Properly build specific Torch image

This commit is contained in:
Matthias 2023-04-08 16:59:17 +02:00
parent 69b9b35a08
commit bed51fa790
3 changed files with 10 additions and 2 deletions

View File

@ -12,6 +12,7 @@ TAG=$(echo "${BRANCH_NAME}" | sed -e "s/\//_/g")
TAG_PLOT=${TAG}_plot
TAG_FREQAI=${TAG}_freqai
TAG_FREQAI_RL=${TAG_FREQAI}rl
TAG_FREQAI_TORCH=${TAG_FREQAI}torch
TAG_PI="${TAG}_pi"
TAG_ARM=${TAG}_arm
@ -84,6 +85,10 @@ docker manifest push -p ${IMAGE_NAME}:${TAG_FREQAI}
docker manifest create ${IMAGE_NAME}:${TAG_FREQAI_RL} ${CACHE_IMAGE}:${TAG_FREQAI_RL} ${CACHE_IMAGE}:${TAG_FREQAI_RL_ARM}
docker manifest push -p ${IMAGE_NAME}:${TAG_FREQAI_RL}
# Create special Torch tag - which is identical to the RL tag.
docker manifest create ${IMAGE_NAME}:${TAG_FREQAI_TORCH} ${CACHE_IMAGE}:${TAG_FREQAI_RL} ${CACHE_IMAGE}:${TAG_FREQAI_RL_ARM}
docker manifest push -p ${IMAGE_NAME}:${TAG_FREQAI_TORCH}
# copy images to ghcr.io
alias crane="docker run --rm -i -v $(pwd)/.crane:/home/nonroot/.docker/ gcr.io/go-containerregistry/crane"
@ -93,6 +98,7 @@ chmod a+rwx .crane
echo "${GHCR_TOKEN}" | crane auth login ghcr.io -u "${GHCR_USERNAME}" --password-stdin
crane copy ${IMAGE_NAME}:${TAG_FREQAI_RL} ${GHCR_IMAGE_NAME}:${TAG_FREQAI_RL}
crane copy ${IMAGE_NAME}:${TAG_FREQAI_RL} ${GHCR_IMAGE_NAME}:${TAG_FREQAI_TORCH}
crane copy ${IMAGE_NAME}:${TAG_FREQAI} ${GHCR_IMAGE_NAME}:${TAG_FREQAI}
crane copy ${IMAGE_NAME}:${TAG_PLOT} ${GHCR_IMAGE_NAME}:${TAG_PLOT}
crane copy ${IMAGE_NAME}:${TAG} ${GHCR_IMAGE_NAME}:${TAG}

View File

@ -9,7 +9,6 @@ TAG=$(echo "${BRANCH_NAME}" | sed -e "s/\//_/g")
TAG_PLOT=${TAG}_plot
TAG_FREQAI=${TAG}_freqai
TAG_FREQAI_RL=${TAG_FREQAI}rl
TAG_FREQAI_RL=${TAG_FREQAI}torch
TAG_PI="${TAG}_pi"
PI_PLATFORM="linux/arm/v7"
@ -66,7 +65,6 @@ docker build --build-arg sourceimage=freqtrade --build-arg sourcetag=${TAG_FREQA
docker tag freqtrade:$TAG_PLOT ${CACHE_IMAGE}:$TAG_PLOT
docker tag freqtrade:$TAG_FREQAI ${CACHE_IMAGE}:$TAG_FREQAI
docker tag freqtrade:$TAG_FREQAI_RL ${CACHE_IMAGE}:$TAG_FREQAI_RL
docker tag freqtrade:$TAG_FREQAI_RL ${CACHE_IMAGE}:$TAG_FREQAI_TORCH
# Run backtest
docker run --rm -v $(pwd)/config_examples/config_bittrex.example.json:/freqtrade/config.json:ro -v $(pwd)/tests:/tests freqtrade:${TAG} backtesting --datadir /tests/testdata --strategy-path /tests/strategy/strats/ --strategy StrategyTestV3

View File

@ -254,6 +254,7 @@ freqtrade trade --config config_examples/config_freqai.example.json --strategy F
### Structure
#### Model
You can construct your own Neural Network architecture in PyTorch by simply defining your `nn.Module` class inside your custom [`IFreqaiModel` file](#using-different-prediction-models) and then using that class in your `def train()` function. Here is an example of logistic regression model implementation using PyTorch (should be used with nn.BCELoss criterion) for classification tasks.
```python
@ -322,6 +323,7 @@ class MyCoolPyTorchClassifier(BasePyTorchClassifier):
```
#### Trainer
The `PyTorchModelTrainer` performs the idiomatic PyTorch train loop:
Define our model, loss function, and optimizer, and then move them to the appropriate device (GPU or CPU). Inside the loop, we iterate through the batches in the dataloader, move the data to the device, compute the prediction and loss, backpropagate, and update the model parameters using the optimizer.
@ -330,6 +332,7 @@ In addition, the trainer is responsible for the following:
- converting the data from `pandas.DataFrame` to `torch.Tensor`.
#### Integration with Freqai module
Like all freqai models, PyTorch models inherit `IFreqaiModel`. `IFreqaiModel` declares three abstract methods: `train`, `fit`, and `predict`. we implement these methods in three levels of hierarchy.
From top to bottom:
@ -340,6 +343,7 @@ From top to bottom:
![image](assets/freqai_pytorch-diagram.png)
#### Full example
Building a PyTorch regressor using MLP (multilayer perceptron) model, MSELoss criterion, and AdamW optimizer.
```python