Update app.py

app.py

@@ -5,9 +5,9 @@ from clarifai_grpc.grpc.api.status import status_code_pb2
 import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer
 
-
-model = AutoModelForCausalLM.from_pretrained("microsoft/phi-2", torch_dtype="auto", trust_remote_code=True)
-tokenizer = AutoTokenizer.from_pretraine
+torch.set_default_device("cpu")
+model = AutoModelForCausalLM.from_pretrained("microsoft/phi-2", torch_dtype="auto", trust_remote_code=True).to(device)
+tokenizer = AutoTokenizer.from_pretrained("microsoft/phi-2", trust_remote_code=True)
 
 # GPT-4 credentials
 PAT_GPT4 = "3ca5bd8b0f2244eb8d0e4b2838fc3cf1"
@@ -37,86 +37,80 @@ APP_ID_NEWSGUARDIAN = "your_app_id"
 MODEL_ID_NEWSGUARDIAN = "your_model_id"
 MODEL_VERSION_ID_NEWSGUARDIAN = "your_model_version_id"
 
+#
 
 # Set up gRPC channel for NewsGuardian model
 channel_tts = ClarifaiChannel.get_grpc_channel()
 stub_tts = service_pb2_grpc.V2Stub(channel_tts)
 metadata_tts = (('authorization', 'Key ' + PAT_TTS),)
-userDataObject_tts = resources_pb2.UserAppIDSet(user_id=USER_ID_TTS, app_id=APP_ID_TTS)
+userDataObject_tts = resources_pb2.UserAppIDSet(user_id=USER_ID_TTS, app_id=APP_ID_TTS,)
 
 # Streamlit app
-st.title("NewsGuardian and phi-2 Text Generation")
-
-# Function to generate text using the "microsoft/phi-2" model
-def generate_phi2_text(input_text):
-    inputs = tokenizer_phi2(input_text, return_tensors="pt", return_attention_mask=False)
-    outputs = model_phi2.generate(**inputs, max_length=200)
-    generated_text = tokenizer_phi2.batch_decode(outputs)[0]
-    return generated_text
-
-# User input for phi-2 model
-raw_text_phi2 = st.text_area("Enter text for phi-2 model")
-
-# Button to generate result using "microsoft/phi-2" model
-if st.button("Generate text with phi-2 model"):
-    if raw_text_phi2:
-        generated_text_phi2 = generate_phi2_text(raw_text_phi2)
-        st.text("Generated text with phi-2 model")
-        st.text(generated_text_phi2)
-    else:
-        st.warning("Please enter text for phi-2 model")
-
-# User input for selecting the model
-model_type = st.selectbox("Select Model", ["NewsGuardian model", "DALL-E", "phi-2"])
-raw_text_news_guardian = st.text_area("This news is real or fake?")
-image_upload_news_guardian = st.file_uploader("Upload Image", type=["jpg", "jpeg", "png"])
-
-# Button to generate result for the selected model
-if st.button("Generate Result"):
+st.title("NewsGuardian")
+
+
+# Inserting logo
+st.image("https://encrypted-tbn0.gstatic.com/images?q=tbn:ANd9GcTdA-MJ_SUCRgLs1prqudpMdaX4x-x10Zqlwp7cpzXWCMM9xjBAJYWdJsDlLoHBqNpj8qs&usqp=CAU")
+# Function to get gRPC channel for NewsGuardian model
+def get_tts_channel():
+    channel_tts = ClarifaiChannel.get_grpc_channel()
+    return channel_tts, channel_tts.metadata
+
+
+
+# User input
+model_type = st.selectbox("Select Model", ["NewsGuardian model","NewsGuardian model"])
+raw_text = st.text_area("This news is real or fake?")
+image_upload = st.file_uploader("Upload Image", type=["jpg", "jpeg", "png"])
+
+# Button to generate result
+if st.button("NewsGuardian News Result"):
     if model_type == "NewsGuardian model":
         # Set up gRPC channel for NewsGuardian model
-        channel_news_guardian = ClarifaiChannel.get_grpc_channel()
-        stub_news_guardian = service_pb2_grpc.V2Stub(channel_news_guardian)
-        metadata_news_guardian = (('authorization', 'Key ' + PAT_NEWSGUARDIAN),)
-        userDataObject_news_guardian = resources_pb2.UserAppIDSet(user_id=USER_ID_NEWSGUARDIAN, app_id=APP_ID_NEWSGUARDIAN)
+        channel_gpt4 = ClarifaiChannel.get_grpc_channel()
+        stub_gpt4 = service_pb2_grpc.V2Stub(channel_gpt4)
+        metadata_gpt4 = (('authorization', 'Key ' + PAT_GPT4),)
+        userDataObject_gpt4 = resources_pb2.UserAppIDSet(user_id=USER_ID_GPT4, app_id=APP_ID_GPT4)
 
