Model save
Browse files- README.md +2 -4
- all_results.json +4 -9
- config.json +1 -1
- model.safetensors +1 -1
- runs/Jan31_22-35-38_ip-26-0-169-239/events.out.tfevents.1738363002.ip-26-0-169-239.841204.0 +3 -0
- train_results.json +4 -4
- trainer_state.json +871 -151
- training_args.bin +2 -2
README.md
CHANGED
|
@@ -1,11 +1,9 @@
|
|
| 1 |
---
|
| 2 |
base_model: deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B
|
| 3 |
-
datasets: AI-MO/NuminaMath-TIR
|
| 4 |
library_name: transformers
|
| 5 |
model_name: Qwen2.5-1.5B-Open-R1-GRPO
|
| 6 |
tags:
|
| 7 |
- generated_from_trainer
|
| 8 |
-
- open-r1
|
| 9 |
- trl
|
| 10 |
- grpo
|
| 11 |
licence: license
|
|
@@ -13,7 +11,7 @@ licence: license
|
|
| 13 |
|
| 14 |
# Model Card for Qwen2.5-1.5B-Open-R1-GRPO
|
| 15 |
|
| 16 |
-
This model is a fine-tuned version of [deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B](https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B)
|
| 17 |
It has been trained using [TRL](https://github.com/huggingface/trl).
|
| 18 |
|
| 19 |
## Quick start
|
|
@@ -29,7 +27,7 @@ print(output["generated_text"])
|
|
| 29 |
|
| 30 |
## Training procedure
|
| 31 |
|
| 32 |
-
[<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/huggingface/huggingface/runs/
|
| 33 |
|
| 34 |
|
| 35 |
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).
|
|
|
|
| 1 |
---
|
| 2 |
base_model: deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B
|
|
|
|
| 3 |
library_name: transformers
|
| 4 |
model_name: Qwen2.5-1.5B-Open-R1-GRPO
|
| 5 |
tags:
|
| 6 |
- generated_from_trainer
|
|
|
|
| 7 |
- trl
|
| 8 |
- grpo
|
| 9 |
licence: license
|
|
|
|
| 11 |
|
| 12 |
# Model Card for Qwen2.5-1.5B-Open-R1-GRPO
|
| 13 |
|
| 14 |
+
This model is a fine-tuned version of [deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B](https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B).
|
| 15 |
It has been trained using [TRL](https://github.com/huggingface/trl).
|
| 16 |
|
| 17 |
## Quick start
|
|
|
|
| 27 |
|
| 28 |
## Training procedure
|
| 29 |
|
| 30 |
+
[<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/huggingface/huggingface/runs/t256wwde)
|
| 31 |
|
| 32 |
|
| 33 |
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).
|
all_results.json
CHANGED
|
@@ -1,13 +1,8 @@
|
|
| 1 |
{
|
| 2 |
-
"eval_loss": 0.0400250144302845,
|
| 3 |
-
"eval_runtime": 45.4853,
|
| 4 |
-
"eval_samples": 99,
|
| 5 |
-
"eval_samples_per_second": 2.177,
|
| 6 |
-
"eval_steps_per_second": 0.022,
|
| 7 |
"total_flos": 0.0,
|
| 8 |
-
"train_loss": 0.
|
| 9 |
-
"train_runtime":
|
| 10 |
"train_samples": 72441,
|
| 11 |
-
"train_samples_per_second": 1.
|
| 12 |
-
"train_steps_per_second": 0.
|
| 13 |
}
|
|
|
|
| 1 |
{
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
"total_flos": 0.0,
|
| 3 |
+
"train_loss": 0.01602011715643949,
|
| 4 |
+
"train_runtime": 67938.2282,
|
| 5 |
"train_samples": 72441,
|
| 6 |
+
"train_samples_per_second": 1.066,
|
| 7 |
+
"train_steps_per_second": 0.01
|
| 8 |
}
|
config.json
CHANGED
|
@@ -23,7 +23,7 @@
|
|
| 23 |
"tie_word_embeddings": false,
|
| 24 |
"torch_dtype": "bfloat16",
|
| 25 |
"transformers_version": "4.48.2",
|
| 26 |
-
"use_cache":
|
| 27 |
"use_mrope": false,
|
| 28 |
"use_sliding_window": false,
|
| 29 |
"vocab_size": 151936
|
|
|
|
| 23 |
"tie_word_embeddings": false,
|
| 24 |
"torch_dtype": "bfloat16",
|
| 25 |
"transformers_version": "4.48.2",
|
| 26 |
+
"use_cache": false,
|
| 27 |
"use_mrope": false,
|
| 28 |
"use_sliding_window": false,
|
| 29 |
"vocab_size": 151936
|
model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 3554214752
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6bf1d4882b9a83e72271397bc79ecb7db9444234d1c6e30017d7791b27bb18aa
|
| 3 |
size 3554214752
|
runs/Jan31_22-35-38_ip-26-0-169-239/events.out.tfevents.1738363002.ip-26-0-169-239.841204.0
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3a55ad917e8e212f1627efdda6cfb148342496636a53269d145add11584f9535
|
| 3 |
+
size 16637
|
train_results.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"total_flos": 0.0,
|
| 3 |
-
"train_loss": 0.
|
| 4 |
-
"train_runtime":
|
| 5 |
"train_samples": 72441,
|
| 6 |
-
"train_samples_per_second": 1.
|
| 7 |
-
"train_steps_per_second": 0.
|
| 8 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"total_flos": 0.0,
|
| 3 |
+
"train_loss": 0.01602011715643949,
|
| 4 |
+
"train_runtime": 67938.2282,
|
| 5 |
"train_samples": 72441,
|
| 6 |
+
"train_samples_per_second": 1.066,
|
| 7 |
+
"train_steps_per_second": 0.01
|
| 8 |
}
|
trainer_state.json
CHANGED
|
@@ -1,227 +1,947 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
-
"epoch": 0.
|
| 5 |
"eval_steps": 100,
|
| 6 |
-
"global_step":
|
| 7 |
"is_hyper_param_search": false,
|
| 8 |
"is_local_process_zero": true,
|
| 9 |
"is_world_process_zero": true,
|
| 10 |
"log_history": [
|
| 11 |
{
|
| 12 |
-
"completion_length":
|
| 13 |
-
"epoch": 0.
|
| 14 |
-
"grad_norm": 0.
|
| 15 |
-
"kl": 0.
|
| 16 |
-
"learning_rate":
|
| 17 |
-
"loss": 0.
|
| 18 |
-
"reward": 0.
|
| 19 |
-
"reward_std": 0.
|
| 20 |
-
"rewards/accuracy_reward": 0.
|
| 21 |
"rewards/format_reward": 0.0,
|
| 22 |
"step": 10
|
| 23 |
},
|
| 24 |
{
|
| 25 |
-
"completion_length":
|
| 26 |
-
"epoch": 0.
|
| 27 |
-
"grad_norm": 0.
|
| 28 |
-
"kl": 0.
|
| 29 |
-
"learning_rate":
|
| 30 |
-
"loss": 0.
|
| 31 |
-
"reward": 0.
|
| 32 |
-
"reward_std": 0.
|
| 33 |
-
"rewards/accuracy_reward": 0.
|
| 34 |
"rewards/format_reward": 0.0,
|
| 35 |
"step": 20
|
| 36 |
},
|
| 37 |
{
|
| 38 |
-
"completion_length":
|
| 39 |
-
"epoch": 0.
|
| 40 |
-
"grad_norm": 0.
|
| 41 |
-
"kl": 0.
|
| 42 |
-
"learning_rate":
|
| 43 |
-
"loss": 0.
|
| 44 |
-
"reward": 0.
|
| 45 |
-
"reward_std": 0.
|
| 46 |
-
"rewards/accuracy_reward": 0.
|
| 47 |
"rewards/format_reward": 0.0,
|
| 48 |
"step": 30
|
| 49 |
},
|
| 50 |
{
|
| 51 |
-
"completion_length":
|
| 52 |
-
"epoch": 0.
|
| 53 |
-
"grad_norm": 0.
|
| 54 |
-
"kl": 0.
|
| 55 |
-
"learning_rate": 1.
|
| 56 |
-
"loss": 0.
|
| 57 |
-
"reward": 0.
|
| 58 |
-
"reward_std": 0.
|
| 59 |
-
"rewards/accuracy_reward": 0.
|
| 60 |
"rewards/format_reward": 0.0,
|
| 61 |
"step": 40
|
| 62 |
},
|
| 63 |
{
|
| 64 |
-
"completion_length":
|
| 65 |
-
"epoch": 0.
|
| 66 |
-
"grad_norm": 0.
|
| 67 |
-
"kl": 0.
|
| 68 |
-
"learning_rate": 1.
|
| 69 |
-
"loss": 0.
|
| 70 |
-
"reward": 0.
|
| 71 |
-
"reward_std": 0.
