Additional metadata fields!

app.py

@@ -153,8 +153,10 @@ def main():
     # EVAL RESULTS
     ############################
     st.markdown("### Evaluation results")
-    st.markdown("To edit the metrics, you can either use the YAML editor below, or add new metrics using the handy "
-                "form under it.")
+    st.markdown(
+        "To edit the metrics, you can either use the YAML editor below, or add new metrics using the handy "
+        "form under it."
+    )
     if "results" not in metadata["model-index"][0]:
         metadata["model-index"][0]["results"] = []
 
@@ -166,35 +168,80 @@ def main():
         results_yaml = st_ace(value=results_yaml, language="yaml")
         metadata["model-index"][0]["results"] = try_parse_yaml(results_yaml)
 
+    dataset_path_kwargs = {}
+    dataset_name_kwargs = {}
+    if (
+        len(metadata["model-index"][0]["results"]) > 0
+        and "dataset" in metadata["model-index"][0]["results"]
+    ):
+        if "type" in metadata["model-index"][0]["results"]["dataset"]:
+            dataset_path_kwargs["value"] = metadata["model-index"][0]["results"][
+                "dataset"
+            ]["type"]
+        if "name" in metadata["model-index"][0]["results"]["dataset"]:
+            dataset_name_kwargs["value"] = metadata["model-index"][0]["results"][
+                "dataset"
+            ]["type"]
+
     with st.form(key="eval_form"):
-        dataset_name = st.text_input(
-            label="Full name of the dataset", placeholder="Common Voice 8.0"
-        )
         dataset_path = st.text_input(
-            label="Dataset path / id", placeholder="mozilla-foundation/common_voice_8_0"
+            label="Dataset path / id",
+            placeholder="mozilla-foundation/common_voice_8_0",
+            **dataset_path_kwargs,
+        )
+        dataset_name = st.text_input(
+            label="A pretty name for the dataset. Examples: 'Common Voice 9.0 (French)', 'LibriSpeech (clean)'",
+            placeholder="Common Voice 8.0 (French)",
+            **dataset_name_kwargs,
         )
         dataset_config = st.text_input(
-            label="Dataset config (language). Examples: en, pt-BR, clean",
+            label="Dataset configuration. Examples: clean, other, en, pt-BR",
             placeholder="en",
         )
-        metric_name = st.text_input(label="Metric name", placeholder="Test WER (+LM)")
+        dataset_language = st.text_input(
+            label="Dataset language. Examples: en, pt-BR",
+            value=languages[0],
+            placeholder="en",
+        )
+        dataset_split = st.text_input(
+            label="Dataset split. Examples: test, validation",
+            value="test",
+            placeholder="test",
+        )
         metric2name = {"wer": "Word Error Rate", "cer": "Character Error Rate"}
         metric_type = st.selectbox(
             label="Metric",
             options=["wer", "cer"],
             format_func=lambda key: metric2name[key],
         )
+        metric_name = st.text_input(
+            label="A pretty name for the metric. Example: Test WER (+LM)",
+            placeholder="Test WER",
+            value="Test WER",
+        )
         metric_value = st.text_input(
-            label="Metric value (0.0 - 100.0)",
+            label="Metric value. Use values in range 0.0 to 100.0.",
             placeholder="12.34",
         )
-        try:
-            metric_value = float(metric_value)
-        except ValueError:
-            st.error(f"Couldn't parse `{metric_value}`. Make sure it's a number from 0.0 to 100.0")
+        # try:
+        #    metric_value = float(metric_value)
+        # except ValueError:
+        #    st.error(
+        #        f"Couldn't parse `{metric_value}`. Make sure it's a number from 0.0 to 100.0"
+        #    )
 
-        submitted = st.form_submit_button("Submit")
-        if submitted:
+        submitted = st.form_submit_button("Add metric")
+        if (
+            submitted
+            and dataset_name
+            and dataset_path
+            and dataset_config
+            and dataset_split
+            and dataset_language
+            and metric_name
+            and metric_type
+            and metric_value
+        ):
             metric = {
                 "name": metric_name,
                 "type": metric_type,
@@ -206,7 +253,10 @@ def main():
                 existing_dataset = existing_result["dataset"]
                 if (
                     existing_dataset["type"] == dataset_path
-                    and existing_dataset["args"] == dataset_config
+                    and "config" in existing_dataset
+                    and existing_dataset["config"] == dataset_config
+                    and "split" in existing_dataset
+                    and existing_dataset["split"] == dataset_split
                 ):
                     if "metrics" not in existing_result:
                         existing_result["metrics"] = []
@@ -223,7 +273,9 @@ def main():
                     "dataset": {
                         "name": dataset_name,
                         "type": dataset_path,
-                        "args": dataset_config,
+                        "config": dataset_config,
+                        "split": dataset_split,
+                        "args": {"language": dataset_language},
                     },
                     "metrics": [metric],
                 }
@@ -238,8 +290,14 @@ def main():
                 )
                 results_yaml = st_ace(value=results_yaml, language="yaml")
                 metadata["model-index"][0]["results"] = try_parse_yaml(results_yaml)
-            st.success(f"Added the metric for {dataset_path} - {dataset_config}! "
-                       f"Check the result in the YAML editor above.")
+            st.success(
+                f"Added the metric for {dataset_path} - {dataset_config}! "
+                f"Check the result in the YAML editor above."
+            )
+        elif submitted:
+            st.error(
+                f"Make sure that you've filled the whole form before clicking 'Add metric'!"
+            )
 
     ############################
     # FINAL YAML
@@ -247,8 +305,10 @@ def main():
     st.markdown("## 3. Copy the generated metadata")
     st.markdown(
         "Copy the YAML from below and replace the metadata at the top of your model's README.md here: "
-        f"https://huggingface.co/{model_id}/blob/main/README.md"
+        f"https://huggingface.co/{model_id}/edit/main/README.md"
     )
+    st.markdown("For mor info on the metadata schema please refer to "
+                "https://raw.githubusercontent.com/huggingface/hub-docs/main/modelcard.md")
 
     new_yaml = yaml.dump(metadata, sort_keys=False, line_break="\n")
     st.markdown(f"```yaml\n---\n{new_yaml}---\n```")

requirements.txt

@@ -1,2 +1,3 @@
 streamlit-tags
-streamlit-ace
+streamlit-ace
+streamlit==1.5