File size: 10,455 Bytes
598baca fd50dea e1f3912 fd50dea e1f3912 fd50dea e1f3912 fd50dea e1f3912 fd50dea e1f3912 fd50dea 7923492 fd50dea e1f3912 fd50dea e1f3912 fd50dea e1f3912 |
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 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 |
---
license: llama3
language:
- en
- fa
pipeline_tag: text-generation
---
# Model Details
This Repository is a 4-bit quantized version of [Dorna-Llama3-8B-Instruct](https://huggingface.co/PartAI/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](https://arxiv.org/abs/2307.08691) 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.
```Python
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 |
-->
<div style="overflow-x: auto;">
<table>
<thead>
<tr style="vertical-align: middle;">
<th style="white-space: nowrap; vertical-align: middle;"><strong>Model Pairs</strong></th>
<th style="white-space: nowrap; vertical-align: middle;"><strong>Parameters</strong></th>
<th style="white-space: nowrap; vertical-align: middle;"><strong>Bool Complex</strong></th>
<th style="white-space: nowrap; vertical-align: middle;"><strong>Bool Easy</strong></th>
<th style="white-space: nowrap; vertical-align: middle;"><strong>Code Gen</strong></th>
<th style="white-space: nowrap; vertical-align: middle;"><strong>General Long Response</strong></th>
<th style="white-space: nowrap; vertical-align: middle;"><strong>Historical Long Response</strong></th>
<th style="white-space: nowrap; vertical-align: middle;"><strong>Math Complex</strong></th>
<th style="white-space: nowrap; vertical-align: middle;"><strong>Math Easy</strong></th>
<th style="white-space: nowrap; vertical-align: middle;"><strong>News QA Complex</strong></th>
<th style="white-space: nowrap; vertical-align: middle;"><strong>News QA Easy</strong></th>
<th style="white-space: nowrap; vertical-align: middle;"><strong>Paraphrasing</strong></th>
<th style="white-space: nowrap; vertical-align: middle;"><strong>General Knowledge Easy</strong></th>
<th style="white-space: nowrap; vertical-align: middle;"><strong>General Knowledge Hard</strong></th>
<th style="white-space: nowrap; vertical-align: middle;"><strong>Summarization</strong></th>
</tr>
</thead>
<tbody>
<tr>
<td style="white-space: nowrap; vertical-align: middle;">Dorna-8B-it <strong>vs.</strong> Meta-Llama-3-8B-Instruct</td>
<td>8B</td>
<td>0.25/0.25/0.5</td>
<td>0.28/<strong>0.35</strong>/0.38</td>
<td><strong>0.6</strong>/0.1/0.3</td>
<td><strong>0.8</strong>/0.08/0.12</td>
<td><strong>0.4</strong>/0.3/0.3</td>
<td><strong>0.28</strong>/0.08/0.65</td>
<td><strong>0.47</strong>/0.00/0.53</td>
<td><strong>0.55</strong>/0.07/0.38</td>
<td><strong>0.43</strong>/0.15/0.42</td>
<td><strong>0.1</strong>/0.05/0.85</td>
<td><strong>0.31</strong>/0.2/0.49</td>
<td><strong>0.59</strong>/0.13/0.28</td>
<td><strong>0.28</strong>/0.2/0.53</td>
</tr>
<tr>
<td style="white-space: nowrap; vertical-align: middle;">Dorna-8B-it <strong>vs.</strong> GPT 3.5 turbo-1106</td>
<td>N.A.</td>
<td>0.35/0.35/0.3</td>
<td>0.3/0.3/0.4</td>
<td>0.1/<strong>0.3</strong>/.06</td>
<td>0.2/<strong>0.45</strong>/0.35</td>
<td><strong>0.46</strong>/0.27/0.27</td>
<td><strong>0.25</strong>/0.1/0.65</td>
<td>0.05/<strong>0.1</strong>/0.85</td>
<td>0.12/<strong>0.35</strong>/0.53</td>
<td><strong>0.15</strong>/0.1/0.75</td>
<td><strong>0.25</strong>/0.15/0.6</td>
<td>0.3/<strong>0.32</strong>/0.38</td>
<td>0.22/<strong>0.53</strong>/0.25</td>
<td>0.35/<strong>0.55</strong>/0.1</td>
</tr>
<tr>
<td style="white-space: nowrap; vertical-align: middle;">Dorna-8B-it <strong>vs.</strong> Persian Mind</td>
<td>7B</td>
<td><strong>0.47</strong>/0.25/0.28</td>
<td><strong>0.57</strong>/0.15/0.28</td>
<td><strong>0.9</strong>/0.1/0.0</td>
<td><strong>0.82</strong>/0.08/0.1</td>
<td><strong>0.4</strong>/0.17/0.42</td>
<td><strong>0.3</strong>/0.0/0.7</td>
<td><strong>0.22</strong>/0.08/0.7</td>
<td><strong>0.72</strong>/0.07/0.2</td>
<td><strong>0.7</strong>/0.0/0.3</td>
<td><strong>0.7</strong>/0.05/0.25</td>
<td><strong>0.51</strong>/0.12/0.37</td>
<td><strong>0.61</strong>/0.1/0.29</td>
<td><strong>0.93</strong>/0.0/0.07</td>
</tr>
</tbody>
</table>
</div>
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](https://huggingface.co/amirMohammadi/Dorna-Llama3-8B-Instruct-Quantized4Bit/discussions) on Hugging Face. |