-
It seems that where there is a strong correlation between the style and inpu…
-
**Describe the bug**
We are experiencing an issue where jobs scheduled to run at regular intervals (e.g., once per minute using cron syntax * * * * *) occasionally trigger far too many times, seeming…
-
In case user passes in interleaved tensor for conv activations, we should automatically select best sharding scheme given input tensors for the convolution.
Primary motivation of this work is to reduc…
-
-
# Activations
激活函数可以通过设置单独的激活层实现，也可以在构造层对象时通过传递activation参数实现
## 预定义激活函数
- [ ] softmax
- [ ] elu
- [ ] selu
- [ ] softplus
- [ ] softsign
- [ ] relu
- [ ] tanh
- [ ] sigmoid
- [ ] hard_s…
-
The initial task + namespace pull request had feedback about deadlines not being fully implemented. To help reduce friction in that review and get some changes landed. I removed the partially complete…
-
TODO decide on approach
**Given** I have an environment eval'd in my `.zshrc`
**I need** to add `--start-services`
**So I can** have services running in my default terminal
-
As explained in the function reformatting [open task](https://lets-unify.ai/docs/ivy/overview/contributing/open\_tasks.html#function-formatting), each function in the [activations](https://github.com/…
-
## Issue
Currently the conv2d API expects channels last activations, this is totally fine and expected, but then it is mixed with channels first weights layout (i.e. PyTorch style weights).
## Pro…
-
`get_activations = K.function([model.layers[0].input, K.learning_phase()],[model.layers[0].output,])
activations = get_activations([trainFeatures,0])[0]`
ghost updated
2 years ago