File size: 6,837 Bytes
cbea437 efd8812 cbea437 efd8812 cbea437 efd8812 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 |
---
language:
- en
license: apache-2.0
tags:
- mistral
- instruct
- finetune
- chatml
- gpt4
- synthetic data
- distillation
- dpo
- rlhf
datasets:
- mlabonne/chatml_dpo_pairs
base_model: teknium/OpenHermes-2.5-Mistral-7B
model-index:
- name: NeuralHermes-2.5-Mistral-7B
results:
- task:
type: text-generation
name: Text Generation
dataset:
name: AI2 Reasoning Challenge (25-Shot)
type: ai2_arc
config: ARC-Challenge
split: test
args:
num_few_shot: 25
metrics:
- type: acc_norm
value: 66.55
name: normalized accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=mlabonne/NeuralHermes-2.5-Mistral-7B
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: HellaSwag (10-Shot)
type: hellaswag
split: validation
args:
num_few_shot: 10
metrics:
- type: acc_norm
value: 84.9
name: normalized accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=mlabonne/NeuralHermes-2.5-Mistral-7B
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: MMLU (5-Shot)
type: cais/mmlu
config: all
split: test
args:
num_few_shot: 5
metrics:
- type: acc
value: 63.32
name: accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=mlabonne/NeuralHermes-2.5-Mistral-7B
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: TruthfulQA (0-shot)
type: truthful_qa
config: multiple_choice
split: validation
args:
num_few_shot: 0
metrics:
- type: mc2
value: 54.93
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=mlabonne/NeuralHermes-2.5-Mistral-7B
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: Winogrande (5-shot)
type: winogrande
config: winogrande_xl
split: validation
args:
num_few_shot: 5
metrics:
- type: acc
value: 78.3
name: accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=mlabonne/NeuralHermes-2.5-Mistral-7B
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: GSM8k (5-shot)
type: gsm8k
config: main
split: test
args:
num_few_shot: 5
metrics:
- type: acc
value: 61.33
name: accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=mlabonne/NeuralHermes-2.5-Mistral-7B
name: Open LLM Leaderboard
---
<center><img src="https://i.imgur.com/qIhaFNM.png"></center>
# NeuralHermes 2.5 - Mistral 7B
NeuralHermes is based on the [teknium/OpenHermes-2.5-Mistral-7B](https://huggingface.co/teknium/OpenHermes-2.5-Mistral-7B) model that has been further fine-tuned with Direct Preference Optimization (DPO) using the [mlabonne/chatml_dpo_pairs](https://huggingface.co/datasets/mlabonne/chatml_dpo_pairs) dataset. It surpasses the original model on most benchmarks (see results).
It is directly inspired by the RLHF process described by [Intel/neural-chat-7b-v3-1](https://huggingface.co/Intel/neural-chat-7b-v3-1)'s authors to improve performance. I used the same dataset and reformatted it to apply the ChatML template.
The code to train this model is available on [Google Colab](https://colab.research.google.com/drive/15iFBr1xWgztXvhrj5I9fBv20c7CFOPBE?usp=sharing) and [GitHub](https://github.com/mlabonne/llm-course/tree/main). It required an A100 GPU for about an hour.
## Quantized models
* **GGUF**: https://huggingface.co/TheBloke/NeuralHermes-2.5-Mistral-7B-GGUF
* **AWQ**: https://huggingface.co/TheBloke/NeuralHermes-2.5-Mistral-7B-AWQ
* **GPTQ**: https://huggingface.co/TheBloke/NeuralHermes-2.5-Mistral-7B-GPTQ
* **EXL2**:
* 3.0bpw: https://huggingface.co/LoneStriker/NeuralHermes-2.5-Mistral-7B-3.0bpw-h6-exl2
* 4.0bpw: https://huggingface.co/LoneStriker/NeuralHermes-2.5-Mistral-7B-4.0bpw-h6-exl2
* 5.0bpw: https://huggingface.co/LoneStriker/NeuralHermes-2.5-Mistral-7B-5.0bpw-h6-exl2
* 6.0bpw: https://huggingface.co/LoneStriker/NeuralHermes-2.5-Mistral-7B-6.0bpw-h6-exl2
* 8.0bpw: https://huggingface.co/LoneStriker/NeuralHermes-2.5-Mistral-7B-8.0bpw-h8-exl2
## Results
**Update:** NeuralHermes-2.5 became the best Hermes-based model on the Open LLM leaderboard and one of the very best 7b models. 🎉
Teknium (author of OpenHermes-2.5-Mistral-7B) benchmarked the model ([see his tweet](https://twitter.com/Teknium1/status/1729955709377503660)).
Results are improved on every benchmark: **AGIEval** (from 43.07% to 43.62%), **GPT4All** (from 73.12% to 73.25%), and **TruthfulQA**.
### AGIEval
### GPT4All
### TruthfulQA
You can check the Weights & Biases project [here](https://wandb.ai/mlabonne/DPO/runs/axe71gr0?nw=nwusermlabonne).
## Usage
You can run this model using [LM Studio](https://lmstudio.ai/) or any other frontend.
You can also run this model using the following code:
```python
import transformers
from transformers import AutoTokenizer
# Format prompt
message = [
{"role": "system", "content": "You are a helpful assistant chatbot."},
{"role": "user", "content": "What is a Large Language Model?"}
]
tokenizer = AutoTokenizer.from_pretrained(new_model)
prompt = tokenizer.apply_chat_template(message, add_generation_prompt=True, tokenize=False)
# Create pipeline
pipeline = transformers.pipeline(
"text-generation",
model=new_model,
tokenizer=tokenizer
)
# Generate text
sequences = pipeline(
prompt,
do_sample=True,
temperature=0.7,
top_p=0.9,
num_return_sequences=1,
max_length=200,
)
print(sequences[0]['generated_text'])
```
## Training hyperparameters
**LoRA**:
* r=16
* lora_alpha=16
* lora_dropout=0.05
* bias="none"
* task_type="CAUSAL_LM"
* target_modules=['k_proj', 'gate_proj', 'v_proj', 'up_proj', 'q_proj', 'o_proj', 'down_proj']
**Training arguments**:
* per_device_train_batch_size=4
* gradient_accumulation_steps=4
* gradient_checkpointing=True
* learning_rate=5e-5
* lr_scheduler_type="cosine"
* max_steps=200
* optim="paged_adamw_32bit"
* warmup_steps=100
**DPOTrainer**:
* beta=0.1
* max_prompt_length=1024
* max_length=1536
|