-
# [RFC] Aten Operators in Triton for Multi-backend support
## Abstract
This RFC discusses
1. the benefits and challenges of developing dispatch functions for Aten operators in Triton.
2. a…
-
## 🐛 Bug
Tracing a model multiple times seems to increase the memory usage.
Reported by [alepack](https://discuss.pytorch.org/u/alepack) in [this post](https://discuss.pytorch.org/t/possible-memor…
-
### Issue type
Performance
### Have you reproduced the bug with TensorFlow Nightly?
Yes
### Source
binary
### TensorFlow version
2.18.0-dev20240919
### Custom code
Yes
### OS platform and di…
-
**Version**
Name: flash-attn
Version: 2.6.3
Name: transformer-engine
Version: 1.11.0+c27ee60
Name: flashattn-hopper
Version: 3.0.0b1
**Bug report**
The bug occurs in this function:
@jit_fuser
def …
-
Hello,
I am trying to train a RL agent in JAX for Asterix-Minatar, however, I get the following error:
` File "/opt/miniconda3/envs/minasa/lib/python3.11/site-packages/gymnax/wrappers/gym.py"…
-
Starting: yolov9s_fp.pt
Opening YOLOv9 model
YOLOv9s summary (fused): 486 layers, 7,167,862 parameters, 0 gradients, 26.7 GFLOPs
Creating labels.txt file
Exporting the model to ONNX
Traceback (…
-
Hi ptitseb
Has box64 a plan to dev a complete tracing jit compiler? so that it can eliminate the redundancy brought by X86 code gen.
current implementation could improve the code layout, but a t…
jhe33 updated
8 months ago
-
I wanted to share the method to export detic model to ONNX format with custom vocabulary.
**Step 1)** First of all, comment out this line `box_features = _ScaleGradient.apply(box_features, 1.0 / se…
-
## 🐛 Bug
I want to convert a Python function that might contain PyTorch calls into a Thunder function inside the lookaside function.
I wasn't successful at using `thunder.core.interpreter.interpre…
-
#57138 demonstrates a case in which checking the trace doesn't work because running the module to trace it modifies some internal state that affects verification. We should see if we can copy/clone th…