Update app.py

app.py

@@ -356,40 +356,49 @@ def create_cluster_dataframes(processed_df):
 
 
 
-
+from random import uniform
 from transformers import GPTNeoForCausalLM, GPT2Tokenizer
 def generate_project_proposal(prompt): # Generate the proposal
     # model_Name = "EleutherAI/gpt-neo-2.7B"
+    # tempareCHUR = uniform(0.3,0.6)
+    
     model_Name = "EleutherAI/gpt-neo-1.3B"
+    tempareCHUR = uniform(0.5,0.8)
 
     consoleMessage_and_Print(f"Trying to access {model_Name} model. The Prompt is: \n{prompt}")
     
     model = GPTNeoForCausalLM.from_pretrained(model_Name)
     tokenizer = GPT2Tokenizer.from_pretrained(model_Name)
-    model_max_token_limit = 2048
+    model_max_token_limit = 2047
         
     try:
         # input_ids = tokenizer.encode(prompt, return_tensors="pt")
         # Truncate the prompt to fit within the model's input limits
         # Adjust as per your model's limit
-        input_ids = tokenizer.encode(prompt, return_tensors="pt", truncation=True, max_length = model_max_token_limit/2)
+        input_ids = tokenizer.encode(prompt, return_tensors="pt", truncation=True, max_length = int(model_max_token_limit/2) )
         
         
         print("Input IDs shape:", input_ids.shape)
+        
+        pad_tokenId = tokenizer.pad_token_id if tokenizer.pad_token_id is not None else tokenizer.eos_token_id # Padding with EOS token may always be great
+        attentionMask = input_ids.ne(pad_tokenId).long()
+
         # Generate the output
         output = model.generate(
             input_ids,
+            min_length = int(model_max_token_limit/3),  # minimum length of the generated output
             max_new_tokens = model_max_token_limit,
             num_return_sequences=1,
             no_repeat_ngram_size=2,
-            temperature=0.5,
-            pad_token_id=tokenizer.eos_token_id  # Ensure padding with EOS token
+            temperature=tempareCHUR,
+            attention_mask=attentionMask, # This was previously not being used
+            pad_token_id=pad_tokenId
             )
         print("Output shape:", output.shape)
 
 
         # Decode the output to text
-        full_returned_segment = tokenizer.decode(output[0], skip_special_tokens=True)
+        # full_returned_segment = tokenizer.decode(output[0], skip_special_tokens=True)
 
         # Slice off the input part if the input length is known
         input_length = input_ids.shape[1]
@@ -410,10 +419,9 @@ def generate_project_proposal(prompt): # Generate the proposal
 
 
 
-import copy
-
 
 
+import copy
 def create_project_proposals(budget_cluster_df, problem_cluster_df, location_clusters, problem_clusters):
     consoleMessage_and_Print("\n Starting function: create_project_proposals")
     proposals = {}