app.py

import os, shutil, json, re
import torch
import torch.nn as nn
import torch.nn.functional as F
import gradio as gr
from transformers import AutoTokenizer, AutoModel

# ==========================
# 🚧 0. 基础设置与缓存清理 (保持不变)
# ==========================
os.environ["HF_HOME"] = "/tmp/hf_cache"
os.environ["TRANSFORMERS_CACHE"] = "/tmp/hf_cache"
os.environ["HF_HUB_DISABLE_SYMLINKS_WARNING"] = "1"

for path in ["/tmp/hf_cache", os.path.expanduser("~/.cache/huggingface")]:
    shutil.rmtree(path, ignore_errors=True)
    os.makedirs(path, exist_ok=True)

# ==========================
# 1. Model Definition (保持不变)
# ==========================
class AttentionPooling(nn.Module):
    def __init__(self, d_model):
        super().__init__()
        self.attention_net = nn.Linear(d_model, 1)

    def forward(self, x, mask):
        attn_logits = self.attention_net(x).squeeze(2)
        attn_logits.masked_fill_(mask == 0, -float('inf'))
        attn_weights = F.softmax(attn_logits, dim=1)
        return torch.bmm(attn_weights.unsqueeze(1), x).squeeze(1)

class ProtDualBranchEnhancedClassifier(nn.Module):
    def __init__(self, d_model, projection_dim, num_classes, dropout, kernel_size):
        super().__init__()
        self.cls_projector = nn.Linear(d_model, projection_dim)
        self.token_refiner = nn.Sequential(
            nn.Conv1d(d_model, d_model, kernel_size, padding='same'),
            nn.ReLU()
        )
        self.attention_pooling = AttentionPooling(d_model)
        self.tok_projector = nn.Linear(d_model, projection_dim)
        fused_dim = projection_dim * 2
        self.gate = nn.Sequential(
            nn.Linear(fused_dim, fused_dim),
            nn.Sigmoid()
        )
        self.classifier_head = nn.Sequential(
            nn.LayerNorm(fused_dim),
            nn.Linear(fused_dim, fused_dim * 2),
            nn.ReLU(),
            nn.Dropout(dropout),
            nn.Linear(fused_dim * 2, num_classes)
        )

    def forward(self, cls_embedding, token_embeddings, mask):
        z_cls = self.cls_projector(cls_embedding)
        tok_emb_permuted = token_embeddings.permute(0, 2, 1)
        refined_tok_emb = self.token_refiner(tok_emb_permuted).permute(0, 2, 1)
        z_tok_pooled = self.attention_pooling(refined_tok_emb, mask)
        z_tok = self.tok_projector(z_tok_pooled)
        z_fused_concat = torch.cat([z_cls, z_tok], dim=1)
        gate_values = self.gate(z_fused_concat)
        z_fused_gated = z_fused_concat * gate_values
        return self.classifier_head(z_fused_gated)

# ==========================
# 2. Load Models (保持不变)
# ==========================
DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
PLM_MODEL_NAME = "facebook/esm2_t30_150M_UR50D"  
CLASSIFIER_PATH = "best_model_esm2_t30_150M_UR50D.pth"
LABEL_MAP_PATH = "label_map.json"

if not os.path.exists(LABEL_MAP_PATH):
    raise FileNotFoundError(f"Error: Missing '{LABEL_MAP_PATH}'.")
with open(LABEL_MAP_PATH, 'r') as f:
    label_to_idx = json.load(f)
    idx_to_label = {v: k for k, v in label_to_idx.items()}

NUM_CLASSES = len(idx_to_label)
D_MODEL = 640

print("🔹 Loading models...")
tokenizer = AutoTokenizer.from_pretrained(PLM_MODEL_NAME)
plm_model = AutoModel.from_pretrained(PLM_MODEL_NAME).to(DEVICE)
plm_model.eval()

classifier = ProtDualBranchEnhancedClassifier(
    d_model=D_MODEL, projection_dim=32, num_classes=NUM_CLASSES,
    dropout=0.3, kernel_size=3
).to(DEVICE)

if not os.path.exists(CLASSIFIER_PATH):
    raise FileNotFoundError(f"Error: Could not find '{CLASSIFIER_PATH}'.")

classifier.load_state_dict(torch.load(CLASSIFIER_PATH, map_location=DEVICE))
classifier.eval()
print("✅ Ready.")

