Update app.py

Gen AI text generation with GPT2!

app.py

@@ -131,19 +131,51 @@
 # However, in contrast with zero-shot classification, few-shot classification makes use of very few labeled samples during the training process. 
 # The implementation of the few-shot classification methods can be found in OpenAI, where the GPT3 classifier is a well-known example of a few-shot classifier.
 
-from transformers import pipeline
+# Deploying the following code works but comes with a warning: "No model was supplied, defaulted to facebook/bart-large-mnli and revision c626438 (https://huggingface.co/facebook/bart-large-mnli).
+# Using a pipeline without specifying a model name and revision in production is not recommended."
+
+# from transformers import pipeline
+# import gradio as grad
+
+# zero_shot_classifier = pipeline("zero-shot-classification")
+
+# def classify(text,labels):
+#     classifer_labels = labels.split(",")
+#     #["software", "politics", "love", "movies", "emergency", "advertisment","sports"]
+#     response = zero_shot_classifier(text,classifer_labels)
+#     return response
+
+# txt=grad.Textbox(lines=1, label="English", placeholder="text to be classified")
+# labels=grad.Textbox(lines=1, label="Labels", placeholder="comma separated labels")
+# out=grad.Textbox(lines=1, label="Classification")
+
+# grad.Interface(classify, inputs=[txt,labels], outputs=out).launch()
+
+#-----------------------------------------------------------------------------------
+# 8. Text Generation Task/Models
+# The earliest text generation models were based on Markov chains . Markov chains are like a state machine wherein 
+# using only the previous state, the next state is predicted. This is similar also to what we studied in bigrams.
+
+# Post the Markov chains, recurrent neural networks (RNNs) , which were capable of retaining a greater context of the text, were introduced. 
+# They are based on neural network architectures that are recurrent in nature. RNNs are able to retain a greater context of the text that was introduced. 
+# Nevertheless, the amount of information that these kinds of networks are able to remember is constrained, and it is also difficult to train them, 
+# which means that they are not effective at generating lengthy texts. To counter this issue with RNNs, LSTM architectures were evolved, 
+# which could capture long-term dependencies in text. Finally, we came to transformers, whose decoder architecture became popular for generative models 
+# used for generating text as an example.
+
+from transformers import GPT2LMHeadModel,GPT2Tokenizer
 import gradio as grad
 
-zero_shot_classifier = pipeline("zero-shot-classification")
+mdl = GPT2LMHeadModel.from_pretrained('gpt2')
+gpt2_tkn=GPT2Tokenizer.from_pretrained('gpt2')
 
-def classify(text,labels):
-    classifer_labels = labels.split(",")
-    #["software", "politics", "love", "movies", "emergency", "advertisment","sports"]
-    response = zero_shot_classifier(text,classifer_labels)
+def generate(starting_text):
+    tkn_ids = gpt2_tkn.encode(starting_text, return_tensors = 'pt')
+    gpt2_tensors = mdl.generate(tkn_ids)
+    response = gpt2_tensors
     return response
 
-txt=grad.Textbox(lines=1, label="English", placeholder="text to be classified")
-labels=grad.Textbox(lines=1, label="Labels", placeholder="comma separated labels")
-out=grad.Textbox(lines=1, label="Classification")
+txt=grad.Textbox(lines=1, label="English", placeholder="English Text here")
+out=grad.Textbox(lines=1, label="Generated Tensors")
 
-grad.Interface(classify, inputs=[txt,labels], outputs=out).launch()
+grad.Interface(generate, inputs=txt, outputs=out).launch()