Search results

Expert parallelism

Status: Open.

#34 In huggingface/picotron;

· kiankyars opened on Jun 22, 2025

Why do we need to call <code>init_model_parameters</code> after <code>load_state_dict</code>

Status: Open.

#32 In huggingface/picotron;

· GeeeekExplorer opened on May 13, 2025

No such file or directory: "./hf_model_safetensors/model.safetensors"

Status: Open.

#30 In huggingface/picotron;

· SeekPoint opened on Apr 21, 2025

can we support 2080ti?

Status: Open.

#29 In huggingface/picotron;

· SeekPoint opened on Apr 21, 2025

Expert parallelism

Status: Open.

#19 In huggingface/picotron;

· sutyum opened on Jan 7, 2025

Change process group order to optimize inter-node communication

Status: Open.

#17 In huggingface/picotron;

· cailun01 opened on Dec 22, 2024

suggestion: train.py启动参数优化

enhancementNew feature or requestNew feature or request

Status: Open.

#14 In huggingface/picotron;

· yuanzhoulvpi2017 opened on Dec 19, 2024

suggestion: import的时候，改为相对引用

enhancementNew feature or requestNew feature or request

Status: Open.

#13 In huggingface/picotron;

· yuanzhoulvpi2017 opened on Dec 19, 2024

Suggestion: Integrate DeepSpeed-Ulysses with Head Dimensional Splitting to Form a 5D Parallelism Scheme

featuresNew featuresNew features

Status: Open.

#12 In huggingface/picotron;

· feifeibear opened on Dec 19, 2024