AresMicro-10k final checkpoint

README.md

@@ -20,6 +20,6 @@ tags:
 - Hidden: 32
 - Heads: 4
 - MLP: 64
-- PT target: 50,000,000,000
-- RL target: 50,000,000,000
+- PT target: 5,000,000,000
+- RL target: 5,000,000,000
 - Benchmark file: purearith-1-answer-only.jsonl

checkpoint.pkl

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ddd45726dfb0b0519a8e7ed1fb0767a00bc2576a00409231a1f10128c3fa9443
+oid sha256:51c713bc6da5976d133f09280bbb61753e5ad9d9eb51e27a9f268c1efc4a2c8b
 size 134063

config.json

@@ -20,7 +20,7 @@
   "rl_temperature": 0.7,
   "pt_batch_size": 8192,
   "rl_batch_size": 8192,
-  "pt_target_tokens": 50000000000,
-  "rl_target_tokens": 50000000000,
+  "pt_target_tokens": 5000000000,
+  "rl_target_tokens": 5000000000,
   "benchmark_file": "purearith-1-answer-only.jsonl"
 }

training_state.json

@@ -3,6 +3,6 @@
   "pt_tokens": 8043888640,
   "rl_tokens": 8043626496,
   "total_tokens": 16087515136,
-  "elapsed_seconds": 57327.947034835815,
-  "final": false
+  "elapsed_seconds": 57327.94744849205,
+  "final": true
 }