Rearranged folders (#21)

* fixed error in selection of time_sku features * fixed error in force save agent * fixed parameter name for dataset * accelerated getitem method * minor fix in predict function * minor fix in predict function * enables out-of-sample experiments * print steps while testing * fixed error in normalization for oos products * fixed error in normalizing oos demand * fixed error in normalization of oos SKUs * fixed error in normalization of oos SKUs * added option for run_experiment to return score * fixed average calc tune! * implemented xgb * renaming * renamed to ddopai * updated requirements * updated requirements * moved some imports to extra cells * rearranged
d3group · Sep 30, 2024 · 83ca13b · 83ca13b
1 parent d98888e
commit 83ca13b
Show file tree

Hide file tree

Showing 85 changed files with 1,573 additions and 3,111 deletions.
diff --git a/README.md b/README.md
@@ -1,14 +1,14 @@
-# ddopnew
+# ddopai
 
 
 <!-- WARNING: THIS FILE WAS AUTOGENERATED! DO NOT EDIT! -->
 
 ## Install
 
 ``` sh
-pip install ddopnew
+pip install ddopai
 ```
 
-## What is ddopnew?
+## What is ddopai?
 
 To be written.
diff --git a/ddopnew/RL_approximators.py → ddopai/RL_approximators.py b/ddopnew/RL_approximators.py → ddopai/RL_approximators.py
@@ -6,7 +6,10 @@
 __all__ = ['RNNWrapper', 'BaseApproximator', 'BaseApproximatorMLP', 'RNNMLPHybrid', 'BaseApproximatorRNN', 'MLPStateAction',
            'MLPState', 'MLPActor', 'RNNActor', 'RNNStateAction']
 
-# %% ../nbs/60_approximators/21_critic_networks.ipynb 4
+# %% ../nbs/60_approximators/21_critic_networks.ipynb 3
+# import logging
+# logging_level = logging.DEBUG
+
 from abc import ABC, abstractmethod
 from typing import Union, Tuple, List
 import numpy as np
@@ -18,7 +21,7 @@
 
 import time
 
-# %% ../nbs/60_approximators/21_critic_networks.ipynb 5
+# %% ../nbs/60_approximators/21_critic_networks.ipynb 4
 class RNNWrapper(nn.Module):
     def __init__(self, rnn_cell_class, *args, **kwargs):
         """
@@ -52,7 +55,7 @@ def __init__(self, *args, **kwargs):
 
         return SpecificRNNWrapper
 
-# %% ../nbs/60_approximators/21_critic_networks.ipynb 6
+# %% ../nbs/60_approximators/21_critic_networks.ipynb 5
 class BaseApproximator():
 
     """ Some basic functions for approximators """
@@ -142,7 +145,7 @@ def forward(self, x):
         """ Forward pass through the network - overwrite this if necessary """
         return self.model(x)
 
-# %% ../nbs/60_approximators/21_critic_networks.ipynb 7
+# %% ../nbs/60_approximators/21_critic_networks.ipynb 6
 class BaseApproximatorMLP(BaseApproximator, nn.Module):
 
     """ Some basic functions for approximators """
@@ -195,7 +198,7 @@ def build_MLP(  self,
 
         self.model = model
 
-# %% ../nbs/60_approximators/21_critic_networks.ipynb 8
+# %% ../nbs/60_approximators/21_critic_networks.ipynb 7
 class RNNMLPHybrid(nn.Module, BaseApproximator):
 
     """ A hybrid model combining an RNN and an MLP """
@@ -294,7 +297,7 @@ def forward(self, x_rnn, x_mlp=None):
 
         return x
 
-# %% ../nbs/60_approximators/21_critic_networks.ipynb 9
+# %% ../nbs/60_approximators/21_critic_networks.ipynb 8
 class BaseApproximatorRNN(BaseApproximator, nn.Module):
 
     """ Some basic functions for approximators """
@@ -351,7 +354,7 @@ def build_RNN(  self,
 
 
 
