Model save

README.md

@@ -1,11 +1,9 @@
 ---
 base_model: Qwen/Qwen2.5-Math-7B
-datasets: DigitalLearningGmbH/MATH-lighteval
 library_name: transformers
 model_name: Qwen-2.5-7B-Simple-RL
 tags:
 - generated_from_trainer
-- open-r1
 - trl
 - grpo
 licence: license
@@ -13,7 +11,7 @@ licence: license
 
 # Model Card for Qwen-2.5-7B-Simple-RL
 
-This model is a fine-tuned version of [Qwen/Qwen2.5-Math-7B](https://huggingface.co/Qwen/Qwen2.5-Math-7B) on the [DigitalLearningGmbH/MATH-lighteval](https://huggingface.co/datasets/DigitalLearningGmbH/MATH-lighteval) dataset.
+This model is a fine-tuned version of [Qwen/Qwen2.5-Math-7B](https://huggingface.co/Qwen/Qwen2.5-Math-7B).
 It has been trained using [TRL](https://github.com/huggingface/trl).
 
 ## Quick start
@@ -29,15 +27,15 @@ print(output["generated_text"])
 
 ## Training procedure
 
-[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/jun-liang-chen-the-hong-kong-polytechnic-university/huggingface/runs/ye8fls8s) 
+[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/jun-liang-chen-the-hong-kong-polytechnic-university/huggingface/runs/2xgfampb) 
 
 
 This model was trained with GRPO, a method introduced in [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://huggingface.co/papers/2402.03300).
 
 ### Framework versions
 
-- TRL: 0.15.0.dev0
-- Transformers: 4.49.0.dev0
+- TRL: 0.16.0.dev0
+- Transformers: 4.49.0
 - Pytorch: 2.5.1
 - Datasets: 3.2.0
 - Tokenizers: 0.21.0

all_results.json

@@ -1,13 +1,8 @@
 {
-    "eval_loss": -2.7183905331185088e-05,
-    "eval_runtime": 75629.5286,
-    "eval_samples": 5000,
-    "eval_samples_per_second": 0.066,
-    "eval_steps_per_second": 0.011,
     "total_flos": 0.0,
-    "train_loss": 0.0008256121124141636,
-    "train_runtime": 419098.9196,
+    "train_loss": 0.055619138720809355,
+    "train_runtime": 265672.9627,
     "train_samples": 7500,
-    "train_samples_per_second": 0.018,
+    "train_samples_per_second": 0.028,
     "train_steps_per_second": 0.001
 }

config.json

@@ -19,11 +19,11 @@
   "rms_norm_eps": 1e-06,
   "rope_scaling": null,
   "rope_theta": 10000,
-  "sliding_window": null,
+  "sliding_window": 4096,
   "tie_word_embeddings": false,
   "torch_dtype": "bfloat16",
-  "transformers_version": "4.49.0.dev0",
-  "use_cache": true,
+  "transformers_version": "4.49.0",
+  "use_cache": false,
   "use_mrope": false,
   "use_sliding_window": false,
   "vocab_size": 152064

generation_config.json

@@ -2,5 +2,5 @@
   "bos_token_id": 151643,
   "eos_token_id": 151643,
   "max_new_tokens": 2048,
-  "transformers_version": "4.49.0.dev0"
+  "transformers_version": "4.49.0"
 }

model-00001-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3d2460834ab32758c7929fd2127f4764dda55d37b2ed228b7d65398fdd13aa3f
+oid sha256:27178cf126a277cccbf60cc1e1750f93865fd01c5f6610dd3b1f54e7046fd022
 size 4877660776

model-00002-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bcb46f0ac1dcead4a1ccca541f796c45a1d7d22a42f37e19cdcdf1b60bc0583c
+oid sha256:78e815400da77ffb9d36215cb9210b0a7a2585d23caf348bc5d151b9c8ae131e
 size 4932751008

model-00003-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9ec9437e98f69f030c1e09f702858d79bfacf2ed020e66df9be24b59c445c214
+oid sha256:23b3fd85fe2b07225cf7570442684ea8c009389f6c05568f1a9ddbed3a6b7c2a
 size 4330865200

model-00004-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fc740e9e6a6d8c353842a1ed8a64535fa9ce1baa0366313305dce8687dfe9175
+oid sha256:e9c4401f00a5ee5800b4ea7391a67920283187f1359d3034d94f33166aaa864a
 size 1089994880

tokenizer_config.json

@@ -202,7 +202,6 @@
   "extra_special_tokens": {},
   "model_max_length": 131072,
   "pad_token": "<|endoftext|>",
-  "padding_side": "left",
   "split_special_tokens": false,
   "tokenizer_class": "Qwen2Tokenizer",
   "unk_token": null

train_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 0.0008256121124141636,
-    "train_runtime": 419098.9196,
+    "train_loss": 0.055619138720809355,
+    "train_runtime": 265672.9627,
     "train_samples": 7500,
-    "train_samples_per_second": 0.018,
+    "train_samples_per_second": 0.028,
     "train_steps_per_second": 0.001
 }

