initial commit

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+marian_finetuned_kde4_enfr.pth filter=lfs diff=lfs merge=lfs -text

app.py

@@ -0,0 +1,47 @@
+import torch
+import gradio as gr
+
+from model import create_marian_enfr
+
+# Setup model and tokenizer
+model, tokenizer = create_marian_enfr()
+
+# Load state dict from model
+model.load_state_dict(
+    torch.load(
+        f="marian_finetuned_kde4_enfr.pth",
+        map_location=torch.device("cpu")
+    ))
+
+# Predict function
+def predict(text:str):
+
+  # Tokenize inputs and get model outputs
+  input = tokenizer(text, 
+                    max_length=128,
+                    truncation=True,
+                    return_tensors="pt")
+  output_tokens = model.generate(**input)
+  output_text = tokenizer.batch_decode(output_tokens, 
+                                       skip_special_tokens=True)
+  
+  return output_text
+
+# Create examples list
+examples_list = ['What a beautiful day',
+                 'I love music']
+
+# Create gradio app
+title = "English to French translator"
+description = "Marian model finetuned for english to french translation on the kde4 dataset."
+
+demo = gr.Interface(fn=predict,
+                    inputs=gr.inputs.Textbox(label="Input",
+                                             placeholder="Enter sentence here..."),
+                    outputs="text",
+                    examples=examples_list,
+                    title=title,
+                    description=description)
+
+# Launch gradio
+demo.launch()

marian_finetuned_kde4_enfr.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ca3fb97610cb79aa65b0f377810a298556309c239679cfac8222deca6bff682f
+size 300875111

model.py

@@ -0,0 +1,11 @@
+from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
+
+def create_marian_enfr():
+  """
+  Initializes model and tokenizer.
+  """
+  checkpoint = 'Helsinki-NLP/opus-mt-en-fr'
+  tokenizer = AutoTokenizer.from_pretrained(checkpoint, return_tensors='pt')
+  model = AutoModelForSeq2SeqLM.from_pretrained(checkpoint)
+  
+  return model, tokenizer

requirements.txt

@@ -0,0 +1,3 @@
+torch==1.12.0
+gradio==3.1.4
+transformers==4.33.1