Fix Deepspeed Zero3 Config (#791)

* Update zero3.json

Take away CPU Offload by default (Slows things down horribly, better off reducing batchsize), and changes LR Scheduler to a properly decaying one

* Update zero3.json

fix something

deepspeed/zero3.json

@@ -1,14 +1,6 @@
 {
   "zero_optimization": {
     "stage": 3,
-    "offload_optimizer": {
-      "device": "cpu",
-      "pin_memory": true
-    },
-    "offload_param": {
-      "device": "cpu",
-      "pin_memory": true
-    },
     "overlap_comm": true,
     "contiguous_gradients": true,
     "sub_group_size": 0,
@@ -41,12 +33,13 @@
     }
   },
   "scheduler": {
-    "type": "WarmupLR",
+    "type": "WarmupDecayLR",
     "params": {
       "warmup_min_lr": "auto",
       "warmup_max_lr": "auto",
       "warmup_num_steps": "auto",
-      "warmup_type": "linear"
+      "warmup_type": "linear",
+      "total_num_steps": "auto"
     }
   },
   "gradient_accumulation_steps": "auto",