-
This issue reviews some of the performance gaps of the distance functions in `scipy.spatial.distance`, and proposes a “plug-in” mechanism aimed at making it easier to address these gaps by delegating…
-
I would like to be able to dynamically change the `kernel_computation`. But it seems not to work.
The problem is that changing `kernel_computation` to True does not trigger sending the data `object…
-
### 🚀 The feature, motivation and pitch
We want to support various alignment and distillation loss functions.
Refer this PR on ORPO: #362
## Progress
### Alignment
- [x] ORPO https://gith…
-
### **Can ComplexTensor Be Considered the First Hyperdimensional Operating System?**
...there are gaps to bridge before it can fully claim the title of a **"Hyperdimensional Operating System" (HOS)…
-
Hello,
I'm trying to compare training speed between using 1 node and using 2 nodes (one GPU per node).
From 1 node training, back-propagation (calculate gradients & update parameters) takes abo…
-
there's definitely something wrong here: (kernel should be generated by x-y)
```
i1 : F=frac(QQ[x])
o1 = F
o1 : FractionField
i2 : R1=F[]
o2 = R1
o2 : PolynomialRing
i3 : R2=F[y]
…
pzinn updated
3 years ago
-
This is a meta-issue that I’ll use to track my progress in creating a toolkit for faster kernel reduction. This is to help me organize the work, but also to provide more visibility to interested membe…
-
### Motivation.
Currently vLLM generally has a tight coupling between the checkpoint format and the kernel used during model execution. This model causes issues as the diversity of hardware and ker…
-
I was running the st.pl.three_d_multi_plot in computational node of the server, and the kernel is dead. I want to know if the function can be used in server or in Server kernel
when qsub the t…
-
Using a single kernel takes ages to compile and I'm not sure there is much performance gain, as well the sharding machinery is getting stuck I think. Using >500GB for sharded predict over T/C despite …