spacegoing

What init strategy used when extending 2B model to 5B?

1

### Feature request / 功能建议 Would love to know the team's experience extending 2B model to 5B. Including init methods, training stages etc.

[bug] potential `CPU` memory leak OOM

6

I'm running a 8-node, 64gpu training task. the cuda memory is fine. the problem is with CPU memory: as can be seen from the figure, there seems to be a...

I'm training 93x480p model with videos less than 93 frames, and need to pad frames / mask attention weights. However, https://github.com/PKU-YuanGroup/Open-Sora-Plan/blob/294993ca78bf65dec1c3b6fb25541432c545eda9/opensora/utils/dataset_utils.py#L155-L156 As can be seen here, `collate` function assert no...

Videos are sliced by: https://github.com/PKU-YuanGroup/Open-Sora-Plan/blob/294993ca78bf65dec1c3b6fb25541432c545eda9/opensora/dataset/t2v_datasets.py#L104-L112 and https://github.com/PKU-YuanGroup/Open-Sora-Plan/blob/294993ca78bf65dec1c3b6fb25541432c545eda9/opensora/utils/dataset_utils.py#L120-L159 Yet none of these two parts handles `sp_size` properly, which will leads to assertion fail: https://github.com/PKU-YuanGroup/Open-Sora-Plan/blob/294993ca78bf65dec1c3b6fb25541432c545eda9/opensora/acceleration/communications.py#L124

python layer lsp mode, setting pylsp plugin vars takes no effect

3

I set these vars in dotspacemacs: ``` elisp (defun dotspacemacs/user-config () (with-eval-after-load 'lsp-mode (lsp-register-custom-settings '(("pylsp.plugins.ruff.enabled" t t) ("pylsp.plugins.ruff.executable" "/root/.venv/base/bin/ruff") ("pylsp.plugins.ruff.config" "/root/myCodeLab/GitConfig/python/.config/ruff/ruff.toml") ("pylsp.plugins.ruff.formatEnabled" t t) ("pylsp.plugins.ruff.severities" (("D212" . "I"))) ) ```...