Merge pull request #3 from opentensor/mvp-dashboard

opendashboards/assets/inspect.py

@@ -3,7 +3,6 @@ import streamlit as st
 import pandas as pd
 import opendashboards.utils.utils as utils
 
-
 @st.cache_data
 def explode_data(df):
     list_cols = utils.get_list_col_lengths(df)

opendashboards/assets/io.py

@@ -5,13 +5,12 @@ import streamlit as st
 
 import  opendashboards.utils.utils as utils
 
-BASE_DIR = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
 
 @st.cache_data
 def load_runs(project, filters, min_steps=10):
     runs = []
     msg = st.empty()
-    for run in utils.get_runs(project, filters):
+    for run in utils.get_runs(project, filters, api_key=st.secrets['WANDB_API_KEY']):
         step = run.summary.get('_step',0)
         if step < min_steps:
             msg.warning(f'Skipped run `{run.name}` because it contains {step} events (<{min_steps})')
@@ -48,17 +47,19 @@ def load_data(selected_runs, load=True, save=False):
 
     frames = []
     n_events = 0
+    successful = 0
     progress = st.progress(0, 'Loading data')
     info = st.empty()
+    if not os.path.exists('data/'):
+        os.makedirs('data/')
     for i, idx in enumerate(selected_runs.index):
         run = selected_runs.loc[idx]
-        prog_msg = f'Loading data {i/len(selected_runs)*100:.0f}% ({i}/{len(selected_runs)} runs, {n_events} events)'
+        prog_msg = f'Loading data {i/len(selected_runs)*100:.0f}% ({successful}/{len(selected_runs)} runs, {n_events} events)'
 
-        rel_path = os.path.join('data',f'history-{run.id}.csv')
-        file_path = os.path.join(BASE_DIR,rel_path)
+        file_path = os.path.join('data',f'history-{run.id}.csv')
 
         if load and os.path.exists(file_path):
-            progress.progress(i/len(selected_runs),f'{prog_msg}... **reading** `{rel_path}`')
+            progress.progress(i/len(selected_runs),f'{prog_msg}... **reading** `{file_path}`')
             try:
                 df = utils.load_data(file_path)
             except Exception as e:
@@ -70,9 +71,8 @@ def load_data(selected_runs, load=True, save=False):
             try:
                 # Download the history from wandb
                 df = utils.download_data(run.path)
+                # Add metadata to the dataframe
                 df.assign(**run.to_dict())
-                if not os.path.exists('data/'):
-                    os.makedirs(file_path)
 
                 if save and run.state != 'running':
                     df.to_csv(file_path, index=False)
@@ -84,6 +84,7 @@ def load_data(selected_runs, load=True, save=False):
 
         frames.append(df)
         n_events += df.shape[0]
+        successful += 1
 
     progress.empty()
     if not frames:

opendashboards/assets/metric.py

@@ -1,5 +1,3 @@
-import os
-import re
 import time
 import pandas as pd
 import streamlit as st

opendashboards/assets/plot.py

@@ -1,6 +1,6 @@
 
 import streamlit as st
-import utils.plotting as plotting
+import opendashboards.utils.plotting as plotting
 
 # @st.cache_data
 def uid_diversty(df, rm_failed=True):

opendashboards/utils/plotting.py

@@ -251,7 +251,7 @@ def plot_leaderboard(
 
     print(f"Using top {ntop} {group_on} by {agg_col}: \n{rankings}")
     return px.bar(
-        x=rankings,
+        x=rankings.astype(float),
         y=index,
         color=rankings,
         orientation="h",

opendashboards/utils/utils.py

@@ -24,7 +24,7 @@ from pandas.api.types import is_list_like
 from typing import List, Dict, Any, Union
 
 
-def get_runs(project: str = "openvalidators", filters: Dict[str, Any] = None, return_paths: bool = False) -> List:
+def get_runs(project: str = "openvalidators", filters: Dict[str, Any] = None, return_paths: bool = False, api_key: str = None) -> List:
     """Download runs from wandb.
 
     Args:
@@ -35,8 +35,8 @@ def get_runs(project: str = "openvalidators", filters: Dict[str, Any] = None, re
     Returns:
         List[wandb.apis.public.Run]: List of runs or run paths (List[str]).
     """
-    api = wandb.Api()
-    wandb.login()
+    api = wandb.Api(api_key=api_key)
+    wandb.login(anonymous="allow")
 
     runs = api.runs(project, filters=filters)
     if return_paths:
@@ -45,7 +45,7 @@ def get_runs(project: str = "openvalidators", filters: Dict[str, Any] = None, re
         return runs
 
 
-def download_data(run_path: Union[str, List] = None, timeout: float = 600) -> pd.DataFrame:
+def download_data(run_path: Union[str, List] = None, timeout: float = 600, api_key: str = None) -> pd.DataFrame:
     """Download data from wandb.
 
     Args:
@@ -55,8 +55,8 @@ def download_data(run_path: Union[str, List] = None, timeout: float = 600) -> pd
     Returns:
         pd.DataFrame: Dataframe of event log.
     """
-    api = wandb.Api(timeout=timeout)
-    wandb.login()
+    api = wandb.Api(api_key=api_key, timeout=timeout)
+    wandb.login(anonymous="allow")
 
     if isinstance(run_path, str):
         run_path = [run_path]