Using our own models.

mi_clase.py

@@ -17,8 +17,12 @@ from nltk.corpus import stopwords
 from nltk.tokenize import word_tokenize
 from transformers import AutoTokenizer, AutoModel
 
-QUERY_MODEL = "/kaggle/input/bge-small-faiss/"
-GENERATOR_MODEL="/kaggle/input/training-model-2/model_v2"
+
+GENERATOR_MODEL = "JosueElias/pipeline_generator_model"
+GENERATOR_TOKENIZER = "JosueElias/pipeline_generator_tokenizer"
+QUERY_MODEL = "JosueElias/pipeline_query_model"
+QUERY_TOKENIZER = "JosueElias/pipeline_query_tokenizer"
+
 DEVICE = "cpu" # cpu or cuda
 
 class Pipeline:
@@ -27,9 +31,9 @@ class Pipeline:
     
     def __init__(self):
         self.model = AutoModelForMultipleChoice.from_pretrained(GENERATOR_MODEL)
-        self.tokenizer = AutoTokenizer.from_pretrained(GENERATOR_MODEL)
+        self.tokenizer = AutoTokenizer.from_pretrained(GENERATOR_TOKENIZER)
         self.semModel = AutoModel.from_pretrained(QUERY_MODEL)   
-        self.semTokenizer = AutoTokenizer.from_pretrained(QUERY_MODEL)
+        self.semTokenizer = AutoTokenizer.from_pretrained(QUERY_TOKENIZER)
         self.device = torch.device(DEVICE)
 
         self.semModel.to(self.device)