Philip Turner
Philip Turner
In https://github.com/openmm/openmm/issues/3937, I was discussing a hypothetical hipSYCL backend for OpenMM. The framework uses JIT compilation extensively, mostly to implement custom force fields (injecting code blobs into otherwise already-known shaders)....
This is a progress tracker for implementation of the hipSYCL backend. I figured it's most appropriate to extract it into a separate thread, and not flood #460. Preliminary Research: -...
I've [reverse-engineered](https://github.com/philipturner/metal-benchmarks) the M1 GPU entirely, and it has some *really* small cache sizes. L1D = 8 KB, L1I = 12 KB, L2 = 256 KB (M1 Pro), 512 KB...
I was working on a large structural piece (600,000 atoms, 50 nm x 120 nm), and found a bottleneck in my workflow. I am trying to utilize a GPU cluster...
For smaller models, quantization causes more quality loss than large models. Could the repository try 6-bit / 128 groups for stuff like LLaMa-7B? This could be most useful for some...
| | | |------------------|-----------------| |Previous ID | SR-15793 | |Radar | None | |Original Reporter | @philipturner | |Type | Bug | Additional Detail from JIRA | | | |------------------|-----------------|...
| | | |------------------|-----------------| |Previous ID | SR-15823 | |Radar | None | |Original Reporter | @philipturner | |Type | Bug | Attachment: [Download](https://user-images.githubusercontent.com/2727770/164964449-0329857b-939e-43da-bdff-233537fcc6f7.gz) Additional Detail from JIRA | |...
https://github.com/philipturner/ARHeadsetKit
Need a bit of cleaning up before it’s mergeable, but here is the code. Resolves #4
This repository could be much faster by encoding each frame in parallel. This is necessary to achieve acceptable latency in a production renderer, where videos are currently encoded in the...