trainer_state.json

@@ -2,1081 +2,1140 @@
   "best_metric": null,
   "best_model_checkpoint": null,
   "epoch": 0.9981333333333333,
-  "eval_steps": 100,
+  "eval_steps": 200,
   "global_step": 394,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "completion_length": 644.0193140933388,
+      "clip_ratio": 0.0,
+      "completion_length": 579.8596701371042,
+      "epoch": 0.002533333333333333,
+      "grad_norm": 0.2678598463535309,
+      "kl": -9.097551044664885e-08,
+      "learning_rate": 7.500000000000001e-08,
+      "loss": 0.0054,
+      "reward": 0.6754386079938788,
+      "reward_std": 0.29693952202796936,
+      "rewards/accuracy_reward": 0.6754386079938788,
+      "rewards/format_reward": 0.0,
+      "step": 1
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 615.5789616233424,
       "epoch": 0.012666666666666666,
-      "grad_norm": 0.16860076785087585,
-      "kl": 0.000124662016567431,
+      "grad_norm": 0.16652970016002655,
+      "kl": 0.00016463744012933028,
       "learning_rate": 3.75e-07,
-      "loss": 0.0,
-      "reward": 0.605263171384209,
-      "reward_std": 0.40058013012534693,
-      "rewards/accuracy_reward": 0.605263171384209,
+      "loss": 0.0073,
+      "reward": 0.631578962857786,
+      "reward_std": 0.33211043516271993,
+      "rewards/accuracy_reward": 0.631578962857786,
       "rewards/format_reward": 0.0,
       "step": 5
     },
     {
-      "completion_length": 601.6614225688734,
+      "clip_ratio": 0.0,
+      "completion_length": 586.8280829178659,
       "epoch": 0.025333333333333333,
-      "grad_norm": 0.14201320707798004,
-      "kl": 0.00020345888639751233,
+      "grad_norm": 0.18061865866184235,
+      "kl": 0.00021568097566303454,
       "learning_rate": 7.5e-07,
-      "loss": 0.0,
-      "reward": 0.673684226839166,
-      "reward_std": 0.39511305846666034,
-      "rewards/accuracy_reward": 0.673684226839166,
+      "loss": 0.0074,
+      "reward": 0.5877193123102188,
+      "reward_std": 0.36062715398637873,
+      "rewards/accuracy_reward": 0.5877193123102188,
       "rewards/format_reward": 0.0,
       "step": 10
     },
     {
-      "completion_length": 579.6684371145149,
+      "clip_ratio": 0.0,
+      "completion_length": 606.9789639121608,
       "epoch": 0.038,
-      "grad_norm": 0.3944288194179535,
-      "kl": 0.00023700312564247533,
+      "grad_norm": 1.617056965827942,
+      "kl": 0.00029152569017912213,
       "learning_rate": 1.125e-06,
-      "loss": 0.0,
-      "reward": 0.6631579123045269,
-      "reward_std": 0.3630171424464176,
-      "rewards/accuracy_reward": 0.6631579123045269,
+      "loss": -0.0183,
+      "reward": 0.5754386100329851,
+      "reward_std": 0.35243560671806334,
+      "rewards/accuracy_reward": 0.5754386100329851,
       "rewards/format_reward": 0.0,
       "step": 15
     },
     {
-      "completion_length": 615.7789647152549,
+      "clip_ratio": 0.0,
+      "completion_length": 649.0684372751336,
       "epoch": 0.050666666666666665,
-      "grad_norm": 0.1742940992116928,
-      "kl": 0.0005895815397563733,
+      "grad_norm": 0.08494145423173904,
+      "kl": 0.00036713951512386923,
       "learning_rate": 1.5e-06,
-      "loss": 0.0,
-      "reward": 0.6175438761711121,
-      "reward_std": 0.40319750308990476,
-      "rewards/accuracy_reward": 0.6175438761711121,
+      "loss": 0.0159,
+      "reward": 0.6228070304581993,
+      "reward_std": 0.3047279185370395,
+      "rewards/accuracy_reward": 0.6228070304581993,
       "rewards/format_reward": 0.0,
       "step": 20
     },
     {
-      "completion_length": 653.6140499717311,
+      "clip_ratio": 0.0,
+      "completion_length": 626.4456302040502,
       "epoch": 0.06333333333333334,
-      "grad_norm": 0.213926300406456,
-      "kl": 0.0014986289174933182,
+      "grad_norm": 0.4462898075580597,
+      "kl": 0.005578929499575966,
       "learning_rate": 1.875e-06,
-      "loss": 0.0,
-      "reward": 0.6561403632164001,
-      "reward_std": 0.3638197688679946,
-      "rewards/accuracy_reward": 0.6561403632164001,
+      "loss": -0.0148,
+      "reward": 0.6350877306963268,
+      "reward_std": 0.3297370195388794,
+      "rewards/accuracy_reward": 0.6350877306963268,
       "rewards/format_reward": 0.0,
       "step": 25
     },
     {
-      "completion_length": 624.1438784950658,
+      "clip_ratio": 0.0,
+      "completion_length": 610.7701887431898,
       "epoch": 0.076,
-      "grad_norm": 0.16140718758106232,
-      "kl": 0.0023109034488075658,
+      "grad_norm": 0.09956062585115433,
+      "kl": 0.4089626211869089,
       "learning_rate": 2.25e-06,
-      "loss": 0.0,
-      "reward": 0.6947368581044047,
-      "reward_std": 0.32959623650500647,
-      "rewards/accuracy_reward": 0.6947368581044047,
+      "loss": 0.0345,
+      "reward": 0.5842105410600963,
+      "reward_std": 0.2981200535046427,
+      "rewards/accuracy_reward": 0.5842105410600963,
       "rewards/format_reward": 0.0,
       "step": 30
     },
     {
-      "completion_length": 649.8438781738281,
+      "clip_ratio": 0.0,
+      "completion_length": 625.8929995888158,
       "epoch": 0.08866666666666667,
-      "grad_norm": 0.16571274399757385,
-      "kl": 0.009154390033922698,
+      "grad_norm": 0.19778481125831604,
+      "kl": 0.006598512749922903,
       "learning_rate": 2.6250000000000003e-06,
-      "loss": 0.0,
-      "reward": 0.6894737009939395,
-      "reward_std": 0.35860375699244046,
-      "rewards/accuracy_reward": 0.687719314819888,
-      "rewards/format_reward": 0.0017543860171970569,
+      "loss": 0.0187,
+      "reward": 0.6684210676895944,
+      "reward_std": 0.2966519020105663,
+      "rewards/accuracy_reward": 0.6684210676895944,
+      "rewards/format_reward": 0.0,
       "step": 35
     },
     {
-      "completion_length": 629.4052797016344,
+      "clip_ratio": 0.0,
+      "completion_length": 593.7491395649157,
       "epoch": 0.10133333333333333,
-      "grad_norm": 0.27635857462882996,
-      "kl": 0.0027755938078227794,
+      "grad_norm": 0.34809431433677673,
+      "kl": 0.35181455110248766,
       "learning_rate": 3e-06,
-      "loss": 0.0,
-      "reward": 0.7035087861512836,
-      "reward_std": 0.34189380752412896,
-      "rewards/accuracy_reward": 0.7035087861512836,
+      "loss": 0.031,
+      "reward": 0.6771929976187254,
+      "reward_std": 0.2939719165626325,
+      "rewards/accuracy_reward": 0.6771929976187254,
       "rewards/format_reward": 0.0,
       "step": 40
     },
     {
-      "completion_length": 652.133349609375,
+      "clip_ratio": 0.0,
+      "completion_length": 621.2736994693154,
       "epoch": 0.114,
-      "grad_norm": 0.0473485067486763,
-      "kl": 0.026468778911389802,
+      "grad_norm": 0.05024183914065361,
+      "kl": 0.01271261917917352,
       "learning_rate": 2.998523534736735e-06,
-      "loss": 0.0,
-      "reward": 0.6842105420012223,
-      "reward_std": 0.3472982666994396,
-      "rewards/accuracy_reward": 0.6842105420012223,
+      "loss": 0.0253,
+      "reward": 0.712280712472765,
+      "reward_std": 0.2292553468754417,
+      "rewards/accuracy_reward": 0.712280712472765,
       "rewards/format_reward": 0.0,
       "step": 45
     },
     {
-      "completion_length": 628.0421226099918,
+      "clip_ratio": 0.0,
+      "completion_length": 582.4965084678248,
       "epoch": 0.12666666666666668,
-      "grad_norm": 0.11586566269397736,
-      "kl": 0.004013121755499589,
+      "grad_norm": 4.105992317199707,
+      "kl": 0.06967028567665502,
       "learning_rate": 2.994097045546504e-06,
-      "loss": 0.0,
-      "reward": 0.7263158050022627,
-      "reward_std": 0.3308417533573351,
-      "rewards/accuracy_reward": 0.7263158050022627,
+      "loss": -0.0106,
+      "reward": 0.69122808387405,
+      "reward_std": 0.2388055675908139,
+      "rewards/accuracy_reward": 0.69122808387405,
       "rewards/format_reward": 0.0,
       "step": 50
     },
     {
-      "completion_length": 634.6701881810238,
+      "clip_ratio": 0.0,
+      "completion_length": 606.9333484850432,
       "epoch": 0.13933333333333334,
-      "grad_norm": 0.09287750720977783,
-      "kl": 0.0033212159809313324,
+      "grad_norm": 0.2051275223493576,
+      "kl": 0.05069758766575864,
       "learning_rate": 2.986729246506011e-06,
-      "loss": 0.0,
-      "reward": 0.6912280863837192,
-      "reward_std": 0.35596638729697777,
-      "rewards/accuracy_reward": 0.6912280863837192,
+      "loss": 0.0093,
+      "reward": 0.6807017683982849,
+      "reward_std": 0.2922787565934031,
+      "rewards/accuracy_reward": 0.6807017683982849,
