Model save
Browse files- README.md +2 -4
- all_results.json +5 -5
- train_results.json +5 -5
- trainer_state.json +1618 -19
README.md
CHANGED
|
@@ -1,11 +1,9 @@
|
|
| 1 |
---
|
| 2 |
base_model: deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B
|
| 3 |
-
datasets: open-r1/OpenR1-Math-220k
|
| 4 |
library_name: transformers
|
| 5 |
model_name: DeepSeek-R1-Distill-Qwen-1.5B-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 DeepSeek-R1-Distill-Qwen-1.5B-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/yeshugno-microsoft/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: DeepSeek-R1-Distill-Qwen-1.5B-GRPO
|
| 5 |
tags:
|
| 6 |
- generated_from_trainer
|
|
|
|
| 7 |
- trl
|
| 8 |
- grpo
|
| 9 |
licence: license
|
|
|
|
| 11 |
|
| 12 |
# Model Card for DeepSeek-R1-Distill-Qwen-1.5B-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/yeshugno-microsoft/huggingface/runs/0cfrj5s4)
|
| 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
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"total_flos": 0.0,
|
| 3 |
-
"train_loss":
|
| 4 |
-
"train_runtime":
|
| 5 |
-
"train_samples":
|
| 6 |
-
"train_samples_per_second": 0.
|
| 7 |
-
"train_steps_per_second": 0.
|
| 8 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"total_flos": 0.0,
|
| 3 |
+
"train_loss": 0.0003328805177226313,
|
| 4 |
+
"train_runtime": 21588.6366,
|
| 5 |
+
"train_samples": 1000,
|
| 6 |
+
"train_samples_per_second": 0.046,
|
| 7 |
+
"train_steps_per_second": 0.006
|
| 8 |
}
|
train_results.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"total_flos": 0.0,
|
| 3 |
-
"train_loss":
|
| 4 |
-
"train_runtime":
|
| 5 |
-
"train_samples":
|
| 6 |
-
"train_samples_per_second": 0.
|
| 7 |
-
"train_steps_per_second": 0.
|
| 8 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"total_flos": 0.0,
|
| 3 |
+
"train_loss": 0.0003328805177226313,
|
| 4 |
+
"train_runtime": 21588.6366,
|
| 5 |
+
"train_samples": 1000,
|
| 6 |
+
"train_samples_per_second": 0.046,
|
| 7 |
+
"train_steps_per_second": 0.006
|
| 8 |
}
|
trainer_state.json
CHANGED
|
@@ -1,51 +1,1650 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
-
"epoch": 0
|
| 5 |
"eval_steps": 500,
|
| 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": 0.0,
|
| 15 |
"kl": 0.0,
|
| 16 |
"learning_rate": 0.0,
|
| 17 |
"loss": 0.0,
|
| 18 |
-
"reward": 0.
|
| 19 |
-
"reward_std": 0.
|
| 20 |
-
"rewards/accuracy_reward": 0.
|
| 21 |
"rewards/format_reward": 0.0,
|
| 22 |
"step": 1
|
| 23 |
},
|
| 24 |
{
|
| 25 |
-
"completion_length":
|
| 26 |
-
"epoch": 0.
|
| 27 |
"grad_norm": 0.0,
|
| 28 |
"kl": 0.0,
|
| 29 |
-
"learning_rate":
|
| 30 |
"loss": 0.0,
|
| 31 |
-
"reward": 0.
|
| 32 |
-
"reward_std": 0.
|
| 33 |
-
"rewards/accuracy_reward": 0.
|
| 34 |
"rewards/format_reward": 0.0,
|
| 35 |
"step": 2
|
| 36 |
},
|
| 37 |
{
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 38 |
"epoch": 0.8,
|
| 39 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 40 |
"total_flos": 0.0,
|
| 41 |
-
"train_loss":
|
| 42 |
-
"train_runtime":
|
| 43 |
-
"train_samples_per_second": 0.
|
| 44 |
-
"train_steps_per_second": 0.
