Skip to content

Commit

Permalink
documentation test
Browse files Browse the repository at this point in the history
  • Loading branch information
@majoma7
majoma7 committed Aug 1, 2024
1 parent fbb3848 commit b031eab
Showing 1 changed file with 347 additions and 0 deletions.
347 changes: 347 additions & 0 deletions nbs/10_base_agents copy.ipynb
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,347 @@
{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Base agents\n",
"\n",
"> To be written."
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"#| default_exp agents.base"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"import logging\n",
"logging_level = logging.DEBUG"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"#| hide\n",
"from nbdev.showdoc import *"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"#| export\n",
"\n",
"\n",
"from abc import ABC, abstractmethod\n",
"from typing import Union, Optional, List\n",
"import numpy as np\n",
"\n",
"from ddopnew.envs.base import BaseEnvironment\n",
"from ddopnew.utils import MDPInfo\n",
"\n",
"# # TEMPORARY\n",
"# from sklearn.utils.validation import check_array\n",
"# import numbers"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"#| export\n",
"\n",
"class BaseAgent():\n",
"\n",
" train_mode = \"direct_fit\" # or \"epochs_fit\" or \"env_interaction\"\n",
" \n",
" def __init__(self,\n",
" environment_info: MDPInfo,\n",
" preprocessors: Optional[List[object]] = None, # default is empty list\n",
" postprocessors: Optional[List[object]] = None # default is empty list\n",
" ):\n",
" \n",
" \"\"\"\n",
" Initialize a BaseAgent.\n",
"\n",
" Args:\n",
" environment_info (MDPInfo): Information about the environment (MDP).\n",
" preprocessors (Optional[List[object]]): A list of preprocessors to apply to input data.\n",
" postprocessors (Optional[List[object]]): A list of postprocessors to apply to output data.\n",
" \"\"\"\n",
"\n",
" self.preprocessors = preprocessors or []\n",
" self.postprocessors = postprocessors or []\n",
"\n",
" self.environment_info = environment_info\n",
" self.mode = \"train\"\n",
" self.print = False # Can be used for debugging\n",
" self.receive_batch_dim = False\n",
"\n",
" @abstractmethod\n",
" def draw_action_(self, observation):\n",
" pass\n",
"\n",
" def draw_action(self, observation):\n",
"\n",
" observation = self.add_batch_dim(observation)\n",
"\n",
" for preprocessor in self.preprocessors:\n",
" observation = preprocessor(observation)\n",
"\n",
" action = self.draw_action_(observation)\n",
" \n",
" for postprocessor in self.postprocessors:\n",
" action = postprocessor(action)\n",
" return action\n",
"\n",
" def add_preprocessor(self, preprocessor):\n",
" self.preprocessors.append(preprocessor)\n",
" \n",
" def add_postprocessor(self, postprocessor):\n",
" self.postprocessors.append(postprocessor)\n",
"\n",
" def train(self):\n",
" self.mode = \"train\"\n",
" \n",
" def eval(self):\n",
" self.mode = \"eval\"\n",
" \n",
" def add_batch_dim(self, input: np.ndarray) -> np.ndarray:\n",
" \n",
" \"\"\"\n",
" Add a batch dimension to the input array if it doesn't already have one.\n",
"\n",
" Args:\n",
" input (np.ndarray): The input array that may need a batch dimension added.\n",
"\n",
" Returns:\n",
" np.ndarray: The input array with an added batch dimension, if required.\n",
" \"\"\"\n",
"\n",
" if self.receive_batch_dim:\n",
" # If the batch dimension is expected, return the input as is\n",
" return input\n",
" else:\n",
" # Add a batch dimension by expanding the dimensions of the input\n",
" return np.expand_dims(input, axis=0)\n",
" \n",
" def flatten_X(self, X):\n",
"\n",
" \"\"\"\n",
"\n",
" Function to flatten the time-dimension of the feature matrix\n",
" for agents that require a 2D input.\n",
" \n",
" \n",
" Args:\n",
" X (np.ndarray): The input data to be flattened.\n",
"\n",
" Returns:\n",
" _type_: _description_\n",
" \"\"\"\n",
"\n",
" if X.ndim == 3:\n",
" return X.reshape(X.shape[0], -1)\n",
" else:\n",
" return X\n",
" \n",
" def save(self):\n",
" raise NotImplementedError(\"This agent does not have a save method implemented.\")\n",
"\n",
" def load(self):\n",
