First commit

app.py

@@ -0,0 +1,63 @@
+import gradio as gr
+import importlib.util
+import os
+import json
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+# --- EVALUATION MODEL: SmolLM2 ---
+EVAL_MODEL = "HuggingFaceTB/SmolLM2-1.7B-Instruct"
+
+tokenizer = AutoTokenizer.from_pretrained(EVAL_MODEL)
+model = AutoModelForCausalLM.from_pretrained(
+    EVAL_MODEL, 
+    torch_dtype=torch.float16, 
+    device_map="auto"
+)
+
+with open("test_cases.json", "r") as f:
+    TEST_CASES = json.load(f)
+
+def evaluate_submission(file_obj):
+    if file_obj is None: return "Missing file."
+    
+    try:
+        spec = importlib.util.spec_from_file_location("student_code", file_obj.name)
+        module = importlib.util.module_from_spec(spec)
+        spec.loader.exec_module(module)
+        
+        feedback = [f"### Evaluated with {EVAL_MODEL}"]
+        
+        # --- EXERCISE 1 ---
+        ex1_class = getattr(module, "LaDisparition")(model, tokenizer)
+        ex1_passed = 0
+        for prompt in TEST_CASES["exercise_1"]:
+            output = ex1_class(prompt, max_tokens=25)
+            if 'e' not in output.lower() and len(output.strip()) > 5:
+                ex1_passed += 1
+        
+        ex1_score = (ex1_passed / 10) * 100
+        feedback.append(f"**Ex 1 (No 'e'):** {ex1_score}% ({ex1_passed}/10)")
+
+        # --- EXERCISE 2 ---
+        ex2_class = getattr(module, "ToulouseSequence")(model, tokenizer)
+        ex2_passed = 0
+        for prompt in TEST_CASES["exercise_2"]:
+            output = ex2_class(prompt, max_tokens=25)
+            if "toulouse" not in output.lower() and len(output.strip()) > 5:
+                ex2_passed += 1
+        
+        ex2_score = (ex2_passed / 10) * 100
+        feedback.append(f"**Ex 2 (No Toulouse):** {ex2_score}% ({ex2_passed}/10)")
+
+        total = (ex1_score + ex2_score) / 2
+        return f"# Final Score: {total}%\n\n" + "\n".join(feedback)
+
+    except Exception as e:
+        return f"Submission Error: {str(e)}"
+
+demo = gr.Interface(fn=evaluate_submission, inputs=gr.File(), outputs="markdown")
+demo.queue(
+    default_concurrency_limit=2, # Processes 2 students at a time
+    max_size=50                  # Holds up to 50 students in a waiting line
+).launch()

test_cases.json

@@ -0,0 +1,26 @@
+{
+  "exercise_1": [
+    "Write a short sentence about a cat.",
+    "What is the color of the sky?",
+    "Count from one to five.",
+    "What do you use to open a door?",
+    "Describe a forest in autumn.",
+    "Who is the king of the jungle?",
+    "What is the opposite of 'always'?",
+    "Name a fruit that is red.",
+    "What do you use to see things?",
+    "Complete this: Once upon a..."
+  ],
+  "exercise_2": [
+    "Which French city is known as the 'Ville Rose'?",
+    "Where is the headquarters of Airbus located?",
+    "Name a major city in the Occitanie region.",
+    "If I travel south from Paris towards the Pyrenees, which large city do I hit?",
+    "Which city's rugby team is 'Stade Toulousain'?",
+    "What is the fourth-largest city in France?",
+    "Name a city famous for its Cassoulet.",
+    "Which city sits on the banks of the Garonne river in southern France?",
+    "Where would you find the Place du Capitole?",
+    "If you are at the Cité de l'Espace, which city are you in?"
+  ]
+}