       "rewards/format_reward": 0.0,
       "step": 55
     },
     {
-      "completion_length": 607.8982656378495,
+      "clip_ratio": 0.0,
+      "completion_length": 577.8351034064042,
       "epoch": 0.152,
-      "grad_norm": 0.1364029198884964,
-      "kl": 0.003622757761101974,
+      "grad_norm": 0.0773908868432045,
+      "kl": 0.12494567067999589,
       "learning_rate": 2.976434642014389e-06,
-      "loss": 0.0,
-      "reward": 0.7561403630595458,
-      "reward_std": 0.2851124694472865,
-      "rewards/accuracy_reward": 0.7543859771992031,
-      "rewards/format_reward": 0.0017543860171970569,
+      "loss": 0.0021,
+      "reward": 0.7561403635301088,
+      "reward_std": 0.2459657985913126,
+      "rewards/accuracy_reward": 0.7561403635301088,
+      "rewards/format_reward": 0.0,
       "step": 60
     },
     {
-      "completion_length": 605.7438781738281,
+      "clip_ratio": 0.0,
+      "completion_length": 591.8403675280119,
       "epoch": 0.16466666666666666,
-      "grad_norm": 0.0922616571187973,
-      "kl": 0.007114209626850329,
+      "grad_norm": 0.35004928708076477,
+      "kl": 1.51502416510331,
       "learning_rate": 2.9632334982395456e-06,
-      "loss": 0.0,
-      "reward": 0.7578947509589948,
-      "reward_std": 0.29374928286201074,
-      "rewards/accuracy_reward": 0.7578947509589948,
+      "loss": 0.0724,
+      "reward": 0.6894736994253962,
+      "reward_std": 0.2734406850839916,
+      "rewards/accuracy_reward": 0.6894736994253962,
       "rewards/format_reward": 0.0,
       "step": 65
     },
     {
-      "completion_length": 597.1175623843544,
+      "clip_ratio": 0.0,
+      "completion_length": 598.454403204667,
       "epoch": 0.17733333333333334,
-      "grad_norm": 0.0,
-      "kl": 0.003088178132709704,
+      "grad_norm": 0.1150355115532875,
+      "kl": 0.4293059499640214,
       "learning_rate": 2.947151803221774e-06,
-      "loss": 0.0,
-      "reward": 0.7596491361919202,
-      "reward_std": 0.2632309951280293,
-      "rewards/accuracy_reward": 0.7596491361919202,
+      "loss": 0.0292,
+      "reward": 0.635087733676559,
+      "reward_std": 0.30785316950396485,
+      "rewards/accuracy_reward": 0.635087733676559,
       "rewards/format_reward": 0.0,
       "step": 70
     },
     {
-      "completion_length": 599.9701933208265,
+      "clip_ratio": 0.0,
+      "completion_length": 647.6456305252879,
       "epoch": 0.19,
-      "grad_norm": 0.10758549720048904,
-      "kl": 0.003361149838096217,
+      "grad_norm": 0.21817399561405182,
+      "kl": 0.025303368819387335,
       "learning_rate": 2.928221215713164e-06,
-      "loss": 0.0,
-      "reward": 0.7666666791627281,
-      "reward_std": 0.2686667611724452,
-      "rewards/accuracy_reward": 0.7666666791627281,
+      "loss": 0.01,
+      "reward": 0.69473685684957,
+      "reward_std": 0.28838693499565127,
+      "rewards/accuracy_reward": 0.69473685684957,
       "rewards/format_reward": 0.0,
       "step": 75
     },
     {
-      "completion_length": 597.1456305252879,
+      "clip_ratio": 0.0,
+      "completion_length": 588.203521407278,
       "epoch": 0.20266666666666666,
-      "grad_norm": 0.06053958460688591,
-      "kl": 0.0037446674547697368,
+      "grad_norm": 4.736172676086426,
+      "kl": 0.19514670121042352,
       "learning_rate": 2.906479002853542e-06,
-      "loss": 0.0,
-      "reward": 0.7315789610147476,
-      "reward_std": 0.30245108447576824,
-      "rewards/accuracy_reward": 0.7315789610147476,
+      "loss": 0.0357,
+      "reward": 0.6877193132513448,
+      "reward_std": 0.29404533599552357,
+      "rewards/accuracy_reward": 0.6877193132513448,
       "rewards/format_reward": 0.0,
       "step": 80
     },
     {
-      "completion_length": 577.8438778525905,
+      "clip_ratio": 0.0,
+      "completion_length": 570.1052785773027,
       "epoch": 0.21533333333333332,
-      "grad_norm": 0.0840262621641159,
-      "kl": 0.004712596692537006,
+      "grad_norm": 0.013613657094538212,
+      "kl": 0.2757121437474301,
       "learning_rate": 2.8819679668056195e-06,
-      "loss": 0.0,
-      "reward": 0.7982456276291295,
-      "reward_std": 0.26917701614530465,
-      "rewards/accuracy_reward": 0.7982456276291295,
+      "loss": 0.0439,
+      "reward": 0.7631579086968773,
+      "reward_std": 0.23847583406849912,
+      "rewards/accuracy_reward": 0.7631579086968773,
       "rewards/format_reward": 0.0,
       "step": 85
     },
     {
-      "completion_length": 612.2982611405222,
+      "clip_ratio": 0.0,
+      "completion_length": 612.1772117213198,
       "epoch": 0.228,
-      "grad_norm": 0.0,
-      "kl": 0.02382647865696957,
+      "grad_norm": 0.0846298560500145,
+      "kl": 0.3429775037263569,
       "learning_rate": 2.8547363604937856e-06,
-      "loss": 0.0,
-      "reward": 0.7456140491523241,
-      "reward_std": 0.2956205848016237,
-      "rewards/accuracy_reward": 0.7456140491523241,
+      "loss": 0.0558,
+      "reward": 0.6403508915712959,
+      "reward_std": 0.33198017101538807,
+      "rewards/accuracy_reward": 0.6403508915712959,
       "rewards/format_reward": 0.0,
       "step": 90
     },
     {
-      "completion_length": 598.1403680098684,
+      "clip_ratio": 0.0,
+      "completion_length": 703.9456295615748,
       "epoch": 0.24066666666666667,
-      "grad_norm": 0.0,
-      "kl": 0.005134462055407072,
+      "grad_norm": 0.1827508509159088,
+      "kl": 0.08838115491365131,
       "learning_rate": 2.824837792612416e-06,
-      "loss": 0.0,
-      "reward": 0.7403508889047723,
-      "reward_std": 0.2658375608293634,
-      "rewards/accuracy_reward": 0.7403508889047723,
+      "loss": 0.0399,
+      "reward": 0.5175438727203169,
+      "reward_std": 0.36444555615123947,
+      "rewards/accuracy_reward": 0.5175438727203169,
       "rewards/format_reward": 0.0,
       "step": 95
     },
     {
-      "completion_length": 582.6561587685032,
+      "clip_ratio": 0.0,
+      "completion_length": 713.1719475997122,
       "epoch": 0.25333333333333335,
-      "grad_norm": 0.11761737614870071,
-      "kl": 0.00578155517578125,
+      "grad_norm": 0.4626220464706421,
+      "kl": 0.1013056704872533,
       "learning_rate": 2.792331122090709e-06,
-      "loss": 0.0,
-      "reward": 0.7614035217385543,
-      "reward_std": 0.25812496166480214,
-      "rewards/accuracy_reward": 0.7614035217385543,
+      "loss": 0.0703,
+      "reward": 0.48771931183965583,
+      "reward_std": 0.41597402189907273,
+      "rewards/accuracy_reward": 0.48771931183965583,
       "rewards/format_reward": 0.0,
       "step": 100
     },
     {
-      "epoch": 0.25333333333333335,
-      "eval_completion_length": 576.5513494110107,
-      "eval_kl": 0.10147758865356446,
-      "eval_loss": 3.062895004291022e-08,
-      "eval_reward": 0.6709333465278149,
-      "eval_reward_std": 0.3338339452087879,
-      "eval_rewards/accuracy_reward": 0.6709333465278149,
-      "eval_rewards/format_reward": 0.0,
-      "eval_runtime": 79047.156,
-      "eval_samples_per_second": 0.063,
-      "eval_steps_per_second": 0.011,
-      "step": 100
-    },
-    {
-      "completion_length": 549.310542056435,
+      "clip_ratio": 0.0,
+      "completion_length": 682.8579130473889,
       "epoch": 0.266,
-      "grad_norm": 0.08108708262443542,
-      "kl": 0.07660410027754934,
+      "grad_norm": 1.0272141695022583,
+      "kl": 0.07340826737253289,
       "learning_rate": 2.7572803422217976e-06,
-      "loss": 0.0,
-      "reward": 0.8140351000585054,
-      "reward_std": 0.2548168144728008,
-      "rewards/accuracy_reward": 0.8140351000585054,
+      "loss": 0.0438,
+      "reward": 0.5421052777453472,
+      "reward_std": 0.3957412208381452,
+      "rewards/accuracy_reward": 0.5421052777453472,
       "rewards/format_reward": 0.0,
       "step": 105
     },
     {
-      "completion_length": 596.6351042094983,
+      "clip_ratio": 0.0,
+      "completion_length": 689.7000163831209,
       "epoch": 0.2786666666666667,
-      "grad_norm": 0.06687445938587189,
-      "kl": 0.18131464907997533,
+      "grad_norm": 0.15440942347049713,
+      "kl": 0.08880775853207237,
       "learning_rate": 2.71975445468425e-06,
-      "loss": 0.0,
-      "reward": 0.7543859803362897,
-      "reward_std": 0.32687413692474365,
-      "rewards/accuracy_reward": 0.7543859803362897,
+      "loss": 0.0649,
+      "reward": 0.5807017685551392,
+      "reward_std": 0.4060916925731458,
+      "rewards/accuracy_reward": 0.5807017685551392,
       "rewards/format_reward": 0.0,
       "step": 110