|
| 45 |
}
|
| 46 |
],
|
| 47 |
"logging_steps": 1,
|
| 48 |
-
"max_steps":
|
| 49 |
"num_input_tokens_seen": 0,
|
| 50 |
"num_train_epochs": 1,
|
| 51 |
"save_steps": 500,
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 1.0,
|
| 5 |
"eval_steps": 500,
|
| 6 |
+
"global_step": 125,
|
| 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": 1721.0417022705078,
|
| 13 |
+
"epoch": 0.008,
|
| 14 |
"grad_norm": 0.0,
|
| 15 |
"kl": 0.0,
|
| 16 |
"learning_rate": 0.0,
|
| 17 |
"loss": 0.0,
|
| 18 |
+
"reward": 0.17708333488553762,
|
| 19 |
+
"reward_std": 0.09763014316558838,
|
| 20 |
+
"rewards/accuracy_reward": 0.17708333488553762,
|
| 21 |
"rewards/format_reward": 0.0,
|
| 22 |
"step": 1
|
| 23 |
},
|
| 24 |
{
|
| 25 |
+
"completion_length": 1944.875015258789,
|
| 26 |
+
"epoch": 0.016,
|
| 27 |
"grad_norm": 0.0,
|
| 28 |
"kl": 0.0,
|
| 29 |
+
"learning_rate": 7.692307692307692e-08,
|
| 30 |
"loss": 0.0,
|
| 31 |
+
"reward": 0.09375,
|
| 32 |
+
"reward_std": 0.05653337761759758,
|
| 33 |
+
"rewards/accuracy_reward": 0.09375,
|
| 34 |
"rewards/format_reward": 0.0,
|
| 35 |
"step": 2
|
| 36 |
},
|
| 37 |
{
|
| 38 |
+
"completion_length": 1621.6250228881836,
|
| 39 |
+
"epoch": 0.024,
|
| 40 |
+
"grad_norm": 0.0,
|
| 41 |
+
"kl": 0.0,
|
| 42 |
+
"learning_rate": 7.692307692307692e-08,
|
| 43 |
+
"loss": 0.0,
|
| 44 |
+
"reward": 0.14583333861082792,
|
| 45 |
+
"reward_std": 0.16199623048305511,
|
| 46 |
+
"rewards/accuracy_reward": 0.14583333861082792,
|
| 47 |
+
"rewards/format_reward": 0.0,
|
| 48 |
+
"step": 3
|
| 49 |
+
},
|
| 50 |
+
{
|
| 51 |
+
"completion_length": 1667.333351135254,
|
| 52 |
+
"epoch": 0.032,
|
| 53 |
+
"grad_norm": 0.0,
|
| 54 |
+
"kl": 0.0,
|
| 55 |
+
"learning_rate": 7.692307692307692e-08,
|
| 56 |
+
"loss": 0.0,
|
| 57 |
+
"reward": 0.09375000558793545,
|
| 58 |
+
"reward_std": 0.11302226781845093,
|
| 59 |
+
"rewards/accuracy_reward": 0.0729166716337204,
|
| 60 |
+
"rewards/format_reward": 0.02083333395421505,
|
| 61 |
+
"step": 4
|
| 62 |
+
},
|
| 63 |
+
{
|
| 64 |
+
"completion_length": 1555.8854598999023,
|
| 65 |
+
"epoch": 0.04,
|
| 66 |
+
"grad_norm": 0.0,
|
| 67 |
+
"kl": 0.0,
|
| 68 |
+
"learning_rate": 1.5384615384615385e-07,
|
| 69 |
+
"loss": 0.0,
|
| 70 |
+
"reward": 0.0625,
|
| 71 |
+
"reward_std": 0.06527911871671677,
|
| 72 |
+
"rewards/accuracy_reward": 0.0625,
|
| 73 |
+
"rewards/format_reward": 0.0,
|
| 74 |
+
"step": 5
|
| 75 |
+
},
|
| 76 |
+
{
|
| 77 |
+
"completion_length": 1598.6354522705078,
|
| 78 |
+
"epoch": 0.048,
|
| 79 |
+
"grad_norm": 0.0,
|
| 80 |
+
"kl": 0.0,
|
| 81 |
+
"learning_rate": 2.3076923076923078e-07,
|
| 82 |
+
"loss": 0.0,
|
| 83 |
