View difference between Paste ID: QiMFhVLi and 6iGDgyz5

 #!/bin/bash
 # Copyright (c) Microsoft Corporation.
 # SPDX-License-Identifier: Apache-2.0
 
 # DeepSpeed Team
 ACTOR_MODEL_PATH=$1
 CRITIC_MODEL_PATH=$2
 ACTOR_ZERO_STAGE=$3
 CRITIC_ZERO_STAGE=$4
 OUTPUT=$5
-if [ "$OUTPUT" == "" ]; then
+
-    OUTPUT=/nobackup/jirayu/models/opt-1.3b-ppo
+
 fi
 if [ "$CRITIC_ZERO_STAGE" == "" ]; then
     CRITIC_ZERO_STAGE=2
 fi
 
 # if actor and critic model names are not provided, then use the publicly available AdamG012/chat-opt-1.3b-sft-deepspeed and AdamG012/chat-opt-350m-reward-deepspeed
 
 mkdir -p $OUTPUT
 
-if [ "$ACTOR_MODEL_PATH" == "" ]; then
+
-    ACTOR_MODEL_PATH=/nobackup/jirayu/models/opt-1.3b-sft
+
 Actor_Lr=1e-5
-if [ "$CRITIC_MODEL_PATH" == "" ]; then
+
-    CRITIC_MODEL_PATH=/nobackup/jirayu/models/opt-350m-rm
+
 deepspeed --master_port 12346 main.py \
    --data_path stanfordnlp/SHP \
    --data_split 0,0,10 \
    --actor_model_name_or_path $ACTOR_MODEL_PATH \
    --critic_model_name_or_path $CRITIC_MODEL_PATH \
    --num_padding_at_beginning 1 \
    --per_device_generation_batch_size 8 \
    --per_device_training_batch_size 8 \
    --generation_batches 1 \
    --ppo_epochs 1 \
    --max_answer_seq_len 256 \
    --max_prompt_seq_len 256 \
    --actor_learning_rate ${Actor_Lr} \
    --critic_learning_rate ${Critic_Lr} \
    --num_train_epochs 1 \
    --lr_scheduler_type cosine \
    --gradient_accumulation_steps 1 \
    --disable_actor_dropout \
    --num_warmup_steps 100 \
    --deepspeed --seed 1234 \
    --enable_hybrid_engine \
    --actor_zero_stage $ACTOR_ZERO_STAGE \
    --critic_zero_stage $CRITIC_ZERO_STAGE \
    --enable_ema \
    --output_dir $OUTPUT \
    --enable_tensorboard \
    --tensorboard_path $OUTPUT \
     | tee $OUTPUT/training.log