|
|
--- |
|
|
pipeline_tag: text-generation |
|
|
library_name: transformers |
|
|
language: en |
|
|
license: mit |
|
|
tags: |
|
|
- text-generation |
|
|
- tinyllama |
|
|
- smilyai |
|
|
- sam-large |
|
|
- speacil |
|
|
--- |
|
|
|
|
|
# THIS IS THE GPU EDITION# |
|
|
**CPU VERISON HERE(https://huggingface.co/Smilyai-labs/Sam-large-v1-speacil-v1-cpu)** |
|
|
# 🧠 Sam-large-v1-speacil |
|
|
|
|
|
**Model Author:** [Smilyai-labs](https://huggingface.co/Smilyai-labs) |
|
|
**Model Size:** \~1.1B parameters |
|
|
**Architecture:** Decoder-only Transformer |
|
|
**Base Model:** based on TinyLLaMA |
|
|
**License:** MIT |
|
|
**Language:** English |
|
|
**Tags:** #text-generation, #chatbot, #instruction-tuned, #smilyai, #sam |
|
|
|
|
|
## 📝 Model Summary |
|
|
|
|
|
**Sam-large-v1-speacil** is a customized large language model developed by Smilyai-labs for conversational AI, instruction-following tasks, and general-purpose text generation. It is a fine-tuned and enhanced variant of the `Sam-large-v1` model, with special improvements in reasoning, identity handling, and emotional response learning. |
|
|
|
|
|
This model is trained to represent the persona “Sam,” an intelligent and slightly chaotic AI assistant with unique behavior traits, making it suitable for role-play bots, experimental dialogue systems, and character-driven applications. |
|
|
|
|
|
--- |
|
|
|
|
|
## 🧠 Intended Use |
|
|
|
|
|
* Instruction-based text generation |
|
|
* Character chat and roleplay |
|
|
* Experimental alignment behaviors |
|
|
* Creative writing and scenario building |
|
|
* Local deployment for private assistant usage |
|
|
|
|
|
--- |
|
|
|
|
|
## 🚫 Limitations |
|
|
|
|
|
* May hallucinate facts or invent information |
|
|
* Can produce unexpected outputs when prompted ambiguously |
|
|
* Not suitable for production environments without safety layers |
|
|
* Behavior is tuned to have personality traits (like mischief) that may not suit all applications |
|
|
|
|
|
--- |
|
|
|
|
|
## 📚 Training Details |
|
|
|
|
|
* Fine-tuned on synthetic and curated datasets using LoRA/full fine-tuning |
|
|
* Special prompt styles were introduced to enhance behavior |
|
|
* Dataset includes: |
|
|
|
|
|
* Multi-step reasoning samples |
|
|
* Emotionally reactive instruction responses |
|
|
* SmilyAI-specific identity alignment examples |
|
|
|
|
|
--- |
|
|
|
|
|
## 🔧 How to Use |
|
|
|
|
|
```python |
|
|
from transformers import AutoTokenizer, AutoModelForCausalLM |
|
|
|
|
|
model = AutoModelForCausalLM.from_pretrained("Smilyai-labs/Sam-large-v1-speacil") |
|
|
tokenizer = AutoTokenizer.from_pretrained("Smilyai-labs/Sam-large-v1-speacil") |
|
|
|
|
|
input_text = "You are Sam. Who are you?" |
|
|
inputs = tokenizer(input_text, return_tensors="pt") |
|
|
outputs = model.generate(**inputs, max_new_tokens=100) |
|
|
print(tokenizer.decode(outputs[0], skip_special_tokens=True)) |
|
|
``` |
|
|
|
|
|
--- |
|
|
|
|
|
## 🤝 Citation |
|
|
|
|
|
If you use this model in your work, please cite it as: |
|
|
|
|
|
```bibtex |
|
|
@misc{samlargev1speacil2025, |
|
|
title={Sam-large-v1-speacil}, |
|
|
author={Smilyai-labs}, |
|
|
year={2025}, |
|
|
publisher={Hugging Face}, |
|
|
howpublished={\url{https://huggingface.co/Smilyai-labs/Sam-large-v1-speacil}} |
|
|
} |
|
|
``` |
