1. fed_stats is broken due to the simulator changes, change the works…

…pace directory in the notebooks and examples (#2691) 2. for df_stats job, notice the new datasets has empty line at end, causing pandas to mistaken numerical data type to string type. We change the prepare_data.py to remove empty line 3. update df_stats notebook to add panda data table format. Co-authored-by: Sean Yang <[email protected]>
NVIDIA · Jul 12, 2024 · dd248ca · dd248ca
1 parent f7ca641
commit dd248ca
Show file tree

Hide file tree

Showing 7 changed files with 95 additions and 205 deletions.
diff --git a/examples/advanced/federated-statistics/df_stats.ipynb b/examples/advanced/federated-statistics/df_stats.ipynb
@@ -13,12 +13,7 @@
     "\n",
     "## Set Up NVFLARE\n",
     "\n",
-    "Follow [Getting Started](https://nvflare.readthedocs.io/en/main/getting_started.html) to set up a virtual environment and install NVFLARE.\n",
-    "\n",
-    "You can also follow this [notebook](../../nvflare_setup.ipynb) to get set up.\n",
-    "\n",
-    "\n",
-    "> make sure you have installed nvflare from the terminal first"
+    "Follow [Getting Started](https://nvflare.readthedocs.io/en/main/getting_started.html) to set up a virtual environment and install NVFLARE.\n"
    ]
   },
   {
@@ -77,6 +72,63 @@
     "prepare_data(data_root_dir = \"/tmp/nvflare/df_stats/data\")"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "c5444d8f-4938-4759-bd43-831013043c23",
+   "metadata": {},
+   "source": [
+    "#### Let's take a look at the data"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "a1cf37d0-7555-4818-9963-ca7342161a4d",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import pandas as pd\n",
+    "data_path =\"/tmp/nvflare/df_stats/data/site-1/data.csv\"\n",
+    "data_features = [\n",
+    "            \"Age\",\n",
+    "            \"Workclass\",\n",
+    "            \"fnlwgt\",\n",
+    "            \"Education\",\n",
+    "            \"Education-Num\",\n",
+    "            \"Marital Status\",\n",
+    "            \"Occupation\",\n",
+    "            \"Relationship\",\n",
+    "            \"Race\",\n",
+    "            \"Sex\",\n",
+    "            \"Capital Gain\",\n",
+    "            \"Capital Loss\",\n",
+    "            \"Hours per week\",\n",
+    "            \"Country\",\n",
+    "            \"Target\",\n",
+    "        ]\n",
+    "\n",
+    "        # the original dataset has no header,\n",
+    "        # we will use the adult.train dataset for site-1, the adult.test dataset for site-2\n",
+    "        # the adult.test dataset has incorrect formatted row at 1st line, we will skip it.\n",
+    "skip_rows = {\n",
+    "            \"site-1\": [],\n",
+    "            \"site-2\": [0],\n",
+    "        }\n",
+    "\n",
+    "df= pd.read_csv(data_path, names=data_features, sep=r\"\\s*,\\s*\", skiprows=skip_rows, engine=\"python\", na_values=\"?\")\n",
+    "df\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "81f6d572-7dc0-4cec-8382-f25555f52af9",
+   "metadata": {
+    "jp-MarkdownHeadingCollapsed": true
+   },
+   "source": [
+    "> Note **We will only calculate the statistics of numerical features, categorical features will be skipped**"
+   ]
+  },
   {
    "cell_type": "markdown",
    "id": "f00de5e4-4360-4fc5-a819-4eb156e56341",

diff --git a/examples/advanced/federated-statistics/df_stats/README.md b/examples/advanced/federated-statistics/df_stats/README.md
@@ -50,12 +50,12 @@ With FL simulator, we can just run the example with CLI command
 
 ```
 cd NVFlare/examples/advanced/federated-statistics
-nvflare simulator df_stats/jobs/df_stats -w /tmp/nvflare/df_stats -n 2 -t 2
+nvflare simulator df_stats/jobs/df_stats -w /tmp/nvflare/workspace/df_stats -n 2 -t 2
 ```
 
 The results are stored in workspace "/tmp/nvflare"
 ```
-/tmp/nvflare/df_stats/simulate_job/statistics/adults_stats.json
+/tmp/nvflare/workspace/df_stats/simulate_job/statistics/adults_stats.json
 ```
 
