Comment
throughput/tflops
throughput/tflops
MODEL_SIZE: GPT33B, machine: ThetaGPU, world_size: 32, env.SP_TYPE: megatron, micro_batch_size: 1, seq_length: 300000, env.GAS: 1, global_batch_size: 1, zero_stage: 1, env.MPSIZE: 32, env.PPSIZE: 1, env.SPSIZE: 1, use_flash_attn: true
MODEL_SIZE: GPT33B, machine: ThetaGPU, world_size: 32, env.SP_TYPE: megatron, micro_batch_size: 1, seq_length: 288000, env.GAS: 1, global_batch_size: 1, zero_stage: 1, env.MPSIZE: 32, env.PPSIZE: 1, env.SPSIZE: 1, use_flash_attn: true
MODEL_SIZE: GPT33B, machine: ThetaGPU, world_size: 32, env.SP_TYPE: megatron, micro_batch_size: 1, seq_length: 200000, env.GAS: 1, global_batch_size: 1, zero_stage: 1, env.MPSIZE: 32, env.PPSIZE: 1, env.SPSIZE: 1, use_flash_attn: true
MODEL_SIZE: GPT33B, machine: ThetaGPU, world_size: 32, env.SP_TYPE: megatron, micro_batch_size: 1, seq_length: 192000, env.GAS: 1, global_batch_size: 1, zero_stage: 1, env.MPSIZE: 32, env.PPSIZE: 1, env.SPSIZE: 1, use_flash_attn: true
Run set
5
Add a comment
Created with ❤️ on Weights & Biases.
https://wandb.ai/l2hmc-qcd/Megatron-DS-Benchmarking/reports/throughput-tflops-23-09-16-07-28-43---Vmlldzo1NDExODEx