DCAgent/g1_weighted_31600_8b_orig
DCAgent/g1_weighted_31600_8b_orig is an 8 billion parameter language model fine-tuned from Qwen/Qwen3-8B. This model was specifically trained on a weighted dataset derived from 'g1_min_episodes_e1_weighted_top4_31600_glm47_traces' for specialized applications. It leverages a 32K context window, making it suitable for tasks requiring processing of longer sequences of text. The fine-tuning process focused on specific data, suggesting optimization for particular domain-specific or agentic behaviors.
Loading preview...
Model Overview
DCAgent/g1_weighted_31600_8b_orig is an 8 billion parameter language model, fine-tuned from the base Qwen/Qwen3-8B architecture. This model was developed by DCAgent through a specialized fine-tuning process, utilizing a unique dataset /e/scratch/jureap59/raoof1/sft_data/hf_hub/datasets--DCAgent--g1_min_episodes_e1_weighted_top4_31600_glm47_traces. It supports a context length of 32,768 tokens.
Key Characteristics
- Base Model: Qwen3-8B, a robust foundation for language understanding and generation.
- Fine-tuning Focus: Trained on a specific, weighted dataset, indicating an optimization for particular tasks or domain-specific interactions, likely related to agentic or trace-based data.
- Context Window: Features a substantial 32,768-token context length, enabling the model to process and understand extensive inputs.
Training Details
The model was trained with a learning rate of 4e-05, using a total batch size of 48 across 48 devices. The training spanned 7 epochs, employing an AdamW optimizer with cosine learning rate scheduling and a 0.1 warmup ratio. This configuration suggests a focused and efficient fine-tuning approach for its intended specialized use.