vikhyatk/moondream2
Space has been updated to the new model if you want to try it out! vikhyatk/moondream2
reacted to
vikhyatk's
post with π₯π
16 days ago
Post
1237
Just released a dataset with 7000+ hours of synthetically generated lo-fi music.
vikhyatk/lofi
loved it β€οΈ
reacted to
merve's
post with β€οΈππ₯
16 days ago
Post
5077
Hugging Face Hub Python library now comes with easy inference for vision language models! β¨
$ pip install huggingface_hub π€
$ pip install huggingface_hub π€
-
- 1 reply
reacted to
nroggendorff's
post with π
17 days ago
Post
3287
@echo off
echo hello world
pause
- 3 replies
reacted to
fdaudens's
post with β€οΈ
18 days ago
Post
2749
π€― Plot twist: Size isn't everything in AI! A lean 32B parameter model just showed up to the party and outperformed a 70B one. Efficiency > Scale? The AI world just got more interesting...
Cohere For AI released Aya Expanse, a new family of multilingual models (8B and 32B) spanning 23 popular languages.
Models: CohereForAI/c4ai-aya-expanse-671a83d6b2c07c692beab3c3
Blog post: https://huggingface.co/blog/aya-expanse
Demo: CohereForAI/aya_expanse
Cohere For AI released Aya Expanse, a new family of multilingual models (8B and 32B) spanning 23 popular languages.
Models: CohereForAI/c4ai-aya-expanse-671a83d6b2c07c692beab3c3
Blog post: https://huggingface.co/blog/aya-expanse
Demo: CohereForAI/aya_expanse
reacted to
merve's
post with β€οΈπ
19 days ago
Post
2404
Lotus πͺ· is a new foundation model on monocular depth estimation β¨
Compared to previous diffusion-based MDE models, Lotus is modified for dense prediction tasks
Authors also released a model for normal prediction π€
Find everything in this collection merve/lotus-6718fb957dc1c85a47ca1210
Compared to previous diffusion-based MDE models, Lotus is modified for dense prediction tasks
Authors also released a model for normal prediction π€
Find everything in this collection merve/lotus-6718fb957dc1c85a47ca1210
reacted to
thomwolf's
post with πβ€οΈ
19 days ago
Post
3883
Parents in the 1990: Teach the kids to code
Parents now: Teach the kids to fix the code when it starts walking around π€β¨
Parents now: Teach the kids to fix the code when it starts walking around π€β¨
- 2 replies
reacted to
fdaudens's
post with π
20 days ago
Post
1341
Just watched
@thomwolf
tear down the over-hyped AGI narrative in 30 seconds - and it's refreshingly grounded.
No wild speculation about superintelligence timelines or consciousness. Just practical insights from someone who really understands the technology.
This is the kind of level-headed perspective that helps us focus on what AI can actually do today (which is already transformative) rather than getting lost in AGI fantasy. Worth your time if you want to understand AI progress without the hype.
Watch the full interview at CogX here: https://www.youtube.com/watch?v=IjL_6Th6Ea0
No wild speculation about superintelligence timelines or consciousness. Just practical insights from someone who really understands the technology.
This is the kind of level-headed perspective that helps us focus on what AI can actually do today (which is already transformative) rather than getting lost in AGI fantasy. Worth your time if you want to understand AI progress without the hype.
Watch the full interview at CogX here: https://www.youtube.com/watch?v=IjL_6Th6Ea0