Finance Lora Deepseek R1 1.5b

Finance Lora Deepseek R1 1.5b is a LoRA adapter designed for use with the DeepSeek-R1-Distill-Qwen-1.5B-4bit model. The adapter is fine-tuned on the gbharti/finance-alpaca financial instruction dataset, with the goal of improving performance on financial instruction tasks. The model card notes that test perplexity decreased from 10.846 (base) to 9.632 (tuned) across three seeds, indicating an improvement after fine-tuning. The tool can be used with MLX, and instructions are provided for downloading the model from the Hugging Face Hub and integrating it into local apps or environments such as Google Colab, Kaggle, and LM Studio. Usage involves loading the base model and applying the LoRA adapter via the provided adapter path. The evidence does not specify the intended audience, licensing terms, or pricing model. There is no mention of additional features, integrations, or supported platforms beyond those listed. Finance Lora Deepseek R1 1.5b is positioned as a model adapter within the foundation-models class, specifically tailored for financial instruction datasets. Further details about its broader capabilities or user scenarios are not provided in the available evidence.