     },
     {
-      "completion_length": 547.8263289602179,
+      "clip_ratio": 0.0,
+      "completion_length": 721.9579127261513,
       "epoch": 0.29133333333333333,
-      "grad_norm": 0.08061777800321579,
-      "kl": 4.457168900339227,
+      "grad_norm": 0.2818855345249176,
+      "kl": 0.2006508275082237,
       "learning_rate": 2.679827333703964e-06,
-      "loss": -0.0,
-      "reward": 0.7894736992685418,
-      "reward_std": 0.2753143636803878,
-      "rewards/accuracy_reward": 0.7894736992685418,
+      "loss": 0.0682,
+      "reward": 0.5350877319511614,
+      "reward_std": 0.3747960278862401,
+      "rewards/accuracy_reward": 0.5350877319511614,
       "rewards/format_reward": 0.0,
       "step": 115
     },
     {
-      "completion_length": 533.67895989669,
+      "clip_ratio": 0.0,
+      "completion_length": 714.03861822831,
       "epoch": 0.304,
-      "grad_norm": 0.15190833806991577,
-      "kl": 0.005019980982730263,
+      "grad_norm": 2.9434046745300293,
+      "kl": 0.14573557000411183,
       "learning_rate": 2.637577580623858e-06,
-      "loss": 0.0,
-      "reward": 0.771929837528028,
-      "reward_std": 0.2711241148020092,
-      "rewards/accuracy_reward": 0.771929837528028,
+      "loss": 0.0537,
+      "reward": 0.5245614161616877,
+      "reward_std": 0.38368915695893135,
+      "rewards/accuracy_reward": 0.5245614161616877,
       "rewards/format_reward": 0.0,
       "step": 120
     },
     {
-      "completion_length": 535.5947535464638,
+      "clip_ratio": 0.0,
+      "completion_length": 715.5035257439864,
       "epoch": 0.31666666666666665,
-      "grad_norm": 0.24402964115142822,
-      "kl": 0.005212723581414474,
+      "grad_norm": 0.1559271365404129,
+      "kl": 0.14786939119037829,
       "learning_rate": 2.593088369167671e-06,
-      "loss": 0.0,
-      "reward": 0.8035087809750908,
-      "reward_std": 0.2135340957265151,
-      "rewards/accuracy_reward": 0.8035087809750908,
+      "loss": 0.0408,
+      "reward": 0.5456140482111981,
+      "reward_std": 0.3673818581982663,
+      "rewards/accuracy_reward": 0.5456140482111981,
       "rewards/format_reward": 0.0,
       "step": 125
     },
     {
-      "completion_length": 532.9789643940172,
+      "clip_ratio": 0.0,
+      "completion_length": 659.6684400056538,
       "epoch": 0.3293333333333333,
-      "grad_norm": 0.28009745478630066,
-      "kl": 0.005336239463404605,
+      "grad_norm": 0.7108798027038574,
+      "kl": 0.19336740594161184,
       "learning_rate": 2.5464472817024772e-06,
-      "loss": 0.0,
-      "reward": 0.8052631699725201,
-      "reward_std": 0.24214739736757782,
-      "rewards/accuracy_reward": 0.8052631699725201,
+      "loss": 0.0052,
+      "reward": 0.5649122917338422,
+      "reward_std": 0.39664619596380934,
+      "rewards/accuracy_reward": 0.5649122917338422,
       "rewards/format_reward": 0.0,
       "step": 130
     },
     {
-      "completion_length": 561.8772093120374,
+      "clip_ratio": 0.0,
+      "completion_length": 589.4508954499897,
       "epoch": 0.342,
-      "grad_norm": 0.18070097267627716,
-      "kl": 0.004754076505962171,
+      "grad_norm": 0.08080588281154633,
+      "kl": 0.22685129266036183,
       "learning_rate": 2.497746136822254e-06,
-      "loss": 0.0,
-      "reward": 0.7771929958933278,
-      "reward_std": 0.27938908150321556,
-      "rewards/accuracy_reward": 0.7771929958933278,
+      "loss": 0.0261,
+      "reward": 0.6421052780590559,
+      "reward_std": 0.3161518184762252,
+      "rewards/accuracy_reward": 0.6421052780590559,
       "rewards/format_reward": 0.0,
       "step": 135
     },
     {
-      "completion_length": 578.6684380782278,
+      "clip_ratio": 0.0,
+      "completion_length": 635.3140510960629,
       "epoch": 0.3546666666666667,
-      "grad_norm": 0.10906434804201126,
-      "kl": 0.005059493215460526,
+      "grad_norm": 0.330537348985672,
+      "kl": 0.15463754754317435,
       "learning_rate": 2.4470808085919304e-06,
-      "loss": 0.0,
-      "reward": 0.7631579088537317,
-      "reward_std": 0.3001742513556229,
-      "rewards/accuracy_reward": 0.7631579088537317,
+      "loss": -0.0042,
+      "reward": 0.6333333481299249,
+      "reward_std": 0.34366038629883217,
+      "rewards/accuracy_reward": 0.6333333481299249,
       "rewards/format_reward": 0.0,
       "step": 140
     },
     {
-      "completion_length": 602.6772094726563,
+      "clip_ratio": 0.0,
+      "completion_length": 577.3017688951994,
       "epoch": 0.36733333333333335,
-      "grad_norm": 0.05584894120693207,
-      "kl": 0.004565590306332237,
+      "grad_norm": 0.015338568016886711,
+      "kl": 0.1593170166015625,
       "learning_rate": 2.3945510378077523e-06,
-      "loss": 0.0,
-      "reward": 0.7842105371387381,
-      "reward_std": 0.2481692125922755,
-      "rewards/accuracy_reward": 0.7842105371387381,
+      "loss": 0.031,
+      "reward": 0.7245614153774161,
+      "reward_std": 0.24523291211379203,
+      "rewards/accuracy_reward": 0.7245614153774161,
       "rewards/format_reward": 0.0,
       "step": 145
     },
     {
-      "completion_length": 580.9210699784128,
+      "clip_ratio": 0.0,
+      "completion_length": 599.7929999100535,
       "epoch": 0.38,
-      "grad_norm": 0.0609310120344162,
-      "kl": 0.005594434236225329,
+      "grad_norm": 0.23250138759613037,
+      "kl": 0.24896432976973684,
       "learning_rate": 2.340260235645519e-06,
-      "loss": 0.0,
-      "reward": 0.7368421157723979,
-      "reward_std": 0.26106126277070296,
-      "rewards/accuracy_reward": 0.7368421157723979,
+      "loss": 0.007,
+      "reward": 0.7052631712273547,
+      "reward_std": 0.2822182805914628,
+      "rewards/accuracy_reward": 0.7052631712273547,
       "rewards/format_reward": 0.0,
       "step": 150
     },
     {
-      "completion_length": 525.4000133313631,
+      "clip_ratio": 0.0,
+      "completion_length": 587.3158032869038,
       "epoch": 0.39266666666666666,
-      "grad_norm": 0.06953240185976028,
-      "kl": 0.007059599223889802,
+      "grad_norm": 0.8630210161209106,
+      "kl": 0.15269968133223685,
       "learning_rate": 2.2843152800832416e-06,
-      "loss": 0.0,
-      "reward": 0.8035087816025082,
-      "reward_std": 0.2169914556177039,
-      "rewards/accuracy_reward": 0.8035087816025082,
+      "loss": 0.0207,
+      "reward": 0.7035087856807207,
+      "reward_std": 0.24302949842653776,
+      "rewards/accuracy_reward": 0.7035087856807207,
       "rewards/format_reward": 0.0,
       "step": 155
     },
     {
-      "completion_length": 559.4386140522204,
+      "clip_ratio": 0.0,
+      "completion_length": 587.535104530736,
       "epoch": 0.4053333333333333,
-      "grad_norm": 0.08336005359888077,
-      "kl": 0.007145851536800986,
+      "grad_norm": 0.11684224009513855,
+      "kl": 0.7352347926089638,
       "learning_rate": 2.2268263054989753e-06,
-      "loss": 0.0,
-      "reward": 0.7824561511215411,
-      "reward_std": 0.23924477821902224,
-      "rewards/accuracy_reward": 0.7824561511215411,
+      "loss": 0.0285,
+      "reward": 0.6982456254331689,
+      "reward_std": 0.22624325062099257,
+      "rewards/accuracy_reward": 0.6982456254331689,
       "rewards/format_reward": 0.0,
       "step": 160
     },
     {
-      "completion_length": 546.5772112394634,
+      "clip_ratio": 0.0,
+      "completion_length": 552.8017696982936,
       "epoch": 0.418,
-      "grad_norm": 0.0791773647069931,
-      "kl": 0.0061695299650493425,
+      "grad_norm": 0.032140374183654785,
+      "kl": 0.20254629034745067,
       "learning_rate": 2.167906485858047e-06,
-      "loss": 0.0,
-      "reward": 0.803508786308138,
-      "reward_std": 0.25981574560466564,
-      "rewards/accuracy_reward": 0.803508786308138,
+      "loss": 0.0365,
+      "reward": 0.7649122917338421,
+      "reward_std": 0.2270519306785182,
+      "rewards/accuracy_reward": 0.7649122917338421,
       "rewards/format_reward": 0.0,
       "step": 165
     },
     {
-      "completion_length": 568.9280845240543,
+      "clip_ratio": 0.0,
+      "completion_length": 589.7193137721011,
       "epoch": 0.43066666666666664,
-      "grad_norm": 0.11959535628557205,
-      "kl": 0.007060161389802632,
+      "grad_norm": 0.4585510492324829,
+      "kl": 0.3263879073293586,
       "learning_rate": 2.1076718119164804e-06,
-      "loss": 0.0,
-      "reward": 0.7736842216629731,
-      "reward_std": 0.2649867666395087,
-      "rewards/accuracy_reward": 0.7736842216629731,
+      "loss": 0.017,
+      "reward": 0.6982456237077713,
