🐉 LongGuang-A1-7B

7 B bilingual causal LM with built-in causal discovery, metacognition & long-term memory.
Maintained by Golden Loong AS
[Apache 2.0] – free for commercial use.

---

🔧 Quick Facts

Item	Value
Parameters	~7 B
Context Length	32 k
Vocab	64 k bilingual BPE
Languages	Chinese & English
License	Apache 2.0

---

📦 Files

File	Purpose
config.json	Public hyper-parameters
vocab.json + merges.json	BPE tokenizer
pytorch_model.bin	Model weights (Git-LFS)
examples/	Usage & fine-tune scripts
requirements.txt	Verified dependencies

---

🚀 Usage

1. Install

pip install -r requirements.txt

2. Load

from transformers import AutoTokenizer, AutoModelForCausalLM
tok = AutoTokenizer.from_pretrained("GoldenLoong/LongGuang-A1-7B")
model = AutoModelForCausalLM.from_pretrained("GoldenLoong/LongGuang-A1-7B",
                                             torch_dtype="auto",
                                             device_map="auto")

3. Chat

agent = CausalAgent(model, tok)  # provided in examples/
print(agent("为什么天空是蓝色的？"))

---

🧠 Built-in Skills

• Causal Q&A – “为什么…导致…”
• Counterfactual – “如果…会怎样”
• Memory Recall – remembers prior turns
• Live Search – optional Bing integration
• Code Runner – inline Python/JS sandbox

---

🛡️ Security Notice

• Only public config & tokenizer are open.
• Internal architecture & training code remain confidential.
• Do not redistribute decrypted weights.
• Please specify the author and copyright owner before use.

---

🙋 Citation

@misc{longguang-a1-7b,
  title={LongGuang-A1-7B: A Bilingual Causal Language Model},
  author={Golden Loong AS},
  year={2025},
  url={https://huggingface.co/GoldenLoong/LongGuang-A1-7B},
  license={Apache-2.0}
}