+
"reward": 0.06250000186264515,
|
| 84 |
+
"reward_std": 0.11020193248987198,
|
| 85 |
+
"rewards/accuracy_reward": 0.0416666679084301,
|
| 86 |
+
"rewards/format_reward": 0.02083333395421505,
|
| 87 |
+
"step": 6
|
| 88 |
+
},
|
| 89 |
+
{
|
| 90 |
+
"completion_length": 1771.7916870117188,
|
| 91 |
+
"epoch": 0.056,
|
| 92 |
+
"grad_norm": 0.0,
|
| 93 |
+
"kl": 0.0,
|
| 94 |
+
"learning_rate": 3.076923076923077e-07,
|
| 95 |
+
"loss": 0.0,
|
| 96 |
+
"reward": 0.0520833358168602,
|
| 97 |
+
"reward_std": 0.06436608731746674,
|
| 98 |
+
"rewards/accuracy_reward": 0.0520833358168602,
|
| 99 |
+
"rewards/format_reward": 0.0,
|
| 100 |
+
"step": 7
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"completion_length": 1735.2916870117188,
|
| 104 |
+
"epoch": 0.064,
|
| 105 |
+
"grad_norm": 0.0,
|
| 106 |
+
"kl": 0.0,
|
| 107 |
+
"learning_rate": 3.8461538461538463e-07,
|
| 108 |
+
"loss": 0.0,
|
| 109 |
+
"reward": 0.07291666977107525,
|
| 110 |
+
"reward_std": 0.11302226781845093,
|
| 111 |
+
"rewards/accuracy_reward": 0.07291666977107525,
|
| 112 |
+
"rewards/format_reward": 0.0,
|
| 113 |
+
"step": 8
|
| 114 |
+
},
|
| 115 |
+
{
|
| 116 |
+
"completion_length": 1848.2916870117188,
|
| 117 |
+
"epoch": 0.072,
|
| 118 |
+
"grad_norm": 0.0,
|
| 119 |
+
"kl": 0.0,
|
| 120 |
+
"learning_rate": 4.6153846153846156e-07,
|
| 121 |
+
"loss": -0.0,
|
| 122 |
+
"reward": 0.010416666977107525,
|
| 123 |
+
"reward_std": 0.03608439117670059,
|
| 124 |
+
"rewards/accuracy_reward": 0.010416666977107525,
|
| 125 |
+
"rewards/format_reward": 0.0,
|
| 126 |
+
"step": 9
|
| 127 |
+
},
|
| 128 |
+
{
|
| 129 |
+
"completion_length": 1668.5312805175781,
|
| 130 |
+
"epoch": 0.08,
|
| 131 |
+
"grad_norm": 0.0,
|
| 132 |
+
"kl": 0.0,
|
| 133 |
+
"learning_rate": 4.6153846153846156e-07,
|
| 134 |
+
"loss": 0.0,
|
| 135 |
+
"reward": 0.08333333674818277,
|
| 136 |
+
"reward_std": 0.14910665899515152,
|
| 137 |
+
"rewards/accuracy_reward": 0.08333333674818277,
|
| 138 |
+
"rewards/format_reward": 0.0,
|
| 139 |
+
"step": 10
|
| 140 |
+
},
|
| 141 |
+
{
|
| 142 |
+
"completion_length": 1871.9687805175781,
|
| 143 |
+
"epoch": 0.088,
|
| 144 |
+
"grad_norm": 0.0,
|
| 145 |
+
"kl": 0.0,
|
| 146 |
+
"learning_rate": 5.384615384615384e-07,
|
| 147 |
+
"loss": 0.0,
|
| 148 |
+
"reward": 0.0,
|
| 149 |
+
"reward_std": 0.0,
|
| 150 |
+
"rewards/accuracy_reward": 0.0,
|
| 151 |
+
"rewards/format_reward": 0.0,
|
| 152 |
+
"step": 11
|
| 153 |
+
},
|
| 154 |
+
{
|
| 155 |
+
"completion_length": 1863.7500305175781,
|
| 156 |
+
"epoch": 0.096,
|
| 157 |
+
"grad_norm": 0.0,
|
| 158 |
+
"kl": 0.0,
|
| 159 |
+
"learning_rate": 5.384615384615384e-07,
