1.8 KiB
1.8 KiB
datasets, language
| datasets | language | ||||
|---|---|---|---|---|---|
|
|
Introduction
CodeKobzar13B is a generative model that was trained on Ukrainian Wikipedia data and Ukrainian language rules. It has knowledge of Ukrainian history, language, literature and culture.
Model Information
This model is based on vicuna-13b-v1.5.
Model Usage
Use the following prompt template:
USER: {input} ASSISTANT:
We recommend using next configurations:
Temperature: 0.8
Top-p: 0.95
Inference
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM
model_path="ponoma16/CodeKobzar13B"
tokenizer = AutoTokenizer.from_pretrained(model_path)
model = AutoModelForCausalLM.from_pretrained(
model_load_path,
low_cpu_mem_usage=True,
torch_dtype=torch.float16,
load_in_8bit=True,
device_map='auto',
)
model.eval()
prompt = "Яке місто в Україні називають найромантичнішим?"
PROMPT_TEMPLATE = """USER: {prompt} ASSISTANT: """
input_ids = tokenizer(
prompt,
return_tensors="pt",
truncation=True,
).input_ids.cuda()
outputs = model.generate(
input_ids=input_ids,
do_sample=True,
top_p=0.95,
max_new_tokens=150,
temperature=0.5,
)
prediction = tokenizer.batch_decode(outputs.cpu().numpy(), skip_special_tokens=True)[0]
print(prediction)
Contact
If you have any inquiries, please feel free to raise an issue or reach out to us via email at: mariiaponomarenko10@gmail.com, benjamin.ye@me.com. We're here to assist you!"