gpu

🚀 The feature, motivation and pitch

The current implementation of Zero Redundancy optimizer has its own implementation of object broadcasting.

We should replace it with c10d [broadcast_object_list](https://pytorch.org/docs/stable/distributed.html#torch.distributed.broadcast_object_lis

This is a nice first issue:

Add types/docments to the Learner.get_preds function. This function is essential to any fastai user and has almost no documentation. Add types and text to the variables as we have in many places now.

The taichi.lang.util.warning function just prints the warning without consulting the current state of pythons standard library warnings module.

For example:

import warnings
import taichi as ti

with warnings.catch_warnings():
    warnings.simplefilter("ignore

At this moment relu_layer op doesn't allow threshold configuration, and legacy RELU op allows that.
We should add configuration option to relu_layer.

Is your feature request related to a problem? Please describe.
On MacOS, if Neovide is in the background, and I click on its window, that brings Neovide to foreground and also moves the cursor to where I clicked.

Describe the solution you'd like
On MacOS, it's common for apps to receive focus via mouse click, without the click being interpreted as an interaction with the UI elemen

Hi ,

I have tried out both loss.backward() and model_engine.backward(loss) for my code. There are several subtle differences that I have observed , for one retain_graph = True does not work for model_engine.backward(loss) . This is creating a problem since buffers are not being retained every time I run the code for some reason.

Please look into this if you could.

Problem:

_catboost.pyx in _catboost._set_features_order_data_pd_data_frame()

_catboost.pyx in _catboost.get_cat_factor_bytes_representation()

CatBoostError: Invalid type for cat_feature[non-default value idx=1,feature_idx=336]=2.0 : cat_features must be integer or string, real number values and NaN values should be converted to string.

Could you also print a feature name, not o

Our users are often confused by the output from programs such as zip2john sometimes being very large (multi-gigabyte). Maybe we should identify and enhance these programs to output a message to stderr to explain to users that it's normal for the output to be very large - maybe always or maybe only when the output size is above a threshold (e.g., 1 million bytes?)

Description

https://numpy.org/doc/stable/reference/generated/numpy.corrcoef.html

https://docs.cupy.dev/en/stable/reference/generated/cupy.corrcoef.html

Seems args are different

Additional Information

dtype argument added in NumPy version 1.20.

Is your feature request related to a problem? Please describe.
In time series plotting module, lot of plots are customized at the end - template, fig size, etc. Since the same code is repeated in all these plots, maybe this could be modularized and reused.

with fig.batch_update():
    template = _resolve_dict_keys(
        dict_=fig_kwargs, key="template", defaults=fig_default

https://github.com/brendangregg/FlameGraph

There are a number of items in wgpu whose documentation contain examples using GLSL syntax or other references to GLSL elements. Since WGSL is now the standard shading language for WebGPU, it would be beneficial to readers if these examples were presented first in WGSL. (Keeping the GLSL would still be helpful for new users arriving from WebGL.)

Relevant occurrences of the text "GLSL" in docu

Now we are going to have set operations (rapidsai/cudf#11043). To be consistent with other libraries/framework (like Presto: https://prestodb.io/docs/current/functions/array.html), we should rename lists::drop_list_duplicates into lists::distinct. The implementation should be moved into set_operations.hpp|cu to be easily located and for consistency, as mentioned above

如何导入pytorch训练好的模型或者权重文件？