# ==========================
# 3. Predict Logic (保持不变)
# ==========================
def predict(sequence_input):
    if not sequence_input or sequence_input.isspace():
        raise gr.Error("Sequence cannot be empty.")
    
    sequence = "".join(sequence_input.split('\n')[1:]) if sequence_input.startswith('>') else sequence_input
    sequence = re.sub(r'[^A-Z]', '', sequence.upper())

    if not sequence:
        raise gr.Error("Invalid sequence.")

    with torch.no_grad():
        inputs = tokenizer(sequence, return_tensors="pt", truncation=True, max_length=1024).to(DEVICE)
        outputs = plm_model(**inputs)
        hidden_states = outputs.last_hidden_state
        cls_embedding = hidden_states[:, 0, :]
        token_embeddings = hidden_states[:, 1:-1, :]
        token_mask = inputs['attention_mask'][:, 1:-1]
        
        logits = classifier(cls_embedding, token_embeddings, token_mask)
        probabilities = F.softmax(logits, dim=1)[0]

    confidences = {idx_to_label[i]: float(prob) for i, prob in enumerate(probabilities)}
    return confidences

# ==========================
# 4. Ultra-Modern UI Design
# ==========================

# 极简现代风 CSS
modern_css = """
@import url('https://fonts.googleapis.com/css2?family=Inter:wght@300;400;600;800&display=swap');

body {
    font-family: 'Inter', sans-serif !important;
    background-color: #f8fafc;
}

/* 1. 顶部 Hero Section */
.hero-container {
    text-align: center;
    padding: 3rem 1rem;
    margin-bottom: 1rem;
}
.hero-title {
    font-size: 3rem;
    font-weight: 800;
    margin-bottom: 0.5rem;
    background: -webkit-linear-gradient(45deg, #0f172a, #334155);
    -webkit-background-clip: text;
    -webkit-text-fill-color: transparent;
    letter-spacing: -1px;
}
.hero-subtitle {
    font-size: 1.25rem;
    color: #64748b;
    font-weight: 300;
    max-width: 600px;
    margin: 0 auto;
}

/* 2. 卡片风格 */
.modern-card {
    background: white;
    border-radius: 16px;
    padding: 24px;
    border: 1px solid #e2e8f0;
    box-shadow: 0 4px 6px -1px rgba(0, 0, 0, 0.05), 0 2px 4px -1px rgba(0, 0, 0, 0.03);
    transition: all 0.3s ease;
}
.modern-card:hover {
    box-shadow: 0 10px 15px -3px rgba(0, 0, 0, 0.1), 0 4px 6px -2px rgba(0, 0, 0, 0.05);
}

/* 3. 输入框优化 - 模仿代码编辑器 */
textarea {
    font-family: 'SF Mono', 'Menlo', 'Monaco', 'Courier New', monospace !important;
    font-size: 14px !important;
    background-color: #f8fafc !important;
    border: 1px solid #e2e8f0 !important;
    border-radius: 8px !important;
}

/* 4. 按钮优化 */
button.primary {
    background: linear-gradient(135deg, #2563eb 0%, #1d4ed8 100%) !important;
    border: none !important;
    font-weight: 600 !important;
    letter-spacing: 0.5px !important;
    transition: transform 0.1s ease-in-out !important;
}
button.primary:hover {
    transform: translateY(-2px);
    box-shadow: 0 4px 12px rgba(37, 99, 235, 0.3);
}

/* 5. 标签页优化 */
.tabs {
    border: none !important;
    background: transparent !important;
}
.tab-nav {
    border-bottom: 1px solid #e2e8f0;
    margin-bottom: 20px;
}
.tab-nav button {
    font-weight: 600;
    color: #64748b;
}
.tab-nav button.selected {
    color: #2563eb;
    border-bottom: 2px solid #2563eb;
}

