Downloads · 30 days
48
0% of all-time downloads
CHIH-HUNG/llama-2-13b-FINETUNE5_4w-r4-q_k_v_o
llama-2-13b-FINETUNE5_4w-r4-q_k_v_o is a text generation model from CHIH-HUNG. Use it when you need the model to write or continue text. It is set up for transformers. The card lists the license as llama2.
在llama-2-13b上使用huangyt/FINETUNE5資料集進行訓練,總資料筆數約4w
Downloads · 30 days
48
0% of all-time downloads
All-time downloads
62.4K
Public
Repo size
52.1 GB
Likes
0
Public
Click a slice to open those files.
.bin26 GB · 100%
From the Hugging Face model README
在llama-2-13b上使用huangyt/FINETUNE5資料集進行訓練,總資料筆數約4w
| Model | Average | ARC | HellaSwag | MMLU | TruthfulQA | Time (s) |
|---|---|---|---|---|---|---|
| FINETUNE5_4w-r4-q_k_v_o | 56.09 | 54.35 | 79.24 | 54.01 | 36.75 | 22095 |
| FINETUNE5_4w-r8-q_k_v_o | 57.55 | 55.38 | 79.57 | 54.03 | 41.21 | 22127 |
| FINETUNE5_4w-r16-q_k_v_o | 57.26 | 54.35 | 79.74 | 52.29 | 42.68 | 22153 |
| FINETUNE5_4w-r4-gate_up_down | 56.51 | 52.82 | 79.13 | 52.83 | 41.28 | 22899 |
| FINETUNE5_4w-r8-gate_up_down | 56.10 | 52.73 | 79.14 | 52.56 | 39.99 | 22926 |
| FINETUNE5_4w-r16-gate_up_down | 56.23 | 52.39 | 79.48 | 53.42 | 39.62 | 22963 |
| FINETUNE5_4w-r4-q_k_v_o_gate_up_down | 56.06 | 52.56 | 79.21 | 51.67 | 40.80 | 24303 |
| FINETUNE5_4w-r8-q_k_v_o_gate_up_down | 56.35 | 51.88 | 79.42 | 52.00 | 42.10 | 24376 |
| FINETUNE5_4w-r16-q_k_v_o_gate_up_down | 56.73 | 54.18 | 79.53 | 52.77 | 40.46 | 24439 |
import json
from datasets import load_dataset
# 讀取數據集,take可以取得該數據集前n筆資料
dataset = load_dataset("huangyt/FINETUNE5", split="train", streaming=True)
# 提取所需欄位並建立新的字典列表
extracted_data = []
for example in dataset:
extracted_example = {
"instruction": example["instruction"],
"input": example["input"],
"output": example["output"]
}
extracted_data.append(extracted_example)
# 指定 JSON 文件名稱
json_filename = "FINETUNE5.json"
# 寫入 JSON 文件
with open(json_filename, "w") as json_file:
json.dump(extracted_data, json_file, indent=4)
print(f"數據已提取並保存為 {json_filename}")