phi-2-chat-turkish-GGUF/README.md

base_model, datasets, inference, language, model_creator, model_name, pipeline_tag, quantized_by, tags

base_model	datasets	inference	language	model_creator	model_name	pipeline_tag	quantized_by	tags
malhajar/phi-2-chat-turkish	TFLai/Turkish-Alpaca	false	tr	malhajar	phi-2-chat-turkish	text-generation	afrideva	gguf ggml quantized q2_k q3_k_m q4_k_m q5_k_m q6_k q8_0

malhajar/phi-2-chat-turkish-GGUF

Quantized GGUF model files for phi-2-chat-turkish from malhajar

Name	Quant method	Size
phi-2-chat-turkish.fp16.gguf	fp16	5.56 GB
phi-2-chat-turkish.q2_k.gguf	q2_k	1.17 GB
phi-2-chat-turkish.q3_k_m.gguf	q3_k_m	1.48 GB
phi-2-chat-turkish.q4_k_m.gguf	q4_k_m	1.79 GB
phi-2-chat-turkish.q5_k_m.gguf	q5_k_m	2.07 GB
phi-2-chat-turkish.q6_k.gguf	q6_k	2.29 GB
phi-2-chat-turkish.q8_0.gguf	q8_0	2.96 GB

Original Model Card:

Model Card for Model ID

malhajar/phi-2-chat-turkish is a finetuned version of phi-2 using SFT Training. This model can answer information in turkish language as it is finetuned on a turkish dataset specifically Turkish-Alpaca

Model Description

• Developed by: Mohamad Alhajar
• Language(s) (NLP): Turkish
• Finetuned from model: microsoft/phi-2

Prompt Template

### Instruction:

<prompt> (without the <>)

### Response:

How to Get Started with the Model

Use the code sample provided in the original post to interact with the model.

from transformers import AutoTokenizer,AutoModelForCausalLM
 
model_id = "malhajar/phi-2-chat-turkish"
model = AutoModelForCausalLM.from_pretrained(model_name_or_path,
                                             device_map="auto",
                                             torch_dtype=torch.float16,
                                             revision="main")

tokenizer = AutoTokenizer.from_pretrained(model_id)

question: "Türkiyenin en büyük şehir nedir?"
# For generating a response
prompt = f'''
### Instruction:  {question} ### Response:
'''
input_ids = tokenizer(prompt, return_tensors="pt").input_ids
output = model.generate(inputs=input_ids,max_new_tokens=512,pad_token_id=tokenizer.eos_token_id,top_k=50, do_sample=True,repetition_penalty=1.3
        top_p=0.95,trust_remote_code=True,)
response = tokenizer.decode(output[0])

print(response)