-# %% ../nbs/60_approximators/21_critic_networks.ipynb 10
+# %% ../nbs/60_approximators/21_critic_networks.ipynb 9
 class MLPStateAction(BaseApproximatorMLP):
 
     """Multilayer perceptron model for critic networks that take
@@ -396,7 +399,7 @@ def forward(self, state, action):
         # return q
         return torch.squeeze(q)
 
-# %% ../nbs/60_approximators/21_critic_networks.ipynb 11
+# %% ../nbs/60_approximators/21_critic_networks.ipynb 10
 class MLPState(BaseApproximatorMLP):
 
     """Multilayer perceptron model for critic networks that take
@@ -436,7 +439,7 @@ def forward(self, state):
         # return q.squeeze()
         return q
 
-# %% ../nbs/60_approximators/21_critic_networks.ipynb 12
+# %% ../nbs/60_approximators/21_critic_networks.ipynb 11
 class MLPActor(BaseApproximatorMLP):
 
     """Multilayer perceptron model for critic networks that take
@@ -475,7 +478,7 @@ def forward(self, state):
 
         return a
 
-# %% ../nbs/60_approximators/21_critic_networks.ipynb 13
+# %% ../nbs/60_approximators/21_critic_networks.ipynb 12
 class RNNActor(BaseApproximatorRNN):
 
     """Multilayer perceptron model for critic networks that take
@@ -577,7 +580,7 @@ def forward_without_batch(self, state):
 
         return self.model(rnn_input, mlp_input)
 
-# %% ../nbs/60_approximators/21_critic_networks.ipynb 14
+# %% ../nbs/60_approximators/21_critic_networks.ipynb 13
 class RNNStateAction(BaseApproximatorRNN):
 
     """Multilayer perceptron model for critic networks that take

diff --git a/ddopnew/__init__.py → ddopai/__init__.py b/ddopnew/__init__.py → ddopai/__init__.py
diff --git a/ddopai/_modidx.py b/ddopai/_modidx.py
diff --git a/ddopnew/agents/__init__.py → ddopai/agents/__init__.py b/ddopnew/agents/__init__.py → ddopai/agents/__init__.py
diff --git a/ddopnew/agents/base.py → ddopai/agents/base.py b/ddopnew/agents/base.py → ddopai/agents/base.py
@@ -5,7 +5,10 @@
 # %% auto 0
 __all__ = ['BaseAgent']
 
-# %% ../../nbs/40_base_agents/10_base_agents.ipynb 4
+# %% ../../nbs/40_base_agents/10_base_agents.ipynb 3
+# import logging
+# logging_level = logging.DEBUG
+
 from abc import ABC, abstractmethod
 from typing import Union, Optional, List, Tuple
 import numpy as np
@@ -18,7 +21,7 @@
 # from sklearn.utils.validation import check_array
 # import numbers
 
-# %% ../../nbs/40_base_agents/10_base_agents.ipynb 5
+# %% ../../nbs/40_base_agents/10_base_agents.ipynb 4
 class BaseAgent():
 
     """  

