File size: 1,994 Bytes
97122e6 6b89e35 5363d8e d99ae72 97122e6 aab72ce 97122e6 aab72ce 97122e6 |
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 60 61 62 63 64 65 66 |
---
license: apache-2.0
language:
- tr
---
<img src="https://huggingface.co/TURKCELL/Turkcell-LLM-7b-v1/resolve/main/icon.jpeg"
alt="Turkcell LLM" width="300"/>
# Turkcell-LLM-7b-v1
This model is an extended version of a Mistral-based Large Language Model (LLM) for Turkish. It was trained on a cleaned Turkish raw dataset containing 5 billion tokens. The training process involved using the DORA method initially. Following this, we utilized Turkish instruction sets created from various open-source and internal resources for fine-tuning with the LORA method.
## Model Details
- **Base Model**: Mistral 7B based LLM
- **Tokenizer Extension**: Specifically extended for Turkish
- **Training Dataset**: Cleaned Turkish raw data with 5 billion tokens
- **Training Method**: Initially with DORA, followed by fine-tuning with LORA using custom Turkish instruction sets
### DORA Configuration
- `lora_alpha`: 128
- `lora_dropout`: 0.05
- `r`: 64
- `target_modules`: "all-linear"
### LORA Fine-Tuning Configuration
- `lora_alpha`: 128
- `lora_dropout`: 0.05
- `r`: 256
- `target_modules`: "all-linear"
## Usage Examples
```python
from transformers import AutoModelForCausalLM, AutoTokenizer
device = "cuda" # the device to load the model onto
model = AutoModelForCausalLM.from_pretrained("TURKCELL/Turkcell-LLM-7b-v1")
tokenizer = AutoTokenizer.from_pretrained("TURKCELL/Turkcell-LLM-7b-v1")
messages = [
{"role": "user", "content": "Türkiye'nin başkenti neresidir?"},
]
encodeds = tokenizer.apply_chat_template(messages, return_tensors="pt")
eos_token = tokenizer("<|im_end|>",add_special_tokens=False)["input_ids"][0]
model_inputs = encodeds.to(device)
model.to(device)
generated_ids = model.generate(model_inputs,
max_new_tokens=1024,
do_sample=True,
eos_token_id=eos_token)
decoded = tokenizer.batch_decode(generated_ids)
print(decoded[0])
|