Clement Chan
Clement Chan
I went through the code about spectrogram extraction in the code base and found some duplication at padding. For a simple case, applying stft to an audio, it is common...
### PR types Bug fixes ### PR changes Others ### Describe 1. change the codegen code to avoid conversion from heterogeneous 'initializer list' to tuple, which fails on gcc 5.4;...
Glad to find global writer has been added to tensorboardx. It means alot to me: With a global writer, easily log without passing the summary writer and global step as...
I've noticed a detail in decoder prenet that, it also uses dropout at inference. There is a comment saying that "# use dropout also in inference for positional encoding relevance"....
I noticed that in the documentation, there is a session "Intermediate result" in "common pitfalls" . Using intermediate variables in a function will cause an error because the `tmp` is...
update the API to fluid v1.1 1. Use `ParallelExecutor` instead of deprecated `fluid.layers.get_places` and `fluid.layers.ParallelDo`. 2. Use `paddle.batch` instead of `paddle.v2.batch`, which drops the only batch by default even when...
按照 README 中的脚本去进行 eval 的时候会出现 ``` Traceback (most recent call last): File "sentiment_classify.py", line 288, in main(args) File "sentiment_classify.py", line 273, in main args.model_path) File "sentiment_classify.py", line 210, in eval_net...
The code of softmax below is coppied from tutorials to demonstrate that we cannot pass tensors on devices other than "cuda:0" to triton kernel. Errors are: ```text ValueError: Pointer argument...
support grouped query attention(GQA) for flash_attn(related kernels: fwd, bwd, split_kv, total_attention) The GQA paper > Ainslie, Joshua, James Lee-Thorp, Michiel de Jong, Yury Zemlyanskiy, Federico Lebrón, and Sumit Sanghai. “GQA:...
I notice that it is not possible to pass a JITFunction as the parameter to another JITFunction(just call it higher order JITFunction for now). The code below is an example...