-
在精簡深度學習的方式上 : 卷積類神經 (CNN) 採用像素遠近,而遞歸類神經 (RNN) 採用著則是時間遠近
那麼,既然有著類似的設計精神,兩者是否有可能互換應用呢?
-
Please study the RNN model and its important variations (e.g. attention mechanism, transformer) carefully, then using your own idea (not replicate existing paper) to implement it for our bus arrival t…
-
@nicholas-leonard
I am trying to predict the next number in a sequence, where each sequence is of length say 5.
For example: input is `{1,2,3,4,5}`
target is `{2,3,4,5,6}`
Trainin…
-
Hi.
I was trying to work though your example here:
https://github.com/kieranbrowne/clojure-tensorflow/blob/master/examples/rnn/src/rnn/core.clj#L29
however I hit the error of `Unable to resolve…
-
Hi @jcjohnson,
So I work on the [Element-Research/rnn](https://github.com/Element-Research/rnn) package. I have a couple of questions:
1. why did you chose to build your own LSTM/RNN implementations…
-
RWKV is an RNN with Transformer-level LLM performance, which can also be directly trained like a GPT transformer (parallelizable). And it's 100% attention-free. You only need the hidden state at posit…
-
In clang tidy(from rocm 4.1), there is a warning about the functions in rewrite_rnn that exceed the cognitive complexity. There is description of how the congnitive complexity is caclulated [here](htt…
-
- MRU / MFU
- NB (individual)
- DNN (user, time vector together)
- RNN + Attention only (user, time vector together)
- Markov chain (individual)
-
I am getting valuerror on my Juypter Lab and Vscode. This only work on Coblab.
ValueError: Unrecognized keyword arguments passed to Embedding: {'batch_input_shape': [32, None]}
def build_model(…
-
segfaults when training rnn's with text. compiled only with debugging enabled. testing previous commits produces same results. tried tjluyao-master branch, works fine, no segfault. generating text …