## Intro
- Neural Network models for trading: MLP, CNN, Transformer
## Usage
```bash
pip install trading-models
```
```py
import torch as tc
from torch.utils.data import DataLoader
from trading_models.simple_models import CNN, MLP, Transformer
from trading_models.utils import WindowDataset, model_size
"""
trading models
input: x.shape = (T, F)
output: a.shape = (T-W+1, A)
T: time
F: features
W: window length
A: actions
at each time step t in range(W-1, T),
the model looks at data in the window
x[t+1-W : t+1, :]
to make A actions
"""
T, F, W, A = 100, 2, 50, 1
device = tc.device("cuda" if tc.cuda.is_available() else "cpu")
x = tc.randn(T, F).to(device)
dataset = WindowDataset(x, W)
dataloader = DataLoader(dataset, batch_size=32, shuffle=False)
net1 = MLP([W * F, 64, 64, A]).to(device)
net2 = CNN([F, 64, 64, A]).to(device)
net3 = Transformer(W, F, A, d_model=64, d_ff=64, n_head=2, n_layer=2).to(device)
for net in [net1, net2, net3]:
outputs = []
for batch in dataloader:
outputs.append(net(batch.to(device)).detach())
output = tc.cat(outputs, dim=0)
print(net, output.shape, model_size(net), "\n")
```
- Output
```bash
MLP(
(mlp): Sequential(
(0): Linear(in_features=100, out_features=64, bias=True)
(1): SiLU()
(2): Linear(in_features=64, out_features=64, bias=True)
(3): SiLU()
(4): Linear(in_features=64, out_features=1, bias=True)
)
) torch.Size([51, 1]) trainable: 10689/10689
CNN(
(cnn): Sequential(
(0): Conv1d(2, 64, kernel_size=(3,), stride=(1,), padding=(1,))
(1): SiLU()
(2): Conv1d(64, 64, kernel_size=(3,), stride=(1,), padding=(1,))
(3): SiLU()
(4): AdaptiveAvgPool1d(output_size=1)
)
(fc): Linear(in_features=64, out_features=1, bias=True)
) torch.Size([51, 1]) trainable: 12865/12865
Transformer(
(proj): Linear(in_features=2, out_features=64, bias=True)
(trans): TransformerEncoder(
(layers): ModuleList(
(0-1): 2 x TransformerEncoderLayer(
(self_attn): MultiheadAttention(
(out_proj): NonDynamicallyQuantizableLinear(in_features=64, out_features=64, bias=True)
)
(linear1): Linear(in_features=64, out_features=64, bias=True)
(dropout): Dropout(p=0.1, inplace=False)
(linear2): Linear(in_features=64, out_features=64, bias=True)
(norm1): LayerNorm((64,), eps=1e-05, elementwise_affine=True)
(norm2): LayerNorm((64,), eps=1e-05, elementwise_affine=True)
(dropout1): Dropout(p=0.1, inplace=False)
(dropout2): Dropout(p=0.1, inplace=False)
)
)
)
(fc): Linear(in_features=64, out_features=1, bias=True)
) torch.Size([51, 1]) trainable: 50689/50689
```
Raw data
{
"_id": null,
"home_page": "https://github.com/SerenaTradingResearch/trading-models",
"name": "trading-models",
"maintainer": null,
"docs_url": null,
"requires_python": ">=3.8",
"maintainer_email": null,
"keywords": "trading models, neural networks, time-series, MLP, CNN, Transformer",
"author": "Ricky Ding",
"author_email": "e0134117@u.nus.edu",
"download_url": "https://files.pythonhosted.org/packages/9b/fb/b40a22769f17da03b2adf7e69be1a27f25703a1151dce1ffd1269a45d2c7/trading_models-0.1.1.tar.gz",
"platform": null,
"description": "\n## Intro\n\n- Neural Network models for trading: MLP, CNN, Transformer\n\n## Usage\n\n```bash\npip install trading-models\n```\n\n```py\nimport torch as tc\nfrom torch.utils.data import DataLoader\n\nfrom trading_models.simple_models import CNN, MLP, Transformer\nfrom trading_models.utils import WindowDataset, model_size\n\n\"\"\"\ntrading models\n\ninput: x.shape = (T, F)\noutput: a.shape = (T-W+1, A)\n\nT: time\nF: features\nW: window length\nA: actions\n\nat each time step t in range(W-1, T),\nthe model looks at data in the window\nx[t+1-W : t+1, :]\nto make A actions\n\"\"\"\n\nT, F, W, A = 100, 2, 50, 1\ndevice = tc.device(\"cuda\" if tc.cuda.is_available() else \"cpu\")\nx = tc.randn(T, F).to(device)\n\ndataset = WindowDataset(x, W)\ndataloader = DataLoader(dataset, batch_size=32, shuffle=False)\n\nnet1 = MLP([W * F, 64, 64, A]).to(device)\nnet2 = CNN([F, 64, 