diff --git a/ddopnew/agents/basic.py → ddopai/agents/basic.py b/ddopnew/agents/basic.py → ddopai/agents/basic.py
diff --git a/ddopai/agents/class_names.py b/ddopai/agents/class_names.py
@@ -0,0 +1,33 @@
+"""Dict of agent classes and (standard) agent names to allow for dynamic loading of agents."""
+
+# AUTOGENERATED! DO NOT EDIT! File to edit: ../../nbs/40_base_agents/10_AGENT_CLASSES.ipynb.
+
+# %% auto 0
+__all__ = ['AGENT_CLASSES']
+
+# %% ../../nbs/40_base_agents/10_AGENT_CLASSES.ipynb 3
+AGENT_CLASSES = {
+    "RandomAgent": "ddopai.agents.saa.SAA",
+
+    "SAA": "ddopai.agents.newsvendor.saa.NewsvendorSAAagent",
+    "wSAA": "ddopai.agents.newsvendor.saa.NewsvendorRFwSAAagent",
+    "RFwSAA": "ddopai.agents.newsvendor.saa.NewsvendorRFwSAAagent",
+
+    "XGB": "ddopai.agents.newsvendor.erm.NewsvendorXGBAgent",
+
+    "lERM": "ddopai.agents.newsvendor.erm.NewsvendorlERMAgent",
+    "DLNV": "ddopai.agents.newsvendor.erm.NewsvendorDLAgent",
+    "DLNVRNN": "ddopai.agents.newsvendor.erm.NewsvendorDLRNNAgent",
+    "DLNVTransformer": "ddopai.agents.newsvendor.erm.NewsvendorDLTransformerAgent",
+
+    "lERMMeta": "ddopai.agents.newsvendor.erm.NewsvendorlERMMetaAgent",
+    "DLNVMeta": "ddopai.agents.newsvendor.erm.NewsvendorDLMetaAgent",
+    "DLNVRNNMeta": "ddopai.agents.newsvendor.erm.NewsvendorDLRNNMetaAgent",
+    "DLNVTransformerMeta": "ddopai.agents.newsvendor.erm.NewsvendorDLTransformerMetaAgent",
+
+    "SAC": "ddopai.agents.rl.sac.SACAgent",
+    "SACRNN": "ddopai.agents.rl.sac.SACRNNAgent",
+
+    "TD3": "ddopai.agents.rl.td3.TD3Agent",
+    "PPO": "ddopai.agents.rl.ppo.PPOAgent",
+}
diff --git a/ddopnew/agents/newsvendor/__init__.py → ddopai/agents/newsvendor/__init__.py b/ddopnew/agents/newsvendor/__init__.py → ddopai/agents/newsvendor/__init__.py
diff --git a/ddopnew/agents/newsvendor/erm.py → ddopai/agents/newsvendor/erm.py b/ddopnew/agents/newsvendor/erm.py → ddopai/agents/newsvendor/erm.py
@@ -641,7 +641,7 @@ def set_model(self, input_shape, output_shape):
 
         """Set the model for the agent to a linear model"""
 
-        from ddopnew.approximators import LinearModel
+        from ddopai.approximators import LinearModel
 
         # flatten time dim of input
         print("input shape", input_shape)
@@ -650,7 +650,7 @@ def set_model(self, input_shape, output_shape):
 
         self.model = LinearModel(input_size=input_size, output_size=output_size, **self.model_params)
 
-# %% ../../../nbs/41_NV_agents/11_NV_erm_agents.ipynb 30
+# %% ../../../nbs/41_NV_agents/11_NV_erm_agents.ipynb 31
 class NewsvendorDLAgent(NVBaseAgent):
 
     """
@@ -719,10 +719,10 @@ def set_model(self, input_shape, output_shape):
         input_size = np.prod(input_shape)
         output_size = output_shape[0]
 
-        from ddopnew.approximators import MLP
+        from ddopai.approximators import MLP
         self.model = MLP(input_size=input_size, output_size=output_size, **self.model_params)
 
-# %% ../../../nbs/41_NV_agents/11_NV_erm_agents.ipynb 36
+# %% ../../../nbs/41_NV_agents/11_NV_erm_agents.ipynb 37
 class BaseMetaAgent():
 
     def set_meta_dataloader(
@@ -738,7 +738,7 @@ def set_meta_dataloader(
 
         self.dataloader = torch.utils.data.DataLoader(dataset, **dataloader_params)
 
-# %% ../../../nbs/41_NV_agents/11_NV_erm_agents.ipynb 37
+# %% ../../../nbs/41_NV_agents/11_NV_erm_agents.ipynb 38
 class NewsvendorlERMMetaAgent(NewsvendorlERMAgent, BaseMetaAgent):
 
     """
@@ -793,7 +793,7 @@ def __init__(self,
             loss_function=loss_function,
         )
 
-# %% ../../../nbs/41_NV_agents/11_NV_erm_agents.ipynb 38
+# %% ../../../nbs/41_NV_agents/11_NV_erm_agents.ipynb 39
 class NewsvendorDLMetaAgent(NewsvendorDLAgent, BaseMetaAgent):
 