/* 6. Footer */
.footer-text {
    text-align: center;
    color: #94a3b8;
    font-size: 0.8rem;
    margin-top: 40px;
    padding-bottom: 20px;
}
"""

# 使用极简主题作为底子
theme = gr.themes.Soft(
    primary_hue="blue",
    radius_size="lg",
    font=[gr.themes.GoogleFont("Inter"), "system-ui", "sans-serif"]
)

with gr.Blocks(theme=theme, css=modern_css, title="LocPred-Prok") as app:
    
    # --- Hero Section ---
    with gr.Column(elem_classes="hero-container"):
        gr.HTML("""
            <div class="hero-title">LocPred-Prok</div>
            <div class="hero-subtitle">
                Next-generation prokaryotic subcellular localization using dual-branch protein language models.
            </div>
        """)

    # --- Main Content ---
    with gr.Tabs():
        
        # === TAB 1: Predict ===
        with gr.TabItem("Predict", id="tab-predict"):
            with gr.Row():
                # Input Column
                with gr.Column(scale=3, elem_classes="modern-card"):
                    gr.Markdown("### Sequence Input")
                    sequence_input = gr.Textbox(
                        lines=12,
                        placeholder="> Paste FASTA sequence here...",
                        show_label=False,
                        container=False
                    )
                    
                    with gr.Row():
                        clear_btn = gr.ClearButton(components=[sequence_input], value="Clear")
                        submit_btn = gr.Button("Analyze Sequence", variant="primary", scale=2)

                # Output Column
                with gr.Column(scale=2, elem_classes="modern-card"):
                    gr.Markdown("### Analysis Result")
                    # 隐藏 Label 自身的文字标签，保持界面干净
                    output_label = gr.Label(num_top_classes=NUM_CLASSES, show_label=False)
                    
                    gr.HTML("""
                        <div style="margin-top: 20px; padding: 10px; background: #eff6ff; border-radius: 8px; font-size: 0.85rem; color: #1e40af;">
                            ℹ️ <b>Model Insight:</b> Prediction is based on the fusion of global semantic features (ESM-2) and local structural refinements.
                        </div>
                    """)

        # === TAB 2: Methodology ===
        with gr.TabItem("Methodology", id="tab-about"):
            with gr.Column(elem_classes="modern-card"):
                gr.Markdown("### The Architecture")
                gr.Markdown(
                    """
                    **LocPred-Prok** moves beyond the "bigger is better" paradigm. Instead of relying solely on massive parameter counts, we engineered a specialized **Dual-Branch Architecture**:
                    
                    * **Global Branch:** Leverages the `ESM-2 (150M)` foundation model to capture deep semantic dependencies.
                    * **Local Branch:** Utilizes convolutional refinement and attention pooling to detect subtle signal motifs often missed by global pooling.
                    
                    This synergy allows for precise identification of challenging localization sites, particularly in **Cell Wall** and **Outer Membrane** regions.
                    """
                )
                
        # === TAB 3: Cite ===
        with gr.TabItem("Cite", id="tab-cite"):
            with gr.Column(elem_classes="modern-card"):
                gr.Markdown("### BibTeX Reference")
                gr.Code(
                    value="""@article{LocPredProk2025,
  title={LocPred-Prok: Prokaryotic protein subcellular localization prediction with a dual-branch architecture},
  author={Your Name et al.},
  journal={Bioinformatics},
  year={2025}
}""",
                    label=None,
                    language=None, # 防止之前的报错
                    interactive=False
                )

    # --- Footer ---
    gr.HTML("""
        <div class="footer-text">
            © 2025 iSysLab HUST &nbsp;|&nbsp; Powered by PyTorch & ESM-2
        </div>
    """)

    # Logic
    submit_btn.click(fn=predict, inputs=sequence_input, outputs=output_label)
    clear_btn.click(lambda: None, outputs=[output_label])

app.launch()