Model Overview

Concyclics/PeoplesDaily-Qwen3-4B-Base is a 4 billion parameter language model built upon the Qwen3 architecture, designed with a substantial context length of 40960 tokens. This model has been specifically enhanced through Supervised Fine-Tuning (SFT) using the Concyclics/PeoplesDaily dataset.

Training Details

The SFT process involved:

• Batch Size: 96
• Epochs: 2
• Learning Rate: 1.0e-5
• LR Scheduler Type: Cosine
• Warmup Ratio: 0.1
• Total FLOPs: 483 TFlops
• Final Training Loss: 1.646

Potential Use Cases

Given its fine-tuning on the PeoplesDaily dataset, this model is likely well-suited for applications requiring:

• Processing and generating text related to news and public discourse, particularly from the PeoplesDaily domain.
• Tasks involving analysis or summarization of content similar to the training data.
• Applications where a model with a large context window and specific domain adaptation is beneficial.