|
| 72 |
-
"rewards/accuracy_reward": 0.
|
| 73 |
"rewards/format_reward": 0.0,
|
| 74 |
"step": 50
|
| 75 |
},
|
| 76 |
{
|
| 77 |
-
"completion_length":
|
| 78 |
-
"epoch": 0.
|
| 79 |
-
"grad_norm": 0.
|
| 80 |
-
"kl": 0.
|
| 81 |
-
"learning_rate": 1.
|
| 82 |
-
"loss": 0.
|
| 83 |
-
"reward": 0.
|
| 84 |
-
"reward_std": 0.
|
| 85 |
-
"rewards/accuracy_reward": 0.
|
| 86 |
"rewards/format_reward": 0.0,
|
| 87 |
"step": 60
|
| 88 |
},
|
| 89 |
{
|
| 90 |
-
"completion_length":
|
| 91 |
-
"epoch": 0.
|
| 92 |
-
"grad_norm": 0.
|
| 93 |
-
"kl": 0.
|
| 94 |
-
"learning_rate": 1.
|
| 95 |
-
"loss": 0.
|
| 96 |
-
"reward": 0.
|
| 97 |
-
"reward_std": 0.
|
| 98 |
-
"rewards/accuracy_reward": 0.
|
| 99 |
"rewards/format_reward": 0.0,
|
| 100 |
"step": 70
|
| 101 |
},
|
| 102 |
{
|
| 103 |
-
"completion_length":
|
| 104 |
-
"epoch": 0.
|
| 105 |
-
"grad_norm": 0.
|
| 106 |
-
"kl": 0.
|
| 107 |
-
"learning_rate":
|
| 108 |
-
"loss": 0.
|
| 109 |
-
"reward": 0.
|
| 110 |
-
"reward_std": 0.
|
| 111 |
-
"rewards/accuracy_reward": 0.
|
| 112 |
"rewards/format_reward": 0.0,
|
| 113 |
"step": 80
|
| 114 |
},
|
| 115 |
{
|
| 116 |
-
"completion_length":
|
| 117 |
-
"epoch": 0.
|
| 118 |
-
"grad_norm": 0.
|
| 119 |
-
"kl": 0.
|
| 120 |
-
"learning_rate":
|
| 121 |
-
"loss": 0.
|
| 122 |
-
"reward": 0.
|
| 123 |
-
"reward_std": 0.
|
| 124 |
-
"rewards/accuracy_reward": 0.
|
| 125 |
"rewards/format_reward": 0.0,
|
| 126 |
"step": 90
|
| 127 |
},
|
| 128 |
{
|
| 129 |
-
"completion_length":
|
| 130 |
-
"epoch": 0.
|
| 131 |
-
"grad_norm": 0.
|
| 132 |
-
"kl": 0.
|
| 133 |
-
"learning_rate":
|
| 134 |
-
"loss": 0.
|
| 135 |
-
"reward": 0.
|
| 136 |
-
"reward_std": 0.
|
| 137 |
-
"rewards/accuracy_reward": 0.
|
| 138 |
"rewards/format_reward": 0.0,
|
| 139 |
"step": 100
|
| 140 |
},
|
| 141 |
{
|
| 142 |
-
"epoch": 0.
|
| 143 |
-
"eval_completion_length":
|
| 144 |
-
"eval_kl": 0.
|
| 145 |
-
"eval_loss": 0.
|
| 146 |
-
"eval_reward": 0.
|
| 147 |
-
"eval_reward_std": 0.
|
| 148 |
-
"eval_rewards/accuracy_reward": 0.
|
| 149 |
"eval_rewards/format_reward": 0.0,
|
| 150 |
-
"eval_runtime":
|
| 151 |
-
"eval_samples_per_second": 1.
|
| 152 |
-
"eval_steps_per_second": 0.
|
| 153 |
"step": 100
|
| 154 |
},
|
| 155 |
{
|
| 156 |
-
"completion_length":
|
| 157 |
-
"epoch": 0.
|
| 158 |
-
"grad_norm": 0.
|
| 159 |
-
"kl":
|
| 160 |
-
"learning_rate":
|
| 161 |
-
"loss": 0.
|
| 162 |
-
"reward": 0.
|
| 163 |
-
"reward_std": 0.
|
| 164 |
-
"rewards/accuracy_reward": 0.
|
| 165 |
"rewards/format_reward": 0.0,
|
| 166 |
"step": 110
|
| 167 |
},
|
| 168 |
{
|
| 169 |
-
"completion_length":
|
| 170 |
-
"epoch": 0.
|
| 171 |
-
"grad_norm": 0.
|
| 172 |
-
"kl": 0.
|
| 173 |
-
"learning_rate": 1.
|
| 174 |
-
"loss": 0.
|
| 175 |
-
"reward": 0.
|
| 176 |
-
"reward_std": 0.
|
| 177 |
-
"rewards/accuracy_reward": 0.
|
| 178 |
"rewards/format_reward": 0.0,
|
| 179 |
"step": 120
|
| 180 |
},
|
| 181 |
{
|
| 182 |
-
"completion_length":
|
| 183 |
-
"epoch": 0.
|
| 184 |
-
"grad_norm": 0.
|
| 185 |
-
"kl": 0.
|
| 186 |
-
"learning_rate":
|
| 187 |
-
"loss": 0.
|
| 188 |
-
"reward": 0.
|
| 189 |
-
"reward_std": 0.
|
| 190 |
-
"rewards/accuracy_reward": 0.
|
| 191 |
"rewards/format_reward": 0.0,
|
| 192 |
"step": 130
|
| 193 |
},
|
| 194 |
{
|
| 195 |
-
"completion_length":
|
| 196 |
-
"epoch": 0.
|
| 197 |
-
"grad_norm": 0.
|
| 198 |
-
"kl":
|
| 199 |
-
"learning_rate":
|
| 200 |
-
"loss": 0.
|
| 201 |
-
"reward": 0.
|
| 202 |
-
"reward_std": 0.
|
| 203 |
-
"rewards/accuracy_reward": 0.
|
| 204 |
"rewards/format_reward": 0.0,
|
| 205 |
"step": 140
|
| 206 |
},
|
| 207 |
{
|
| 208 |
-
"completion_length":
|
| 209 |
-
"epoch": 0.
|
| 210 |
-
"
|
| 211 |
-
"
|
| 212 |
-
"
|
| 213 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 214 |
"rewards/format_reward": 0.0,
|
| 215 |
-
"step":
|
| 216 |
"total_flos": 0.0,
|
| 217 |
-
"train_loss": 0.
|
| 218 |
-
"train_runtime":
|
| 219 |
-
"train_samples_per_second": 1.
|
| 220 |
-
"train_steps_per_second": 0.
