-
Hi,
Thank you for your excellent work! I have some questions when testing the inference efficiency of TransformerBlock and VSSBlock on a V100 gpu.
The input size is set to the same that [16,16,128,…
-
| | |
|--------------------|----|
| Bugzilla Link | [PR27694](https://bugs.llvm.org/show_bug.cgi?id=27694) |
| Status | NEW |
| Importance | P normal |
|…
-
the size of the model trained on my own data is about 165M, and the inference time, including post-processing, is approximately 237ms
-
| | |
| --- | --- |
| Bugzilla Link | [27694](https://llvm.org/bz27694) |
| Version | trunk |
| OS | Linux |
| CC | @sunfishcode |
## Extended Description
I compared the wasm backend's speed to th…
-
_From [lar...@gmail.com](https://code.google.com/u/106380900043315593284/) on October 02, 2012 19:08:50_
It's possible to generate a Levenshtein automaton in O(n), and to run comparison with another …
-
One does not really need to call GMP for these comparisons, so we could save some extra cycles by not calling GMP.
-
I tried the `test_training_speed.py` script and got the following errors (using tensorflow):
```
Build started
Building huffman tree...
100 %Huffman tree build complete
Setting weights...
Done…
-
```
It's possible to generate a Levenshtein automaton in O(n), and to run
comparison with another string in O(n), for a fixed max edit distance.
This paper has more info:
http://citeseerx.ist.psu.…
-
I performed the comparison five times in a row on an Intel i7 machine with 32GB RAM on Ubuntu 22.04, and `speedtest-rs` consistently shows approximately half the download speed of `speedtest-go`. Any…
-
[act.json](https://github.com/user-attachments/files/17709525/act.json)
[dd.json](https://github.com/user-attachments/files/17709526/dd.json)
[ini.json](https://github.com/user-attachments/files/177…