 ## 3. Visualization
@@ -66,7 +66,7 @@ The results are stored in workspace "/tmp/nvflare"
    assuming NVFLARE_HOME env variable point to the GitHub project location (NVFlare) which contains current example. 
 
 ```bash
-    cp /tmp/nvflare/df_stats/simulate_job/advanced/statistics/adults_stats.json $NVFLARE_HOME/examples/advanced/federated-statistics/df_stats/demo/.
+    cp /tmp/nvflare/workspace/df_stats/simulate_job/advanced/statistics/adults_stats.json $NVFLARE_HOME/examples/advanced/federated-statistics/df_stats/demo/.
 
     cd $NVFLARE_HOME/examples/advanced/federated-statistics/df_stats/demo
 

diff --git a/examples/advanced/federated-statistics/df_stats/jobs/df_stats/app/custom/df_statistics.py b/examples/advanced/federated-statistics/df_stats/jobs/df_stats/app/custom/df_statistics.py
@@ -46,6 +46,10 @@ def __init__(self, data_path):
             "Country",
             "Target",
         ]
+
+        # the original dataset has no header,
+        # we will use the adult.train dataset for site-1, the adult.test dataset for site-2
+        # the adult.test dataset has incorrect formatted row at 1st line, we will skip it.
         self.skip_rows = {
             "site-1": [],
             "site-2": [0],

diff --git a/examples/advanced/federated-statistics/df_stats/utils/prepare_data.py b/examples/advanced/federated-statistics/df_stats/utils/prepare_data.py
@@ -67,7 +67,10 @@ def prepare_data(data_root_dir: str):
             writer = csv.writer(f)
             r = requests.get(url, allow_redirects=True)
             for line in r.iter_lines():
-                writer.writerow(line.decode("utf-8").split(","))
+                if line:
+                    writer.writerow(line.decode("utf-8").split(","))
+                else:
+                    print("skip empty line\n")
     print("\ndone with prepare data")
 
 

diff --git a/examples/advanced/federated-statistics/image_stats.ipynb b/examples/advanced/federated-statistics/image_stats.ipynb
@@ -11,11 +11,7 @@
     "\n",
     "## Setup NVFLARE\n",
     "\n",
-    "Follow [Getting Started](https://nvflare.readthedocs.io/en/main/getting_started.html) to set up a virtual environment and install NVFLARE.\n",
-    "\n",
-    "You can also follow this [notebook](../../nvflare_setup.ipynb) to get set up.\n",
-    "\n",
-    "> Make sure you have installed nvflare from **terminal** \n"
+    "Follow [Getting Started](https://nvflare.readthedocs.io/en/main/getting_started.html) to set up a virtual environment and install NVFLARE.\n"
    ]
   },
   {
@@ -147,7 +143,7 @@
    "outputs": [],
    "source": [
     "from nvflare.private.fed.app.simulator.simulator_runner import SimulatorRunner\n",
-    "runner = SimulatorRunner(job_folder=\"image_stats/jobs/image_stats\", workspace=\"/tmp/nvflare/image_stats\", n_clients = 4, threads=4)\n",
+    "runner = SimulatorRunner(job_folder=\"image_stats/jobs/image_stats\", workspace=\"/tmp/nvflare/workspace/image_stats\", n_clients = 4, threads=4)\n",
     "runner.run()"
    ]
   },
@@ -163,17 +159,14 @@
     "From a **terminal** one can also the following equivallent CLI\n",
     "\n",
     "```\n",
-    "nvflare simulator image_stats/jobs/image_stats -w /tmp/nvflare/image_stats -n 4 -t 4\n",
+    "nvflare simulator image_stats/jobs/image_stats -w /tmp/nvflare/workspace/image_stats -n 4 -t 4\n",
     "\n",
     "```\n",
     "\n",
     "assuming the nvflare is installed from a **terminal**. doing pip install from the notebook cell directory with bash command (! or %%bash) may or may not work depending on which python runtime kernel selected. Also %pip install or %pip install from notebook cell doesn't register the console_scripts in the PATH.   \n",
     "\n",
     "\n",
-    "## Examine the result\n",
-    "\n",
-    "\n",
-    "\n"
+    "## Examine the result\n"
    ]
   },
   {
@@ -194,7 +187,7 @@
    },
    "outputs": [],
    "source": [
-    "! ls -al /tmp/nvflare/image_stats/simulate_job/statistics/image_statistics.json"
+    "! ls -al /tmp/nvflare/workspace/image_stats/simulate_job/statistics/image_statistics.json"
    ]
   },
   {
@@ -217,7 +210,7 @@
    },
    "outputs": [],
    "source": [
-    "! cp /tmp/nvflare/image_stats/simulate_job/statistics/image_statistics.json image_stats/demo/."
+    "! cp /tmp/nvflare/workspace/image_stats/simulate_job/statistics/image_statistics.json image_stats/demo/."
    ]
   },
   {
@@ -266,7 +259,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.8.17"
+   "version": "3.10.2"
   }
  },
  "nbformat": 4,

diff --git a/examples/advanced/federated-statistics/image_stats/demo/visualization.ipynb b/examples/advanced/federated-statistics/image_stats/demo/visualization.ipynb
diff --git a/nvflare/app_common/workflows/statistics_controller.py b/nvflare/app_common/workflows/statistics_controller.py
@@ -309,6 +309,7 @@ def results_cb(self, client_task: ClientTask, fl_ctx: FLContext):
 
         result = client_task.result
         rc = result.get_return_code()
+        ds_features = None
         if rc == ReturnCode.OK:
             self.log_info(fl_ctx, f"Received result entries from client:{client_name}, " f"for task {task_name}")
             dxo = from_shareable(result)