Overview

This model, exp-uns-r2egym-4_2x_glm_4_7_traces_jupiter, is an 8 billion parameter language model derived from the Qwen/Qwen3-8B architecture. It has been fine-tuned on a specific dataset located at /data/cat/ws/befe330h-befe330h-otagent/huggingface/hub/datasets--DCAgent--exp-uns-r2egym-4_2x_glm_4.7_traces_jupiter/snapshots/755851ab1bce4a626f500aaf4e6827f1642f1699_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, leading to a total train batch size of 16
• Optimizer: AdamW_Torch_Fused with betas=(0.9, 0.98) and epsilon=1e-08
• LR Scheduler: Cosine type with a warmup ratio of 0.1
• Epochs: 7.0

Framework Versions

The model was trained using:

• Transformers 4.57.6
• Pytorch 2.9.0+cu128
• Datasets 4.4.1
• Tokenizers 0.22.2

Intended Use

While specific details on intended uses and limitations are not provided, its fine-tuning on a specialized dataset suggests it is optimized for tasks related to that data. Users should consider the nature of the training data when evaluating its suitability for their specific applications.