changing visualise function to host results of zeno client instead of…

… localserver

zeno_build/reporting/visualize.py

@@ -4,10 +4,14 @@
 from typing import Any
 
 import pandas as pd
-from zeno import ModelReturn, ZenoParameters, model, zeno
-
+from zeno import ModelReturn, ZenoParameters, model
+from zeno.classes.base import ZenoColumnType
 from zeno_build.experiments.experiment_run import ExperimentRun
+from zeno.backend import ZenoBackend
+from zeno_client import ZenoClient, ZenoMetric
 
+import os
+import time
 
 def visualize(
     df: pd.DataFrame,
@@ -17,6 +21,7 @@ def visualize(
     data_column: str,
     functions: list[Callable],
     zeno_config: dict = {},
+    project_name:str = ""
 ) -> None:
     """Run Zeno to visualize the results of a parameter search run.
 
@@ -33,6 +38,16 @@ def visualize(
         raise ValueError("Length of data and labels must be equal.")
     if data_column not in df.columns:
         raise ValueError(f"Data column {data_column} not in DataFrame.")
+    if "ZENO_CLIENT_KEY" not in os.environ:
+        raise ValueError(
+            "ZENO_CLIENT_KEY environment variable must be set to visualize results."
+        )
+
+    zeno_client = ZenoClient(os.environ["ZENO_CLIENT_KEY"])
+    if project_name == "":
+        project_name = time.strftime("%Y-%m-%d %H-%M-%S")
+        print(f"Creating project {project_name}")
+
     model_results: dict[str, ExperimentRun] = {x.name: x for x in results}
 
     @model
@@ -61,4 +76,28 @@ def pred(df, ops):
     )
     config = config.copy(update=zeno_config)
 
-    zeno(config)
+    zeno_instance = ZenoBackend(config)
+
+    # wait for the backend to complete processing
+    zeno_thread = zeno_instance.start_processing()
+    zeno_thread.join()
+
+    project = zeno_client.create_project(
+                    name=project_name,
+                    view=view,
+                    metrics=[ZenoMetric(id=idx,name=metric, type='mean', columns=[metric[4:]]) for idx,metric in enumerate(zeno_instance.metric_functions.keys())])
+
+    # upload the dataset
+    dataset_columns = ["index", data_column, "label"]
+    dataset_df = zeno_instance.df[dataset_columns]
+    for predistill_function in zeno_instance.predistill_functions.keys():
+        dataset_df[predistill_function] = zeno_instance.df[ZenoColumnType.PREDISTILL+predistill_function]
+    project.upload_dataset(dataset_df, id_column="index", data_column=data_column, label_column="label")
+
+    # upload the model results
+    for model_name in model_results:
+        df_model = pd.DataFrame({"index": dataset_df["index"]})
+        df_model["output"] = zeno_instance.df[ZenoColumnType.OUTPUT+"output" + model_name]
+        for postdistill_function in zeno_instance.postdistill_functions.keys():
+            df_model[postdistill_function] = zeno_instance.df[ZenoColumnType.POSTDISTILL + postdistill_function + model_name]
+        project.upload_system(df_model, name=model_name, id_column="index", output_column="output")