OsmosisProofling-SFT: A Fine-Tuned Qwen3-8B Model

The xiaolesu/OsmosisProofling-SFT model is an 8 billion parameter language model built upon the Qwen/Qwen3-8B architecture. It has been fine-tuned by xiaolesu using the Axolotl training framework, incorporating several Liger optimizations such as liger_rope, liger_rms_norm, liger_glu_activation, liger_layer_norm, and liger_fused_linear_cross_entropy for enhanced efficiency and performance.

Key Characteristics

• Base Model: Qwen/Qwen3-8B, an 8 billion parameter model.
• Training Framework: Utilizes Axolotl 0.16.0.dev0 with specific Liger plugins.
• Dataset: Fine-tuned on the xiaolesu/OsmosisProofling-SFT dataset for both training and validation.
• Performance Metrics: Achieved a validation loss of 0.3543 and a perplexity (Ppl) of 1.4252 on its evaluation set.
• Training Configuration: Employed a sequence length of 4096, sample packing, and flex attention, with an AdamW optimizer and cosine learning rate scheduler.

Intended Use

This model is specifically tailored by its fine-tuning on the xiaolesu/OsmosisProofling-SFT dataset. Developers should consider its strong performance on this particular dataset when evaluating its suitability for tasks that align with the nature of the training data.