[Cleanup] Remove libuv from run_llama_train.sh

libuv is now enabled by default. we can proably do without the educational blurb there, and don't need the env either since the default has landed. ghstack-source-id: 68c8d2abe7eb0777e2add8df7634367c31b7ec06 Pull Request resolved: pytorch#453
YerevaNN · Jul 15, 2024 · 535acf6 · 535acf6
1 parent 0ddf49b
commit 535acf6
Show file tree

Hide file tree

Showing 3 changed files with 0 additions and 5 deletions.
diff --git a/create_seed_checkpoint.sh b/create_seed_checkpoint.sh
@@ -18,7 +18,6 @@
 
 set -ex
 
-export USE_LIBUV=1
 TRAINER_DIR=${1:-/home/$USER/local/torchtitan}
 NGPU=1
 LOG_RANK=0

diff --git a/multinode_trainer.slurm b/multinode_trainer.slurm
@@ -53,7 +53,6 @@ export NCCL_SOCKET_IFNAME="eth0,en,eth,em,bond"
 export NCCL_BUFFSIZE=2097152
 #export TORCH_DIST_INIT_BARRIER=1
 export FI_EFA_SET_CUDA_SYNC_MEMOPS=0
-#export USE_LIBUV=1
 CONFIG_FILE=${CONFIG_FILE:-"./train_configs/llama2_13b.toml"}
 
 dcgmi profile --pause

diff --git a/run_llama_train.sh b/run_llama_train.sh
@@ -7,9 +7,6 @@
 
 set -ex
 
-# libUV is a scalable backend for TCPStore which is used in processGroup
-# rendezvous. This is the recommended backend for distributed training.
-export USE_LIBUV=1
 TRAINER_DIR=${TRAINER_DIR:-/home/$USER/local/torchtitan}
 
 # use envs as local overrides for convenience
-Original file line number
+Diff line change
@@ Expand Up / @@ -18,7 +18,6 @@ @@
     set -ex
-    export USE_LIBUV=1
     TRAINER_DIR=${1:-/home/$USER/local/torchtitan}
     NGPU=1
     LOG_RANK=0
@@ Expand Down @@