-
**Is your feature request related to a problem? Please describe.**
I see quite some GPT-2 implementations, but I am missing a GPT-Neo/GPT-J implementation. This is a variant of GPT-2 which is quite s…
-
Observed while debugging opt-counter/compactor crash:
```
At commit 0b46639335a8d6203bb810fc6e835936f441a27b
out/ReleaseX64/dart --code-collection-interval-in-us=30 --log-code-drop --always-dro…
-
In Glow, the quantization is done in two passes:
- profile generation
- quantization profile loading, and quantized model inference
Our understanding is that this mechanism works with the assumpt…
-
CompiledClassMetadataFactory was added in #29117 by @fbourigault, but it's not used anywhere in the codebase.
I'm wondering if we shouldn't deprecate it (and it's related classes, like the cache wa…
-
## Problem:
Given an FC operator with the input of [M;K], weights of [K;N] and bias of [N] dimensions.
For certain ASIC architectures it's beneficial to split weights into several chunks and distrib…
-
* Generate docstrings for API documentation
* Create separate `client/` folder for `RalfClient`
* Remove schemas
* Set primary key (or select columns) for Tables
* Move simulation code to sepa…
-
Hi @wenbin-lin,
Thanks for releasing the motion completion module in OcclusionFusion.
I had a doubt regarding the implementation of graph-based ARAP deformation. The paper mentions the use of ARAP…
-
Code coverage reports from the VM are based on invocation counters on functions (Function) and call sites (ICData). If the invocation counter is non-zero, the position is a hit, otherwise a miss. When…
-
Let's start with an example. Building a query as below
```c#
QueryBuilder.SelectAll()
.Graph("urn:graph:uri", c => c.Where(t => t.Subject("s").Predicate("p").Object("o")))
.Filter(e => e.Var…
-
I'm using fx2ait to load an onnx graph. After optimization, the results are not good.
BS: 1, PT Eager time per iter: 0.01654841552734375ms, PT Eager QPS: 60.43, FX2AIT time per iter: 0.02410858642…