changed stuff

.gitignore

@@ -0,0 +1,2 @@
+__pycache__/app.cpython-311.pyc
+flagged/log.csv

app.py

@@ -1,45 +1,65 @@
 import gradio as gr
-
-# Import the pipeline
 from transformers import pipeline
 
-# Define the pipeline
-# Note: This pipeline is hosted on the Hugging Face model hub
-# https://huggingface.co/Helsinki-NLP/opus-mt-en-he
-# You can replace this with any other translation pipeline
-# https://huggingface.co/models?filter=translation
-pipe = pipeline("translation", model="Helsinki-NLP/opus-mt-en-he")
+# Using the latest version of Gradio and Transformers
+# We want to expand the interface to include a reverse translation
+# We want to use the Helsinki-NLP/opus-mt-tc-big-he-en model for the reverse translation
 
-# Define a pipeline for reverse translation
-# Note: This pipeline is hosted on the Hugging Face model hub
-# https://huggingface.co/Helsinki-NLP/opus-mt-he-en
-# You can replace this with any other translation pipeline
-# https://huggingface.co/models?filter=translation
-pipe_reverse = pipeline("translation", model="Helsinki-NLP/opus-mt-he-en")
+# A dropdown menu for selecting the model
+model_names = ["Helsinki-NLP/opus-mt-en-he", "Helsinki-NLP/opus-mt-tc-big-he-en"]
+model_name = gr.inputs.Dropdown(model_names, label="Model")
+# Name the dropdown options
+model_name.choices = ["English to Hebrew", "Hebrew to English"]
 
+# An output text box displaying the translated text and reverse translated text
+translation = gr.outputs.Textbox(label="Translation")
+reverse_translation = gr.outputs.Textbox(label="Reverse Translation")
 
-# Define the function
-def predict(text):
-  # Return the translation
-  return pipe(text)[0]["translation_text"]
+# A function for translating text
+def translate(model_name, text):
+    # Create a pipeline for translating from English to Hebrew
+    pipe = pipeline("translation", model=model_name)
 
-def predict_reverse(text):
     # Return the translation
-    return pipe_reverse(text)[0]["translation_text"]
+    return pipe(text)[0]["translation_text"]
 
-# Define the interface
-iface = gr.Interface(
-  fn=predict, 
-  fn_reverse=predict_reverse,
-  inputs='text',
-  outputs='text',
-    title="English to Hebrew Translator",
-  description="Translate English to Hebrew",
-  examples=[["Hello! My name is Bob."], ["I like to eat apples and banana"]]
-)
+# Create an interface for translating text
+from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
+import torch
 
+tokenizer = AutoTokenizer.from_pretrained("Helsinki-NLP/opus-mt-en-he")
+model = AutoModelForSeq2SeqLM.from_pretrained("Helsinki-NLP/opus-mt-en-he")
 
+def translate(model_name, text):
+    # Create a pipeline for translating from English to Hebrew
+    #Console out the model name
+    print(model_name)
+    if model_name == "English to Hebrew":
+        forward_tokenizer = AutoTokenizer.from_pretrained("Helsinki-NLP/opus-mt-en-he")
+        forward_model = AutoModelForSeq2SeqLM.from_pretrained("Helsinki-NLP/opus-mt-en-he")
+        reverse_tokenizer = AutoTokenizer.from_pretrained("Helsinki-NLP/opus-mt-tc-big-he-en")
+        reverse_model = AutoModelForSeq2SeqLM.from_pretrained("Helsinki-NLP/opus-mt-tc-big-he-en")
+    elif model_name == "Hebrew to English":
+        forward_tokenizer = AutoTokenizer.from_pretrained("Helsinki-NLP/opus-mt-tc-big-he-en")
+        forward_model = AutoModelForSeq2SeqLM.from_pretrained("Helsinki-NLP/opus-mt-tc-big-he-en")
+        reverse_tokenizer = AutoTokenizer.from_pretrained("Helsinki-NLP/opus-mt-en-he")
+        reverse_model = AutoModelForSeq2SeqLM.from_pretrained("Helsinki-NLP/opus-mt-en-he")
+    else:
+        raise ValueError("Invalid model name")  
+    
+    # Forward translation
+    forward_input_ids = forward_tokenizer.encode(text, return_tensors="pt")
+    forward_outputs = forward_model.generate(forward_input_ids)
+    forward_translation = forward_tokenizer.decode(forward_outputs[0], skip_special_tokens=True)
 
-# Launch the interface
-iface.launch()
+    # Reverse translation
+    reverse_input_ids = reverse_tokenizer.encode(forward_translation, return_tensors="pt")
+    reverse_outputs = reverse_model.generate(reverse_input_ids)
+    reverse_translation = reverse_tokenizer.decode(reverse_outputs[0], skip_special_tokens=True)
+    
+    return forward_translation, reverse_translation
 
+iface = gr.Interface(fn=translate, inputs=[model_name, "text"], outputs=[translation, reverse_translation])
+
+# Launch the interface
+iface.launch(share=True)

requirements.txt

@@ -1,3 +1,4 @@
 gradio
-transformers==4.28.0
-torch
+git+https://github.com/huggingface/transformers
+torch
+sentencepiece