Warning: Training went fine until about 2.4k steps in and then it declined and started glitching. Either overcooked or in its experimental phase.

Uploaded model

  • Developed by: Lambent
  • License: apache-2.0
  • Finetuned from model : Lambent/danube3.1-4b-Reasoning-Light

This llama model was trained 2x faster with Unsloth and Huggingface's TRL library.

Downloads last month
3
Safetensors
Model size
3.96B params
Tensor type
BF16
·
Inference Providers NEW
This model is not currently available via any of the supported Inference Providers.

Model tree for Lambent/danube3.1-4b-Reasoning-1Epoch

Finetuned
(1)
this model