From 330e46e9e808ba7106e9618da9dd9790dd404cbd Mon Sep 17 00:00:00 2001 From: sysuyy <2607384125@qq.com> Date: Mon, 11 Mar 2024 14:45:23 +0000 Subject: [PATCH] [feat]: use accelerate on multi-node Former-commit-id: 0b12fc31f1d2a63f9da38821174ab63d9b3a2d72 --- scripts/accelerate_configs/hostfile | 2 ++ .../accelerate_configs/multi_node_example.yaml | 18 ++++++++++++++++++ 2 files changed, 20 insertions(+) create mode 100644 scripts/accelerate_configs/hostfile create mode 100644 scripts/accelerate_configs/multi_node_example.yaml diff --git a/scripts/accelerate_configs/hostfile b/scripts/accelerate_configs/hostfile new file mode 100644 index 000000000..5e384ed0f --- /dev/null +++ b/scripts/accelerate_configs/hostfile @@ -0,0 +1,2 @@ +server1 slots=8 # your server name and GPU in total +server2 slots=8 diff --git a/scripts/accelerate_configs/multi_node_example.yaml b/scripts/accelerate_configs/multi_node_example.yaml new file mode 100644 index 000000000..c462d62cc --- /dev/null +++ b/scripts/accelerate_configs/multi_node_example.yaml @@ -0,0 +1,18 @@ +compute_environment: LOCAL_MACHINE +distributed_type: DEEPSPEED +deepspeed_config: + deepspeed_config_file: scripts/accelerate_configs/zero3_offload.json + deepspeed_hostfile: # FullPath here +fsdp_config: {} +machine_rank: 0 +main_process_ip: # main process ip +main_process_port: 29501 +main_training_function: main +num_machines: 2 +num_processes: 16 +rdzv_backend: static +same_network: true +tpu_env: [] +tpu_use_cluster: false +tpu_use_sudo: false +use_cpu: false