     """
@@ -852,7 +852,7 @@ def __init__(self,
         )
 
 
-# %% ../../../nbs/41_NV_agents/11_NV_erm_agents.ipynb 39
+# %% ../../../nbs/41_NV_agents/11_NV_erm_agents.ipynb 40
 class NewsvendorDLTransformerAgent(NVBaseAgent):
 
     """
@@ -929,10 +929,10 @@ def set_model(self, input_shape, output_shape):
 
         output_size = output_shape[0]
 
-        from ddopnew.approximators import Transformer
+        from ddopai.approximators import Transformer
         self.model = Transformer(input_size=input_shape, output_size=output_size, **self.model_params)
 
-# %% ../../../nbs/41_NV_agents/11_NV_erm_agents.ipynb 40
+# %% ../../../nbs/41_NV_agents/11_NV_erm_agents.ipynb 41
 class NewsvendorDLTransformerMetaAgent(NewsvendorDLTransformerAgent, BaseMetaAgent):
 
     """

diff --git a/ddopnew/agents/newsvendor/saa.py → ddopai/agents/newsvendor/saa.py b/ddopnew/agents/newsvendor/saa.py → ddopai/agents/newsvendor/saa.py
@@ -22,7 +22,6 @@
 from sklearn.ensemble import RandomForestRegressor
 from sklearn.utils.validation import check_array
 
-
 # %% ../../../nbs/41_NV_agents/10_NV_saa_agents.ipynb 4
 class BaseSAAagent(BaseAgent):
 

diff --git a/ddopnew/agents/rl/__init__.py → ddopai/agents/rl/__init__.py b/ddopnew/agents/rl/__init__.py → ddopai/agents/rl/__init__.py
diff --git a/ddopnew/agents/rl/mushroom_rl.py → ddopai/agents/rl/mushroom_rl.py b/ddopnew/agents/rl/mushroom_rl.py → ddopai/agents/rl/mushroom_rl.py
diff --git a/ddopnew/agents/rl/ppo.py → ddopai/agents/rl/ppo.py b/ddopnew/agents/rl/ppo.py → ddopai/agents/rl/ppo.py
diff --git a/ddopnew/agents/rl/sac.py → ddopai/agents/rl/sac.py b/ddopnew/agents/rl/sac.py → ddopai/agents/rl/sac.py
@@ -349,7 +349,7 @@ def __init__(self,
             network_critic_params=network_critic_params,
         ) 
 
-# %% ../../../nbs/51_RL_agents/10_SAC_agents.ipynb 8
+# %% ../../../nbs/51_RL_agents/10_SAC_agents.ipynb 9
 class SACRNNAgent(SACBaseAgent):
 
     """

diff --git a/ddopnew/agents/rl/td3.py → ddopai/agents/rl/td3.py b/ddopnew/agents/rl/td3.py → ddopai/agents/rl/td3.py
diff --git a/ddopnew/approximators.py → ddopai/approximators.py b/ddopnew/approximators.py → ddopai/approximators.py
@@ -6,7 +6,10 @@
 __all__ = ['BaseModule', 'LinearModel', 'MLP', 'Transformer', 'LlamaRotaryEmbedding', 'rotate_half', 'apply_rotary_pos_emb',
            'CausalSelfAttention', 'find_multiple', 'MLP_block', 'RMSNorm', 'Block']
 
-# %% ../nbs/60_approximators/11_approximators.ipynb 4
+# %% ../nbs/60_approximators/11_approximators.ipynb 3
+# import logging
+# logging_level = logging.DEBUG
+
 from abc import ABC, abstractmethod
 from typing import Union, Dict, Literal
 import numpy as np
@@ -18,7 +21,7 @@
 
 import time
 
-# %% ../nbs/60_approximators/11_approximators.ipynb 5
+# %% ../nbs/60_approximators/11_approximators.ipynb 4
 # TODO: Merge with base from RL networks to avoid duplicaiotn of code
 
 class BaseModule(nn.Module):
@@ -44,7 +47,7 @@ def select_activation(activation):
         else:
             raise ValueError(f"Activation function {activation} not recognized")
 
