Model Details

This Repository is a 4-bit quantized version of Dorna-Llama3-8B-Instruct model for efficient memory usage. Dorna model is a decoder-only model, specifically trained/fine-tuned on Persian data. Flash Attention 2 is also integrated for faster inference.

Benefits

  • Reduced Memory Usage: 4-bit quantization lowers memory requirements.
  • Faster Inference: Flash Attention 2 speeds up processing.
  • Easy Deployment: No need for additional libraries like LlamaCPP or Candle.
  • Ready to Use: Compatible with Langchain, Haystack, LlamaIndex 2, and more.
  • Google Colab Friendly: Can run on Google Colab free tier with T4 GPU (less than 15 GB of GPU RAM).

How to use

You can run conversational inference using the Transformers Auto classes with the generate() function. Let's look at an example.

import torch
import transformers
from transformers import AutoTokenizer, AutoModelForCausalLM

model_path = "amirMohammadi/Dorna-Llama3-8B-Instruct-Quantized4Bit"

tokenizer = AutoTokenizer.from_pretrained(model_path)
model = AutoModelForCausalLM.from_pretrained(
    model_path,
    torch_dtype=torch.bfloat16,
    device_map="auto",
)

messages = [
    {"role": "system",
     "content": "You are a helpful Persian assistant. Please answer questions in the asked language."},
    {"role": "user", "content": "اصفهان بزرگ تر است یا قم؟"},
]

input_ids = tokenizer.apply_chat_template(
    messages,
    add_generation_prompt=True,
    return_tensors="pt"
).to(model.device)

terminators = [
    tokenizer.eos_token_id,
    tokenizer.convert_tokens_to_ids("<|eot_id|>")
]

outputs = model.generate(
    input_ids,
    max_new_tokens=256,
    eos_token_id=terminators,
    do_sample=True,
    temperature=0.6,
    top_p=0.9,
)
response = outputs[0][input_ids.shape[-1]:]
print(tokenizer.decode(response, skip_special_tokens=True))

Evaluation of Non-Quantized version

This model is evaluated on questions across various tasks, including Boolean Questions, Code Generation, Long Response, Math, News QA, Paraphrasing, General Knowledge, and Summarization. Most categories typically have two main difficulty levels: Hard and Easy.

Both human evaluation and automatic evaluation (with GPT-4 as the judge) are performed.

In both tables, Dorna-8B-it is used as an abbreviated form of Dorna-Llama3-8B-Instruct.

Overall human evaluation results are as follows:

Model Pairs Parameters Win % Lose % Tie %
Dorna-8B-it vs. Meta-Llama-3-8B-Instruct 8B 36.94 17.39 45.67
Dorna-8B-it vs. GPT 3.5 turbo-1106 N.A. 32.01 26.94 41.05
Dorna-8B-it vs. Persian Mind 7B 55.77 10.49 33.74

Category-based human evaluation results are as follows:

Win/Lose/Tie % is reported for each category.

Model Pairs Parameters Bool Complex Bool Easy Code Gen General Long Response Historical Long Response Math Complex Math Easy News QA Complex News QA Easy Paraphrasing General Knowledge Easy General Knowledge Hard Summarization
Dorna-8B-it vs. Meta-Llama-3-8B-Instruct 8B 0.25/0.25/0.5 0.28/0.35/0.38 0.6/0.1/0.3 0.8/0.08/0.12 0.4/0.3/0.3 0.28/0.08/0.65 0.47/0.00/0.53 0.55/0.07/0.38 0.43/0.15/0.42 0.1/0.05/0.85 0.31/0.2/0.49 0.59/0.13/0.28 0.28/0.2/0.53
Dorna-8B-it vs. GPT 3.5 turbo-1106 N.A. 0.35/0.35/0.3 0.3/0.3/0.4 0.1/0.3/.06 0.2/0.45/0.35 0.46/0.27/0.27 0.25/0.1/0.65 0.05/0.1/0.85 0.12/0.35/0.53 0.15/0.1/0.75 0.25/0.15/0.6 0.3/0.32/0.38 0.22/0.53/0.25 0.35/0.55/0.1
Dorna-8B-it vs. Persian Mind 7B 0.47/0.25/0.28 0.57/0.15/0.28 0.9/0.1/0.0 0.82/0.08/0.1 0.4/0.17/0.42 0.3/0.0/0.7 0.22/0.08/0.7 0.72/0.07/0.2 0.7/0.0/0.3 0.7/0.05/0.25 0.51/0.12/0.37 0.61/0.1/0.29 0.93/0.0/0.07

Automatic evaluation results are as follows:

Model Pairs Parameters Overall Win Rate % Easy Win Rate % Hard Win Rate %
Dorna-8B-it vs. Llama 3 base 8B 58.96 56.00 64.49
Dorna-8B-it vs. Part Mistral 7B 77.20 73.00 85.05
Dorna-8B-it vs. Persian Mind 7B 90.88 87.50 97.20
Dorna-8B-it vs. Neuraorca Gemma 7b 7B 86.32 86.50 85.98
Dorna-8B-it vs. Maral 7b 7B 97.39 97.00 98.13
Dorna-8B-it vs. PersianLlama 7b 7B 98.70 98.00 100.00
Dorna-8B-it vs. Aya-23-8B 8B 52.77 56.50 45.79
Dorna-8B-it vs. Aya-23-35B 35B 45.93 54.00 30.84
Dorna-8B-it vs. Command R 35B 58.63 61.00 54.21

Contact us

If you have any questions regarding this model, you can reach us via the community on Hugging Face.

Downloads last month
34
Safetensors
Model size
4.65B params
Tensor type
BF16
·
F32
·
U8
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Space using amirMohammadi/Dorna-Llama3-8B-Instruct-Quantized4Bit 1