-
Ability to write Prompt in more than 100 languages.
Kandinsky 2.0
https://github.com/ai-forever/Kandinsky-2.0
https://huggingface.co/sberbank-ai/Kandinsky_2.0
Model architecture:
It is a late…
-
Hello, I tried running the pruning example at [examples/pipeline_pruning/pipeline_pruning.py](https://github.com/airaria/TextPruner/blob/main/examples/pipeline_pruning/pipeline_pruning.py). I am not s…
-
I am learning at https://pytorch.org/text/stable/tutorials/sst2_classification_non_distributed.html#sphx-glr-tutorials-sst2-classification-non-distributed-py . Source code Install pre-requistes
```…
-
TorchBench CI has detected a performance signal or runtime regression.
Base PyTorch commit: 8a24a912a5f545d18059b59629aa3598f3783f25
Affected PyTorch commit: 8a24a912a5f545d18059b59629aa3598f3783f2…
-
## 🚀 Feature
Allow users to download only the COMET checkpoint without the XLM-R checkpoint from huggingface hub.
### Motivation
Currently, all encoders in COMET repo are loaded by `XXModel.f…
-
~~~
./torchbench.py --no-skip -k fambench_xlmr
~~~
First error comes while `deepcopy`ing the model. And its a known issue - https://githubhot.com/repo/pytorch/fairseq/issues/4185
After removin…
-
copying fairseq\examples\wav2vec\unsupervised\kaldi_self_train\st\local\prepare_lm.sh -> build\lib.win-amd64-cpython-310\fairseq\examples\wav2vec\unsupervised\kaldi_self_train\st\local
cop…
-
1、初始化代码：
auto_loader = AutoLoader(
task_name="txt_img_matching",
model_dir="./checkpoints",
model_name="AltCLIP-XLMR-L" # Load the checkpoints from Modelhub(model.baai.ac.cn/models)…
-
TorchBench CI has detected a performance signal.
Affected Tests:
- eval-cuda-fp32:
- attention_is_all_you_need_pytorch[dynamo-blade (latency)] 5.893 -> 6.688, -13.4906%
- attention_is_all_you_nee…
-
TorchBench CI has detected a performance signal.
Affected Tests:
- eval-cuda-fp32:
- attention_is_all_you_need_pytorch[dynamo-blade (latency)] 5.893 -> 6.611, -12.1839%
- attention_is_all_you_nee…