Hamid Shojanazeri

ml @pytorch model optimization, distributed training.

[Hamid Shojanazeri @Nazeri2010]

@AmericaSpoof yes!

[Hamid Shojanazeri @Nazeri2010]

@AmericaSpoof D

[Hamid Shojanazeri @Nazeri2010]

@trump_dona19543 Yes!

[Hamid Shojanazeri @Nazeri2010]

@Karolinleavit5 Yes!

[Hamid Shojanazeri @Nazeri2010]

@dylan522p He is a legend!

[Hamid Shojanazeri @Nazeri2010]

Here is a quick demo on Llama 4 long context!

code can be found here! https://t.co/Z8N5P8lZ4h

[Sanyam Bhutani @bhutanisanyam1]

Llama 4 takes 43 seconds to analyse 900k tokens!

[Hamid Shojanazeri @Nazeri2010]

Llama 4 now supports 10M context length — yes, 10 million tokens!

We just ran a 900k-token GitHub repo through it and generated a guide in under 3 minutes. This can unlock many applications!

Checkout the code: https://t.co/Z8N5P8lZ4h

[Amirhossein Kazemnejad @a_kazemnejad]

Exciting times for "doing actual RL" with Tulu3 and likely Llama 3.3! Heading to #NeurIPS to present VinePPO at MATHAI, where we built the right RL framework for LLM reasoning. Let’s chat about
- RL training of LLMs
- Learning to search
- Test-time scaling

DM me if interested!

[Daniel Han @danielhanchen]

Llama 3.2 Multimodal benchmarks

MMMU for 11B 60.3 vs Claude Haiku 50.2
MMMU for 90B 60.3 vs GPT 4o mini 59.4
90B looks extremely powerful!

[Hamid Shojanazeri @Nazeri2010]

🔥 New Fine-tuning Recipes are here! 🔧

⚡ TorchTune: https://t.co/01vqxkCYWN

📄 Llama Recipes: https://t.co/2EAufcog4f
Drop your feedback on the repos!

Big thanks to @huggingface, @vllm_project, and all our incredible launch partners! 🙌

[Hamid Shojanazeri @Nazeri2010]

🚀 We’re live!

The next-gen image reasoning capabilities with powerful 11B & 90B models, plus sleek on-device 1B & 3B SOTA models!

Explore how to use the Llama 3.2 models now 👇

📄 Llama Vision Guide: https://t.co/9oMdAhLxuW
📲 Llama Stack Apps: https://t.co/QFBVhNBrTN

[AI at Meta @AIatMeta]

📣 Introducing Llama 3.2: Lightweight models for edge devices, vision models and more!

What’s new?
• Llama 3.2 1B & 3B models deliver state-of-the-art capabilities for their class for several on-device use cases — with support for @Arm, @MediaTek & @Qualcomm on day one.
• Llama 3.2 11B & 90B vision models deliver performance competitive with leading closed models — and can be used as drop-in replacements for Llama 3.1 8B & 70B.
• New Llama Guard models to support multimodal use cases and edge deployments.
• The first official distro of Llama Stack simplifies and supercharges the way developers & enterprises can build around Llama to support agentic applications and more.

Details in the full announcement ➡️ https://t.co/1bnEeLY9qf
Download Llama 3.2 models ➡️ https://t.co/DZoTQvESbG

These models are available to download now directly from Meta and @HuggingFace — and will be available across offerings from 25+ partners that are rolling out starting today, including @accenture, @awscloud, @AMD, @azure, @Databricks, @Dell, @Deloitte, @FireworksAI_HQ, @GoogleCloud, @GroqInc, @IBMwatsonx, @Infosys, @Intel, @kaggle, @NVIDIA, @OracleCloud, @PwC, @scale_AI, @snowflakeDB, @togethercompute and more.

With Llama 3.2 we’re making it possible to run Llama in even more places, with even more flexible capabilities. We’ve said it before and we’ll say it again: open source AI is how we ensure that these innovations reflect the global community they’re built for and benefit everyone. We’re continuing our drive to make open source the standard with Llama 3.2.

[Hamid Shojanazeri @Nazeri2010]

@RickLamers Congrats @RickLamers!! Great work! Would love to land a tutorial on https://t.co/SiXnhhTrq0 as well!

[Hamid Shojanazeri @Nazeri2010]

The focus on #agentic workflows was a positive indicator of the community's organic direction, unlocking numerous opportunities!

Thanks to all the partners! @cerebral_valley @awscloud,@GroqInc,@wandb,@LambdaAPI,
@huggingface,@llama_index, @ollama,@brave,
@OctoAICloud

[Hamid Shojanazeri @Nazeri2010]

It was so inspiring to see the enthusiasm surrounding #Llama3, thanks everyone for hacking with us over the weekend in #Llama3Hackathon! so many cool projects!
https://t.co/I0dNeJ3HJM

[Max Jaderberg @maxjaderberg]

Super excited to be releasing AlphaFold 3 today, developed by @IsomorphicLabs and @GoogleDeepMind: our next generation AI model for predicting the biomolecular structures and interactions of proteins, DNA, RNA, small molecules, and more: https://t.co/F4qSUELVLA
1/

[Hamid Shojanazeri @Nazeri2010]

@HamelHusain Great to hear that @HamelHusain! Thanks for sharing! wondering if you have any insights into 8B as well?

[Philipp Schmid @_philschmid]

While wrapping up my FSDP + QLora Llama 3 blog, I noticed that ~80 samples (stacked to 3k sequence length) are enough for Llama 3 70B to "converge" to the instruction templates. Thats impressive!🤯

If you are using fine-tuned GPT-3.5 models, you might want to look into Llama 3 70B! 🚀

[Philipp Schmid @_philschmid]

Easily Fine-tune @AIatMeta Llama 3 70B! 🦙 I am excited to share a new guide on how to fine-tune Llama 3 70B with @PyTorch FSDP, Q-Lora, and Flash Attention 2 (SDPA) using @huggingface build for consumer-size GPUs (4x 24GB). 🚀

Blog: https://t.co/rpmnwlTnjw

The blog covers:
👨‍💻 Instructions to set up the environment & preprocess the dataset
🥇 Uses @PyTorch FSDP, Q-LoRA, and Flash Attention v2 (SDPA) for efficient distributed training
🧠 Build with @huggingface TRL, Transformers, PEFT, and Datasets
💾 Tested on @nvidia H100 and A10G GPUs (4x 24GB)
🚀 Can be adjusted to other open LLMs, like Mixtral

What are you building today? 🤗

[Hamid Shojanazeri @Nazeri2010]

@jeremyphoward @AIatMeta Congrats @jeremyphoward ! this is a awesome! it unlocks many potentials!

[Hamid Shojanazeri @Nazeri2010]

@art_zucker @rayzarion @JustinLin610 pretty cool! thanks @art_zucker for sharing! is there any further update on upstreaming changes in `generate`?