|
| 221 |
}
|
| 222 |
],
|
| 223 |
"logging_steps": 10,
|
| 224 |
-
"max_steps":
|
| 225 |
"num_input_tokens_seen": 0,
|
| 226 |
"num_train_epochs": 1,
|
| 227 |
"save_steps": 500,
|
|
@@ -238,7 +958,7 @@
|
|
| 238 |
}
|
| 239 |
},
|
| 240 |
"total_flos": 0.0,
|
| 241 |
-
"train_batch_size":
|
| 242 |
"trial_name": null,
|
| 243 |
"trial_params": null
|
| 244 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.9987438399845395,
|
| 5 |
"eval_steps": 100,
|
| 6 |
+
"global_step": 646,
|
| 7 |
"is_hyper_param_search": false,
|
| 8 |
"is_local_process_zero": true,
|
| 9 |
"is_world_process_zero": true,
|
| 10 |
"log_history": [
|
| 11 |
{
|
| 12 |
+
"completion_length": 890.5919452667237,
|
| 13 |
+
"epoch": 0.015460430959512996,
|
| 14 |
+
"grad_norm": 0.0048114829798690025,
|
| 15 |
+
"kl": 0.0004779815673828125,
|
| 16 |
+
"learning_rate": 3.0769230769230774e-06,
|
| 17 |
+
"loss": 0.0,
|
| 18 |
+
"reward": 0.2079081600648351,
|
| 19 |
+
"reward_std": 0.18389849485829474,
|
| 20 |
+
"rewards/accuracy_reward": 0.2079081600648351,
|
| 21 |
"rewards/format_reward": 0.0,
|
| 22 |
"step": 10
|
| 23 |
},
|
| 24 |
{
|
| 25 |
+
"completion_length": 815.0035549163819,
|
| 26 |
+
"epoch": 0.03092086191902599,
|
| 27 |
+
"grad_norm": 0.0037700873930097754,
|
| 28 |
+
"kl": 0.005376839637756347,
|
| 29 |
+
"learning_rate": 6.153846153846155e-06,
|
| 30 |
+
"loss": 0.0002,
|
| 31 |
+
"reward": 0.3485969334375113,
|
| 32 |
+
"reward_std": 0.20335620292462409,
|
| 33 |
+
"rewards/accuracy_reward": 0.3485969334375113,
|
| 34 |
"rewards/format_reward": 0.0,
|
| 35 |
"step": 20
|
| 36 |
},
|
| 37 |
{
|
| 38 |
+
"completion_length": 783.1976894378662,
|
| 39 |
+
"epoch": 0.04638129287853899,
|
| 40 |
+
"grad_norm": 0.003633377980625143,
|
| 41 |
+
"kl": 0.008260059356689452,
|
| 42 |
+
"learning_rate": 9.230769230769232e-06,
|
| 43 |
+
"loss": 0.0003,
|
| 44 |
+
"reward": 0.41823978765169156,
|
| 45 |
+
"reward_std": 0.214983982546255,
|
| 46 |
+
"rewards/accuracy_reward": 0.41823978765169156,
|
| 47 |
"rewards/format_reward": 0.0,
|
| 48 |
"step": 30
|
| 49 |
},
|
| 50 |
{
|
| 51 |
+
"completion_length": 796.1503639221191,
|
| 52 |
+
"epoch": 0.06184172383805198,
|
| 53 |
+
"grad_norm": 0.004207792051316878,
|
| 54 |
+
"kl": 0.013314247131347656,
|
| 55 |
+
"learning_rate": 1.230769230769231e-05,
|
| 56 |
+
"loss": 0.0005,
|
| 57 |
+
"reward": 0.45267856353893876,
|
| 58 |
+
"reward_std": 0.22209063512273133,
|
| 59 |
+
"rewards/accuracy_reward": 0.45267856353893876,
|
| 60 |
"rewards/format_reward": 0.0,
|
| 61 |
"step": 40
|
| 62 |
},
|
| 63 |
{
|
| 64 |
+
"completion_length": 707.4483289718628,
|
| 65 |
+
"epoch": 0.07730215479756498,
|
| 66 |
+
"grad_norm": 0.0039377851262065435,
|
| 67 |
+
"kl": 0.023092269897460938,
|
| 68 |
+
"learning_rate": 1.5384615384615387e-05,
|
| 69 |
+
"loss": 0.0009,
|
| 70 |
+
"reward": 0.5378826450556516,
|
| 71 |
+
"reward_std": 0.2228247532621026,
|
| 72 |
+
"rewards/accuracy_reward": 0.5378826450556516,
|
| 73 |
"rewards/format_reward": 0.0,
|
| 74 |
"step": 50
|
| 75 |
},
|
| 76 |
{
|
| 77 |
+
"completion_length": 738.098327255249,
|
| 78 |
+
"epoch": 0.09276258575707798,
|
| 79 |
+
"grad_norm": 0.004922000110022058,
|
| 80 |
+
"kl": 0.03480720520019531,
|
| 81 |
+
"learning_rate": 1.8461538461538465e-05,
|
| 82 |
+
"loss": 0.0014,
|
| 83 |
+
"reward": 0.5255101950955577,
|
| 84 |
+
"reward_std": 0.21384936766698956,
|
| 85 |
+
"rewards/accuracy_reward": 0.5255101950955577,
|
| 86 |
"rewards/format_reward": 0.0,
|
| 87 |
"step": 60
|
| 88 |
},
|
| 89 |
{
|
| 90 |
+
"completion_length": 714.9747295379639,
|
| 91 |
+
"epoch": 0.10822301671659097,
|
| 92 |
+
"grad_norm": 0.025813508862428863,
|
| 93 |
+
"kl": 0.11458740234375,
|
| 94 |
+
"learning_rate": 1.999634547413886e-05,
|
| 95 |
+
"loss": 0.0046,
|
| 96 |
+
"reward": 0.5369897866621614,
|
| 97 |
+
"reward_std": 0.24942327085882426,
|
| 98 |
+
"rewards/accuracy_reward": 0.5369897866621614,
|
| 99 |
"rewards/format_reward": 0.0,
|
| 100 |
"step": 70
|
| 101 |
},
|
| 102 |
{
|
| 103 |
+
"completion_length": 729.6705226898193,
|
| 104 |
+
"epoch": 0.12368344767610397,
|
| 105 |
+
"grad_norm": 0.004934077059395142,
|
| 106 |
+
"kl": 0.1241455078125,
|
| 107 |
+
"learning_rate": 1.9967125291968495e-05,
|
| 108 |
+
"loss": 0.005,
|
| 109 |
+
"reward": 0.5182397849857807,
|
| 110 |
+
"reward_std": 0.22350065293721855,
|
| 111 |
+
"rewards/accuracy_reward": 0.5182397849857807,
|
| 112 |
"rewards/format_reward": 0.0,
|
| 113 |
"step": 80
|
| 114 |
},
|
| 115 |
{
|
| 116 |
+
"completion_length": 752.231746673584,
|
| 117 |
+
"epoch": 0.13914387863561697,
|
| 118 |
+
"grad_norm": 0.004490175226607394,
|
| 119 |
+
"kl": 0.1960784912109375,
|
| 120 |
+
"learning_rate": 1.990877034074683e-05,
|
| 121 |
+
"loss": 0.0078,
|
| 122 |
+
"reward": 0.47576529716607185,
|
| 123 |
+
"reward_std": 0.21953069823794066,
|
| 124 |
+
"rewards/accuracy_reward": 0.47576529716607185,
|
| 125 |
"rewards/format_reward": 0.0,
|
| 126 |
"step": 90
|
| 127 |
},
|
| 128 |
{
|
| 129 |
+
"completion_length": 735.3539382934571,
|
| 130 |
+
"epoch": 0.15460430959512997,
|
| 131 |
+
"grad_norm": 0.11798471666893984,
|
| 132 |
+
"kl": 0.30410003662109375,
|
| 133 |
+
"learning_rate": 1.9821451197042028e-05,
|
| 134 |
+
"loss": 0.0122,
|
| 135 |
+
"reward": 0.4843112153466791,
|
| 136 |
+
"reward_std": 0.22822934831492603,
|
| 137 |
+
"rewards/accuracy_reward": 0.4843112153466791,
|
| 138 |
"rewards/format_reward": 0.0,
|
| 139 |
"step": 100
|
| 140 |
},
|
| 141 |
{
|
| 142 |
+
"epoch": 0.15460430959512997,
|
| 143 |
+
"eval_completion_length": 649.0448852539063,
|
| 144 |
+
"eval_kl": 0.27952473958333335,
|
| 145 |
+
"eval_loss": 0.009933823719620705,
|
| 146 |
+
"eval_reward": 0.45850338935852053,
|
| 147 |
+
"eval_reward_std": 0.26114755471547446,
|
| 148 |
+
"eval_rewards/accuracy_reward": 0.45850338935852053,
|
| 149 |
"eval_rewards/format_reward": 0.0,
|
| 150 |
+
"eval_runtime": 88.3141,
|
| 151 |
+
"eval_samples_per_second": 1.121,
|
| 152 |
+
"eval_steps_per_second": 0.17,
|
| 153 |
"step": 100
|
| 154 |
},
|
| 155 |
{
|
| 156 |
+
"completion_length": 732.8517738342285,
|
| 157 |
+
"epoch": 0.17006474055464296,
|
