Y Hacker News new | ask | show | jobs

user: kashifr created: 2011-03-11 karma: 750 https://github.com/kashif https://twitter.com/krasul submissions: Carbon: Autoregressive Genomic Foundation Model 7 points | 1 comments The ultimate guide to RL environments: building and scaling them in the LLM era 7 points | 0 comments Distilling 100B+ Models 40x Faster with TRL 13 points | 0 comments 0 points | 0 comments Keep the Tokens Flowing: Lessons from 16 Open-Source RL Libraries 2 points | 0 comments Transformers V5 is out! 10 points | 0 comments The Smol Training Playbook: The Secrets to Building World-Class LLMs 265 points | 19 comments Unlocking On-Policy Distillation for Any Model Family 6 points | 1 comments Transformers 4.55 New OpenAI GPT OSS 2 points | 1 comments Smollm3: Smol, multilingual, long-context reasoner LLM 388 points | 79 comments 0 points | 0 comments Epic vs. Apple 7 points | 0 comments AIMO (AI Math Olympiad) progress prize winning solution 9 points | 0 comments 0 points | 0 comments MaPO: A reference-free alignment technique for diffusion models 2 points | 1 comments 0 points | 0 comments OpenHermesPreferences: Dataset of ~1M AI preferences from teknium/OpenHermes-2.5 7 points | 1 comments HuggingFace Training Cluster as a Service 101 points | 45 comments HuggingFace 235M series D at a $4.5B valuation 3 points | 0 comments Fine-tune Llama 2 with DPO 3 points | 0 comments 0 points | 0 comments QLoRA 4-bit finetuning of LLMs 7 points | 1 comments 0 points | 0 comments 0 points | 0 comments StackLlama: A hands-on guide to train LlaMa with RLHF 165 points | 38 comments 0 points | 0 comments HuggingFace Diffusers 0.2 with Stable Diffusion pipeline 2 points | 1 comments 0 points | 0 comments Diffusers: Modular Diffusion model library from HuggingFace 47 points | 5 comments