Getting the following error:
I’m trying to profile llama3-8b model training in a docker container (nvcr.io/nvidia/pytorch 24.12-py3) with root enabled using llama-cookbook
This is the command I’m running:
nsys profile -t cuda,opengl,nvtx,osrt -w true -x true -o out/llama_8b_h100_singlegpu --sample=none --cpuctxsw=none --gpu-metrics-devices=all --cuda-memory-usage=true -y 60 --duration 600 --wait all python -m finetuning --use_peft --peft_method lora --quantization 4bit --quantization_config.quant_type nf4 --batch_size_training 1 --context_length 2048 --dataset samsum_dataset --model_name “meta-llama/Llama-3.1-8B” --output_dir out/
Can anyone please help me fix this issue?