adds model repo dataset id to the model card

interface.py

@@ -439,6 +439,7 @@ def start_voxtral_training(
     freeze_audio_tower: bool,
     push_to_hub: bool,
     deploy_demo: bool,
+    dataset_repo_name_value: str = "",
 ) -> str:
     """Start Voxtral training and return collected logs as a string."""
     env = os.environ.copy()
@@ -526,6 +527,13 @@ def start_voxtral_training(
                 "--max-epochs", str(epochs),
                 "--trackio-url", env.get("TRACKIO_URL", "N/A"),
             ]
+            # If user provided a Hub dataset repo id, include it for the model card
+            try:
+                ds_name = (dataset_repo_name_value or "").strip()
+                if ds_name and "/" in ds_name and " " not in ds_name and len(ds_name.split("/")) == 2:
+                    push_args += ["--dataset-name", ds_name]
+            except Exception:
+                pass
             all_logs.append(f"📤 Pushing model to Hugging Face Hub: {full_repo_name}")
             push_code = collect_logs_with_code(run_command_stream(push_args, env))
             if push_code != 0:

templates/model_card.md

@@ -5,6 +5,8 @@ tags:
 - asr
 - speech-to-text
 - fine-tuning
+- tonic
+
 pipeline_tag: automatic-speech-recognition
 base_model: {{base_model}}
 {{#if has_hub_dataset_id}}
@@ -38,6 +40,15 @@ max_seq_length: {{max_seq_length}}
 {{#if hardware_info}}
 hardware: "{{hardware_info}}"
 {{/if}}
+language:
+- hi
+- en
+- fr
+- de
+- it
+- pt
+- nl
+library_name: peft
 ---
 
 # {{model_name}}