-# %% ../nbs/60_approximators/11_approximators.ipynb 6
+# %% ../nbs/60_approximators/11_approximators.ipynb 5
 class LinearModel(BaseModule):
     """Linear regression model"""
 
@@ -64,7 +67,7 @@ def forward(self,x):
         out=self.final_activation(out)
         return out
 
-# %% ../nbs/60_approximators/11_approximators.ipynb 7
+# %% ../nbs/60_approximators/11_approximators.ipynb 6
 class MLP(BaseModule):
 
     """ Multilayer perceptron model """
@@ -103,7 +106,7 @@ def __init__(self,
     def forward(self, x):
         return self.model(x)
 
-# %% ../nbs/60_approximators/11_approximators.ipynb 8
+# %% ../nbs/60_approximators/11_approximators.ipynb 7
 class Transformer(BaseModule):
 
     """ Multilayer perceptron model """
@@ -168,7 +171,7 @@ def forward(    self,
 
         return output
 
-# %% ../nbs/60_approximators/11_approximators.ipynb 9
+# %% ../nbs/60_approximators/11_approximators.ipynb 8
 class LlamaRotaryEmbedding(torch.nn.Module):
 
     """
@@ -234,7 +237,7 @@ def apply_rotary_pos_emb(q, k, cos, sin, position_ids):
     k_embed = (k * cos) + (rotate_half(k) * sin)
     return q_embed, k_embed
 
-# %% ../nbs/60_approximators/11_approximators.ipynb 10
+# %% ../nbs/60_approximators/11_approximators.ipynb 9
 class CausalSelfAttention(nn.Module):
 
     """ Causeal self-attention module
@@ -331,7 +334,7 @@ def find_multiple(n: int, k: int) -> int:
         return n
     return n + k - (n % k)
 
-# %% ../nbs/60_approximators/11_approximators.ipynb 11
+# %% ../nbs/60_approximators/11_approximators.ipynb 10
 class MLP_block(nn.Module):
     def __init__(self, n_embd_per_head, n_head, min_multiple = 256, gating = True) -> None:
         super().__init__()
@@ -361,7 +364,7 @@ def forward(self, x: torch.Tensor) -> torch.Tensor:
         x = self.c_proj(x)
         return x
 
-# %% ../nbs/60_approximators/11_approximators.ipynb 12
+# %% ../nbs/60_approximators/11_approximators.ipynb 11
 class RMSNorm(nn.Module):
     """Root Mean Square Layer Normalization as implemented in https://github.com/time-series-foundation-models/lag-llama.
 
@@ -381,7 +384,7 @@ def forward(self, x: torch.Tensor) -> torch.Tensor:
         output = (self.scale * x_normed).type_as(x)
         return output
 
-# %% ../nbs/60_approximators/11_approximators.ipynb 13
+# %% ../nbs/60_approximators/11_approximators.ipynb 12
 class Block(nn.Module):
     def __init__(self, n_embd_per_head, n_head, block_size, dropout, min_multiple = 256, gating=True) -> None:
         super().__init__()

diff --git a/ddopnew/dataloaders/__init__.py → ddopai/dataloaders/__init__.py b/ddopnew/dataloaders/__init__.py → ddopai/dataloaders/__init__.py
diff --git a/ddopnew/dataloaders/base.py → ddopai/dataloaders/base.py b/ddopnew/dataloaders/base.py → ddopai/dataloaders/base.py
diff --git a/ddopnew/dataloaders/distribution.py → ddopai/dataloaders/distribution.py b/ddopnew/dataloaders/distribution.py → ddopai/dataloaders/distribution.py
diff --git a/ddopnew/dataloaders/tabular.py → ddopai/dataloaders/tabular.py b/ddopnew/dataloaders/tabular.py → ddopai/dataloaders/tabular.py
diff --git a/ddopnew/datasets/__init__.py → ddopai/datasets/__init__.py b/ddopnew/datasets/__init__.py → ddopai/datasets/__init__.py
diff --git a/ddopnew/datasets/default_datasets.py → ddopai/datasets/default_datasets.py b/ddopnew/datasets/default_datasets.py → ddopai/datasets/default_datasets.py
@@ -18,7 +18,7 @@
 