| 158 |
+
"grad_norm": 0.02615303400330413,
|
| 159 |
+
"kl": 1.07412109375,
|
| 160 |
+
"learning_rate": 1.9705423102261324e-05,
|
| 161 |
+
"loss": 0.043,
|
| 162 |
+
"reward": 0.3991071363911033,
|
| 163 |
+
"reward_std": 0.2730441292747855,
|
| 164 |
+
"rewards/accuracy_reward": 0.3991071363911033,
|
| 165 |
"rewards/format_reward": 0.0,
|
| 166 |
"step": 110
|
| 167 |
},
|
| 168 |
{
|
| 169 |
+
"completion_length": 792.610315322876,
|
| 170 |
+
"epoch": 0.18552517151415596,
|
| 171 |
+
"grad_norm": 0.0287118341101585,
|
| 172 |
+
"kl": 0.542120361328125,
|
| 173 |
+
"learning_rate": 1.956102521655831e-05,
|
| 174 |
+
"loss": 0.0217,
|
| 175 |
+
"reward": 0.3727040741709061,
|
| 176 |
+
"reward_std": 0.26482684616930785,
|
| 177 |
+
"rewards/accuracy_reward": 0.3727040741709061,
|
| 178 |
"rewards/format_reward": 0.0,
|
| 179 |
"step": 120
|
| 180 |
},
|
| 181 |
{
|
| 182 |
+
"completion_length": 742.854447555542,
|
| 183 |
+
"epoch": 0.20098560247366895,
|
| 184 |
+
"grad_norm": 0.026940670011693296,
|
| 185 |
+
"kl": 0.4710235595703125,
|
| 186 |
+
"learning_rate": 1.9388679627438486e-05,
|
| 187 |
+
"loss": 0.0188,
|
| 188 |
+
"reward": 0.46033162334933875,
|
| 189 |
+
"reward_std": 0.24005382088944316,
|
| 190 |
+
"rewards/accuracy_reward": 0.46033162334933875,
|
| 191 |
"rewards/format_reward": 0.0,
|
| 192 |
"step": 130
|
| 193 |
},
|
| 194 |
{
|
| 195 |
+
"completion_length": 747.0299587249756,
|
| 196 |
+
"epoch": 0.21644603343318194,
|
| 197 |
+
"grad_norm": 0.044131709279297326,
|
| 198 |
+
"kl": 0.3888519287109375,
|
| 199 |
+
"learning_rate": 1.9188890115960967e-05,
|
| 200 |
+
"loss": 0.0156,
|
| 201 |
+
"reward": 0.4734693782404065,
|
| 202 |
+
"reward_std": 0.2484902088996023,
|
| 203 |
+
"rewards/accuracy_reward": 0.4734693782404065,
|
| 204 |
"rewards/format_reward": 0.0,
|
| 205 |
"step": 140
|
| 206 |
},
|
| 207 |
{
|
| 208 |
+
"completion_length": 719.1231998443603,
|
| 209 |
+
"epoch": 0.23190646439269494,
|
| 210 |
+
"grad_norm": 0.04685629239096149,
|
| 211 |
+
"kl": 0.6114959716796875,
|
| 212 |
+
"learning_rate": 1.8962240684142923e-05,
|
| 213 |
+
"loss": 0.0245,
|
| 214 |
+
"reward": 0.4772959094494581,
|
| 215 |
+
"reward_std": 0.26747574456967416,
|
| 216 |
+
"rewards/accuracy_reward": 0.4772959094494581,
|
| 217 |
+
"rewards/format_reward": 0.0,
|
| 218 |
+
"step": 150
|
| 219 |
+
},
|
| 220 |
+
{
|
| 221 |
+
"completion_length": 752.0598068237305,
|
| 222 |
+
"epoch": 0.24736689535220793,
|
| 223 |
+
"grad_norm": 0.00786304561850755,
|
| 224 |
+
"kl": 0.27758941650390623,
|
| 225 |
+
"learning_rate": 1.8709393847871146e-05,
|
| 226 |
+
"loss": 0.0111,
|
| 227 |
+
"reward": 0.4918367238715291,
|
| 228 |
+
"reward_std": 0.24101508525200188,
|
| 229 |
+
"rewards/accuracy_reward": 0.4918367238715291,
|
| 230 |
+
"rewards/format_reward": 0.0,
|
| 231 |
+
"step": 160
|
| 232 |
+
},
|
| 233 |
+
{
|
| 234 |
+
"completion_length": 713.9837879180908,
|
| 235 |
+
"epoch": 0.26282732631172095,
|
| 236 |
+
"grad_norm": 0.005284057529020878,
|
| 237 |
+
"kl": 0.22295989990234374,
|
| 238 |
+
"learning_rate": 1.8431088700310846e-05,
|
| 239 |
+
"loss": 0.0089,
|
| 240 |
+
"reward": 0.5107142773456872,
|
| 241 |
+
"reward_std": 0.2484118543099612,
|
| 242 |
+
"rewards/accuracy_reward": 0.5107142773456872,
|
| 243 |
+
"rewards/format_reward": 0.0,
|
| 244 |
+
"step": 170
|
| 245 |
+
},
|
| 246 |
+
{
|
| 247 |
+
"completion_length": 721.1976871490479,
|
| 248 |
+
"epoch": 0.27828775727123395,
|
| 249 |
+
"grad_norm": 0.0059488370043247725,
|
| 250 |
+
"kl": 0.27528076171875,
|
| 251 |
+
"learning_rate": 1.8128138751472432e-05,
|
| 252 |
+
"loss": 0.011,
|
| 253 |
+
"reward": 0.4794642778113484,
|
| 254 |
+
"reward_std": 0.28188897627405823,
|
| 255 |
+
"rewards/accuracy_reward": 0.4794642778113484,
|
| 256 |
+
"rewards/format_reward": 0.0,
|
| 257 |
+
"step": 180
|
| 258 |
+
},
|
| 259 |
+
{
|
| 260 |
+
"completion_length": 685.0909267425537,
|
| 261 |
+
"epoch": 0.29374818823074694,
|
| 262 |
+
"grad_norm": 0.024776439473654245,
|
| 263 |
+
"kl": 0.3252960205078125,
|
| 264 |
+
"learning_rate": 1.780142955025139e-05,
|
| 265 |
+
"loss": 0.013,
|
| 266 |
+
"reward": 0.49030611482448877,
|
| 267 |
+
"reward_std": 0.29441971494816244,
|
| 268 |
+
"rewards/accuracy_reward": 0.49030611482448877,
|
| 269 |
+
"rewards/format_reward": 0.0,
|
| 270 |
+
"step": 190
|
| 271 |
+
},
|
| 272 |
+
{
|
| 273 |
+
"completion_length": 688.9476896286011,
|
| 274 |
+
"epoch": 0.30920861919025994,
|
| 275 |
+
"grad_norm": 0.018511152998280257,
|
| 276 |
+
"kl": 0.407830810546875,
|
| 277 |
+
"learning_rate": 1.745191609589231e-05,
|
| 278 |
+
"loss": 0.0163,
|
| 279 |
+
"reward": 0.4463010121136904,
|
| 280 |
+
"reward_std": 0.29204082568176093,
|
| 281 |
+
"rewards/accuracy_reward": 0.4463010121136904,
|
| 282 |
+
"rewards/format_reward": 0.0,
|
| 283 |
+
"step": 200
|
| 284 |
+
},
|
| 285 |
+
{
|
| 286 |
+
"epoch": 0.30920861919025994,
|
| 287 |
+
"eval_completion_length": 684.5197184244792,
|
| 288 |
+
"eval_kl": 0.31907552083333335,
|
| 289 |
+
"eval_loss": 0.013225244358181953,
|
| 290 |
+
"eval_reward": 0.5034013529618581,
|
| 291 |
+
"eval_reward_std": 0.26736749211947125,
|
| 292 |
+
"eval_rewards/accuracy_reward": 0.5034013529618581,
|
| 293 |
+
"eval_rewards/format_reward": 0.0,
|
| 294 |
+
"eval_runtime": 88.6591,
|
| 295 |
+
"eval_samples_per_second": 1.117,
|
| 296 |
+
"eval_steps_per_second": 0.169,
|
| 297 |
+
"step": 200
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"completion_length": 688.203684425354,
|
| 301 |
+
"epoch": 0.32466905014977293,
|
| 302 |
+
"grad_norm": 0.04166474923994624,
|
| 303 |
+
"kl": 0.4008056640625,
|
| 304 |
+
"learning_rate": 1.7080620046443503e-05,
|
| 305 |
+
"loss": 0.016,
|
| 306 |
+
"reward": 0.45382652347907426,
|
| 307 |
+
"reward_std": 0.27682951451279225,
|
| 308 |
+
"rewards/accuracy_reward": 0.45382652347907426,
|
| 309 |
+
"rewards/format_reward": 0.0,
|
| 310 |
+
"step": 210
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"completion_length": 671.052919960022,
|
| 314 |
+
"epoch": 0.3401294811092859,
|
| 315 |
+
"grad_norm": 0.03352718681096581,
|
| 316 |
+
"kl": 0.3749359130859375,
|
| 317 |
