RT @PyTorch: The Kubeflow Trainer project has been integrated into the PyTorch Ecosystem! This integration ensures that Kubeflow Trainer al….

RT @PyTorch: PyTorch and vLLM are both critical to the AI ecosystem and are increasingly being used together for cutting edge generative AI….

RT @LysandreJik: The Transformers library is undergoing it's largest pivot to date 🙌. It now cements its role as the central model definiti….

RT @PyTorch: PyTorch Foundation has expanded into an umbrella foundation. @vllm_project and @DeepSpeedAI have been accepted as hosted proje….

🚀 Bamba v2 (9B) is here: faster, stronger, and smarter!.A leaderboard model in just 3T tokens!!. Bamba v1 +1T tokens of training. Outperforms Llama 3.1 8B on L1 & L2 benchmark scores 📈. 2–2.5× faster inference with vLLM than standard transformer based models 🏎️. Open weights +.

RT @tri_dao: Very strong 8B and 56B Mamba hybrid models trained to 20T tokens, on 6K H100s, with FP8!.This answers many of the open questio….

Job alert: . Red Hat team is expanding folks in PyTorch! You will be working side-by-side with the inference experts @neuralmagic !.

RT @danielhanchen: Excited to share that @UnslothAI now supports:. • Full fine-tuning + 8bit.• Nearly any model like Mixtral, Cohere, Grani….

RT @PyTorch: Optimize your model training with smarter memory management!. Check out our latest blog post to learn how PyTorch’s new activa….

@Thom_Wolf @eliebakouch Yup, we are working on that!.

@Thom_Wolf @eliebakouch ⬆️.

Super excited about the release of GneissWeb dataset from IBM:

For the MoE experts, I am curious as to how many experts are triggered in a batch of sequences? If I have a batch of 16, would it be possible to trigger (16 x num active parameters) in one forward pass?. Could not find any study either 😢.@deepseek_ai @MistralAI.

@hsu_byron interestingly Mamba2 has many triton kernels :).

🚀 Exciting News! 🚀. In a joint effort between IBM Research, Princeton, CMU, and UIUC, we are thrilled to announce the release of our high-performing hybrid Mamba2 model! This model is trained entirely on open datasets, and we’re releasing intermediate and final checkpoints to

RT @PyTorch: Presenting HadaCore: Tensor Core Accelerated Hadamard Transform Kernel 🔍 Take a look at how we achieve state-of-the-art perfor….

Super kicked by this work from @PyTorch team at @AIatMeta and @IBMResearch team. @TheZachMueller , now we just need to get this into @huggingface accelerate and enable the community!. Let’s go and make training even faster!!! 🔥🔥🔥.

Great work from @IBM and @PyTorch ! This will allow for native tensor parallel using @huggingface transformers library! Long sequences are now being tamed ;).

RT @aashkaa_: I’m presenting our poster on INDUS: Effective & Efficient Language Models for Scientific Applications at #EMNLP2024 tomorrow….

RT @PyTorch: Learn the inner workings of Triton, the hardware agnostic language for GPU programming and powering TorchInductor: https://t.c….