" raise NotImplementedError(\"This agent does not have a load method implemented.\")\n",
" "
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# #| export\n",
"\n",
"# def check_cu_co(cu, co, n_outputs):\n",
"# \"\"\"Validate under- and overage costs.\n",
"\n",
"# Parameters\n",
"# ----------\n",
"# cu : {ndarray, Number or None}, shape (n_outputs,)\n",
"# The underage costs per unit. Passing cu=None will output an array of ones.\n",
"# co : {ndarray, Number or None}, shape (n_outputs,)\n",
"# The overage costs per unit. Passing co=None will output an array of ones.\n",
"# n_outputs : int\n",
"# The number of outputs.\n",
"# Returns\n",
"# -------\n",
"# cu : ndarray, shape (n_outputs,)\n",
"# Validated underage costs. It is guaranteed to be \"C\" contiguous.\n",
"# co : ndarray, shape (n_outputs,)\n",
"# Validated overage costs. It is guaranteed to be \"C\" contiguous.\n",
"# \"\"\"\n",
"# costs = [[cu, \"cu\"], [co, \"co\"]]\n",
"# costs_validated = []\n",
"# for c in costs:\n",
"# if c[0] is None:\n",
"# cost = np.ones(n_outputs, dtype=np.float64)\n",
"# elif isinstance(c[0], numbers.Number):\n",
"# cost = np.full(n_outputs, c[0], dtype=np.float64)\n",
"# else:\n",
"# cost = check_array(\n",
"# c[0], accept_sparse=False, ensure_2d=False, dtype=np.float64,\n",
"# order=\"C\"\n",
"# )\n",
"# if cost.ndim != 1:\n",
"# raise ValueError(c[1], \"must be 1D array or scalar\")\n",
"\n",
"# if cost.shape != (n_outputs,):\n",
"# raise ValueError(\"{}.shape == {}, expected {}!\"\n",
"# .format(c[1], cost.shape, (n_outputs,)))\n",
"# costs_validated.append(cost)\n",
"# cu = costs_validated[0]\n",
"# co = costs_validated[1]\n",
"# return cu, co\n",
"\n",
"# class NewsvendorSAAagentOLD(BaseAgent):\n",
"\n",
"# def __init__(self, environment_info, cu, co):\n",
"# self.cu = cu\n",
"# self.co = co\n",
"\n",
"# self.sl = cu / (cu + co)\n",
"\n",
"# self.quantiles = np.array([0.0])\n",
"\n",
"# super().__init__(environment_info)\n",
"\n",
"# self.fitted = False\n",
"\n",
"# def _calc_weights(self):\n",
"# weights = np.full(self.n_samples_, 1 / self.n_samples_)\n",
"# return weights\n",
"\n",
"# def fit(self, Y, X, mask=None):\n",
"\n",
"# demand = Y\n",
"# features = X\n",
"\n",
"# self.mask=mask\n",
"# y=demand\n",
"\n",
"# if mask is not None:\n",
"# if demand.shape != mask.shape:\n",
"# if demand.shape[1]==1 & len(mask.shape)==1:\n",
"# mask = mask.reshape((-1,1))\n",
"# self.mask = mask\n",
"# if demand.shape != mask.shape:\n",
"# raise ValueError(\"Shapes of mask and demand do not match\")\n",
"# # check if 1 either in mask.shape or demand.shape, if yes squeeze\n",
"# demand=demand*mask\n",
" \n",
"# y = check_array(y, ensure_2d=False, accept_sparse='csr')\n",
"\n",
"# if y.ndim == 1:\n",
"# y = np.reshape(y, (-1, 1))\n",
"\n",
"# # Training data\n",
"# self.y_ = y\n",
"# self.n_samples_ = y.shape[0]\n",
"\n",
"# # Determine output settings\n",
"# self.n_outputs_ = y.shape[1]\n",
"\n",
"# # Check and format under- and overage costs\n",
"# self.cu_, self.co_ = check_cu_co(self.cu, self.co, self.n_outputs_)\n",
"\n",
"# self.q_star = np.array(self._findQ(self._calc_weights()))\n",
"\n",
"# self.fitted=True\n",
"\n",
"# return self\n",
"\n",
"# def _findQ(self, weights):\n",
"# \"\"\"Calculate the optimal order quantity q\"\"\"\n",
"\n",
"# y = self.y_\n",
"# q = []\n",
"\n",
"# for k in range(self.n_outputs_):\n",
"# alpha = self.cu_[k] / (self.cu_[k] + self.co_[k])\n",
"# y_product = y[:,k]\n",
"# if self.mask is not None:\n",
"# mask_product = self.mask[:,k]\n",
"# y_product = y_product[mask_product.astype(bool)]\n",
"# # print(y_product.shape)\n",
"# q.append(np.quantile(y_product, alpha, interpolation=\"higher\"))\n",
"# print(\"found optimal q:\", q)\n",
"# return q\n",
"\n",
"# def draw_action(self, *args, **kwargs):\n",
"\n",
"# if self.fitted:\n",
"# pred = self.q_star\n",
" \n",
"# else:\n",
"# pred = np.random.rand(1) \n",
"# return pred"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"#| hide\n",
"import nbdev; nbdev.nbdev_export()"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "python3",
"language": "python",
"name": "python3"
}
},
"nbformat": 4,
"nbformat_minor": 4
}

0 comments on commit b031eab

Please sign in to comment.