open-sci/sft__ot30k_Qwen2.5-1.5B-DPO-Tulu3-decontaminated
The open-sci/sft__ot30k_Qwen2.5-1.5B-DPO-Tulu3-decontaminated model is a 1.5 billion parameter language model, fine-tuned from ali-elganzory/Qwen2.5-1.5B-DPO-Tulu3-decontaminated. It was trained on the open_thoughts3-1.2_m_30000_samples dataset, featuring a 32K context length. This model is optimized for tasks aligned with its specific fine-tuning data, making it suitable for applications requiring nuanced language understanding and generation within its training domain.
Loading preview...
Model Overview
This model, open-sci/sft__ot30k_Qwen2.5-1.5B-DPO-Tulu3-decontaminated, is a 1.5 billion parameter language model. It is a fine-tuned variant of the ali-elganzory/Qwen2.5-1.5B-DPO-Tulu3-decontaminated base model, leveraging the Qwen2.5 architecture. The model was specifically fine-tuned on the /gpfs/scratch/ehpc524/ot/hf_hub/datasets/open-thoughts_open_thoughts3-1.2_m_30000_samples/default/0.0.0/f679a5c592c8dffb dataset, indicating a specialization towards the characteristics of this particular data.
Key Training Details
- Base Model:
ali-elganzory/Qwen2.5-1.5B-DPO-Tulu3-decontaminated - Fine-tuning Dataset:
open_thoughts3-1.2_m_30000_samples - Learning Rate: 4e-05
- Batch Size: 1 (train), 8 (eval) with 4 gradient accumulation steps, resulting in a total effective batch size of 128.
- Optimizer: AdamW_Torch_Fused with betas=(0.9, 0.999) and epsilon=1e-08.
- Epochs: 5.0
- Context Length: 32768 tokens
Potential Use Cases
Given its fine-tuning on a specific dataset, this model is likely best suited for:
- Applications requiring language generation or understanding within the domain covered by the
open_thoughts3-1.2_m_30000_samplesdataset. - Research and development exploring the impact of specific dataset fine-tuning on Qwen2.5-1.5B models.
Further details on specific intended uses and limitations would require more information about the open_thoughts3-1.2_m_30000_samples dataset content.