Sherman Siu

icon location Current MMath (CS) student at the University of Waterloo. BMath (CS), minor in stats, minor in combinatorics and optimization. Graduated with distinction.

Results 94 comments of Sherman Siu

Adding support for encoder-decoder models, like T5 or BART

I see you've already answered this in the FAQ here: https://vllm.readthedocs.io/en/latest/models/adding_model.html

Adding support for encoder-decoder models, like T5 or BART

Also, I suppose the encoder cache eviction would be different. i.e. The encoder's cross-attention values would need to be kept as long as the decoding is active for a prompt,...

Adding support for encoder-decoder models, like T5 or BART

(Never mind, for the sake of simplicity, LRU should work just fine)

Pandarallel very slow after loading huge dataframe

Can you check your RAM usage with `htop`? It's possible that your computer slowed down because it was holding a huge dataframe in-memory.

consider taking in global environment variable for nb_workers and possibly other parameters too

I'd prefer to keep the code in the core package minimal to make things easier to maintain. Wouldn't you be able to maintain the same functionality by reading in the...

Pandarallel could stuck without raising any errors when using all the physical cores

I'm glad that your code works if you leave out 4 of your 20 cores! Nevertheless, it will be difficult to fix this issue if you don't provide some minimal...

parallel_apply not working with Pandas >= 2.1

It works just fine for me on Pandas 2.1. Do you have a minimal code example to reproduce your bug? Python: 3.10.13 Pandarallel: 1.6.5 Pandas: 2.1.0 Ubuntu 22.04 ```python import...

Cannot use all CPUs at the same time

"The code gets stuck when using more than 1 CPU." I can't reproduce this, even when using `psutil.cpu_count(logical=False)` workers. Do you have some minimally-working Pandas code that lags when you...

OSX can not parallel urllib

The code snippet works for me in Jupyter as expected, even when `use_memory_fs=False`. Python: 3.10.13 Pandarallel: 1.6.5 Pandas: 2.2.0 Jupyter core: 5.7.1 Jupyterlab: 4.1.0 I'm on Linux though.

fix DataFrame.applymap has been deprecated.

Unless we're dropping support for Pandas versions < 2.1, I'm against this, especially since Pandas 2.1 was released relatively recently. There's also no reason to change the minimum Python version...