super-dainiu
super-dainiu
### What's new? After a discussion with my collaborator @Cypher30, I figured out my misinterpretation of algorithm 3 in https://arxiv.org/abs/1604.06174. This PR can be viewed as an amendment to this...
### What's new? Since all data in `OperationData` are meta tensors, their `tensor.data_ptr()=0`. I replace the original `tensor.data_ptr()` with a customized function generated by `uuid4()`. ### Why data_ptr? Generally speaking,...
### What's new A new version of `ColoTracer`. Currently supports: - [x] torchvision - [x] timm - [x] huggingface - [x] albert - [x] bert - [x] diffusion - [x]...
### What's new? Add more items for meta registrations. e.g. - aten::_cudnn_rnn - aten::_cudnn_rnn_backward - Tensor.cpu() - Tensor.cuda()
This is a draft. ### Usage - [x] Standalone ``MetaInfoProp`` - [ ] Standalone ``CheckpointSolver`` ### Remove redundant code - [x] Remove algorithms #2312 - [ ] Remove tests ###...
Things to hack: - [x] torch.Tensor factory function - [x] eager-mode - [x] nn.Parameter - [x] materialize & traceable ### New features - A naive LazyTensor - LazyInit for a...
### 🐛 Describe the bug A list of `torch.Tensor` on `device='meta'` cannot be printed by `DistributedLogger` ```python logger = get_dist_logger() logger.info(f"{[torch.tensor([1,2,3], device='meta')]}") ``` ### Environment _No response_
## 📌 Checklist before creating the PR - [x] I have created an issue for this PR for traceability - [x] The title follows the standard format: `[doc/gemini/tensor/...]: A concise...
# Analyzer # Overview The Analyzer is a collection of static graph utils including Colossal-AI FX. Features include: - MetaTensor -- enabling: - Ahead-of-time Profiling - Shape Propagation - Ideal...
Dear Authors, I am amazed that the performance of POET outperforms POFO so much. Since PyTorch does not have a traced backward graph, I am curious about which framework you...