Torch
is a scientific computing framework for LuaJIT.
- comes with a large ecosystem of community-driven packages in ML, CV, etc
- is easy to use and efficient
- can build arbitrary graphs of neural networks, and parallelize them over CPUs and GPUs in an efficient manner
- Go through Torch documentation and try using useful functions.
- Go through other useful modules and try to use them.
- Start implementing something real
- Read and process process data as needed by the network.
- Read
Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation
and implement a GRU. - Read
LSTM: A Search Space Odyssey
and implement LSTM. - Use Tree LSTM in place of LSTM.
Paper
Code
Slides
- Go through
Penlight
documentation
- Go through
Moses
- Go through
nn
module documentation and check how some to the functionality is implemented