         # Prepare the request for NewsGuardian model
-        input_data_news_guardian = resources_pb2.Data()
+        input_data_gpt4 = resources_pb2.Data()
 
-        if raw_text_news_guardian:
-            input_data_news_guardian.text.raw = raw_text_news_guardian
+        if raw_text:
+            input_data_gpt4.text.raw = raw_text
 
-        if image_upload_news_guardian is not None:
-            image_bytes_news_guardian = image_upload_news_guardian.read()
-            input_data_news_guardian.image.base64 = image_bytes_news_guardian
+        if image_upload is not None:
+            image_bytes_gpt4 = image_upload.read()
+            input_data_gpt4.image.base64 = image_bytes_gpt4
 
-        post_model_outputs_response_news_guardian = stub_news_guardian.PostModelOutputs(
+        post_model_outputs_response_gpt4 = stub_gpt4.PostModelOutputs(
             service_pb2.PostModelOutputsRequest(
-                user_app_id=userDataObject_news_guardian,
-                model_id=MODEL_ID_NEWSGUARDIAN,
-                version_id=MODEL_VERSION_ID_NEWSGUARDIAN,
-                inputs=[resources_pb2.Input(data=input_data_news_guardian)]
+                user_app_id=userDataObject_gpt4,
+                model_id=MODEL_ID_GPT4,
+                version_id=MODEL_VERSION_ID_GPT4,
+                inputs=[resources_pb2.Input(data=input_data_gpt4)]
             ),
-            metadata=metadata_news_guardian  # Use metadata directly in the gRPC request
+            metadata=metadata_gpt4  # Use metadata directly in the gRPC request
         )
 
         # Check if the request was successful for NewsGuardian model
-        if post_model_outputs_response_news_guardian.status.code != status_code_pb2.SUCCESS:
-            st.error(f"NewsGuardian model API request failed: {post_model_outputs_response_news_guardian.status.description}")
+        if post_model_outputs_response_gpt4.status.code != status_code_pb2.SUCCESS:
+            st.error(f"NewsGuardian model API request failed: {post_model_outputs_response_gpt4.status.description}")
         else:
             # Get the output for NewsGuardian model
-            output_news_guardian = post_model_outputs_response_news_guardian.outputs[0].data
+            output_gpt4 = post_model_outputs_response_gpt4.outputs[0].data
 
             # Display the result for NewsGuardian model
-            if output_news_guardian.HasField("image"):
-                st.image(output_news_guardian.image.base64, caption='Generated Image (NewsGuardian model)', use_column_width=True)
-            elif output_news_guardian.HasField("text"):
+            if output_gpt4.HasField("image"):
+                st.image(output_gpt4.image.base64, caption='Generated Image (NewsGuardian model)', use_column_width=True)
+            elif output_gpt4.HasField("text"):
                 # Display the text result
-                st.text(output_news_guardian.text.raw)
+                st.text(output_gpt4.text.raw)
 