+
"learning_rate": 1.6688626732362192e-05,
|
| 318 |
+
"loss": 0.015,
|
| 319 |
+
"reward": 0.4804846870712936,
|
| 320 |
+
"reward_std": 0.2816257219295949,
|
| 321 |
+
"rewards/accuracy_reward": 0.4804846870712936,
|
| 322 |
+
"rewards/format_reward": 0.0,
|
| 323 |
+
"step": 220
|
| 324 |
+
},
|
| 325 |
+
{
|
| 326 |
+
"completion_length": 643.8374883651734,
|
| 327 |
+
"epoch": 0.3555899120687989,
|
| 328 |
+
"grad_norm": 0.020351629255596172,
|
| 329 |
+
"kl": 0.41683349609375,
|
| 330 |
+
"learning_rate": 1.6277081983999742e-05,
|
| 331 |
+
"loss": 0.0167,
|
| 332 |
+
"reward": 0.4418367271311581,
|
| 333 |
+
"reward_std": 0.28084711018018427,
|
| 334 |
+
"rewards/accuracy_reward": 0.4418367271311581,
|
| 335 |
+
"rewards/format_reward": 0.0,
|
| 336 |
+
"step": 230
|
| 337 |
+
},
|
| 338 |
+
{
|
| 339 |
+
"completion_length": 676.2223068237305,
|
| 340 |
+
"epoch": 0.3710503430283119,
|
| 341 |
+
"grad_norm": 0.05256794474201724,
|
| 342 |
+
"kl": 0.4444091796875,
|
| 343 |
+
"learning_rate": 1.5847188782240473e-05,
|
| 344 |
+
"loss": 0.0178,
|
| 345 |
+
"reward": 0.4517857059370726,
|
| 346 |
+
"reward_std": 0.29922230960801244,
|
| 347 |
+
"rewards/accuracy_reward": 0.4517857059370726,
|
| 348 |
+
"rewards/format_reward": 0.0,
|
| 349 |
+
"step": 240
|
| 350 |
+
},
|
| 351 |
+
{
|
| 352 |
+
"completion_length": 668.4691181182861,
|
| 353 |
+
"epoch": 0.3865107739878249,
|
| 354 |
+
"grad_norm": 0.03902146766301656,
|
| 355 |
+
"kl": 0.5255126953125,
|
| 356 |
+
"learning_rate": 1.5400203742084508e-05,
|
| 357 |
+
"loss": 0.021,
|
| 358 |
+
"reward": 0.4383928496390581,
|
| 359 |
+
"reward_std": 0.2969975466839969,
|
| 360 |
+
"rewards/accuracy_reward": 0.4383928496390581,
|
| 361 |
+
"rewards/format_reward": 0.0,
|
| 362 |
+
"step": 250
|
| 363 |
+
},
|
| 364 |
+
{
|
| 365 |
+
"completion_length": 603.6825130462646,
|
| 366 |
+
"epoch": 0.4019712049473379,
|
| 367 |
+
"grad_norm": 0.02207794461900282,
|
| 368 |
+
"kl": 0.4320556640625,
|
| 369 |
+
"learning_rate": 1.4937433439453465e-05,
|
| 370 |
+
"loss": 0.0173,
|
| 371 |
+
"reward": 0.45331631754525004,
|
| 372 |
+
"reward_std": 0.29558597495779393,
|
| 373 |
+
"rewards/accuracy_reward": 0.45331631754525004,
|
| 374 |
+
"rewards/format_reward": 0.0,
|
| 375 |
+
"step": 260
|
| 376 |
+
},
|
| 377 |
+
{
|
| 378 |
+
"completion_length": 640.7654218673706,
|
| 379 |
+
"epoch": 0.4174316359068509,
|
| 380 |
+
"grad_norm": 0.02435264101678905,
|
| 381 |
+
"kl": 0.510498046875,
|
| 382 |
+
"learning_rate": 1.4460230591956097e-05,
|
| 383 |
+
"loss": 0.0204,
|
| 384 |
+
"reward": 0.4378826460801065,
|
| 385 |
+
"reward_std": 0.2932774598710239,
|
| 386 |
+
"rewards/accuracy_reward": 0.4378826460801065,
|
| 387 |
+
"rewards/format_reward": 0.0,
|
| 388 |
+
"step": 270
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"completion_length": 645.673583984375,
|
| 392 |
+
"epoch": 0.4328920668663639,
|
| 393 |
+
"grad_norm": 0.08816250567178784,
|
| 394 |
+
"kl": 0.658056640625,
|
| 395 |
+
"learning_rate": 1.3969990104777712e-05,
|
| 396 |
+
"loss": 0.0263,
|
| 397 |
+
"reward": 0.42423468651250007,
|
| 398 |
+
"reward_std": 0.2902136994060129,
|
| 399 |
+
"rewards/accuracy_reward": 0.42423468651250007,
|
| 400 |
+
"rewards/format_reward": 0.0,
|
| 401 |
+
"step": 280
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"completion_length": 620.3529205322266,
|
| 405 |
+
"epoch": 0.4483524978258769,
|
| 406 |
+
"grad_norm": 0.03880404727089487,
|
| 407 |
+
"kl": 0.92225341796875,
|
| 408 |
+
"learning_rate": 1.3468144993251735e-05,
|
| 409 |
+
"loss": 0.0369,
|
| 410 |
+
"reward": 0.45433672657236457,
|
| 411 |
+
"reward_std": 0.29660415309481325,
|
| 412 |
+
"rewards/accuracy_reward": 0.45433672657236457,
|
| 413 |
+
"rewards/format_reward": 0.0,
|
| 414 |
+
"step": 290
|
| 415 |
+
},
|
| 416 |
+
{
|
| 417 |
+
"completion_length": 662.0209051132202,
|
| 418 |
+
"epoch": 0.4638129287853899,
|
| 419 |
+
"grad_norm": 0.0580905052695571,
|
| 420 |
+
"kl": 1.16871337890625,
|
| 421 |
+
"learning_rate": 1.295616219403197e-05,
|
| 422 |
+
"loss": 0.0468,
|
| 423 |
+
"reward": 0.4096938706934452,
|
| 424 |
+
"reward_std": 0.3028755730483681,
|
| 425 |
+
"rewards/accuracy_reward": 0.4096938706934452,
|
| 426 |
+
"rewards/format_reward": 0.0,
|
| 427 |
+
"step": 300
|
| 428 |
+
},
|
| 429 |
+
{
|
| 430 |
+
"epoch": 0.4638129287853899,
|
| 431 |
+
"eval_completion_length": 604.8503275553386,
|
| 432 |
+
"eval_kl": 1.15458984375,
|
| 433 |
+
"eval_loss": 0.04879453405737877,
|
| 434 |
+
"eval_reward": 0.4707482943932215,
|
| 435 |
+
"eval_reward_std": 0.27726571063200633,
|
| 436 |
+
"eval_rewards/accuracy_reward": 0.4707482943932215,
|
| 437 |
+
"eval_rewards/format_reward": 0.0,
|
| 438 |
+
"eval_runtime": 85.9413,
|
| 439 |
+
"eval_samples_per_second": 1.152,
|
| 440 |
+
"eval_steps_per_second": 0.175,
|
| 441 |
+
"step": 300
|
| 442 |
+
},
|
| 443 |
+
{
|
| 444 |
+
"completion_length": 618.5189908981323,
|
| 445 |
+
"epoch": 0.47927335974490287,
|
| 446 |
+
"grad_norm": 0.04382094795523676,
|
| 447 |
+
"kl": 0.9635986328125,
|
| 448 |
+
"learning_rate": 1.2435538277109919e-05,
|
| 449 |
+
"loss": 0.0385,
|
| 450 |
+
"reward": 0.44183672638610005,
|
| 451 |
+
"reward_std": 0.28991906996816397,
|
| 452 |
+
"rewards/accuracy_reward": 0.44183672638610005,
|
| 453 |
+
"rewards/format_reward": 0.0,
|
| 454 |
+
"step": 310
|
| 455 |
+
},
|
| 456 |
+
{
|
| 457 |
+
"completion_length": 576.9373609542847,
|
| 458 |
+
"epoch": 0.49473379070441587,
|
| 459 |
+
"grad_norm": 0.03238899986450347,
|
| 460 |
+
"kl": 0.90596923828125,
|
| 461 |
+
"learning_rate": 1.19077950712113e-05,
|
| 462 |
+
"loss": 0.0362,
|
| 463 |
+
"reward": 0.4463010119274259,
|
| 464 |
+
"reward_std": 0.27278245403431356,
|
| 465 |
+
"rewards/accuracy_reward": 0.4463010119274259,
|
| 466 |
+
"rewards/format_reward": 0.0,
|
| 467 |
+
"step": 320
|
| 468 |
+
},
|
| 469 |
+
{
|
| 470 |
+
"completion_length": 535.9521591186524,
|
| 471 |
+
"epoch": 0.5101942216639289,
|
| 472 |
+
"grad_norm": 0.018431343598236234,
|
| 473 |
+
"kl": 0.5312744140625,
|
| 474 |
+
"learning_rate": 1.137447521535908e-05,
|
| 475 |
+
"loss": 0.0213,
|
| 476 |
+