64, A]).to(device)\nnet3 = Transformer(W, F, A, d_model=64, d_ff=64, n_head=2, n_layer=2).to(device)\n\nfor net in [net1, net2, net3]:\n outputs = []\n for batch in dataloader:\n outputs.append(net(batch.to(device)).detach())\n output = tc.cat(outputs, dim=0)\n print(net, output.shape, model_size(net), \"\\n\")\n\n```\n\n- Output\n\n```bash\nMLP(\n (mlp): Sequential(\n (0): Linear(in_features=100, out_features=64, bias=True)\n (1): SiLU()\n (2): Linear(in_features=64, out_features=64, bias=True)\n (3): SiLU()\n (4): Linear(in_features=64, out_features=1, bias=True)\n )\n) torch.Size([51, 1]) trainable: 10689/10689 \n\nCNN(\n (cnn): Sequential(\n (0): Conv1d(2, 64, kernel_size=(3,), stride=(1,), padding=(1,))\n (1): SiLU()\n (2): Conv1d(64, 64, kernel_size=(3,), stride=(1,), padding=(1,))\n (3): SiLU()\n (4): AdaptiveAvgPool1d(output_size=1)\n )\n (fc): Linear(in_features=64, out_features=1, bias=True)\n) torch.Size([51, 1]) trainable: 12865/12865 \n\nTransformer(\n (proj): Linear(in_features=2, out_features=64, bias=True)\n (trans): TransformerEncoder(\n (layers): ModuleList(\n (0-1): 2 x TransformerEncoderLayer(\n (self_attn): MultiheadAttention(\n (out_proj): NonDynamicallyQuantizableLinear(in_features=64, out_features=64, bias=True)\n )\n (linear1): Linear(in_features=64, out_features=64, bias=True)\n (dropout): Dropout(p=0.1, inplace=False)\n (linear2): Linear(in_features=64, out_features=64, bias=True)\n (norm1): LayerNorm((64,), eps=1e-05, elementwise_affine=True)\n (norm2): LayerNorm((64,), eps=1e-05, elementwise_affine=True)\n (dropout1): Dropout(p=0.1, inplace=False)\n (dropout2): Dropout(p=0.1, inplace=False)\n )\n )\n )\n (fc): Linear(in_features=64, out_features=1, bias=True)\n) torch.Size([51, 1]) trainable: 50689/50689 \n```\n",
"bugtrack_url": null,
"license": "MIT",
"summary": "MLP, CNN, Transformer models for time-series trading predictions.",
"version": "0.1.1",
"project_urls": {
"Homepage": "https://github.com/SerenaTradingResearch/trading-models"
},
"split_keywords": [
"trading models",
" neural networks",
" time-series",
" mlp",
" cnn",
" transformer"
],
"urls": [
{
"comment_text": null,
"digests": {
"blake2b_256": "8a5b9fccbd488de90b776edbf45358a4dedef5e1dffd7fb903d34de77bfd461b",
"md5": "c110549aec69c491db9a1cc2bffd0044",
"sha256": "aee97ff67028056122df462cfb5605ddb9ad883bc7c194df59edc0a18953c20f"
},
"downloads": -1,
"filename": "trading_models-0.1.1-py3-none-any.whl",
"has_sig": false,
"md5_digest": "c110549aec69c491db9a1cc2bffd0044",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.8",
"size": 4374,
"upload_time": "2025-08-11T15:33:12",
"upload_time_iso_8601": "2025-08-11T15:33:12.071996Z",
"url": "https://files.pythonhosted.org/packages/8a/5b/9fccbd488de90b776edbf45358a4dedef5e1dffd7fb903d34de77bfd461b/trading_models-0.1.1-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": null,
"digests": {
"blake2b_256": "9bfbb40a22769f17da03b2adf7e69be1a27f25703a1151dce1ffd1269a45d2c7",
"md5": "e14afdf99de17c5f02ce65327e6b3061",
"sha256": "35dff9fa30e1b34191ab484021b3ae0dbd0cd7de8cd9f47cbf67f77577489e04"
},
"downloads": -1,
"filename": "trading_models-0.1.1.tar.gz",
"has_sig": false,
"md5_digest": "e14afdf99de17c5f02ce65327e6b3061",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.8",
"size": 4044,
"upload_time": "2025-08-11T15:33:13",
"upload_time_iso_8601": "2025-08-11T15:33:13.026048Z",
"url": "https://files.pythonhosted.org/packages/9b/fb/b40a22769f17da03b2adf7e69be1a27f25703a1151dce1ffd1269a45d2c7/trading_models-0.1.1.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2025-08-11 15:33:13",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "SerenaTradingResearch",
"github_project": "trading-models",
"travis_ci": false,
"coveralls": false,
"github_actions": false,
"requirements": [
{
"name": "torch",
"specs": []
},
{
"name": "matplotlib",
"specs": []
},
{
"name": "numpy",
"specs": []
}
],
"lcname": "trading-models"
}
JSON
