Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,75 @@
|
|
1 |
-
---
|
2 |
-
license: apache-2.0
|
3 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: apache-2.0
|
3 |
+
---
|
4 |
+
README.md for Hugging Face - MK-LLM-Mistral
|
5 |
+
This README will help contributors, developers, and AI enthusiasts understand your MK-LLM-Mistral project.
|
6 |
+
|
7 |
+
π MK-LLM-Mistral: The First Macedonian LLM
|
8 |
+
π’ MK-LLM-Mistral is the first Macedonian Language Large Language Model π²π°, developed by AI Now - Association for Artificial Intelligence in Macedonia.
|
9 |
+
|
10 |
+
π Website: www.ainow.mk
|
11 |
+
π© Contact: [email protected]
|
12 |
+
π GitHub Repository: MK-LLM Project
|
13 |
+
|
14 |
+
π Model Overview
|
15 |
+
Model Name: MK-LLM-Mistral
|
16 |
+
Base Model: Mistral-7B
|
17 |
+
Language: Macedonian π²π°
|
18 |
+
Fine-tuned on: Wikipedia, news articles, legal documents, and public datasets in Macedonian
|
19 |
+
Tasks: Chatbot, Text Completion, Q&A, Macedonian NLP tasks
|
20 |
+
π How to Use the Model Locally
|
21 |
+
1οΈβ£ Install Required Libraries
|
22 |
+
bash
|
23 |
+
Copy
|
24 |
+
Edit
|
25 |
+
pip install transformers torch huggingface_hub
|
26 |
+
2οΈβ£ Load the Model in Python
|
27 |
+
python
|
28 |
+
Copy
|
29 |
+
Edit
|
30 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
31 |
+
import torch
|
32 |
+
|
33 |
+
# Load Model
|
34 |
+
MODEL_NAME = "ainowmk/MK-LLM-Mistral"
|
35 |
+
tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
|
36 |
+
model = AutoModelForCausalLM.from_pretrained(MODEL_NAME)
|
37 |
+
|
38 |
+
# Move model to GPU if available
|
39 |
+
device = "cuda" if torch.cuda.is_available() else "cpu"
|
40 |
+
model.to(device)
|
41 |
+
|
42 |
+
# Test the Model
|
43 |
+
input_text = "ΠΠ΄ΡΠ°Π²ΠΎ, ΠΊΠ°ΠΊΠΎ ΡΠΈ?"
|
44 |
+
inputs = tokenizer(input_text, return_tensors="pt").to(device)
|
45 |
+
outputs = model.generate(**inputs, max_length=100)
|
46 |
+
|
47 |
+
# Decode and print the result
|
48 |
+
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
|
49 |
+
π Model Files
|
50 |
+
File Name Description
|
51 |
+
pytorch_model.bin The fine-tuned weights of the model
|
52 |
+
config.json Configuration for the model architecture
|
53 |
+
tokenizer.json Tokenizer used for the Macedonian language
|
54 |
+
README.md Documentation for the model
|
55 |
+
.gitattributes Git LFS tracking for large files
|
56 |
+
π Training Details
|
57 |
+
Dataset: Collected Macedonian texts (Wikipedia, news, government websites)
|
58 |
+
Training Compute: GPU-based training on NVIDIA A100
|
59 |
+
Training Time: Estimated XX hours
|
60 |
+
Fine-tuned using: Hugging Face Transformers & PyTorch
|
61 |
+
π Contributing
|
62 |
+
MK-LLM-Mistral is an open-source project, and contributions are welcome! π―
|
63 |
+
|
64 |
+
Open issues on GitHub
|
65 |
+
Submit pull requests for improvements
|
66 |
+
Join discussions on Hugging Face Community
|
67 |
+
π‘ If you want to help in data collection, fine-tuning, or evaluation, reach out at [email protected]
|
68 |
+
|
69 |
+
π License
|
70 |
+
This model is licensed under Apache 2.0.
|
71 |
+
You are free to use, distribute, and modify it, but attribution is required.
|
72 |
+
|
73 |
+
π Letβs build the future of Macedonian AI together! π²π°
|
74 |
+
π AI Now - Association for Artificial Intelligence in Macedonia
|
75 |
+
π© [email protected] | π www.ainow.mk
|