Upload best model checkpoint

metrics.json

@@ -1,30 +1,30 @@
 {
-    "rl_info/A2G": 0.04484192281961441,
-    "rl_info/entropy": 2.499966621398926,
-    "rl_info/total_token": 140.0,
-    "rl_info/advantage_b4_norm": -2400.245361328125,
-    "rl_info/advantage_after_gnorm": -0.29543447494506836,
+    "rl_info/A2G": 0.00036897839163430035,
+    "rl_info/entropy": 2.8974661827087402,
+    "rl_info/total_token": 1058.0,
+    "rl_info/advantage_b4_norm": -662.7789306640625,
+    "rl_info/advantage_after_gnorm": -0.8332476615905762,
     "rl_info/kl_w_ref": 0.0,
-    "train/rl_loss": -4.484442234039307,
-    "train/total_loss": -4.484442234039307,
-    "xsum/rouge1": 0.18106894965442452,
-    "xsum/rouge2": 0.03488567263648188,
-    "xsum/rougeL": 0.14794279968444224,
-    "xsum/rougeLsum": 0.14825366834030124,
-    "xsum/bertscore_precision": 0.7590583597402083,
-    "xsum/bertscore_recall": 0.7190113407984515,
-    "xsum/bertscore_f1": 0.7378082892157234,
-    "eval_agg/avg_all_rougef": 0.12803777257891247,
-    "eval_agg/avg_all_bertf": 0.7378082892157234,
-    "eval_agg/avg_all": 0.43292303089731793,
-    "num_rl_rollout": 107,
+    "train/rl_loss": -0.03718758374452591,
+    "train/total_loss": -0.03718758374452591,
+    "xsum/rouge1": 0.17959011108226783,
+    "xsum/rouge2": 0.034164289805494366,
+    "xsum/rougeL": 0.14642892388250248,
+    "xsum/rougeLsum": 0.14675109876360795,
+    "xsum/bertscore_precision": 0.7542628072753078,
+    "xsum/bertscore_recall": 0.7164772123976614,
+    "xsum/bertscore_f1": 0.7342423905250528,
+    "eval_agg/avg_all_rougef": 0.12673360588346816,
+    "eval_agg/avg_all_bertf": 0.7342423905250528,
+    "eval_agg/avg_all": 0.4304879982042605,
+    "num_rl_rollout": 132,
     "lm_epoch": 0,
     "rl_epoch": 0,
-    "step": 17000,
-    "total_data_token": 1466042,
-    "total_rl_token": 19647438,
+    "step": 21000,
+    "total_data_token": 1816556,
+    "total_rl_token": 22448590,
     "total_lm_token": 0,
-    "total_token": 19647438,
-    "completed_steps": 17000,
-    "tune_objective": 0.43529236723131914
+    "total_token": 22448590,
+    "completed_steps": 21000,
+    "tune_objective": 0.43147928832451976
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:624ffb91772dfc99584c53b703e954c615f57bbb53e41b48eb3086d9ddc9fcea
+oid sha256:06aeea1c4a4f348a8fe8c7605cdfbe1d8457f812c38d2c757b1e1dbde758e40d
 size 497774208

train_configs.json

@@ -52,11 +52,11 @@
         "max_new_tokens": 40
     },
     "rollout_config": {
-        "accuracy_w": 1146.6528200301207,
+        "accuracy_w": 1347.37082407859,
         "len_pen": 1.0,
-        "accuracy_w2": 0.010606301264164078,
+        "accuracy_w2": 0.011313372019312368,
         "len_pen2": 1.0,
-        "threshold": 0.008255360186196097,
+        "threshold": 0.007420025186256454,
         "similarity_fn": "rouge",
         "acc_scale": "log",
         "max_ctx_len": 760