xiaolesu/OsmosisProofling-SFT-NT-GRPO-NT-No-Overlap
TEXT GENERATIONConcurrency Cost:1Model Size:8BQuant:FP8Ctx Length:32kPublished:Apr 8, 2026License:apache-2.0Architecture:Transformer Open Weights Cold

The xiaolesu/OsmosisProofling-SFT-NT-GRPO-NT-No-Overlap is an 8 billion parameter language model fine-tuned from Qwen/Qwen3-8B. This model was trained using Axolotl with specific Liger optimizations for improved performance. It is fine-tuned on the xiaolesu/OsmosisProofling-v3-SFT dataset, achieving a validation loss of 0.3543 and perplexity of 1.4252. This model is suitable for tasks requiring a Qwen3-8B base with specialized fine-tuning.

Loading preview...