DeepScaleR: Surpassing O1-Preview with a 1.5B Model by Scaling RLpretty-radio-b75.notion.site19 points·by mluo·tahun lalu·0 comments
mluo·tahun lalu·discussCheck out one of my prior work: https://stylus-diffusion.github.io/This work scales up selection/routing over many models/LoRAs
mluo·tahun lalu·discussFor quantization, very big impact for small models, can drop at much as 10% on AIME. Our model does best on bfloat16 ;)Come checkout our repo at: https://github.com/agentica-project/deepscaler
mluo·tahun lalu·discussIt's simply bc the model is small (1.5B), making it sensitive to weight perturbations
mluo·tahun lalu·discussThink there are some people who made GGUFs as branches of our model, try it out!https://huggingface.co/models?other=base_model:quantized:age...
mluo·tahun lalu·discussNice, very glad to see it works! Small models are very sensitive to the dtype :(
mluo·tahun lalu·discussWe beat O1-preview and even many other 7B models over many math benchmarks, which was TEST set (not in training set at all).If you want to make the model fully generalist, feel free to train it over coding datasets (such as RL with passing unit tests as reward).
mluo·tahun lalu·discussHi, one of the lead authors for this work.We recommend using Bfloat16 (not fp16), quantization for small models can really hurt performance!
mluo·3 tahun yang lalu·discussWith inflation in mind, wouldn't there a larger gap between Ray's sort and the previous WR holder?
This work scales up selection/routing over many models/LoRAs