Fail early if model requires `trust_remote_code`

So we don't have to download the model only to fail during conversion.

cc

@reach-vb


@awni

app.py

@@ -15,6 +15,7 @@ from gradio_huggingfacehub_search import HuggingfaceHubSearch
 from apscheduler.schedulers.background import BackgroundScheduler
 
 from textwrap import dedent
+from transformers import AutoConfig
 
 import mlx_lm
 from mlx_lm import convert
@@ -108,6 +109,9 @@ def process_model(model_id, q_method, oauth_token: gr.OAuthToken | None):
     model_name = model_id.split('/')[-1]
     username = whoami(oauth_token.token)["name"]
     try:
+        # Fail early if the model requires custom code
+        _ = AutoConfig.from_pretrained(model_id)
+
         if q_method == "FP16":
             upload_repo = f"{username}/{model_name}-mlx-fp16"
             with tempfile.TemporaryDirectory(dir="converted") as tmpdir: