Model Overview

This model, exp-syh-r2egym-swesmith-mixed_glm_4_7_traces_jupiter, is an 8 billion parameter language model derived from the Qwen/Qwen3-8B architecture. It has undergone a specific fine-tuning process using a unique dataset located at /data/cat/ws/befe330h-befe330h-otagent/huggingface/hub/datasets--DCAgent--exp-syh-r2egym-swesmith-mixed_glm_4.7_traces_jupiter/snapshots/97638d480d61a3575e634d80606a58bfc6a0f9e_thinking_preprocessed.

Training Details

The fine-tuning process involved several key hyperparameters:

• Learning Rate: 4e-05
• Batch Size: 1 (train), 8 (eval)
• Gradient Accumulation Steps: 2
• Optimizer: ADAMW_TORCH_FUSED with betas=(0.9, 0.98) and epsilon=1e-08
• LR Scheduler: cosine with 0.1 warmup ratio
• Epochs: 7.0

This model's specific training on a custom dataset suggests it may be optimized for tasks related to the content and structure of that particular data. Further details on its intended uses, limitations, and specific capabilities would require more information about the training dataset's nature.