Merge pull request #5 from NSLS-II-PDF/offline-testing

Offline testing

pdf_agents/agents.py

@@ -14,7 +14,7 @@
 from bluesky_queueserver_api.http import REManagerAPI
 from numpy.typing import ArrayLike
 
-from .utils import OfflineKafka
+from .utils import OfflineKafka, OfflineRedis
 
 logger = getLogger(__name__)
 
@@ -31,13 +31,10 @@ def __init__(
         offline=False,
         **kwargs,
     ):
-        self._rkvs = redis.Redis(host="info.pdf.nsls2.bnl.gov", port=6379, db=0)  # redis key value store
-        self._motor_name = motor_name
-        self._motor_resolution = motor_resolution
-        self._data_key = data_key
-        self._roi_key = roi_key
-        self._roi = roi
-        # Attributes pulled in from Redis
+        if offline:
+            self._rkvs = OfflineRedis()
+        else:
+            self._rkvs = redis.Redis(host="info.pdf.nsls2.bnl.gov", port=6379, db=0)  # redis key value store
         self._exposure = float(self._rkvs.get("PDF:desired_exposure_time").decode("utf-8"))
         self._sample_number = int(self._rkvs.get("PDF:xpdacq:sample_number").decode("utf-8"))
         self._background = np.array(
@@ -46,6 +43,14 @@ def __init__(
                 ast.literal_eval(self._rkvs.get("PDF:bgd:y").decode("utf-8")),
             ]
         )
+
+        self._motor_name = motor_name
+        self._motor_resolution = motor_resolution
+        self._data_key = data_key
+        self._roi_key = roi_key
+        self._roi = roi
+        # Attributes pulled in from Redis
+
         if offline:
             _default_kwargs = self.get_offline_objects()
         else:
@@ -245,7 +250,7 @@ def get_offline_objects() -> dict:
         offline_kafka = OfflineKafka()
         try:
             node = tiled.client.from_profile(f"{beamline_tla}_bluesky_sandbox")
-        except tiled.profile.ProfileNotFound:
+        except tiled.profiles.ProfileNotFound:
             node = None
 
         return dict(

pdf_agents/offline_testing/ask_agent_from_df.ipynb

@@ -0,0 +1,150 @@
+{
+ "cells": [
+  {
+   "attachments": {},
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# Example agent invesitgation\n",
+    "This mimics the scripts in `pdf_agents/startup_scripts` with minimal and perhaps broken functionality due to a lack of redis, tiled, kafka, etc. \n",
+    "\n",
+    "It doesn't look at what an **exact** agent **did**, but what a type of agent **might do**.\n",
+    "\n",
+    "Recreations of other agents could be attempted by following a similar approach. As such multiple agents are shown in the cells below. "
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import sys\n",
+    "\n",
+    "sys.path.append(\"../../\")\n",
+    "import pandas as pd\n",
+    "import numpy as np\n",
+    "from pdf_agents.scientific_value import ScientificValueAgentBase\n",
+    "from pdf_agents.sklearn import ActiveKmeansAgent"
+   ]
+  },
+  {
+   "attachments": {},
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Choose your fighter"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "### GP driven Baysian optimization of a scientific value function ###\n",
+    "# agent = ScientificValueAgentBase(bounds=[-12.0, 46.0], ask_on_tell=False, report_on_tell=False, offline=True)\n",
+    "### GP driven Baysian optimization of a scientific value function ###\n",
+    "\n",
+    "### KMeans driven selection of interesting points for XAFS beamline. ###\n",
+    "## This skips over the MonarchSubject businsiness, and uses the same approach for driving a single beamline\n",
+    "agent = ActiveKmeansAgent(\n",
+    "    k_clusters=4, bounds=[-12.0, 46.0], ask_on_tell=False, report_on_tell=False, offline=True\n",
+    ")\n",
+    "### KMeans driven selection of interesting points for XAFS beamline. ###"
+   ]
+  },
+  {
+   "attachments": {},
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Load and tell the agent about data"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "df = pd.read_pickle(\"~/Downloads/sim_pdf_df_020pts.pckl\")\n",
+    "independent_vars = df.columns.to_numpy()\n",
+    "observables = df.to_numpy().T"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "agent.tell_many(independent_vars, observables)\n",
+    "for x in independent_vars:\n",
+    "    # This is done because the tell is happening without UIDs.\n",
+    "    agent.tell_cache.append(x)"
+   ]
+  },
+  {
+   "attachments": {},
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Ask the agent what it would do next\n",
+    "\n",
+    "- Kmeans agents can suggest multiple points\n",
+    "- The GP Bayesopt agents are designed to suggest one at a time. \n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# docs, suggestions = agent.ask(1) # The argument here is batch size/number of suggestions\n",
+    "docs, suggestions = agent.ask(4)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "docs"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "suggestions"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "venv",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.9.12"
+  },
+  "orig_nbformat": 4
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}

pdf_agents/utils.py

@@ -1,4 +1,5 @@
 import numpy as np
+from pydantic import parse_obj_as
 
 
 class OfflineKafka:
@@ -18,6 +19,14 @@ def stop(self, *args, **kwargs):
         pass
 
 
+class OfflineRedis:
+    def get(self, *args, **kwargs):
+        return b"0"
+
+    def set(self, *args, **kwargs):
+        return b"0"
+
+
 def discretize(value: np.typing.ArrayLike, resolution: np.typing.ArrayLike):
     return np.floor(value / resolution)