Florence 2 FT Lung Cancer Detection

🚀 醫學視覺問答模型 - Florence-2_FT_Lung-Cancer-detection

基於microsoft/Florence-2-base-ft微調的肺癌檢測模型，利用肺部圖像精準識別肺癌類型，為醫學診斷提供高效解決方案。

🚀 快速開始

安裝依賴

! pip install -q "flash_attn==2.6.3" "timm==1.0.8" "einops==0.8.0" "transformers==4.44.0" 

環境配置

device = "cuda:0" if torch.cuda.is_available() else "cpu"
torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32

加載模型和處理器

model = AutoModelForCausalLM.from_pretrained("nirusanan/Florence-2_FT_Lung-Cancer-detection", torch_dtype=torch_dtype, trust_remote_code=True).to(device)
processor = AutoProcessor.from_pretrained("nirusanan/Florence-2_FT_Lung-Cancer-detection", trust_remote_code=True)

運行示例

prompt = "<DocVQA>" + "What is the type of lung cancer?"

url = "https://www.uab.edu/news/images/ct_scan.jpg"
image = Image.open(requests.get(url, stream=True).raw)

inputs = processor(text=prompt, images=image, return_tensors="pt").to(device, torch_dtype)

generated_ids = model.generate(
    input_ids=inputs["input_ids"],
    pixel_values=inputs["pixel_values"],
    max_new_tokens=1024,
    do_sample=False,
    num_beams=3
)
generated_text = processor.batch_decode(generated_ids, skip_special_tokens=False)[0]

parsed_answer = processor.post_process_generation(generated_text, task="<DocVQA>", image_size=(image.width, image.height))

print(parsed_answer)

✨ 主要特性

• 微調模型：該模型是microsoft/Florence-2-base-ft的微調版本，專門針對肺癌檢測任務進行優化。
• 視覺問答：支持視覺問答任務，可根據肺部圖像回答相關問題。

📦 安裝指南

運行以下命令安裝所需依賴：

! pip install -q "flash_attn==2.6.3" "timm==1.0.8" "einops==0.8.0" "transformers==4.44.0" 

💻 使用示例

基礎用法

# 安裝依賴
! pip install -q "flash_attn==2.6.3" "timm==1.0.8" "einops==0.8.0" "transformers==4.44.0" 

# 環境配置
device = "cuda:0" if torch.cuda.is_available() else "cpu"
torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32

# 加載模型和處理器
model = AutoModelForCausalLM.from_pretrained("nirusanan/Florence-2_FT_Lung-Cancer-detection", torch_dtype=torch_dtype, trust_remote_code=True).to(device)
processor = AutoProcessor.from_pretrained("nirusanan/Florence-2_FT_Lung-Cancer-detection", trust_remote_code=True)

# 定義問題和圖像
prompt = "<DocVQA>" + "What is the type of lung cancer?"
url = "https://www.uab.edu/news/images/ct_scan.jpg"
image = Image.open(requests.get(url, stream=True).raw)

# 處理輸入
inputs = processor(text=prompt, images=image, return_tensors="pt").to(device, torch_dtype)

# 生成回答
generated_ids = model.generate(
    input_ids=inputs["input_ids"],
    pixel_values=inputs["pixel_values"],
    max_new_tokens=1024,
    do_sample=False,
    num_beams=3
)
generated_text = processor.batch_decode(generated_ids, skip_special_tokens=False)[0]

# 解析回答
parsed_answer = processor.post_process_generation(generated_text, task="<DocVQA>", image_size=(image.width, image.height))

# 打印回答
print(parsed_answer)

📚 詳細文檔

模型信息

評估指標

測試準確率：99.17%

開發者信息

• 開發者：Nirusanan
• 許可證：未提供
• 基礎模型：microsoft/Florence-2-base-ft