-
I have been trying to fix this error for a while now, and the ongoing threads are of NO help.
I have checked these (and ALL issue on the HF community page for this model):
* https://github.com/Qwe…
-
have a try to reproduce Nvidia's results on using slurm + enroot + pyxis
1. downgrade the transformers and huggingface_hub libs (huggingface_hub==0.23.2 transformers==4.40.2) because the versions…
-
[`aten.sym_size.int`](https://pytorch.org/docs/stable/torch.compiler_ir.html) returns a `SymInt` representing the size at the given dimension. This op does not return a tensor. What should we do wit…
-
I have an ONNX model which takes one float tensor and three int tensors as input. The inference session works well in python version of onnxruntime. However it cannot work with "2.0.0-rc.6" version of…
-
pose_aa_h1 = torch.cat([torch.zeros((1, 1, 3)), H1_ROTATION_AXIS * dof_pos[..., None], torch.zeros((1, 2, 3))], axis = 1)
RuntimeError: The size of tensor a (19) must match the size of tensor b (…
-
### Overview ###
This issue is a tracker for the next steps of enabling the data tiling fusion improvements, which are currently behind the `--iree-dispatch-creation-experimental-data-tiling` flag. O…
-
I'm trying to finetune **Mistral-Nemo-Base-2407** with a `text` dataset of long inputs. Usually, the SFTrainer will truncate it to fit the specified context size.
However, I get an error when using…
-
-
### bug描述 Describe the Bug
File "/data/mlops/Open-Assistant/inference/server/oasst_inference_server/plugins/vectors_db/loaders/data_loader.py", line 383, in path_to_doc1
res = file_to_doc(file, …
-
When the `Triton::MakeTensorPtrOp` has to be rewritten by the `RewriteTensorPointer` pass to use "regular" memory operations, the generated code seems less performant than a code directly written usin…