"reward": 0.5080357047729194,
|
| 477 |
+
"reward_std": 0.2620480744168162,
|
| 478 |
+
"rewards/accuracy_reward": 0.5080357047729194,
|
| 479 |
+
"rewards/format_reward": 0.0,
|
| 480 |
+
"step": 330
|
| 481 |
+
},
|
| 482 |
+
{
|
| 483 |
+
"completion_length": 646.3938673019409,
|
| 484 |
+
"epoch": 0.5256546526234419,
|
| 485 |
+
"grad_norm": 0.02317552951279719,
|
| 486 |
+
"kl": 0.513262939453125,
|
| 487 |
+
"learning_rate": 1.0837137649606241e-05,
|
| 488 |
+
"loss": 0.0205,
|
| 489 |
+
"reward": 0.44579080818220973,
|
| 490 |
+
"reward_std": 0.29469514368101957,
|
| 491 |
+
"rewards/accuracy_reward": 0.44579080818220973,
|
| 492 |
+
"rewards/format_reward": 0.0,
|
| 493 |
+
"step": 340
|
| 494 |
+
},
|
| 495 |
+
{
|
| 496 |
+
"completion_length": 586.0761362075806,
|
| 497 |
+
"epoch": 0.5411150835829549,
|
| 498 |
+
"grad_norm": 0.016226017261457597,
|
| 499 |
+
"kl": 0.3769073486328125,
|
| 500 |
+
"learning_rate": 1.0297353058119209e-05,
|
| 501 |
+
"loss": 0.0151,
|
| 502 |
+
"reward": 0.4859693799167871,
|
| 503 |
+
"reward_std": 0.2713043099734932,
|
| 504 |
+
"rewards/accuracy_reward": 0.4859693799167871,
|
| 505 |
+
"rewards/format_reward": 0.0,
|
| 506 |
+
"step": 350
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"completion_length": 571.9863378524781,
|
| 510 |
+
"epoch": 0.5565755145424679,
|
| 511 |
+
"grad_norm": 0.0504216391704931,
|
| 512 |
+
"kl": 0.599774169921875,
|
| 513 |
+
"learning_rate": 9.756699277932196e-06,
|
| 514 |
+
"loss": 0.024,
|
| 515 |
+
"reward": 0.48443876539822667,
|
| 516 |
+
"reward_std": 0.28054420156404375,
|
| 517 |
+
"rewards/accuracy_reward": 0.48443876539822667,
|
| 518 |
+
"rewards/format_reward": 0.0,
|
| 519 |
+
"step": 360
|
| 520 |
+
},
|
| 521 |
+
{
|
| 522 |
+
"completion_length": 563.5765197753906,
|
| 523 |
+
"epoch": 0.5720359455019809,
|
| 524 |
+
"grad_norm": 0.13990571845181696,
|
| 525 |
+
"kl": 0.52132568359375,
|
| 526 |
+
"learning_rate": 9.216756686793163e-06,
|
| 527 |
+
"loss": 0.0209,
|
| 528 |
+
"reward": 0.45790815523359923,
|
| 529 |
+
"reward_std": 0.26586609268561007,
|
| 530 |
+
"rewards/accuracy_reward": 0.45790815523359923,
|
| 531 |
+
"rewards/format_reward": 0.0,
|
| 532 |
+
"step": 370
|
| 533 |
+
},
|
| 534 |
+
{
|
| 535 |
+
"completion_length": 575.829453086853,
|
| 536 |
+
"epoch": 0.5874963764614939,
|
| 537 |
+
"grad_norm": 0.013870992065908727,
|
| 538 |
+
"kl": 0.330126953125,
|
| 539 |
+
"learning_rate": 8.67910358358298e-06,
|
| 540 |
+
"loss": 0.0132,
|
| 541 |
+
"reward": 0.4734693790320307,
|
| 542 |
+
"reward_std": 0.27416237886063755,
|
| 543 |
+
"rewards/accuracy_reward": 0.4734693790320307,
|
| 544 |
+
"rewards/format_reward": 0.0,
|
| 545 |
+
"step": 380
|
| 546 |
+
},
|
| 547 |
+
{
|
| 548 |
+
"completion_length": 568.1160606384277,
|
| 549 |
+
"epoch": 0.6029568074210069,
|
| 550 |
+
"grad_norm": 0.010163464330427247,
|
| 551 |
+
"kl": 0.2426544189453125,
|
| 552 |
+
"learning_rate": 8.145311574811325e-06,
|
| 553 |
+
"loss": 0.0097,
|
| 554 |
+
"reward": 0.4954081534408033,
|
| 555 |
+
"reward_std": 0.26710083298385146,
|
| 556 |
+
"rewards/accuracy_reward": 0.4954081534408033,
|
| 557 |
+
"rewards/format_reward": 0.0,
|
| 558 |
+
"step": 390
|
| 559 |
+
},
|
| 560 |
+
{
|
| 561 |
+
"completion_length": 665.6186065673828,
|
| 562 |
+
"epoch": 0.6184172383805199,
|
| 563 |
+
"grad_norm": 0.013029334976500356,
|
| 564 |
+
"kl": 0.321246337890625,
|
| 565 |
+
"learning_rate": 7.616940980675004e-06,
|
| 566 |
+
"loss": 0.0128,
|
| 567 |
+
"reward": 0.4474489719606936,
|
| 568 |
+
"reward_std": 0.27854115669615565,
|
| 569 |
+
"rewards/accuracy_reward": 0.4474489719606936,
|
| 570 |
+
"rewards/format_reward": 0.0,
|
| 571 |
+
"step": 400
|
| 572 |
+
},
|
| 573 |
+
{
|
| 574 |
+
"epoch": 0.6184172383805199,
|
| 575 |
+
"eval_completion_length": 627.0203979492187,
|
| 576 |
+
"eval_kl": 0.2732747395833333,
|
| 577 |
+
"eval_loss": 0.011420400813221931,
|
| 578 |
+
"eval_reward": 0.5047619010011355,
|
| 579 |
+
"eval_reward_std": 0.2536137938499451,
|
| 580 |
+
"eval_rewards/accuracy_reward": 0.5047619010011355,
|
| 581 |
+
"eval_rewards/format_reward": 0.0,
|
| 582 |
+
"eval_runtime": 87.816,
|
| 583 |
+
"eval_samples_per_second": 1.127,
|
| 584 |
+
"eval_steps_per_second": 0.171,
|
| 585 |
+
"step": 400
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"completion_length": 646.4711584091186,
|
| 589 |
+
"epoch": 0.6338776693400329,
|
| 590 |
+
"grad_norm": 0.023099135498393764,
|
| 591 |
+
"kl": 0.343701171875,
|
| 592 |
+
"learning_rate": 7.095536274107046e-06,
|
| 593 |
+
"loss": 0.0137,
|
| 594 |
+
"reward": 0.4645408088341355,
|
| 595 |
+
"reward_std": 0.2891133207827806,
|
| 596 |
+
"rewards/accuracy_reward": 0.4645408088341355,
|
| 597 |
+
"rewards/format_reward": 0.0,
|
| 598 |
+
"step": 410
|
| 599 |
+
},
|
| 600 |
+
{
|
| 601 |
+
"completion_length": 587.3729467391968,
|
| 602 |
+
"epoch": 0.6493381002995459,
|
| 603 |
+
"grad_norm": 0.013467292425479267,
|
| 604 |
+
"kl": 0.2858642578125,
|
| 605 |
+
"learning_rate": 6.58262156614881e-06,
|
| 606 |
+
"loss": 0.0114,
|
| 607 |
+
"reward": 0.503188765514642,
|
| 608 |
+
"reward_std": 0.2590713477227837,
|
| 609 |
+
"rewards/accuracy_reward": 0.503188765514642,
|
| 610 |
+
"rewards/format_reward": 0.0,
|
| 611 |
+
"step": 420
|
| 612 |
+
},
|
| 613 |
+
{
|
| 614 |
+
"completion_length": 594.2052188873291,
|
| 615 |
+
"epoch": 0.6647985312590589,
|
| 616 |
+
"grad_norm": 0.013766538000227693,
|
| 617 |
+
"kl": 0.285015869140625,
|
| 618 |
+
"learning_rate": 6.079696150841634e-06,
|
| 619 |
+
"loss": 0.0114,
|
| 620 |
+
"reward": 0.4913265212439001,
|
| 621 |
+
"reward_std": 0.2684762907214463,
|
| 622 |
+
"rewards/accuracy_reward": 0.4913265212439001,
|
| 623 |
+
"rewards/format_reward": 0.0,
|
| 624 |
+
"step": 430
|
| 625 |
+
},
|
| 626 |
+
{
|
| 627 |
+
"completion_length": 631.7939949035645,
|
| 628 |
+
"epoch": 0.6802589622185718,
|
| 629 |
+
"grad_norm": 0.07419237005783043,
|
| 630 |
+
"kl": 0.346197509765625,
|
| 631 |
+
"learning_rate": 5.588230122660672e-06,
|
| 632 |
+
"loss": 0.0138,
|
| 633 |
+
"reward": 0.4860969296656549,
|
| 634 |
+
"reward_std": 0.2773646651767194,
|
| 635 |
+