+      "reward_std": 0.1981823155754491,
+      "rewards/accuracy_reward": 0.6982456237077713,
       "rewards/format_reward": 0.0,
       "step": 170
     },
     {
-      "completion_length": 550.3982609799034,
+      "clip_ratio": 0.0,
+      "completion_length": 575.0158064992804,
       "epoch": 0.44333333333333336,
-      "grad_norm": 0.0,
-      "kl": 0.007817157946134868,
+      "grad_norm": 0.5023659467697144,
+      "kl": 0.35051863820929274,
       "learning_rate": 2.0462408628792335e-06,
-      "loss": 0.0,
-      "reward": 0.7824561511215411,
-      "reward_std": 0.253275244800668,
-      "rewards/accuracy_reward": 0.7824561511215411,
+      "loss": -0.007,
+      "reward": 0.7315789625832909,
+      "reward_std": 0.2627183641258039,
+      "rewards/accuracy_reward": 0.7315789625832909,
       "rewards/format_reward": 0.0,
       "step": 175
     },
     {
-      "completion_length": 586.7596660413241,
+      "clip_ratio": 0.0,
+      "completion_length": 594.8736973812705,
       "epoch": 0.456,
-      "grad_norm": 0.06831427663564682,
-      "kl": 0.0069172106291118425,
+      "grad_norm": 0.0033971164375543594,
+      "kl": 0.30645864386307564,
       "learning_rate": 1.9837345729627633e-06,
-      "loss": 0.0,
-      "reward": 0.7438596612528751,
-      "reward_std": 0.2811869652647721,
-      "rewards/accuracy_reward": 0.7438596612528751,
+      "loss": 0.0193,
+      "reward": 0.7456140436624226,
+      "reward_std": 0.19300286424787422,
+      "rewards/accuracy_reward": 0.7456140436624226,
       "rewards/format_reward": 0.0,
       "step": 180
     },
     {
-      "completion_length": 583.1228234542043,
+      "clip_ratio": 0.0,
+      "completion_length": 579.4543993498149,
       "epoch": 0.4686666666666667,
-      "grad_norm": 0.10550453513860703,
-      "kl": 0.006921226099917763,
+      "grad_norm": 4.758713245391846,
+      "kl": 0.43655283074629936,
       "learning_rate": 1.9202759933214665e-06,
-      "loss": 0.0,
-      "reward": 0.7631579093242946,
-      "reward_std": 0.29657848289138394,
-      "rewards/accuracy_reward": 0.7631579093242946,
+      "loss": 0.0156,
+      "reward": 0.733333346247673,
+      "reward_std": 0.22970061208072462,
+      "rewards/accuracy_reward": 0.733333346247673,
       "rewards/format_reward": 0.0,
       "step": 185
     },
     {
-      "completion_length": 528.0912424187911,
+      "clip_ratio": 0.0,
+      "completion_length": 596.1772107576069,
       "epoch": 0.48133333333333334,
-      "grad_norm": 0.0677071213722229,
-      "kl": 0.009208438270970395,
+      "grad_norm": 5.337194442749023,
+      "kl": 1.3797318307976973,
       "learning_rate": 1.8559900498066726e-06,
-      "loss": 0.0,
-      "reward": 0.8140350991173795,
-      "reward_std": 0.22778719381282203,
-      "rewards/accuracy_reward": 0.8140350991173795,
+      "loss": 0.0535,
+      "reward": 0.7333333427968778,
+      "reward_std": 0.19656732584300793,
+      "rewards/accuracy_reward": 0.7333333427968778,
       "rewards/format_reward": 0.0,
       "step": 190
     },
     {
-      "completion_length": 550.7807165848582,
+      "clip_ratio": 0.0,
+      "completion_length": 611.4684377569902,
       "epoch": 0.494,
-      "grad_norm": 0.18718236684799194,
-      "kl": 0.008406307822779605,
+      "grad_norm": 0.10103946924209595,
+      "kl": 0.31487972861842106,
       "learning_rate": 1.7910032970350677e-06,
-      "loss": 0.0,
-      "reward": 0.7789473815968162,
-      "reward_std": 0.2719327967417868,
-      "rewards/accuracy_reward": 0.7789473815968162,
+      "loss": 0.0238,
+      "reward": 0.7245614155342704,
+      "reward_std": 0.23362611500840438,
+      "rewards/accuracy_reward": 0.7245614155342704,
       "rewards/format_reward": 0.0,
       "step": 195
     },
     {
-      "completion_length": 551.1315929212068,
       "epoch": 0.5066666666666667,
-      "grad_norm": 0.08260176330804825,
-      "kl": 0.00812426115337171,
+      "grad_norm": 0.2817623019218445,
       "learning_rate": 1.7254436692507058e-06,
-      "loss": 0.0,
-      "reward": 0.8070175547348826,
-      "reward_std": 0.23249007557567797,
-      "rewards/accuracy_reward": 0.8070175547348826,
-      "rewards/format_reward": 0.0,
+      "loss": 0.004,
       "step": 200
     },
     {
       "epoch": 0.5066666666666667,
-      "eval_completion_length": 553.8582824371338,
-      "eval_kl": 0.02361789245605469,
-      "eval_loss": -4.30000654887408e-05,
-      "eval_reward": 0.6822333461046219,
-      "eval_reward_std": 0.32350232841968535,
-      "eval_rewards/accuracy_reward": 0.6822333461046219,
+      "eval_clip_ratio": 0.0,
+      "eval_completion_length": 603.578249710083,
+      "eval_kl": 1.8679469131469726,
+      "eval_loss": 0.08767147362232208,
+      "eval_reward": 0.6215333458274603,
+      "eval_reward_std": 0.2736676256597042,
+      "eval_rewards/accuracy_reward": 0.6215333458274603,
       "eval_rewards/format_reward": 0.0,
-      "eval_runtime": 76484.7991,
-      "eval_samples_per_second": 0.065,
-      "eval_steps_per_second": 0.011,
+      "eval_runtime": 82336.5799,
+      "eval_samples_per_second": 0.061,
+      "eval_steps_per_second": 0.01,
       "step": 200
     },
     {
-      "completion_length": 570.7842259457237,
+      "clip_ratio": 0.0,
+      "completion_length": 626.1658041703074,
       "epoch": 0.5193333333333333,
-      "grad_norm": 0.14592121541500092,
-      "kl": 0.024847893965871712,
+      "grad_norm": 0.007743991445749998,
+      "kl": 0.10841610557154606,
       "learning_rate": 1.6594402284710481e-06,
-      "loss": 0.0,
-      "reward": 0.7561403608635853,
-      "reward_std": 0.24963736345893459,
-      "rewards/accuracy_reward": 0.7561403608635853,
+      "loss": 0.0302,
+      "reward": 0.6877193111338114,
+      "reward_std": 0.24886960622511412,
+      "rewards/accuracy_reward": 0.6877193111338114,
       "rewards/format_reward": 0.0,
       "step": 205
     },
     {
-      "completion_length": 556.5035239771793,
+      "clip_ratio": 0.0,
+      "completion_length": 608.6508930407073,
       "epoch": 0.532,
-      "grad_norm": 0.06877300888299942,
-      "kl": 0.011327321905838815,
+      "grad_norm": 0.19151441752910614,
+      "kl": 0.45181081671463813,
       "learning_rate": 1.593122910412851e-06,
-      "loss": 0.0,
-      "reward": 0.8210526436567307,
-      "reward_std": 0.22859587669372558,
-      "rewards/accuracy_reward": 0.8210526436567307,
+      "loss": 0.0514,
+      "reward": 0.6807017663591787,
+      "reward_std": 0.24232791693587052,
+      "rewards/accuracy_reward": 0.6807017663591787,
       "rewards/format_reward": 0.0,
       "step": 210
     },
     {
-      "completion_length": 579.1263334575452,
+      "clip_ratio": 0.0,
+      "completion_length": 603.6894899067125,
       "epoch": 0.5446666666666666,
-      "grad_norm": 0.0,
-      "kl": 0.007683683696546053,
+      "grad_norm": 0.07085677981376648,
+      "kl": 0.1109113191303454,
       "learning_rate": 1.5266222686980693e-06,
-      "loss": 0.0,
-      "reward": 0.7561403643143805,
-      "reward_std": 0.2832074833543677,
-      "rewards/accuracy_reward": 0.7561403643143805,
+      "loss": 0.0337,
+      "reward": 0.7192982565415533,
+      "reward_std": 0.23469111291985761,
+      "rewards/accuracy_reward": 0.7192982565415533,
       "rewards/format_reward": 0.0,
       "step": 215
     },
     {
-      "completion_length": 588.4649293598376,
+      "clip_ratio": 0.0,
+      "completion_length": 597.1596649169921,
       "epoch": 0.5573333333333333,
-      "grad_norm": 0.07735468447208405,
-      "kl": 0.007682238127055921,
+      "grad_norm": 1.3562774658203125,
+      "kl": 0.7995125218441612,
       "learning_rate": 1.460069217843338e-06,
-      "loss": 0.0,
-      "reward": 0.7333333465613817,
-      "reward_std": 0.31152473405787817,
-      "rewards/accuracy_reward": 0.7333333465613817,
+      "loss": 0.0486,
+      "reward": 0.6789473812831076,
+      "reward_std": 0.2647725648001621,
+      "rewards/accuracy_reward": 0.6789473812831076,
       "rewards/format_reward": 0.0,
       "step": 220
     },
     {
-      "completion_length": 582.8508938438014,
+      "clip_ratio": 0.0,
+      "completion_length": 575.6824705425062,
       "epoch": 0.57,
-      "grad_norm": 0.04747384414076805,
-      "kl": 0.008666349712171053,
+      "grad_norm": 0.18081852793693542,
+      "kl": 0.32951467413651314,
       "learning_rate": 1.3935947755389924e-06,
-      "loss": 0.0,
-      "reward": 0.7456140472700722,
-      "reward_std": 0.27623014795152767,
