SoundProvider

[bug] deepspeed zero++ multinode with liger kernel

2

### 🐛 Describe the bug #### deepspeed zero++ config - I ran the training with slrum ``` { "zero_optimization": { "stage": 3, "stage3_gather_16bit_weights_on_model_save": true, "reduce_bucket_size": "auto", "zero_hpz_partition_size": 8, "zero_quantized_weights": true,...

**Describe the bug** Hello thank you for developing Nemo framework. I'm testing `examples/llm/finetune/automodel.py` with llama checkpoint(`NousResearch/Llama-2-7b-hf`). ### baseline - running script ```bash #!/bin/bash DEVICES=8,9,10,11 CUDA_VISIBLE_DEVICES=${DEVICES} torchrun --nproc-per-node=4 \ examples/llm/finetune/automodel.py \...