                 # Convert text to speech and play the audio
+                stub_tts = service_pb2_grpc.V2Stub(channel_gpt4)  # Use the same channel for TTS
+
                 tts_input_data = resources_pb2.Data()
-                tts_input_data.text.raw = output_news_guardian.text.raw
+                tts_input_data.text.raw = output_gpt4.text.raw
 
                 tts_response = stub_tts.PostModelOutputs(
                     service_pb2.PostModelOutputsRequest(
@@ -125,7 +119,7 @@ if st.button("Generate Result"):
                         version_id=MODEL_VERSION_ID_TTS,
                         inputs=[resources_pb2.Input(data=tts_input_data)]
                     ),
-                    metadata=metadata_tts  # Use the same metadata for TTS
+                    metadata=metadata_gpt4  # Use the same metadata for TTS
                 )
 
                 # Check if the TTS request was successful
@@ -133,7 +127,7 @@ if st.button("Generate Result"):
                     tts_output = tts_response.outputs[0].data
                     st.audio(tts_output.audio.base64, format='audio/wav')
                 else:
-                    st.error(f"TTS API request failed: {tts_response.status.description}")
+                    st.error(f"NewsGuardian model API request failed: {tts_response.status.description}")
 
     elif model_type == "DALL-E":
         # Set up gRPC channel for DALL-E
@@ -145,8 +139,8 @@ if st.button("Generate Result"):
         # Prepare the request for DALL-E
         input_data_dalle = resources_pb2.Data()
 
-        if raw_text_news_guardian:
-            input_data_dalle.text.raw = raw_text_news_guardian
+        if raw_text:
+            input_data_dalle.text.raw = raw_text
 
         post_model_outputs_response_dalle = stub_dalle.PostModelOutputs(
             service_pb2.PostModelOutputsRequest(
@@ -171,6 +165,44 @@ if st.button("Generate Result"):
             elif output_dalle.HasField("text"):
                 st.text(output_dalle.text.raw)
 
+    elif model_type == "NewsGuardian model":
+        # Set up gRPC channel for NewsGuardian model
+        channel_tts = ClarifaiChannel.get_grpc_channel()
+        stub_tts = service_pb2_grpc.V2Stub(channel_tts)
+        metadata_tts = (('authorization', 'Key ' + PAT_TTS),)
+        userDataObject_tts = resources_pb2.UserAppIDSet(user_id=USER_ID_TTS, app_id=APP_ID_TTS)
+
+        # Prepare the request for NewsGuardian model
+        input_data_tts = resources_pb2.Data()
+
+        if raw_text:
+            input_data_tts.text.raw = raw_text
+
+        post_model_outputs_response_tts = stub_tts.PostModelOutputs(
+            service_pb2.PostModelOutputsRequest(
+                user_app_id=userDataObject_tts,
+                model_id=MODEL_ID_TTS,
+                version_id=MODEL_VERSION_ID_TTS,
+                inputs=[resources_pb2.Input(data=input_data_tts)]
+            ),
+            metadata=metadata_tts
+        )
+
+        # Check if the request was successful for NewsGuardian model
+        if post_model_outputs_response_tts.status.code != status_code_pb2.SUCCESS:
+            st.error(f"NewsGuardian model API request failed: {post_model_outputs_response_tts.status.description}")
+        else:
+            # Get the output for NewsGuardian model
+            output_tts = post_model_outputs_response_tts.outputs[0].data
+
+            # Display the result for NewsGuardian model
+            if output_tts.HasField("text"):
+                st.text(output_tts.text.raw)
+
+            if output_tts.HasField("audio"):
+                st.audio(output_tts.audio.base64, format='audio/wav')
+
+
 # Add the beautiful social media icon section
 st.markdown("""
   <div align="center">
@@ -193,4 +225,4 @@ st.markdown("""
         <img src="https://user-images.githubusercontent.com/34125851/226601355-ffe0b597-9840-4e10-bbef-43d6c74b5a9e.png" width="2%" alt="" /></a>      
   </div>
   <hr>
-""", unsafe_allow_html=True)
+""", unsafe_allow_html=True)