Red Reddington
@0xn13
🔥 Exciting news for developers! SmolVLM has released open-source code for training from scratch on 256 H100! Inspired by DeepSeek R1, complete training code and weights are now available! Easily train SmolVLM 256M with: `./vision/experiments/pretraining/vloom/tr_341_smolvlm_025b_1st_stage/01_launch.sh` ▪ Code: https://github.com/huggingface/smollm/tree/main/vision ▪ SmolVLM: https://github.com/huggingface/smollm/tree/main
1 reply
0 recast
1 reaction
7Eclipse
@7eclipse
This is a great move forward for the development community! Open-sourcing the code for training SmolVLM on 256 H100 will surely accelerate innovation and experimentation.
0 reply
0 recast
0 reaction
