-
# Motivation
The goal is to enable Kineto to support XPU pathways without altering its current design and with minimal maintenance costs. This will allow Kineto's profiler, designed and implemente…
-
Is there an explanation of this error?
```
/usr/local/lib/python3.8/dist-packages/poptorch/experimental.py in __exit__(self, exc_type, value, traceback)
253 if self._compile_using ==…
-
Hi there,
I have the latest sdk installed, graphcore has been updated to 1.5.x (for reason it insisted on that), the odd thing is that on some machines it still wants to see 1.3.1 instead of the 1.…
-
@franko wrote in #4
> Probably what is important for LGSL on windows is to provide:
> - a set of working binaries
> - install informations for the expert users that may wish to use its own BLAS libra…
-
Our current `ipu_eigh` uses the Jacobi algorithm. It is believed for other hw accelerators that the QR algorithm becomes faster than jacobi for larger `d>=512`. Since we are targeting `d>=512` we are …
-
```class CallGraph(v_graphcore.HierGraph):
'''
A graph which represents procedural branches.
'''
def __init__(self):
v_graphcore.Graph.__init__(self)
```
it's defin…
-
### Search before asking
- [X] I have searched the YOLOv5 [issues](https://github.com/ultralytics/yolov5/issues) and found no similar feature requests.
### Description
I have recently been working…
-
Come up with a name for the struct also. Something like GraphFile?
The function for creating the database for tests will also need to be updated.
-
### Idea
Use int4 as the compression technique to fit larger models onto Navi machines or possibly MI series machines. Weights would be compressed using encoding scheme that would pack two 4 bits n…
-
https://news.ycombinator.com/item?id=32610780
Chip Startups could need compilers!