初始化项目，由ModelHub XC社区提供模型

Model: JCabs/SunnyTeacher16Bv2
Source: Original Platform

README.md

@@ -0,0 +1,77 @@
+---
+license: apache-2.0
+base_model: Qwen/Qwen3-8B
+library_name: transformers
+pipeline_tag: text-generation
+tags:
+  - pytorch
+  - transformers
+  - unsloth
+  - lora
+  - education
+  - child-safe
+  - stem
+  - conversational
+---
+
+# SunnyTeacher-8Bv2
+
+SunnyTeacher-8Bv2 is a child-safe educational assistant model fine-tuned from Qwen3-8B.
+
+## Features
+- STEM teaching for ages 3–17
+- Step-by-step explanations
+- Child-safe responses
+- Emotional support guidance
+- Educational reasoning
+- Creativity and imagination support
+- -v2 added Emotional support guidance. 
+
+## Training
+Fine-tuned using Unsloth LoRA training on curated educational and child-safe datasets.
+
+## Intended Use
+Educational assistance, tutoring, STEM learning, and safe child interaction.
+SunnyTeacher-16B is intended to help children and students with homework, STEM learning, and general educational questions by providing step-by-step explanations instead of only giving direct answers.
+The goal is to help learners understand the process, build confidence, and focus on building critical thinking in an age-appropriate way.
+
+## Base Model
+Qwen3-8B
+
+## License
+Apache-2.0
+
+## Training Framework
+
+SunnyTeacher-8B was trained using:
+
+- PyTorch
+- Hugging Face Transformers
+- Unsloth
+- PEFT (LoRA fine-tuning)
+- TRL (Supervised Fine-Tuning)
+- Safetensors
+
+## Hardware
+
+Training performed on:
+- Offline
+- NVIDIA GeForce RTX 4090
+- CUDA 12.8
+
+## Training Method
+
+The model was fine-tuned using supervised instruction tuning on curated child-safe educational datasets focused on:
+- STEM education
+- Emotional safety
+- Age-appropriate explanations
+- Step-by-step reasoning
+- Child-safe conversational behavior
+- 
+## Disclaimer
+
+SunnyTeacher-8Bv2 is currently a work in progress (WIP) research and educational model.
+While significant effort has been made to encourage safe, age-appropriate, and educational responses, the model may still occasionally generate incorrect, inappropriate, incomplete, or unsafe outputs.
+Parents, guardians, teachers, and developers should review responses before relying on them for child use cases.
+If you encounter a response that appears unsafe, inappropriate, or inaccurate for children, please provide feedback or report the issue so the model can continue improving.
+