File size: 3,486 Bytes
2c963d0 bbf47c2 aa1ab0a 4a23bd1 1c58cfa |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 |
---
license: mit
language:
- en
- fa
library_name: transformers
pipeline_tag: text-generation
---
# ChatFrame-Persian Model Card
## Model Description
ChatFrame-Persian is an instruction-following AI model developed by AIFRAME INC. It is a large language model (LLM) designed for English and Persian language support. This model can be used for a variety of commercial applications and fine-tuning. ChatFrame-Persian demonstrates superior performance when compared to similar models such as Llama-70B, GPT-3.5, and GPT-4.
## Intended Uses & Limitations
ChatFrame-Persian is intended for a wide range of language-based tasks, including but not limited to:
- Text generation and language translation
- Question answering and conversational AI
- Language understanding and analysis
The model has been trained on a diverse dataset, but it is important to note that its performance may vary across different cultural, demographic, and phenotypic groups.
## Model Architecture
ChatFrame-Persian utilizes a transformer-based architecture with a large number of parameters, allowing it to capture complex linguistic patterns and generate human-like responses.
## Training Data and Methodology
The model was trained on a vast dataset containing text data from various sources, including books, websites, and social media. The training process involved advanced techniques such as large-batch training and fine-tuning to optimize performance.
## Performance Metrics
ChatFrame-Persian excels in several key performance metrics, including accuracy, precision, and recall. It has been benchmarked against other leading models and consistently demonstrates superior results, especially in language generation and understanding tasks.
## Potential Biases and Limitations
While ChatFrame-Persian has been trained to minimize biases, it is important to recognize that all language models have limitations. Users should be aware of potential biases related to the training data and the possibility of generating inappropriate or offensive responses. Additionally, the model may struggle with highly specialized or technical language.
## Responsible AI Considerations
ChatFrame-Persian has been developed with a focus on responsible AI practices. The model has been trained to respect user privacy and adhere to ethical guidelines. It is important for users to monitor the model's output and provide feedback to continuously improve its performance and address any potential issues.
## Using pipeline to run model(A100 GPU tested):
```python
from transformers import pipeline
import torch
device = 0 if torch.cuda.is_available() else -1
def generate_response(user_input):
pipe = pipeline("text-generation", model="frameai/ChatFrame-Instruct-Persian-Small", device = device)
messages = [
{"role": "user", "content": user_input},
]
response = pipe(messages, max_length=512)
return response[0]['generated_text'][1]["content"]
ai = generate_response(user_input = "ایران رو برام توصیف کن")
print(ai)
```
## Authors
Trained by Experts: The brainchild of Mohammadmoein Pisoude (CEO) and Alex Romniof (Manager), ChatFrame-Persian is the result of the dedication and expertise of the entire AIFRAME INC team. Their combined efforts in coding and project management have led to the creation of this innovative model. |