"rewards/accuracy_reward": 0.4860969296656549,
|
| 636 |
+
"rewards/format_reward": 0.0,
|
| 637 |
+
"step": 440
|
| 638 |
+
},
|
| 639 |
+
{
|
| 640 |
+
"completion_length": 586.2428462982177,
|
| 641 |
+
"epoch": 0.6957193931780848,
|
| 642 |
+
"grad_norm": 0.015937060852966797,
|
| 643 |
+
"kl": 0.2478759765625,
|
| 644 |
+
"learning_rate": 5.109660079301668e-06,
|
| 645 |
+
"loss": 0.0099,
|
| 646 |
+
"reward": 0.5220663199201226,
|
| 647 |
+
"reward_std": 0.2621162030380219,
|
| 648 |
+
"rewards/accuracy_reward": 0.5220663199201226,
|
| 649 |
+
"rewards/format_reward": 0.0,
|
| 650 |
+
"step": 450
|
| 651 |
+
},
|
| 652 |
+
{
|
| 653 |
+
"completion_length": 641.8402923583984,
|
| 654 |
+
"epoch": 0.7111798241375978,
|
| 655 |
+
"grad_norm": 0.04445799644422213,
|
| 656 |
+
"kl": 0.3908416748046875,
|
| 657 |
+
"learning_rate": 4.64538492238166e-06,
|
| 658 |
+
"loss": 0.0156,
|
| 659 |
+
"reward": 0.4795918288640678,
|
| 660 |
+
"reward_std": 0.28201754316687583,
|
| 661 |
+
"rewards/accuracy_reward": 0.4795918288640678,
|
| 662 |
+
"rewards/format_reward": 0.0,
|
| 663 |
+
"step": 460
|
| 664 |
+
},
|
| 665 |
+
{
|
| 666 |
+
"completion_length": 648.5068748474121,
|
| 667 |
+
"epoch": 0.7266402550971108,
|
| 668 |
+
"grad_norm": 0.018272003157409882,
|
| 669 |
+
"kl": 0.44942626953125,
|
| 670 |
+
"learning_rate": 4.196761768328599e-06,
|
| 671 |
+
"loss": 0.018,
|
| 672 |
+
"reward": 0.45650509353727103,
|
| 673 |
+
"reward_std": 0.2825955556239933,
|
| 674 |
+
"rewards/accuracy_reward": 0.45650509353727103,
|
| 675 |
+
"rewards/format_reward": 0.0,
|
| 676 |
+
"step": 470
|
| 677 |
+
},
|
| 678 |
+
{
|
| 679 |
+
"completion_length": 632.1244766235352,
|
| 680 |
+
"epoch": 0.7421006860566238,
|
| 681 |
+
"grad_norm": 0.03550172747608362,
|
| 682 |
+
"kl": 0.54378662109375,
|
| 683 |
+
"learning_rate": 3.7651019814126656e-06,
|
| 684 |
+
"loss": 0.0218,
|
| 685 |
+
"reward": 0.4636479509063065,
|
| 686 |
+
"reward_std": 0.2770009428262711,
|
| 687 |
+
"rewards/accuracy_reward": 0.4636479509063065,
|
| 688 |
+
"rewards/format_reward": 0.0,
|
| 689 |
+
"step": 480
|
| 690 |
+
},
|
| 691 |
+
{
|
| 692 |
+
"completion_length": 632.8059818267823,
|
| 693 |
+
"epoch": 0.7575611170161368,
|
| 694 |
+
"grad_norm": 0.9437401498024955,
|
| 695 |
+
"kl": 1.1802734375,
|
| 696 |
+
"learning_rate": 3.3516673405151546e-06,
|
| 697 |
+
"loss": 0.0472,
|
| 698 |
+
"reward": 0.45255101229995487,
|
| 699 |
+
"reward_std": 0.2746642493642867,
|
| 700 |
+
"rewards/accuracy_reward": 0.45255101229995487,
|
| 701 |
+
"rewards/format_reward": 0.0,
|
| 702 |
+
"step": 490
|
| 703 |
+
},
|
| 704 |
+
{
|
| 705 |
+
"completion_length": 586.2049638748169,
|
| 706 |
+
"epoch": 0.7730215479756498,
|
| 707 |
+
"grad_norm": 0.11836217149566043,
|
| 708 |
+
"kl": 0.545013427734375,
|
| 709 |
+
"learning_rate": 2.957666350839663e-06,
|
| 710 |
+
"loss": 0.0218,
|
| 711 |
+
"reward": 0.48941325647756456,
|
| 712 |
+
"reward_std": 0.263414288777858,
|
| 713 |
+
"rewards/accuracy_reward": 0.48941325647756456,
|
| 714 |
+
"rewards/format_reward": 0.0,
|
| 715 |
+
"step": 500
|
| 716 |
+
},
|
| 717 |
+
{
|
| 718 |
+
"epoch": 0.7730215479756498,
|
| 719 |
+
"eval_completion_length": 551.6952229817708,
|
| 720 |
+
"eval_kl": 0.28020833333333334,
|
| 721 |
+
"eval_loss": 0.011639236472547054,
|
| 722 |
+
"eval_reward": 0.5387754996617635,
|
| 723 |
+
"eval_reward_std": 0.2534260580937068,
|
| 724 |
+
"eval_rewards/accuracy_reward": 0.5387754996617635,
|
| 725 |
+
"eval_rewards/format_reward": 0.0,
|
| 726 |
+
"eval_runtime": 85.6752,
|
| 727 |
+
"eval_samples_per_second": 1.156,
|
| 728 |
+
"eval_steps_per_second": 0.175,
|
| 729 |
+
"step": 500
|
| 730 |
+
},
|
| 731 |
+
{
|
| 732 |
+
"completion_length": 598.6526651382446,
|
| 733 |
+
"epoch": 0.7884819789351628,
|
| 734 |
+
"grad_norm": 0.014324651971585192,
|
| 735 |
+
"kl": 0.284661865234375,
|
| 736 |
+
"learning_rate": 2.5842507113469307e-06,
|
| 737 |
+
"loss": 0.0114,
|
| 738 |
+
"reward": 0.49579080580733714,
|
| 739 |
+
"reward_std": 0.2849952794611454,
|
| 740 |
+
"rewards/accuracy_reward": 0.49579080580733714,
|
| 741 |
+
"rewards/format_reward": 0.0,
|
| 742 |
+
"step": 510
|
| 743 |
+
},
|
| 744 |
+
{
|
| 745 |
+
"completion_length": 602.975754737854,
|
| 746 |
+
"epoch": 0.8039424098946758,
|
| 747 |
+
"grad_norm": 0.013302847380980731,
|
| 748 |
+
"kl": 0.287591552734375,
|
| 749 |
+
"learning_rate": 2.2325119482391466e-06,
|
| 750 |
+
"loss": 0.0115,
|
| 751 |
+
"reward": 0.4869897892698646,
|
| 752 |
+
"reward_std": 0.28300182512030003,
|
| 753 |
+
"rewards/accuracy_reward": 0.4869897892698646,
|
| 754 |
+
"rewards/format_reward": 0.0,
|
| 755 |
+
"step": 520
|
| 756 |
+
},
|
| 757 |
+
{
|
| 758 |
+
"completion_length": 617.0672046661377,
|
| 759 |
+
"epoch": 0.8194028408541888,
|
| 760 |
+
"grad_norm": 0.011506624405655665,
|
| 761 |
+
"kl": 0.316864013671875,
|
| 762 |
+
"learning_rate": 1.9034782243345074e-06,
|
| 763 |
+
"loss": 0.0127,
|
| 764 |
+
"reward": 0.49005101155489683,
|
| 765 |
+
"reward_std": 0.283530889172107,
|
| 766 |
+
"rewards/accuracy_reward": 0.49005101155489683,
|
| 767 |
+
"rewards/format_reward": 0.0,
|
| 768 |
+
"step": 530
|
| 769 |
+
},
|
| 770 |
+
{
|
| 771 |
+
"completion_length": 608.665803527832,
|
| 772 |
+
"epoch": 0.8348632718137018,
|
| 773 |
+
"grad_norm": 0.016337273376864313,
|
| 774 |
+
"kl": 0.557275390625,
|
| 775 |
+
"learning_rate": 1.5981113336584041e-06,
|
| 776 |
+
"loss": 0.0223,
|
| 777 |
+
"reward": 0.49119897168129684,
|
| 778 |
+
"reward_std": 0.28141105216927825,
|
| 779 |
+
"rewards/accuracy_reward": 0.49119897168129684,
|
| 780 |
+
"rewards/format_reward": 0.0,
|
| 781 |
+
"step": 540
|
| 782 |
+
},
|
| 783 |
+
{
|
| 784 |
+
"completion_length": 625.1159269332886,
|
| 785 |
+
"epoch": 0.8503237027732148,
|
| 786 |
+
"grad_norm": 0.057861981590189904,
|
| 787 |
+
"kl": 0.260540771484375,
|
| 788 |
+
"learning_rate": 1.3173038900362977e-06,
|
| 789 |
+
"loss": 0.0104,
|
| 790 |
+
"reward": 0.4499999931082129,
|
| 791 |
+
"reward_std": 0.27894868138246237,
|
| 792 |
+
"rewards/accuracy_reward": 0.4499999931082129,
|
| 793 |
+
"rewards/format_reward": 0.0,
|
| 794 |
+
"step": 550
