Added max new tokens value

main.py

@@ -1,5 +1,5 @@
 from flask import Flask, request, render_template, jsonify
-from models import tiny as chatbot
+from models import tinystories as chatbot
 
 app = Flask("AI API")
 

models/fast.py

@@ -11,6 +11,6 @@ def load():
 
 def generate(input_text):
   input_ids = tokenizer.encode(input_text, return_tensors="pt")
-  output_ids = model.generate(input_ids, num_beams=5, no_repeat_ngram_size=2)
+  output_ids = model.generate(input_ids, num_beams=5, no_repeat_ngram_size=2, max_new_tokens=100)
   
   return tokenizer.decode(output_ids[0], skip_special_tokens=True)

models/gpt2.py

@@ -16,6 +16,6 @@ def generate(input_text):
   attention_mask = tf.ones_like(input_ids)
   
   # Generate output using the model
-  output_ids = model.generate(input_ids, num_beams=5, no_repeat_ngram_size=2)
+  output_ids = model.generate(input_ids, num_beams=5, no_repeat_ngram_size=2, max_new_tokens=100)
   
   return tokenizer.decode(output_ids[0], skip_special_tokens=True)

models/hermes.py

@@ -2,24 +2,17 @@ from transformers import AutoTokenizer, AutoModelForCausalLM
 
 model_name = "NousResearch/Hermes-2-Pro-Llama-3-8B"
 
-model = None
-tokenizer = None
+model = AutoModelForCausalLM.from_pretrained(model_name)
+tokenizer = AutoTokenizer.from_pretrained(model_name)
 
 # Example messages input
 # messages = [
 # {"role": "system", "content": "You are Hermes 2."},
 # {"role": "user", "content": "Hello, who are you?"}
 #]
-
-def load():
-  global model
-  global tokenizer
-  
-  model = AutoModelForCausalLM.from_pretrained(model_name)
-  tokenizer = AutoTokenizer.from_pretrained(model_name)
   
 def generate(messages):
   gen_input = tokenizer.apply_chat_template(messages, return_tensors="pt")
-  output_ids = model.generate(**gen_input, num_beams=5, no_repeat_ngram_size=2)
+  output_ids = model.generate(**gen_input, num_beams=5, no_repeat_ngram_size=2, max_new_tokens=100)
   
   return tokenizer.decode(output_ids[0], skip_special_tokens=True)

models/llama2.py

@@ -11,6 +11,6 @@ def load():
 
 def generate(input_text):
   input_ids = tokenizer.encode(input_text, return_tensors="pt")
-  output_ids = model.generate(input_ids, num_beams=5, no_repeat_ngram_size=2)
+  output_ids = model.generate(input_ids, num_beams=5, no_repeat_ngram_size=2, max_new_tokens=100)
   
   return tokenizer.decode(output_ids[0], skip_special_tokens=True)

models/llama3.py

@@ -11,6 +11,6 @@ def load():
 
 def generate(input_text):
   input_ids = tokenizer.encode(input_text, return_tensors="pt")
-  output_ids = model.generate(input_ids, num_beams=5, no_repeat_ngram_size=2)
+  output_ids = model.generate(input_ids, num_beams=5, no_repeat_ngram_size=2, max_new_tokens=100)
   
   return tokenizer.decode(output_ids[0], skip_special_tokens=True)

models/llamatiny.py

@@ -0,0 +1,16 @@
+from transformers import AutoTokenizer, AutoModelForCausalLM
+
+model_name = "TinyLlama/TinyLlama-1.1B-Chat-v1.0"
+
+def load():
+  global model
+  global tokenizer
+  
+  model = AutoModelForCausalLM.from_pretrained(model_name)
+  tokenizer = AutoTokenizer.from_pretrained(model_name)
+
+def generate(input_text):
+  input_ids = tokenizer.encode(input_text, return_tensors="pt")
+  output_ids = model.generate(input_ids, num_beams=5, no_repeat_ngram_size=2, max_new_tokens=100)
+  
+  return tokenizer.decode(output_ids[0], skip_special_tokens=True)

models/mamba.py

@@ -11,6 +11,6 @@ def load():
 
 def generate(input_text):
   input_ids = tokenizer.encode(input_text, return_tensors="pt")
-  output_ids = model.generate(input_ids, num_beams=5, no_repeat_ngram_size=2)
+  output_ids = model.generate(input_ids, num_beams=5, no_repeat_ngram_size=2, max_new_tokens=100)
   
   return tokenizer.decode(output_ids[0], skip_special_tokens=True)

models/{tiny.py → tinystories.py}

@@ -11,6 +11,6 @@ def load():
 
 def generate(input_text):
   input_ids = tokenizer.encode(input_text, return_tensors="pt")
-  output_ids = model.generate(input_ids, num_beams=5, no_repeat_ngram_size=2)
+  output_ids = model.generate(input_ids, num_beams=5, no_repeat_ngram_size=2, max_new_tokens=100)
   
   return tokenizer.decode(output_ids[0], skip_special_tokens=True)