Update app.py

app.py

@@ -1,13 +1,60 @@
-import time
 import gradio as gr
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from transformers import StoppingCriteria, StoppingCriteriaList, TextIteratorStreamer
+from threading import Thread
 
-def slow_echo(message, history):
-    for i in range(len(message)):
-        time.sleep(0.05)
-        yield "You typed: " + message[: i+1]
+# Loading the tokenizer and model from Hugging Face's model hub.
+tokenizer = AutoTokenizer.from_pretrained("TinyLlama/TinyLlama-1.1B-Chat-v1.0")
+model = AutoModelForCausalLM.from_pretrained("TinyLlama/TinyLlama-1.1B-Chat-v1.0")
+
+# using CUDA for an optimal experience
+device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+model = model.to(device)
+
+
+# Defining a custom stopping criteria class for the model's text generation.
+class StopOnTokens(StoppingCriteria):
+    def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor, **kwargs) -> bool:
+        stop_ids = [2]  # IDs of tokens where the generation should stop.
+        for stop_id in stop_ids:
+            if input_ids[0][-1] == stop_id:  # Checking if the last generated token is a stop token.
+                return True
+        return False
+
+
+# Function to generate model predictions.
+def predict(message, history):
+    history_transformer_format = history + [[message, ""]]
+    stop = StopOnTokens()
+
+    # Formatting the input for the model.
+    messages = "</s>".join(["</s>".join(["\n<|user|>:" + item[0], "\n<|assistant|>:" + item[1]])
+                        for item in history_transformer_format])
+    model_inputs = tokenizer([messages], return_tensors="pt").to(device)
+    streamer = TextIteratorStreamer(tokenizer, timeout=10., skip_prompt=True, skip_special_tokens=True)
+    generate_kwargs = dict(
+        model_inputs,
+        streamer=streamer,
+        max_new_tokens=1024,
+        do_sample=True,
+        top_p=0.95,
+        top_k=50,
+        temperature=0.7,
+        num_beams=1,
+        stopping_criteria=StoppingCriteriaList([stop])
+    )
+    t = Thread(target=model.generate, kwargs=generate_kwargs)
+    t.start()  # Starting the generation in a separate thread.
+    partial_message = ""
+    for new_token in streamer:
+        partial_message += new_token
+        if '</s>' in partial_message:  # Breaking the loop if the stop token is generated.
+            break
+        yield partial_message
 
 demo = gr.ChatInterface(
-    slow_echo,
+    predict,
     chatbot=gr.Chatbot(height=300),
     textbox=gr.Textbox(placeholder="Try Pragna SFT", container=False, scale=7),
     title="pragna-1b-it",