app.py

from pickle import TRUE
import gradio as gr
from transformers import Qwen2Tokenizer, Qwen2ForCausalLM, AutoModelForCausalLM, AutoTokenizer
import torch

#Labels
sDefaultQuestion = "What is K-means clustering?"
sDescription = "Question & Answer"
sTitle = "Question & Answer application using Qwen"

def ask_question_HFModel1(question, modelHF, tokenizerHF):
    inputs = tokenizerHF.encode('Q: ' + question + ' A:', return_tensors='pt')
    attention_mask = torch.ones(inputs.shape)
    outputs = modelHF.generate(inputs, attention_mask = attention_mask, max_new_tokens=100, num_return_sequences=1, )
    gen_text = tokenizerHF.decode(outputs[0], skip_special_tokens=True)
    question, answer = gen_text.split(' A:')
    return question, answer

def FetchQwenFromSpace(sQ):
    HFModelName ="Qwen/Qwen1.5-0.5B"
    tokenizer = Qwen2Tokenizer.from_pretrained(HFModelName) #, trust_remote_code=bTrust_remote_code, model_max_length=8192) #model_max_length = 8192   #, use_fast=False    #Qwen
    model = Qwen2ForCausalLM.from_pretrained(HFModelName)
    modelHF = model
    tokenizerHF = tokenizer
    '''
    #Using pre-trained model from HF
    #from transformers import AutoModelCasualLM
    sHuggingFacePath = "RohiniPS/Qwen1B-QnA-3-5" #"RohiniPS/Qwen1B-QnA-3-5" 
    modelHF = Qwen2ForCausalLM.from_pretrained(sHuggingFacePath)    #("your_username/my-awesome-model")
    tokenizerHF = Qwen2Tokenizer.from_pretrained(sHuggingFacePath)
    ###'''
    
    sQuestion, sAnswer = ask_question_HFModel1(sQ, modelHF, tokenizerHF)
    return sQuestion, sAnswer
    

def GetAnswer(Question):
    sQuestion, sAnswer = FetchQwenFromSpace(Question)
    # sAnswer = "Qwen 1.5-0.5B"
    return sAnswer


# Call Gradio 
GradioCall = gr.Interface(fn=GetAnswer, 
                    inputs=gr.TextArea(value=sDefaultQuestion,lines=7, max_lines=20, label="Ask a question related to AIML!",autoscroll=True), #Input - Question Asked
                    outputs=[gr.TextArea(label="Qwen's Response - ")], 
                    title=sTitle,
                    description=sDescription,)

# Launch - Qwen's QnA!
GradioCall.launch()


'''
from pickle import TRUE
import gradio as gr
from transformers import Qwen2Tokenizer, Qwen2ForCausalLM

sHuggingFacePath = "RohiniPS/Qwen1B-QnA-3-5" 

# Initialize title and description
sDefaultQuestion = "What is K-means clustering?"
sDescription = "Question & Answer"
sTitle = "Question & Answer application using Qwen"

def ask_question_HFModel1(question, modelHF, tokenizerHF):
    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
    modelHF = modelHF.to(device)
    inputs = tokenizerHF.encode('Q: ' + question + ' A:', return_tensors='pt')
    #inputs = tokenizerHF.encode('Q: ' + question + ' A:', return_tensors='pt').to(device)
    attention_mask = torch.ones(inputs.shape, device=device)
    outputs = modelHF.generate(inputs, attention_mask = attention_mask, max_new_tokens=100, num_return_sequences=1, )
    gen_text = tokenizerHF.decode(outputs[0], skip_special_tokens=True)
    question, answer = gen_text.split(' A:')
    return question, answer

def FetchQwenFromSpace(sQ):
            #Initialize Tokenizer & Model
            modelHF = Qwen2ForCausalLM.from_pretrained(sHuggingFacePath)    
            tokenizerHF = Qwen2Tokenizer.from_pretrained(sHuggingFacePath)
            sQuestion, sAnswer = ask_question_HFModel1(sQ, modelHF, tokenizerHF)
            return sQuestion, sAnswer


def GetAnswer(Question):
    
    sQuestion, sAnswer = FetchQwenFromSpace(Question)
    # sAnswer += "convnet_from_scratch_with_augmentation.keras is our trained Tensorflow model file. app.py contains our Gradio app (similar to the code that launched the app).Note: app.py is the default filename used for Hugging Face Spaces, if you deploy your app there, Spaces will by default look for a file called app.py to run. This is changable in settings. examples/ contains example images to use with our Gradio app. model.py contains the model defintion as well as any transforms assosciated with the model. requirements.txt contains the dependencies to run our app such as tensorflow, numpy and gradio."
    # sAnswer += sAnswer
    return sQuestion, sAnswer


# Call Gradio 
GradioCall = gr.Interface(fn=GetAnswer, 
                    inputs=gr.TextArea(value=sDefaultQuestion,lines=7, max_lines=20, label="Ask a question related to AIML!",autoscroll=True), 
                    outputs=[gr.Label(label="You asked,"), 
                                  gr.TextArea(label="Qwen's Response - ")], 
                    title=sTitle,
                    description=sDescription,)

# Launch the Gradio!
GradioCall.launch(inline=False, share=True)
'''