Update app.py

app.py

@@ -9,38 +9,12 @@ import torch.nn.functional as F
 from Bio.Blast import NCBIWWW, NCBIXML
 from transformers import AutoTokenizer, AutoModelForSequenceClassification, AutoModel
 from huggingface_hub import hf_hub_download
+import spaces  # REQUIRED FOR ZEROGPU
 
 # ===================================
-# 1. LOAD AI MODELS (GLOBALLY CACHED)
+# 1. GLOBAL VARIABLES
 # ===================================
-print("Waking up the Genomic Oracle...\n")
-
-# A. Kadir's Gatekeeper
-print("Loading Logression Model...\n")
-clf_coding = joblib.load("coding_classifier_universal.joblib")
-
-# B. Base DNABERT 
-print("Loading foundational DNABERT Architecture...\n")
-tokenizer_base = AutoTokenizer.from_pretrained("DNABERT_Local", trust_remote_code=True)
-model_base = AutoModel.from_pretrained("DNABERT_Local", trust_remote_code=True)
-model_base.eval()
-
-# C. DNABERT-2 Promoter Model 
-print("Loading DNABERT-2 Neural Network...\n")
-tokenizer_promoter = AutoTokenizer.from_pretrained("llm_promoter_classifier_v2", trust_remote_code=True)
-model_promoter = AutoModelForSequenceClassification.from_pretrained("llm_promoter_classifier_v2", trust_remote_code=True)
-model_promoter.eval()
-
-# D. Multi-Feature LightGBM 
-print("Downloading LightGBM from Model Repository...")
-lgbm_path = hf_hub_download(repo_id="Geonomic/Genomic-Oracle-Weights", filename="dnabert_lightgbm_model_feature_type.pkl")
-lightgbm_model = joblib.load(lgbm_path)
-
-# E. Custom ALiBi Lean/Obese BERT 
-print("Downloading Phenotype BERT from Model Repository...")
-tokenizer_pheno = AutoTokenizer.from_pretrained("Geonomic/Genomic-Oracle-Weights", trust_remote_code=True)
-model_pheno = AutoModelForSequenceClassification.from_pretrained("Geonomic/Genomic-Oracle-Weights", trust_remote_code=True)
-model_pheno.eval()
+print("Waking up the Genomic Oracle... Awaiting sequence for Immediate GPU Access.\n")
 
 # Structural Feature Dictionary
 FEATURE_DICT = {
@@ -51,17 +25,35 @@ FEATURE_DICT = {
 # ==============================================
 # 2. CORE INFERENCE ENGINE (ZeroGPU Accelerated)
 # ==============================================
-import spaces  # Import ZeroGPU here to prevent the PyTorch Meta Device Error
-
-@spaces.GPU  
+@spaces.GPU  # This tells Hugging Face to teleport this specific math to the A100 GPU
 def run_deep_learning_cascade(dna_sequence):
     device = torch.device("cuda")  # Wakes up the A100 connection
     clean_seq = "".join(dna_sequence.split()).upper()
     
+    print("🚀 Allocated GPU is Awake! Loading ALiBi models directly into VRAM...")
+    
+    # --- JUST-IN-TIME MODEL LOADING ---
+    clf_coding = joblib.load("coding_classifier_universal.joblib")
+    
+    tokenizer_base = AutoTokenizer.from_pretrained("DNABERT_Local", trust_remote_code=True)
+    model_base = AutoModel.from_pretrained("DNABERT_Local", trust_remote_code=True).to(device)
+    model_base.eval()
+    
+    tokenizer_promoter = AutoTokenizer.from_pretrained("llm_promoter_classifier_v2", trust_remote_code=True)
+    model_promoter = AutoModelForSequenceClassification.from_pretrained("llm_promoter_classifier_v2", trust_remote_code=True).to(device)
+    model_promoter.eval()
+    
+    lgbm_path = hf_hub_download(repo_id="Geonomic/Genomic-Oracle-Weights", filename="dnabert_lightgbm_model_feature_type.pkl")
+    lightgbm_model = joblib.load(lgbm_path)
+    
+    tokenizer_pheno = AutoTokenizer.from_pretrained("Geonomic/Genomic-Oracle-Weights", trust_remote_code=True)
+    model_pheno = AutoModelForSequenceClassification.from_pretrained("Geonomic/Genomic-Oracle-Weights", trust_remote_code=True).to(device)
+    model_pheno.eval()
+    
     # --- LEVEL 1: Base Embedding & Kadir's Gatekeeper ---
     inputs = tokenizer_base([clean_seq], return_tensors="pt", max_length=300, truncation=True, padding=True)
     inputs = {k: v.to(device) for k, v in inputs.items()}
-
+    
     with torch.no_grad():
         out_base = model_base(**inputs)
         mask = inputs["attention_mask"].unsqueeze(-1)