Update app.py

app.py

@@ -7,14 +7,15 @@ import joblib
 import numpy as np
 import torch.nn.functional as F
 from Bio.Blast import NCBIWWW, NCBIXML
-from transformers import AutoTokenizer, AutoModelForSequenceClassification, AutoModel
+
+# 🚨 Colab Native Imports for the Phenotype model!
+from transformers import AutoTokenizer, AutoModelForSequenceClassification, AutoModel, BertTokenizer, BertForSequenceClassification, AutoConfig
 from huggingface_hub import hf_hub_download
-import spaces 
 
 # ===================================
-# 1. LOAD AI MODELS (GLOBALLY CACHED)
+# 1. LOAD AI MODELS (GLOBALLY CACHED ON CPU)
 # ===================================
-print("Waking up the Genomic Oracle...\n")
+print("Waking up the Genomic Oracle... Loading models safely into CPU RAM.\n")
 
 # A. Kadir's Gatekeeper 
 clf_coding = joblib.load("coding_classifier_universal.joblib")
@@ -33,9 +34,10 @@ 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)
 
-# E. Custom ALiBi Lean/Obese BERT 
-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)
+# E. Custom Lean/Obese Phenotype BERT (🚨 Forced Native Architecture via Colab Fix!)
+tokenizer_pheno = BertTokenizer.from_pretrained("Geonomic/Genomic-Oracle-Weights", do_lower_case=False)
+config_pheno = AutoConfig.from_pretrained("Geonomic/Genomic-Oracle-Weights", trust_remote_code=True)
+model_pheno = BertForSequenceClassification.from_pretrained("Geonomic/Genomic-Oracle-Weights", config=config_pheno)
 model_pheno.eval()
 
 FEATURE_DICT = {
@@ -46,8 +48,12 @@ FEATURE_DICT = {
 # ==============================================
 # 2. CORE INFERENCE ENGINE (ZeroGPU Accelerated)
 # ==============================================
-@spaces.GPU  # ZeroGPU natively manages the global models!
+# 🚨 DELAYED IMPORT: ZeroGPU is activated ONLY AFTER CPU loading is perfectly finished!
+import spaces 
+
+@spaces.GPU  
 def run_deep_learning_cascade(dna_sequence):
+    # We send the inputs to cuda, and ZeroGPU natively handles teleporting the models!
     device = torch.device("cuda")  
     clean_seq = "".join(dna_sequence.split()).upper()