File size: 4,825 Bytes
dc4efc4 |
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 |
---
library_name: peft
license: llama3.1
base_model: meta-llama/Llama-3.1-8B
tags:
- generated_from_trainer
model-index:
- name: outputs/020
results: []
---
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
[<img src="https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/axolotl-ai-cloud/axolotl)
<details><summary>See axolotl config</summary>
axolotl version: `0.5.2`
```yaml
# Название базовой модели, которая будет использоваться
base_model: meta-llama/Llama-3.1-8B
chat_template: llama3
datasets:
- path: /workspace/dataset_200_30_repeats_by_cycles.jsonl # (A,B,C)x30
type: chat_template
field_messages: conversations
message_field_role: role
message_field_content: content
roles:
user: ["user"]
assistant: ["assistant"]
system: ["system"]
roles_to_train: ["assistant", "user"]
train_on_eos: turn # Тренировать EOS на каждом конце реплики для лучшего запоминания
# Путь к директории для сохранения результатов обучения
output_dir: ./outputs/020
# Настройки обучения
gradient_accumulation_steps: 5
micro_batch_size: 2
num_epochs: 1
learning_rate: 0.000002
warmup_steps: 500
logging_steps: 10
# Использование повторного обучения через LoRA
adapter: lora
lora_r: 16 # Увеличенное значение для сохранения памяти и генерации точных ответов
lora_alpha: 32
lora_dropout: 0.1
lora_target_modules:
- q_proj
- k_proj
- v_proj
- o_proj
# Тип модели и токенизатора
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer
# Настройки последовательности
sequence_len: 4096 # Достаточная длина для обработки 7 реплик
sample_packing: false # Отключено для лучшего соответствия тексту
pad_to_sequence_len: true
# Оптимизация
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
weight_decay: 0.01
gradient_checkpointing: true
# Использование BF16 для экономии памяти
bf16: true
# Flash Attention для ускорения
flash_attention: true
# Доля данных для валидации
val_set_size: 0.1
# Настройки сохранения
save_safetensors: true
saves_per_epoch: 3 # Увеличено для промежуточного анализа качества модели
# Настройки метрик
evals_per_epoch: 10
eval_max_new_tokens: 128
# Специальные токены
special_tokens:
pad_token: "<|finetune_right_pad_id|>"
bos_token: "<|begin_of_text|>"
eos_token: "<|end_of_text|>"
# WandB интеграция (если требуется)
wandb_project:
wandb_entity:
wandb_watch:
wandb_log_model:
```
</details><br>
# outputs/020
This model is a fine-tuned version of [meta-llama/Llama-3.1-8B](https://huggingface.co/meta-llama/Llama-3.1-8B) on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 1.0477
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 2e-06
- train_batch_size: 2
- eval_batch_size: 2
- seed: 42
- gradient_accumulation_steps: 5
- total_train_batch_size: 10
- optimizer: Use OptimizerNames.ADAMW_BNB with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 500
- num_epochs: 1
### Training results
| Training Loss | Epoch | Step | Validation Loss |
|:-------------:|:------:|:----:|:---------------:|
| No log | 0.0009 | 1 | 1.5900 |
| 1.4756 | 0.1007 | 106 | 1.5880 |
| 1.4972 | 0.2013 | 212 | 1.5655 |
| 1.4563 | 0.3020 | 318 | 1.5018 |
| 1.3754 | 0.4027 | 424 | 1.4127 |
| 1.271 | 0.5033 | 530 | 1.3056 |
| 1.2054 | 0.6040 | 636 | 1.2009 |
| 1.1065 | 0.7047 | 742 | 1.1182 |
| 1.0592 | 0.8053 | 848 | 1.0689 |
| 1.0322 | 0.9060 | 954 | 1.0477 |
### Framework versions
- PEFT 0.13.2
- Transformers 4.46.3
- Pytorch 2.5.1+cu124
- Datasets 3.1.0
- Tokenizers 0.20.3 |