update config to latest veRL version

examples/veRL/config/ppo_trainer.yaml

@@ -45,7 +45,6 @@ actor_rollout_ref:
         # transformer_layer_cls_to_wrap: None
         min_num_params: 0
       param_offload: False
-      grad_offload: False
       optimizer_offload: False
       fsdp_size: -1
   ref:
@@ -104,7 +103,6 @@ critic:
     use_remove_padding: False
     fsdp_config:
       param_offload: False
-      grad_offload: False
       optimizer_offload: False
       wrap_policy:
         # transformer_layer_cls_to_wrap: None
@@ -158,10 +156,16 @@ trainer:
   project_name: verl_examples
   experiment_name: gsm8k
   logger: [ 'console', 'wandb' ]
+  val_generations_to_log_to_wandb: 0
   nnodes: 1
   n_gpus_per_node: 8
   save_freq: -1
+  # auto: find the last ckpt to resume. If can't find, start from scratch
+  resume_mode: auto # or auto or resume_path if
+  resume_from_path: False
   test_freq: -1
   critic_warmup: 0
-  default_hdfs_dir: ~/experiments/gsm8k/ppo/${trainer.experiment_name}
+  default_hdfs_dir: null
+  remove_previous_ckpt_in_save: False
+  del_local_ckpt_after_load: False
   default_local_dir: checkpoints/${trainer.project_name}/${trainer.experiment_name}