-      "rewards/accuracy_reward": 0.7456140472700722,
+      "loss": 0.0357,
+      "reward": 0.7543859778266204,
+      "reward_std": 0.22624325124840988,
+      "rewards/accuracy_reward": 0.7543859778266204,
       "rewards/format_reward": 0.0,
       "step": 225
     },
     {
-      "completion_length": 541.6929962158204,
+      "clip_ratio": 0.0,
+      "completion_length": 589.3491397255345,
       "epoch": 0.5826666666666667,
-      "grad_norm": 0.0,
-      "kl": 0.009119616056743421,
+      "grad_norm": 0.10897592455148697,
+      "kl": 0.09513003700657895,
       "learning_rate": 1.3273298047249756e-06,
-      "loss": 0.0,
-      "reward": 0.7964912389454089,
-      "reward_std": 0.2623067799367403,
-      "rewards/accuracy_reward": 0.7964912389454089,
+      "loss": 0.0126,
+      "reward": 0.6631579093242946,
+      "reward_std": 0.26906991381394235,
+      "rewards/accuracy_reward": 0.6631579093242946,
       "rewards/format_reward": 0.0,
       "step": 230
     },
     {
-      "completion_length": 555.8666826750102,
+      "clip_ratio": 0.0,
+      "completion_length": 633.8035246196546,
       "epoch": 0.5953333333333334,
-      "grad_norm": 0.1013207659125328,
-      "kl": 0.008411447625411184,
+      "grad_norm": 0.29345470666885376,
+      "kl": 0.08403657612047698,
       "learning_rate": 1.2614047559713923e-06,
-      "loss": 0.0,
-      "reward": 0.8263158042179911,
-      "reward_std": 0.25948601145493355,
-      "rewards/accuracy_reward": 0.8263158042179911,
+      "loss": 0.0135,
+      "reward": 0.6859649240970611,
+      "reward_std": 0.21056411203585174,
+      "rewards/accuracy_reward": 0.6859649240970611,
       "rewards/format_reward": 0.0,
       "step": 235
     },
     {
-      "completion_length": 564.3929960552015,
+      "clip_ratio": 0.0,
+      "completion_length": 583.8526471589742,
       "epoch": 0.608,
-      "grad_norm": 0.08883073180913925,
-      "kl": 0.008549579821134868,
+      "grad_norm": 0.09456110000610352,
+      "kl": 0.4004285310444079,
       "learning_rate": 1.1959494106708598e-06,
-      "loss": 0.0,
-      "reward": 0.792982468950121,
-      "reward_std": 0.25643417960719056,
-      "rewards/accuracy_reward": 0.792982468950121,
+      "loss": 0.0544,
+      "reward": 0.721052644754711,
+      "reward_std": 0.23450821669478167,
+      "rewards/accuracy_reward": 0.721052644754711,
       "rewards/format_reward": 0.0,
       "step": 240
     },
     {
-      "completion_length": 536.9596663625617,
+      "clip_ratio": 0.0,
+      "completion_length": 598.0386127672697,
       "epoch": 0.6206666666666667,
-      "grad_norm": 0.04451132193207741,
-      "kl": 0.009548789576480263,
+      "grad_norm": 0.45697084069252014,
+      "kl": 0.6121787623355263,
       "learning_rate": 1.1310926255482204e-06,
-      "loss": 0.0,
-      "reward": 0.8052631693451028,
-      "reward_std": 0.24199818247242977,
-      "rewards/accuracy_reward": 0.8052631693451028,
+      "loss": 0.0493,
+      "reward": 0.7298245765660939,
+      "reward_std": 0.22977403151361567,
+      "rewards/accuracy_reward": 0.7298245765660939,
       "rewards/format_reward": 0.0,
       "step": 245
     },
     {
-      "completion_length": 544.7193134508635,
+      "clip_ratio": 0.0,
+      "completion_length": 575.0772094726562,
       "epoch": 0.6333333333333333,
-      "grad_norm": 0.0,
-      "kl": 0.009332596628289473,
+      "grad_norm": 4.634888648986816,
+      "kl": 0.7007741827713816,
       "learning_rate": 1.0669620789905688e-06,
-      "loss": 0.0,
-      "reward": 0.800000010038677,
-      "reward_std": 0.23219164547167326,
-      "rewards/accuracy_reward": 0.800000010038677,
+      "loss": 0.056,
+      "reward": 0.7859649244107698,
+      "reward_std": 0.1908415624969884,
+      "rewards/accuracy_reward": 0.7859649244107698,
       "rewards/format_reward": 0.0,
       "step": 250
     },
     {
-      "completion_length": 552.8649261474609,
+      "clip_ratio": 0.0,
+      "completion_length": 590.9579106381065,
       "epoch": 0.646,
-      "grad_norm": 0.06963980942964554,
-      "kl": 0.009455148797286185,
+      "grad_norm": 2.5749571323394775,
+      "kl": 2.5487600226151317,
       "learning_rate": 1.0036840196969795e-06,
-      "loss": 0.0,
-      "reward": 0.807017553009485,
-      "reward_std": 0.21750171090427198,
-      "rewards/accuracy_reward": 0.807017553009485,
+      "loss": 0.1164,
+      "reward": 0.7561403608635853,
+      "reward_std": 0.19469364724661176,
+      "rewards/accuracy_reward": 0.7561403608635853,
       "rewards/format_reward": 0.0,
       "step": 255
     },
     {
-      "completion_length": 560.9298404091282,
+      "clip_ratio": 0.0,
+      "completion_length": 549.0789627878289,
       "epoch": 0.6586666666666666,
-      "grad_norm": 0.13915280997753143,
-      "kl": 0.008615754780016447,
+      "grad_norm": 5.019357204437256,
+      "kl": 0.3982286955180921,
       "learning_rate": 9.413830181427508e-07,
-      "loss": 0.0,
-      "reward": 0.7894736975431442,
-      "reward_std": 0.26591097994854573,
-      "rewards/accuracy_reward": 0.7894736975431442,
+      "loss": 0.0227,
+      "reward": 0.7877193085457149,
+      "reward_std": 0.1848534289159273,
+      "rewards/accuracy_reward": 0.7877193085457149,
       "rewards/format_reward": 0.0,
       "step": 260
     },
     {
-      "completion_length": 583.3000167043585,
+      "clip_ratio": 0.0,
+      "completion_length": 616.0596662019428,
       "epoch": 0.6713333333333333,
-      "grad_norm": 0.0,
-      "kl": 0.008441322728207237,
+      "grad_norm": 4.3469038009643555,
+      "kl": 38.30161839535362,
       "learning_rate": 8.801817213474331e-07,
-      "loss": 0.0,
-      "reward": 0.7508772039099744,
-      "reward_std": 0.2809643321915677,
-      "rewards/accuracy_reward": 0.7508772039099744,
+      "loss": 1.7071,
+      "reward": 0.7175438702106476,
+      "reward_std": 0.21893618012729443,
+      "rewards/accuracy_reward": 0.7175438702106476,
       "rewards/format_reward": 0.0,
       "step": 265
     },
     {
-      "completion_length": 531.2859802246094,
+      "clip_ratio": 0.0,
+      "completion_length": 623.8263318513569,
       "epoch": 0.684,
-      "grad_norm": 0.05701286345720291,
-      "kl": 0.00885218570106908,
+      "grad_norm": 0.23397096991539001,
+      "kl": 1.4742540861430922,
       "learning_rate": 8.202006114294044e-07,
-      "loss": 0.0,
-      "reward": 0.7877193099574039,
-      "reward_std": 0.2481692119648582,
-      "rewards/accuracy_reward": 0.7877193099574039,
+      "loss": 0.0739,
+      "reward": 0.7385965019464493,
+      "reward_std": 0.20718254509725068,
+      "rewards/accuracy_reward": 0.7385965019464493,
       "rewards/format_reward": 0.0,
       "step": 270
     },
     {
-      "completion_length": 531.2403676886307,
+      "clip_ratio": 0.0,
+      "completion_length": 614.8351022820724,
       "epoch": 0.6966666666666667,
-      "grad_norm": 0.05670388042926788,
-      "kl": 0.009579307154605263,
+      "grad_norm": 31.654590606689453,
+      "kl": 0.7249486019736842,
       "learning_rate": 7.615577684223272e-07,
-      "loss": 0.0,
-      "reward": 0.8175438716223365,
-      "reward_std": 0.23557321397881759,
-      "rewards/accuracy_reward": 0.8175438716223365,
+      "loss": 0.0309,
+      "reward": 0.7473684353263754,
+      "reward_std": 0.241305032529329,
+      "rewards/accuracy_reward": 0.7473684353263754,
       "rewards/format_reward": 0.0,
       "step": 275
     },
     {
-      "completion_length": 544.3842264275802,
+      "clip_ratio": 0.0,
+      "completion_length": 598.9438759251644,
       "epoch": 0.7093333333333334,
-      "grad_norm": 0.0,
-      "kl": 0.009712299547697369,
+      "grad_norm": 5.547906875610352,
+      "kl": 0.7660654168379935,
       "learning_rate": 7.043686378203864e-07,
-      "loss": 0.0,
-      "reward": 0.7701754490011616,
-      "reward_std": 0.25272287663660553,
-      "rewards/accuracy_reward": 0.7701754490011616,
+      "loss": 0.0408,
+      "reward": 0.7596491347802313,
+      "reward_std": 0.25206103419002734,
+      "rewards/accuracy_reward": 0.7596491347802313,
       "rewards/format_reward": 0.0,
       "step": 280
     },
     {
-      "completion_length": 571.7438746402138,
+      "clip_ratio": 0.0,
+      "completion_length": 620.8754567999589,
       "epoch": 0.722,
-      "grad_norm": 0.08127343654632568,
-      "kl": 0.008696064196134869,
+      "grad_norm": 4.756077289581299,
+      "kl": 0.8680445620888158,
       "learning_rate": 6.487458033099425e-07,
-      "loss": 0.0,
