Age | Commit message (Expand) | Author |
---|---|---|
2017-10-25 | dynamic gpu shape | Marcin Junczys-Dowmunt |
2017-10-22 | swish | Marcin Junczys-Dowmunt |
2017-10-19 | Re-autoformat after merge | Roman Grundkiewicz |
2017-10-19 | Merge with 'master' branch | Roman Grundkiewicz |
2017-10-19 | uses memory allocator for temp memory in kernels | Marcin Junczys-Dowmunt |
2017-10-15 | Autoformat rest of the code | Roman Grundkiewicz |
2017-10-05 | towards beam search with transformer model | Marcin Junczys-Dowmunt |
2017-09-23 | more changes for transformer model | Marcin Junczys-Dowmunt |
2017-09-19 | more complrex transformer | Marcin Junczys-Dowmunt |
2017-09-17 | better hashing | Marcin Junczys-Dowmunt |
2017-09-10 | towards a transformer-style encoder | Marcin Junczys-Dowmunt |
2017-09-02 | fix rows op | Marcin Junczys-Dowmunt |
2017-07-16 | working aligned memory allocation | Marcin Junczys-Dowmunt |
2017-07-02 | merge with master | Marcin Junczys-Dowmunt |
2017-06-22 | Remove debug from Convolution | Tomasz Dwojak |
2017-06-22 | Add conv operator | Tomasz Dwojak |
2017-06-12 | integrate memory allocator | Marcin Junczys-Dowmunt |
2017-06-05 | Autoformat .h and .cpp files | Roman Grundkiewicz |
2017-06-02 | cleaned up code, more separation of gpu implementation and headers | Marcin Junczys-Dowmunt |
2017-04-28 | add scalar operators | Marcin Junczys-Dowmunt |
2017-04-23 | better memory handling | Marcin Junczys-Dowmunt |
2017-04-20 | revert to single parameter, fix derivative | Marcin Junczys-Dowmunt |
2017-04-20 | trainable hyperparameter eps | Marcin Junczys-Dowmunt |
2017-04-20 | backward step for lexprobs | Marcin Junczys-Dowmunt |
2017-04-19 | lexprobs | Marcin Junczys-Dowmunt |
2017-04-15 | cleaner batching, operator for subselecting columns | Marcin Junczys-Dowmunt |
2017-04-15 | more clean-up | Marcin Junczys-Dowmunt |
2017-02-24 | fixed residual connections, added hashing of nodes to identify duplicate nodes | Marcin Junczys-Dowmunt |
2017-02-15 | small changes in mean node | Marcin Junczys-Dowmunt |
2017-02-13 | bn | Marcin Junczys-Dowmunt |
2017-02-13 | batch normalization | Marcin Junczys-Dowmunt |
2017-02-03 | cleaned-up setCudaDevice | Marcin Junczys-Dowmunt |
2017-01-29 | works on different gpus | Marcin Junczys-Dowmunt |
2017-01-25 | refactored layers | Marcin Junczys-Dowmunt |
2017-01-23 | major refactorting | Marcin Junczys-Dowmunt |
2017-01-21 | faster GRU | Marcin Junczys-Dowmunt |
2017-01-21 | sorted files into folders | Marcin Junczys-Dowmunt |