Shourya Bose commited on
Commit
0aa0c9c
1 Parent(s): 4ad6488

update readme

Browse files
Files changed (1) hide show
  1. README.md +4 -8
README.md CHANGED
@@ -7,13 +7,6 @@ This repository provides model weights to run load forecasting models trained on
7
 
8
  Note that `lookback` is denoted by `L` and `lookahead` by `T` in the weights directory. We provide weights for the following `(L,T)` pairs: `(512,4)`, `(512,48)`, and `(512,96)`, and for `HOM`ogenous and `HET`erogenous datasets.
9
 
10
- ## Data
11
-
12
- When using the companion [dataset](https://huggingface.co/datasets/APPFL/Illinois_load_datasets), the following points must be noted (see the page for more information on configuring the data loaders):
13
-
14
- - All models accept normalized inputs and produce normalized outputs, i.e. set `normalize = True` when generating the datasets.
15
- - For Transformer, Autoformer, Informer, and TimesNet set `transformer = True`, while for LSTM, LSTNet, and PatchTST set `transformer = False`.
16
-
17
  ## Packages
18
 
19
  Executing the code only requires `numpy` and `torch` (PyTorch) packages. You can either have them in your Python base installation, or use a `conda` environment.
@@ -22,7 +15,7 @@ Executing the code only requires `numpy` and `torch` (PyTorch) packages. You can
22
 
23
  In order to see how to use the model definitions and load the weights into them, see `example.py`.
24
 
25
- ## Technical Details
26
 
27
  In input layout of the models are as follows:
28
 
@@ -31,6 +24,7 @@ In input layout of the models are as follows:
31
  - `input` is a tensor of shape `(B,L,num_features)` where `B` is the batch size, `L` is the lookback duration, and `num_features` is 8 for our current application.
32
  - `future_time_idx` is a tensor of shape `(B,T,2)` where `T` is the lookahead and 2 is the number of time index features.
33
  - The time indices in `input` as well as `fut_time_idx` are both normalized.
 
34
  - Non-time features are normalized. The mean and standard deviation of the [companion dataset](https://huggingface.co/datasets/APPFL/Illinois_load_datasets) can be inferred by executing `example_dataset.py` there and looking at `Case 1` and `Case 4`.
35
  - The output shape is `(B,1)` denoting the pointwise forecast `T` steps into the future.
36
  - The `forward()` functions of `Transformer`, `Autoformer`, `Informer`, and `TimesNet` take in two arguments:` forward(input, future_time_idx)`. They are laid out as follows:
@@ -38,12 +32,14 @@ In input layout of the models are as follows:
38
  - `input` is a tensor of shape `(B,L,num_features)` where `B` is the batch size, `L` is the lookback duration, and `num_features` is 8 for our current application.
39
  - `future_time_idx` is a tensor of shape `(B,T,2)` where `T` is the lookahead and 2 is the number of time index features.
40
  - The time indices in `input` as well as `fut_time_idx` are un-normalized to allow for embedding.
 
41
  - Non-time features are normalized. The mean and standard deviation of the [companion dataset](https://huggingface.co/datasets/APPFL/Illinois_load_datasets) can be inferred by executing `example_dataset.py` there and looking at `Case 2` and `Case 5`.
42
  - The output shape is `(B,1)` denoting the pointwise forecast `T` steps into the future.
43
  - The `forward()` functions of `TimesFM` takes in one argument:` forward(input)`. It is laid out as follows:
44
 
45
  - `input` is a tensor of shape `(B,L)` where `B` is the batch size and `L` is the lookback duration. Since it is univariate, there is only one feature.
46
  - The sole feature is normalized. The mean and standard deviation of the [companion dataset](https://huggingface.co/datasets/APPFL/Illinois_load_datasets) can be inferred by executing `example_dataset.py` there and looking at `Case 3` and `Case 6`.
 
