Upload ray_tune_logs/params.json with huggingface_hub

ray_tune_logs/params.json

@@ -22,7 +22,7 @@
     "grad_norm": 0.5,
     "gradient_accumulation_steps": 1,
     "gradient_checkpoint": false,
-    "group_relative_norm": true,
+    "group_relative_norm": false,
     "inference_config": {
       "do_sample": true,
       "max_new_tokens": 250,
@@ -60,11 +60,11 @@
     "push_to_hub": null,
     "ref_role1_name_or_path": "gpt2",
     "ref_role2_name_or_path": "gpt2",
-    "rl_algo": "off_policy",
+    "rl_algo": "on_policy",
     "rl_w": 1.0,
     "rollout_config": {
-      "accuracy_w": 0.348280208702833,
-      "accuracy_w2": 7.505241622349544,
+      "accuracy_w": 0.5456347144827906,
+      "accuracy_w2": 2.5739712587324886,
       "len_pen": 1.0,
       "len_pen2": 1.0,
       "max_ctx_len": 1000,
@@ -83,7 +83,7 @@
         "temperature": 1.0
       },
       "similarity_fn": "rouge",
-      "threshold": 0.007501954443620123
+      "threshold": 0.0010903843281968525
     },
     "rollout_game": "baseline3v2",
     "sample_config": {