Create app.py

app.py

@@ -0,0 +1,75 @@
+import gradio as gr
+from huggingface_hub import InferenceClient
+
+client = InferenceClient("google/gemma-1.1-2b-it")
+client = InferenceClient("mistralai/Mistral-Nemo-Instruct-2407")
+
+def models(Query): 
+    
+    messages = []
+    
+    messages.append({"role": "user", "content": f"[SYSTEM] You are ASSISTANT who answer question asked by user in short and concise manner. [USER] {Query}"})
+
+    Response = ""
+
+    for message in client.chat_completion(
+        messages,
+        max_tokens=2048,
+        stream=True
+    ):
+        token = message.choices[0].delta.content
+
+        Response += token
+        yield Response
+
+def nemo(query):
+    budget = 3
+    message = f"""[INST] [SYSTEM] You are a helpful assistant in normal conversation.
+When given a problem to solve, you are an expert problem-solving assistant.
+Your task is to provide a detailed, step-by-step solution to a given question.
+Follow these instructions carefully:
+1. Read the given question carefully and reset counter between <count> and </count> to {budget} (maximum 3 steps).
+2. Think critically like a human researcher or scientist. Break down the problem using first principles to conceptually understand and answer the question.
+3. Generate a detailed, logical step-by-step solution.
+4. Enclose each step of your solution within <step> and </step> tags.
+5. You are allowed to use at most {budget} steps (starting budget), keep track of it by counting down within tags <count> </count>, STOP GENERATING MORE STEPS when hitting 0, you don't have to use all of them.
+6. Do a self-reflection when you are unsure about how to proceed, based on the self-reflection and reward, decide whether you need to return to the previous steps.
+7. After completing the solution steps, reorganize and synthesize the steps into the final answer within <answer> and </answer> tags.
+8. Provide a critical, honest, and subjective self-evaluation of your reasoning process within <reflection> and </reflection> tags.
+9. Assign a quality score to your solution as a float between 0.0 (lowest quality) and 1.0 (highest quality), enclosed in <reward> and </reward> tags.
+Example format:
+<count> [starting budget] </count>
+<step> [Content of step 1] </step>
+<count> [remaining budget] </count>
+<step> [Content of step 2] </step>
+<reflection> [Evaluation of the steps so far] </reflection>
+<reward> [Float between 0.0 and 1.0] </reward>
+<count> [remaining budget] </count>
+<step> [Content of step 3 or Content of some previous step] </step>
+<count> [remaining budget] </count>
+...
+<step> [Content of final step] </step>
+<count> [remaining budget] </count>
+<answer> [Final Answer] </answer> (must give final answer in this format)
+<reflection> [Evaluation of the solution] </reflection>
+<reward> [Float between 0.0 and 1.0] </reward> [/INST] [INST] [QUERY] {query} [/INST] [ASSISTANT] """
+
+    stream = client.text_generation(message, max_new_tokens=4096, stream=True, details=True, return_full_text=False)
+    output = ""
+
+    for response in stream:
+        output += response.token.text
+    return output
+
+description="# Chat GO\n### Enter your query and Press enter and get lightning fast response"
+
+with gr.Blocks() as demo1:
+    gr.Interface(description=description,fn=models, inputs=["text"], outputs="text")
+with gr.Blocks() as demo2:
+    gr.Interface(description="Very low but critical thinker",fn=nemo, inputs=["text"], outputs="text", api_name="critical_thinker", concurrency_limit=10)
+
+with gr.Blocks() as demo:
+    gr.TabbedInterface([demo1, demo2] , ["Fast", "Critical"])
+
+demo.queue(max_size=300000)
+demo.launch()