47
  - The output shape is `(B,T)` denoting the rolling horizon forecast `T` steps into the future.
48
 
49
  ## Credits
 
7
 
8
  Note that `lookback` is denoted by `L` and `lookahead` by `T` in the weights directory. We provide weights for the following `(L,T)` pairs: `(512,4)`, `(512,48)`, and `(512,96)`, and for `HOM`ogenous and `HET`erogenous datasets.
9
 
 
 
 
 
 
 
 
10
  ## Packages
11
 
12
  Executing the code only requires `numpy` and `torch` (PyTorch) packages. You can either have them in your Python base installation, or use a `conda` environment.
 
15
 
16
  In order to see how to use the model definitions and load the weights into them, see `example.py`.
17
 
18
+ ## Technical Details for Running the Models
19
 
20
  In input layout of the models are as follows:
21
 
 
24
  - `input` is a tensor of shape `(B,L,num_features)` where `B` is the batch size, `L` is the lookback duration, and `num_features` is 8 for our current application.
25
  - `future_time_idx` is a tensor of shape `(B,T,2)` where `T` is the lookahead and 2 is the number of time index features.
26
  - The time indices in `input` as well as `fut_time_idx` are both normalized.
27
+ - The custom `torch.utils.data.Dataset` class for the train, val, and test sets can be generated by executing the `get_data_and_generate_train_val_test_sets` function in the `custom_dataset.py` file in the [companion dataset](https://huggingface.co/datasets/APPFL/Illinois_load_datasets).
28
  - Non-time features are normalized. The mean and standard deviation of the [companion dataset](https://huggingface.co/datasets/APPFL/Illinois_load_datasets) can be inferred by executing `example_dataset.py` there and looking at `Case 1` and `Case 4`.
29
  - The output shape is `(B,1)` denoting the pointwise forecast `T` steps into the future.
30
  - The `forward()` functions of `Transformer`, `Autoformer`, `Informer`, and `TimesNet` take in two arguments:` forward(input, future_time_idx)`. They are laid out as follows:
 
32
  - `input` is a tensor of shape `(B,L,num_features)` where `B` is the batch size, `L` is the lookback duration, and `num_features` is 8 for our current application.
33
  - `future_time_idx` is a tensor of shape `(B,T,2)` where `T` is the lookahead and 2 is the number of time index features.
34
  - The time indices in `input` as well as `fut_time_idx` are un-normalized to allow for embedding.
35
+ - The custom `torch.utils.data.Dataset` class for the train, val, and test sets can be generated by executing the `get_data_and_generate_train_val_test_sets` function in the `custom_dataset.py` file in the [companion dataset](https://huggingface.co/datasets/APPFL/Illinois_load_datasets).
36
  - Non-time features are normalized. The mean and standard deviation of the [companion dataset](https://huggingface.co/datasets/APPFL/Illinois_load_datasets) can be inferred by executing `example_dataset.py` there and looking at `Case 2` and `Case 5`.
37
  - The output shape is `(B,1)` denoting the pointwise forecast `T` steps into the future.
38
  - The `forward()` functions of `TimesFM` takes in one argument:` forward(input)`. It is laid out as follows:
39
 
40
  - `input` is a tensor of shape `(B,L)` where `B` is the batch size and `L` is the lookback duration. Since it is univariate, there is only one feature.
41
  - The sole feature is normalized. The mean and standard deviation of the [companion dataset](https://huggingface.co/datasets/APPFL/Illinois_load_datasets) can be inferred by executing `example_dataset.py` there and looking at `Case 3` and `Case 6`.
42
+ - The custom `torch.utils.data.Dataset` class for the train, val, and test sets can be generated by executing the `get_data_and_generate_train_val_test_sets` function in the `custom_dataset_univariate.py` file in the [companion dataset](https://huggingface.co/datasets/APPFL/Illinois_load_datasets).
43
  - The output shape is `(B,T)` denoting the rolling horizon forecast `T` steps into the future.
44
 
45
  ## Credits