我想要一天分享一點「LLM從底層堆疊的技術」,並且每篇文章長度控制在三分鐘以內,讓大家不會壓力太大,但是又能夠每天成長一點。
- AI說書 - 從0開始 - 585 | F-AGI 套件安裝 & 載入圖片
- AI說書 - 從0開始 - 586 | F-AGI 載入圖片
- AI說書 - 從0開始 - 587 | F-AGI 載入圖片
- AI說書 - 從0開始 - 588 | F-AGI 流程介紹
- AI說書 - 從0開始 - 589 | F-AGI 流程介紹
- AI說書 - 從0開始 - 590 | F-AGI 簡單圖片推論
我們可以下載記錄檔:
!curl L https://raw.githubusercontent.com/Denis2054/Transformers_3rd_Edition/master/Chapter19/HGPT1.json --output "HGPT1.json"
並且撰寫程式進行檢視:
with open("HGPT1.json") as f:
data = json.load(f)
for key, value in data.items():
print(f"Task ID: {value['task']['id']}")
print(f"Task Type: {value['task']['task']}")
print(f"Dependencies: {', '.join(map(str, value['task']['dep']))}")
if 'args' in value['task']:
for arg_key, arg_value in value['task']['args'].items():
print(f" {arg_key.capitalize()}: {arg_value}")
if 'inference result' in value:
print("Inference Result:")
if isinstance(value['inference result'], dict):
for result_key, result_value in value['inference result'].items():
print(f" {result_key.capitalize()}: {result_value}")
else:
for item in value['inference result']:
print(f" Answer: {item['answer']} (Score: {item['score']})")
if 'choose model result' in value:
print(f"Model Chosen: {value['choose model result']['id']}")
print(f"Reason: {value['choose model result']['reason']}")
結果為:






