|
| 795 |
+
},
|
| 796 |
+
{
|
| 797 |
+
"completion_length": 632.5747304916382,
|
| 798 |
+
"epoch": 0.8657841337327278,
|
| 799 |
+
"grad_norm": 0.030466073357798586,
|
| 800 |
+
"kl": 0.293267822265625,
|
| 801 |
+
"learning_rate": 1.0618767179063416e-06,
|
| 802 |
+
"loss": 0.0117,
|
| 803 |
+
"reward": 0.4618622355163097,
|
| 804 |
+
"reward_std": 0.2925746965222061,
|
| 805 |
+
"rewards/accuracy_reward": 0.4618622355163097,
|
| 806 |
+
"rewards/format_reward": 0.0,
|
| 807 |
+
"step": 560
|
| 808 |
+
},
|
| 809 |
+
{
|
| 810 |
+
"completion_length": 641.7084041595459,
|
| 811 |
+
"epoch": 0.8812445646922408,
|
| 812 |
+
"grad_norm": 0.017331692484703218,
|
| 813 |
+
"kl": 0.316583251953125,
|
| 814 |
+
"learning_rate": 8.325764529785851e-07,
|
| 815 |
+
"loss": 0.0127,
|
| 816 |
+
"reward": 0.4483418272808194,
|
| 817 |
+
"reward_std": 0.28196476846933366,
|
| 818 |
+
"rewards/accuracy_reward": 0.4483418272808194,
|
| 819 |
+
"rewards/format_reward": 0.0,
|
| 820 |
+
"step": 570
|
| 821 |
+
},
|
| 822 |
+
{
|
| 823 |
+
"completion_length": 646.2618488311767,
|
| 824 |
+
"epoch": 0.8967049956517538,
|
| 825 |
+
"grad_norm": 0.009327065043797485,
|
| 826 |
+
"kl": 0.303216552734375,
|
| 827 |
+
"learning_rate": 6.300733597542086e-07,
|
| 828 |
+
"loss": 0.0121,
|
| 829 |
+
"reward": 0.44196427753195167,
|
| 830 |
+
"reward_std": 0.28285656329244374,
|
| 831 |
+
"rewards/accuracy_reward": 0.44196427753195167,
|
| 832 |
+
"rewards/format_reward": 0.0,
|
| 833 |
+
"step": 580
|
| 834 |
+
},
|
| 835 |
+
{
|
| 836 |
+
"completion_length": 626.1570028305053,
|
| 837 |
+
"epoch": 0.9121654266112668,
|
| 838 |
+
"grad_norm": 0.010427073776751795,
|
| 839 |
+
"kl": 0.31864013671875,
|
| 840 |
+
"learning_rate": 4.549593722844492e-07,
|
| 841 |
+
"loss": 0.0127,
|
| 842 |
+
"reward": 0.46224488839507105,
|
| 843 |
+
"reward_std": 0.28828581105917694,
|
| 844 |
+
"rewards/accuracy_reward": 0.46224488839507105,
|
| 845 |
+
"rewards/format_reward": 0.0,
|
| 846 |
+
"step": 590
|
| 847 |
+
},
|
| 848 |
+
{
|
| 849 |
+
"completion_length": 628.7367259979249,
|
| 850 |
+
"epoch": 0.9276258575707798,
|
| 851 |
+
"grad_norm": 0.01713962537975438,
|
| 852 |
+
"kl": 0.3144775390625,
|
| 853 |
+
"learning_rate": 3.0774636389618196e-07,
|
| 854 |
+
"loss": 0.0126,
|
| 855 |
+
"reward": 0.4502550953067839,
|
| 856 |
+
"reward_std": 0.2758318264503032,
|
| 857 |
+
"rewards/accuracy_reward": 0.4502550953067839,
|
| 858 |
+
"rewards/format_reward": 0.0,
|
| 859 |
+
"step": 600
|
| 860 |
+
},
|
| 861 |
+
{
|
| 862 |
+
"epoch": 0.9276258575707798,
|
| 863 |
+
"eval_completion_length": 607.1224405924479,
|
| 864 |
+
"eval_kl": 0.2647786458333333,
|
| 865 |
+
"eval_loss": 0.011027935892343521,
|
| 866 |
+
"eval_reward": 0.4816326439380646,
|
| 867 |
+
"eval_reward_std": 0.25309162139892577,
|
| 868 |
+
"eval_rewards/accuracy_reward": 0.4816326439380646,
|
| 869 |
+
"eval_rewards/format_reward": 0.0,
|
| 870 |
+
"eval_runtime": 87.0986,
|
| 871 |
+
"eval_samples_per_second": 1.137,
|
| 872 |
+
"eval_steps_per_second": 0.172,
|
| 873 |
+
"step": 600
|
| 874 |
+
},
|
| 875 |
+
{
|
| 876 |
+
"completion_length": 633.0672080993652,
|
| 877 |
+
"epoch": 0.9430862885302927,
|
| 878 |
+
"grad_norm": 0.022562920761009898,
|
| 879 |
+
"kl": 0.302935791015625,
|
| 880 |
+
"learning_rate": 1.8886465094192895e-07,
|
| 881 |
+
"loss": 0.0121,
|
| 882 |
+
"reward": 0.46020407443866135,
|
| 883 |
+
"reward_std": 0.2828258784487844,
|
| 884 |
+
"rewards/accuracy_reward": 0.46020407443866135,
|
| 885 |
+
"rewards/format_reward": 0.0,
|
| 886 |
+
"step": 610
|
| 887 |
+
},
|
| 888 |
+
{
|
| 889 |
+
"completion_length": 626.6780488967895,
|
| 890 |
+
"epoch": 0.9585467194898057,
|
| 891 |
+
"grad_norm": 0.012428624166763307,
|
| 892 |
+
"kl": 0.3028564453125,
|
| 893 |
+
"learning_rate": 9.866173494794462e-08,
|
| 894 |
+
"loss": 0.0121,
|
| 895 |
+
"reward": 0.447959177242592,
|
| 896 |
+
"reward_std": 0.27789597446098924,
|
| 897 |
+
"rewards/accuracy_reward": 0.447959177242592,
|
| 898 |
+
"rewards/format_reward": 0.0,
|
| 899 |
+
"step": 620
|
| 900 |
+
},
|
| 901 |
+
{
|
| 902 |
+
"completion_length": 630.8656784057617,
|
| 903 |
+
"epoch": 0.9740071504493187,
|
| 904 |
+
"grad_norm": 0.009372661260439265,
|
| 905 |
+
"kl": 0.305780029296875,
|
| 906 |
+
"learning_rate": 3.7401286837214224e-08,
|
| 907 |
+
"loss": 0.0122,
|
| 908 |
+
"reward": 0.4577806035755202,
|
| 909 |
+
"reward_std": 0.27292990586720406,
|
| 910 |
+
"rewards/accuracy_reward": 0.4577806035755202,
|
| 911 |
+
"rewards/format_reward": 0.0,
|
| 912 |
+
"step": 630
|
| 913 |
+
},
|
| 914 |
+
{
|
| 915 |
+
"completion_length": 624.4659309387207,
|
| 916 |
+
"epoch": 0.9894675814088317,
|
| 917 |
+
"grad_norm": 0.012393498840602491,
|
| 918 |
+
"kl": 0.301348876953125,
|
| 919 |
+
"learning_rate": 5.262376196544239e-09,
|
| 920 |
+
"loss": 0.0121,
|
| 921 |
+
"reward": 0.4693877460435033,
|
| 922 |
+
"reward_std": 0.27738194689154627,
|
| 923 |
+
"rewards/accuracy_reward": 0.4693877460435033,
|
| 924 |
+
"rewards/format_reward": 0.0,
|
| 925 |
+
"step": 640
|
| 926 |
+
},
|
| 927 |
+
{
|
| 928 |
+
"completion_length": 626.0862976710001,
|
| 929 |
+
"epoch": 0.9987438399845395,
|
| 930 |
+
"kl": 0.31524658203125,
|
| 931 |
+
"reward": 0.46088434507449466,
|
| 932 |
+
"reward_std": 0.28149245004169643,
|
| 933 |
+
"rewards/accuracy_reward": 0.46088434507449466,
|
| 934 |
"rewards/format_reward": 0.0,
|
| 935 |
+
"step": 646,
|
| 936 |
"total_flos": 0.0,
|
| 937 |
+
"train_loss": 0.01602011715643949,
|
| 938 |
+
"train_runtime": 67938.2282,
|
| 939 |
+
"train_samples_per_second": 1.066,
|
| 940 |
+
"train_steps_per_second": 0.01
|
| 941 |
}
|
| 942 |
],
|
| 943 |
"logging_steps": 10,
|
| 944 |
+
"max_steps": 646,
|
| 945 |
"num_input_tokens_seen": 0,
|
| 946 |
"num_train_epochs": 1,
|
| 947 |
"save_steps": 500,
|
|
|
|
| 958 |
}
|
| 959 |
},
|
| 960 |
"total_flos": 0.0,
|
| 961 |
+
"train_batch_size": 1,
|
| 962 |
"trial_name": null,
|
| 963 |
"trial_params": null
|
| 964 |
}
|
training_args.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c5dda480b3facc8d6e3736863e3482b17102b90e8146c5d30c4543a09d4e2a61
|
| 3 |
+
size 7224
|