fix: skip SP for gemma and qwen plans

Open akoumpa opened this issue 1 month ago • 4 comments

Nov 21 '25 06:11 akoumpa

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

Nov 21 '25 06:11 copy-pr-bot[bot]

/ok to test 830b528173e6ecb5996f963995b9bce9596d9dc0

Nov 21 '25 06:11 akoumpa

@akoumpa

As discussed offline:

qwen3(Qwen3ForCausalLM) --> fails with SP
qwen2(Qwen2ForCausalLM) --> works with SP

We can either keep qwen2 or bypass the following functional test https://github.com/NVIDIA-NeMo/Automodel/blob/main/tests/functional_tests/hf_transformer_finetune/L2_HF_Transformer_PEFT_Benchmark_qwen2_custom.sh

Nov 21 '25 18:11 ZhiyuLi-Nvidia

/ok to test 229bff581ff30d4f37a6d04e68d848fe3c291946

Nov 21 '25 21:11 akoumpa