Model Overview

This model, exp-uns-r2egym-16_8x_glm_4_7_traces_jupiter_cleaned, is an 8 billion parameter language model derived from the Qwen/Qwen3-8B architecture. It has undergone a specific fine-tuning process to adapt its capabilities to a particular dataset.

Key Characteristics

• Base Model: Fine-tuned from Qwen/Qwen3-8B.
• Parameter Count: 8 billion parameters.
• Context Length: Supports a context length of 32768 tokens.
• Training Data: Fine-tuned on the /data/cat/ws/befe330h-befe330h-otagent/huggingface/hub/datasets--DCAgent--exp-uns-r2egym-16_8x_glm_4.7_traces_jupiter_cleaned/snapshots/f33c2e87626104d2406a7182a6a713bf0b337bdb_thinking_preprocessed dataset.

Training Details

The fine-tuning process involved several specific hyperparameters:

• Learning Rate: 4e-05
• Optimizer: ADAMW_TORCH_FUSED with betas=(0.9, 0.98) and epsilon=1e-08.
• Epochs: Trained for 7.0 epochs.
• Batch Size: A total training batch size of 16 (with gradient accumulation steps of 2).

This model is specialized for tasks aligned with its unique training dataset, suggesting potential applications in areas related to the dataset's content.