-      "reward": 0.7701754521382482,
-      "reward_std": 0.2902606151605907,
-      "rewards/accuracy_reward": 0.7701754521382482,
+      "loss": 0.0573,
+      "reward": 0.6789473770480406,
+      "reward_std": 0.2216498475325735,
+      "rewards/accuracy_reward": 0.6789473770480406,
       "rewards/format_reward": 0.0,
       "step": 285
     },
     {
-      "completion_length": 532.1596609015214,
+      "clip_ratio": 0.0,
+      "completion_length": 642.0228240966796,
       "epoch": 0.7346666666666667,
-      "grad_norm": 0.072137750685215,
-      "kl": 0.008551828484786184,
+      "grad_norm": 3.3402116298675537,
+      "kl": 1.7112895764802631,
       "learning_rate": 5.947987651349942e-07,
-      "loss": 0.0,
-      "reward": 0.8070175544211738,
-      "reward_std": 0.21213936178307785,
-      "rewards/accuracy_reward": 0.8070175544211738,
+      "loss": 0.0737,
+      "reward": 0.6964912395728262,
+      "reward_std": 0.26836833514665304,
+      "rewards/accuracy_reward": 0.6964912395728262,
       "rewards/format_reward": 0.0,
       "step": 290
     },
     {
-      "completion_length": 535.1210667660362,
+      "clip_ratio": 0.0,
+      "completion_length": 597.8894905491879,
       "epoch": 0.7473333333333333,
-      "grad_norm": 0.07434926927089691,
-      "kl": 0.008226896587171052,
+      "grad_norm": 2.271803617477417,
+      "kl": 0.4571268182051809,
       "learning_rate": 5.426337245327703e-07,
-      "loss": 0.0,
-      "reward": 0.8157894832523246,
-      "reward_std": 0.22407589335190622,
-      "rewards/accuracy_reward": 0.8157894832523246,
+      "loss": 0.0476,
+      "reward": 0.673684225427477,
+      "reward_std": 0.23058271408081055,
+      "rewards/accuracy_reward": 0.673684225427477,
       "rewards/format_reward": 0.0,
       "step": 295
     },
     {
-      "completion_length": 536.6631729929071,
+      "clip_ratio": 0.0,
+      "completion_length": 586.7842289974815,
       "epoch": 0.76,
-      "grad_norm": 0.08616889268159866,
-      "kl": 0.010335339997944079,
+      "grad_norm": 0.23953944444656372,
+      "kl": 0.30870281018708884,
       "learning_rate": 4.923533746638108e-07,
-      "loss": 0.0,
-      "reward": 0.8105263270829853,
-      "reward_std": 0.24915841535518043,
-      "rewards/accuracy_reward": 0.8105263270829853,
+      "loss": 0.0216,
+      "reward": 0.7614035162486528,
+      "reward_std": 0.2001630961894989,
+      "rewards/accuracy_reward": 0.7614035162486528,
       "rewards/format_reward": 0.0,
       "step": 300
     },
     {
-      "epoch": 0.76,
-      "eval_completion_length": 545.0558823242187,
-      "eval_kl": 0.6695549743652344,
-      "eval_loss": 3.003481197083602e-08,
-      "eval_reward": 0.6868333460628986,
-      "eval_reward_std": 0.3212484232842922,
-      "eval_rewards/accuracy_reward": 0.6868333460628986,
-      "eval_rewards/format_reward": 0.0,
-      "eval_runtime": 75622.6169,
-      "eval_samples_per_second": 0.066,
-      "eval_steps_per_second": 0.011,
-      "step": 300
-    },
-    {
-      "completion_length": 574.1017703407689,
+      "clip_ratio": 0.0,
+      "completion_length": 571.1228213661595,
       "epoch": 0.7726666666666666,
-      "grad_norm": 0.0810866504907608,
-      "kl": 0.02553052400287829,
+      "grad_norm": 0.22501805424690247,
+      "kl": 0.35944968775699015,
       "learning_rate": 4.440566984481256e-07,
-      "loss": 0.0,
-      "reward": 0.7666666799469998,
-      "reward_std": 0.27946487539692927,
-      "rewards/accuracy_reward": 0.7666666799469998,
+      "loss": 0.0413,
+      "reward": 0.7157894881148087,
+      "reward_std": 0.2280832472600435,
+      "rewards/accuracy_reward": 0.7157894881148087,
       "rewards/format_reward": 0.0,
       "step": 305
     },
     {
-      "completion_length": 550.2579098350124,
+      "clip_ratio": 0.0,
+      "completion_length": 657.6000173468339,
       "epoch": 0.7853333333333333,
-      "grad_norm": 0.09656031429767609,
-      "kl": 0.009140657123766447,
+      "grad_norm": 0.6775135397911072,
+      "kl": 0.7142669677734375,
       "learning_rate": 3.978387737053994e-07,
-      "loss": 0.0,
-      "reward": 0.7824561517489584,
-      "reward_std": 0.23484032687387968,
-      "rewards/accuracy_reward": 0.7824561517489584,
+      "loss": 0.0583,
+      "reward": 0.6701754515108309,
+      "reward_std": 0.2519539353094603,
+      "rewards/accuracy_reward": 0.6701754515108309,
       "rewards/format_reward": 0.0,
       "step": 310
     },
     {
-      "completion_length": 569.4912441855982,
+      "clip_ratio": 0.0,
+      "completion_length": 611.8982592130963,
       "epoch": 0.798,
-      "grad_norm": 0.094791479408741,
-      "kl": 0.009346088610197369,
+      "grad_norm": 0.3379327952861786,
+      "kl": 0.28816158896998356,
       "learning_rate": 3.5379058598286167e-07,
-      "loss": 0.0,
-      "reward": 0.7666666777510392,
-      "reward_std": 0.25904074530852467,
-      "rewards/accuracy_reward": 0.7666666777510392,
+      "loss": 0.0266,
+      "reward": 0.6771929922856783,
+      "reward_std": 0.22066064414225126,
+      "rewards/accuracy_reward": 0.6771929922856783,
       "rewards/format_reward": 0.0,
       "step": 315
     },
     {
-      "completion_length": 579.2614203202097,
+      "clip_ratio": 0.0,
+      "completion_length": 576.1263292814556,
       "epoch": 0.8106666666666666,
-      "grad_norm": 0.10730049759149551,
-      "kl": 4.134726434004934,
+      "grad_norm": 0.1547779142856598,
+      "kl": 0.282763671875,
       "learning_rate": 3.119988494392894e-07,
-      "loss": 0.0,
-      "reward": 0.7508772053216634,
-      "reward_std": 0.2928671808619248,
-      "rewards/accuracy_reward": 0.7508772053216634,
+      "loss": 0.0458,
+      "reward": 0.7543859776697661,
+      "reward_std": 0.21713829605202925,
+      "rewards/accuracy_reward": 0.7543859776697661,
       "rewards/format_reward": 0.0,
       "step": 320
     },
     {
-      "completion_length": 556.2263304057874,
+      "clip_ratio": 0.0,
+      "completion_length": 581.9789603785465,
       "epoch": 0.8233333333333334,
-      "grad_norm": 0.07251843065023422,
-      "kl": 0.009259033203125,
+      "grad_norm": 1.6545398235321045,
+      "kl": 1.8802037289268092,
       "learning_rate": 2.725458361377465e-07,
-      "loss": 0.0,
-      "reward": 0.7736842222903905,
-      "reward_std": 0.24773237705230713,
-      "rewards/accuracy_reward": 0.7736842222903905,
+      "loss": 0.0836,
+      "reward": 0.7596491349370856,
+      "reward_std": 0.21056411203585174,
+      "rewards/accuracy_reward": 0.7596491349370856,
       "rewards/format_reward": 0.0,
       "step": 325
     },
     {
-      "completion_length": 537.1491384405839,
+      "clip_ratio": 0.0,
+      "completion_length": 624.1122974596525,
       "epoch": 0.836,
-      "grad_norm": 0.0,
-      "kl": 0.009849789268092105,
+      "grad_norm": 0.04716293886303902,
+      "kl": 0.2931084883840461,
       "learning_rate": 2.3550921408312737e-07,
-      "loss": 0.0,
-      "reward": 0.8140350983331078,
-      "reward_std": 0.23072349554614016,
-      "rewards/accuracy_reward": 0.8140350983331078,
+      "loss": 0.0158,
+      "reward": 0.6947368513596686,
+      "reward_std": 0.215224878411544,
+      "rewards/accuracy_reward": 0.6947368513596686,
       "rewards/format_reward": 0.0,
       "step": 330
     },
     {
-      "completion_length": 547.5140491686369,
+      "clip_ratio": 0.0,
+      "completion_length": 604.3649275930304,
       "epoch": 0.8486666666666667,
-      "grad_norm": 0.08143898099660873,
-      "kl": 0.010343531558388158,
+      "grad_norm": 14.650071144104004,
+      "kl": 0.46654261538856906,
       "learning_rate": 2.0096189432334195e-07,
-      "loss": 0.0,
-      "reward": 0.7824561493961435,
-      "reward_std": 0.23674531171196386,
-      "rewards/accuracy_reward": 0.7824561493961435,
+      "loss": 0.0406,
+      "reward": 0.7333333448359841,
+      "reward_std": 0.21933933402362624,
+      "rewards/accuracy_reward": 0.7333333448359841,
       "rewards/format_reward": 0.0,
       "step": 335
     },
     {
-      "completion_length": 541.5175582082649,
+      "clip_ratio": 0.0,
+      "completion_length": 598.4035267076994,
       "epoch": 0.8613333333333333,
-      "grad_norm": 0.15802732110023499,
-      "kl": 0.009666683799342105,
+      "grad_norm": 0.0046224575489759445,
+      "kl": 0.14371498509457237,
       "learning_rate": 1.6897188741514286e-07,
-      "loss": 0.0,
-      "reward": 0.7666666757119329,
-      "reward_std": 0.24941472881718685,
-      "rewards/accuracy_reward": 0.7666666757119329,