 # %% ../../nbs/80_datasets/default_datasets.ipynb 6
 def get_all_release_tags(token=None):
-    url = "https://api.github.com/repos/opimwue/ddopnew/releases"
+    url = "https://api.github.com/repos/opimwue/ddopai/releases"
     headers = {'Authorization': f'Bearer {token}'} if token else {}
     response = requests.get(url, headers=headers)
 
@@ -43,7 +43,7 @@ def get_release_tag(dataset_type, version, token=None):
     return release_tag
 
 def get_dataset_url(dataset_type, dataset_number, release_tag, token=None):
-    api_url = f"https://api.github.com/repos/opimwue/ddopnew/releases/tags/{release_tag}"
+    api_url = f"https://api.github.com/repos/opimwue/ddopai/releases/tags/{release_tag}"
     headers = {'Authorization': f'Bearer {token}'} if token else {}
     response = requests.get(api_url, headers=headers)
 

diff --git a/ddopnew/datasets/kaggle_m5.py → ddopai/datasets/kaggle_m5.py b/ddopnew/datasets/kaggle_m5.py → ddopai/datasets/kaggle_m5.py
diff --git a/ddopnew/envs/__init__.py → ddopai/envs/__init__.py b/ddopnew/envs/__init__.py → ddopai/envs/__init__.py
diff --git a/ddopnew/envs/base.py → ddopai/envs/base.py b/ddopnew/envs/base.py → ddopai/envs/base.py
diff --git a/ddopnew/envs/inventory/__init__.py → ddopai/envs/inventory/__init__.py b/ddopnew/envs/inventory/__init__.py → ddopai/envs/inventory/__init__.py
diff --git a/ddopnew/envs/inventory/base.py → ddopai/envs/inventory/base.py b/ddopnew/envs/inventory/base.py → ddopai/envs/inventory/base.py
diff --git a/ddopnew/envs/inventory/inventory_utils.py → ddopai/envs/inventory/inventory_utils.py b/ddopnew/envs/inventory/inventory_utils.py → ddopai/envs/inventory/inventory_utils.py
@@ -26,6 +26,7 @@
 
 # %% ../../../nbs/21_envs_inventory/00_inventory_utils.ipynb 4
 class OrderPipeline():
+
     """
     Class to handle the order pipeline in the inventory environments. It is used to keep track of the orders
     that are placed. It can account for fixed and variable lead times.

diff --git a/ddopnew/envs/inventory/multi_period.py → ddopai/envs/inventory/multi_period.py b/ddopnew/envs/inventory/multi_period.py → ddopai/envs/inventory/multi_period.py
diff --git a/ddopnew/envs/inventory/single_period.py → ddopai/envs/inventory/single_period.py b/ddopnew/envs/inventory/single_period.py → ddopai/envs/inventory/single_period.py
@@ -164,7 +164,7 @@ def update_cu_co(self, cu=None, co=None):
             self.set_param("sl", sl, shape=(self.num_SKUs[0],))
 
 
-# %% ../../../nbs/21_envs_inventory/20_single_period_envs.ipynb 14
+# %% ../../../nbs/21_envs_inventory/20_single_period_envs.ipynb 15
 class NewsvendorEnvVariableSL(NewsvendorEnv, ABC):
     def __init__(self,
 
@@ -301,7 +301,7 @@ def get_observation(self):
 
         # print("env mode:", self.mode)
         # print("dataloader mode:", self.dataloader.dataset_type)
-        
+
         X_item, Y_item = self.dataloader[self.index]
 
         # check if any value in X_item is nan.

diff --git a/ddopnew/experiment_functions.py → ddopai/experiment_functions.py b/ddopnew/experiment_functions.py → ddopai/experiment_functions.py
diff --git a/ddopnew/loss_functions.py → ddopai/loss_functions.py b/ddopnew/loss_functions.py → ddopai/loss_functions.py