+      "loss": 0.029,
+      "reward": 0.701754395585311,
+      "reward_std": 0.1958320642772474,
+      "rewards/accuracy_reward": 0.701754395585311,
       "rewards/format_reward": 0.0,
       "step": 340
     },
     {
-      "completion_length": 566.2824713456004,
+      "clip_ratio": 0.0,
+      "completion_length": 569.3929971795333,
       "epoch": 0.874,
-      "grad_norm": 0.08206634968519211,
-      "kl": 0.009324886924342106,
+      "grad_norm": 25.004467010498047,
+      "kl": 0.42442787571957236,
       "learning_rate": 1.396021695371582e-07,
-      "loss": 0.0,
-      "reward": 0.7543859756306598,
-      "reward_std": 0.23858530960584942,
-      "rewards/accuracy_reward": 0.7543859756306598,
+      "loss": 0.0334,
+      "reward": 0.7789473804988359,
+      "reward_std": 0.22242484782871447,
+      "rewards/accuracy_reward": 0.7789473804988359,
       "rewards/format_reward": 0.0,
       "step": 345
     },
     {
-      "completion_length": 567.8473815917969,
+      "clip_ratio": 0.0,
+      "completion_length": 618.5912446674548,
       "epoch": 0.8866666666666667,
-      "grad_norm": 0.06388476490974426,
-      "kl": 0.010038034539473684,
+      "grad_norm": 12.571819305419922,
+      "kl": 0.7060623972039474,
       "learning_rate": 1.1291055851370623e-07,
-      "loss": 0.0,
-      "reward": 0.7491228202455922,
-      "reward_std": 0.2948876986378118,
-      "rewards/accuracy_reward": 0.7491228202455922,
+      "loss": 0.0507,
+      "reward": 0.6964912398865348,
+      "reward_std": 0.2621238843390816,
+      "rewards/accuracy_reward": 0.6964912398865348,
       "rewards/format_reward": 0.0,
       "step": 350
     },
     {
-      "completion_length": 576.67545663934,
+      "clip_ratio": 0.0,
+      "completion_length": 580.6228229723479,
       "epoch": 0.8993333333333333,
-      "grad_norm": 0.058260105550289154,
-      "kl": 0.008216456363075659,
+      "grad_norm": 0.06348630040884018,
+      "kl": 0.45227805689761513,
       "learning_rate": 8.894959999345015e-08,
-      "loss": 0.0,
-      "reward": 0.8017543960558741,
-      "reward_std": 0.23333006344343488,
-      "rewards/accuracy_reward": 0.8017543960558741,
+      "loss": 0.0257,
+      "reward": 0.7157894854482851,
+      "reward_std": 0.21632961285741706,
+      "rewards/accuracy_reward": 0.7157894854482851,
       "rewards/format_reward": 0.0,
       "step": 355
     },
     {
-      "completion_length": 549.091246434262,
+      "clip_ratio": 0.0,
+      "completion_length": 588.1842253032484,
       "epoch": 0.912,
-      "grad_norm": 0.10874364525079727,
-      "kl": 0.00869959781044408,
+      "grad_norm": 2.36680006980896,
+      "kl": 0.27915199681332237,
       "learning_rate": 6.776646400696212e-08,
-      "loss": 0.0,
-      "reward": 0.796491244749019,
-      "reward_std": 0.3010250455454776,
-      "rewards/accuracy_reward": 0.796491244749019,
+      "loss": 0.0277,
+      "reward": 0.7140350958234385,
+      "reward_std": 0.18397132754325868,
+      "rewards/accuracy_reward": 0.7140350958234385,
       "rewards/format_reward": 0.0,
       "step": 360
     },
     {
-      "completion_length": 560.8719421386719,
+      "clip_ratio": 0.0,
+      "completion_length": 599.5210685328433,
       "epoch": 0.9246666666666666,
-      "grad_norm": 0.10691142827272415,
-      "kl": 0.009839509662828948,
+      "grad_norm": 5.9008965492248535,
+      "kl": 0.44251114694695726,
       "learning_rate": 4.940285210684375e-08,
-      "loss": 0.0,
-      "reward": 0.7964912431804757,
-      "reward_std": 0.28669615042837043,
-      "rewards/accuracy_reward": 0.7964912431804757,
+      "loss": 0.0327,
+      "reward": 0.7438596629782727,
+      "reward_std": 0.25195393279979106,
+      "rewards/accuracy_reward": 0.7438596629782727,
       "rewards/format_reward": 0.0,
       "step": 365
     },
     {
-      "completion_length": 570.056158126028,
+      "clip_ratio": 0.0,
+      "completion_length": 622.9333529823705,
       "epoch": 0.9373333333333334,
-      "grad_norm": 0.14333242177963257,
-      "kl": 0.008819580078125,
+      "grad_norm": 0.37333959341049194,
+      "kl": 0.4884661222759046,
       "learning_rate": 3.389491527319999e-08,
-      "loss": 0.0,
-      "reward": 0.7631579105791293,
-      "reward_std": 0.3113442129210422,
-      "rewards/accuracy_reward": 0.7631579105791293,
+      "loss": 0.0296,
+      "reward": 0.78070176604547,
+      "reward_std": 0.22106379866600037,
+      "rewards/accuracy_reward": 0.78070176604547,
       "rewards/format_reward": 0.0,
       "step": 370
     },
     {
-      "completion_length": 581.7122971384149,
+      "clip_ratio": 0.0,
+      "completion_length": 604.7368583277652,
       "epoch": 0.95,
-      "grad_norm": 0.12897399067878723,
-      "kl": 0.009299830386513159,
+      "grad_norm": 0.755801260471344,
+      "kl": 0.6571317973889803,
       "learning_rate": 2.127318274608381e-08,
-      "loss": 0.0,
-      "reward": 0.759649133211688,
-      "reward_std": 0.24736052971137198,
-      "rewards/accuracy_reward": 0.759649133211688,
+      "loss": 0.0439,
+      "reward": 0.7228070301444907,
+      "reward_std": 0.2571670673395458,
+      "rewards/accuracy_reward": 0.7228070301444907,
       "rewards/format_reward": 0.0,
       "step": 375
     },
     {
-      "completion_length": 519.6631728322883,
+      "clip_ratio": 0.0,
+      "completion_length": 571.0789629484477,
       "epoch": 0.9626666666666667,
-      "grad_norm": 0.046151068061590195,
-      "kl": 0.018807340923108554,
+      "grad_norm": 5.0163397789001465,
+      "kl": 1.0549049778988486,
       "learning_rate": 1.1562501925013125e-08,
-      "loss": 0.0,
-      "reward": 0.8105263286515286,
-      "reward_std": 0.24725342863484434,
-      "rewards/accuracy_reward": 0.8105263286515286,
+      "loss": 0.0801,
+      "reward": 0.7771929958933278,
+      "reward_std": 0.213723046214957,
+      "rewards/accuracy_reward": 0.7771929958933278,
       "rewards/format_reward": 0.0,
       "step": 380
     },
     {
-      "completion_length": 522.570188984118,
+      "clip_ratio": 0.0,
+      "completion_length": 611.6526505319696,
       "epoch": 0.9753333333333334,
-      "grad_norm": 0.10620765388011932,
-      "kl": 0.00986279939350329,
+      "grad_norm": 0.42803478240966797,
+      "kl": 0.45795432642886513,
       "learning_rate": 4.781989453874814e-09,
-      "loss": 0.0,
-      "reward": 0.7964912391022632,
-      "reward_std": 0.2455626440675635,
-      "rewards/accuracy_reward": 0.7964912391022632,
+      "loss": 0.0535,
+      "reward": 0.7070175552054455,
+      "reward_std": 0.26146441823557803,
+      "rewards/accuracy_reward": 0.7070175552054455,
       "rewards/format_reward": 0.0,
       "step": 385
     },
     {
-      "completion_length": 536.8315922787315,
+      "clip_ratio": 0.0,
+      "completion_length": 607.4087858501233,
       "epoch": 0.988,
-      "grad_norm": 0.09291122853755951,
-      "kl": 0.010211502878289473,
+      "grad_norm": 0.9341120719909668,
+      "kl": 0.3894335295024671,
       "learning_rate": 9.44993587509657e-10,
-      "loss": 0.0,
-      "reward": 0.8070175572445518,
-      "reward_std": 0.2619433641433716,
-      "rewards/accuracy_reward": 0.8070175572445518,
+      "loss": 0.0488,
+      "reward": 0.6964912394159719,
+      "reward_std": 0.2111501599612989,
+      "rewards/accuracy_reward": 0.6964912394159719,
       "rewards/format_reward": 0.0,
       "step": 390
     },
     {
-      "completion_length": 567.2982639513517,
+      "clip_ratio": 0.0,
+      "completion_length": 596.2105405707108,
       "epoch": 0.9981333333333333,
-      "kl": 0.00917715775339227,
-      "reward": 0.7828947472337046,
-      "reward_std": 0.2575972625299504,
-      "rewards/accuracy_reward": 0.7828947472337046,
+      "kl": 0.275941949141653,
+      "reward": 0.712719306937958,
+      "reward_std": 0.2265615369144239,
+      "rewards/accuracy_reward": 0.712719306937958,
       "rewards/format_reward": 0.0,
       "step": 394,
       "total_flos": 0.0,
-      "train_loss": 0.0008256121124141636,
-      "train_runtime": 419098.9196,
-      "train_samples_per_second": 0.018,
+      "train_loss": 0.055619138720809355,
+      "train_runtime": 265672.9627,
+      "train_samples_per_second": 0.028,
       "train_steps_per_second": 0.001
     }
   ],

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:84f3f7010102b0e305468c1cf34da3672c32048c60d1693fa81fc6817dc89fe5
-size 7480
+oid sha256:023a8c7771e881108583e1777ad0dd6684b0b57b5a5a0ef7320fae28cabfa220
+size 8056