|
| 160 |
+
"loss": 0.0,
|
| 161 |
+
"reward": 0.02083333395421505,
|
| 162 |
+
"reward_std": 0.07216878235340118,
|
| 163 |
+
"rewards/accuracy_reward": 0.02083333395421505,
|
| 164 |
+
"rewards/format_reward": 0.0,
|
| 165 |
+
"step": 12
|
| 166 |
+
},
|
| 167 |
+
{
|
| 168 |
+
"completion_length": 1803.5104370117188,
|
| 169 |
+
"epoch": 0.104,
|
| 170 |
+
"grad_norm": 0.0,
|
| 171 |
+
"kl": 0.0,
|
| 172 |
+
"learning_rate": 6.153846153846154e-07,
|
| 173 |
+
"loss": 0.0,
|
| 174 |
+
"reward": 0.0,
|
| 175 |
+
"reward_std": 0.0,
|
| 176 |
+
"rewards/accuracy_reward": 0.0,
|
| 177 |
+
"rewards/format_reward": 0.0,
|
| 178 |
+
"step": 13
|
| 179 |
+
},
|
| 180 |
+
{
|
| 181 |
+
"completion_length": 1560.7708587646484,
|
| 182 |
+
"epoch": 0.112,
|
| 183 |
+
"grad_norm": 0.0,
|
| 184 |
+
"kl": 0.0,
|
| 185 |
+
"learning_rate": 6.923076923076922e-07,
|
| 186 |
+
"loss": -0.0,
|
| 187 |
+
"reward": 0.14583333395421505,
|
| 188 |
+
"reward_std": 0.04865618050098419,
|
| 189 |
+
"rewards/accuracy_reward": 0.14583333395421505,
|
| 190 |
+
"rewards/format_reward": 0.0,
|
| 191 |
+
"step": 14
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"completion_length": 1918.3646087646484,
|
| 195 |
+
"epoch": 0.12,
|
| 196 |
+
"grad_norm": 0.0,
|
| 197 |
+
"kl": 0.0,
|
| 198 |
+
"learning_rate": 6.923076923076922e-07,
|
| 199 |
+
"loss": -0.0,
|
| 200 |
+
"reward": 0.22916666697710752,
|
| 201 |
+
"reward_std": 0.15789688751101494,
|
| 202 |
+
"rewards/accuracy_reward": 0.22916666697710752,
|
| 203 |
+
"rewards/format_reward": 0.0,
|
| 204 |
+
"step": 15
|
| 205 |
+
},
|
| 206 |
+
{
|
| 207 |
+
"completion_length": 1745.6458740234375,
|
| 208 |
+
"epoch": 0.128,
|
| 209 |
+
"grad_norm": 0.0,
|
| 210 |
+
"kl": 0.0,
|
| 211 |
+
"learning_rate": 6.923076923076922e-07,
|
| 212 |
+
"loss": -0.0,
|
| 213 |
+
"reward": 0.18750000093132257,
|
| 214 |
+
"reward_std": 0.14127394929528236,
|
| 215 |
+
"rewards/accuracy_reward": 0.18750000093132257,
|
| 216 |
+
"rewards/format_reward": 0.0,
|
| 217 |
+
"step": 16
|
| 218 |
+
},
|
| 219 |
+
{
|
| 220 |
+
"completion_length": 1737.7812728881836,
|
| 221 |
+
"epoch": 0.136,
|
| 222 |
+
"grad_norm": 0.0,
|
| 223 |
+
"kl": 0.0,
|
| 224 |
+
"learning_rate": 7.692307692307693e-07,
|
| 225 |
+
"loss": 0.0,
|
| 226 |
+
"reward": 0.0416666679084301,
|
| 227 |
+
"reward_std": 0.06154575198888779,
|
| 228 |
+
"rewards/accuracy_reward": 0.03125,
|
| 229 |
+
"rewards/format_reward": 0.010416666977107525,
|
| 230 |
+
"step": 17
|
| 231 |
+
},
|
| 232 |
+
{
|
| 233 |
+
"completion_length": 1527.2917175292969,
|
| 234 |
+
"epoch": 0.144,
|
| 235 |
+
"grad_norm": 0.0,
|
| 236 |
+
"kl": 0.0,
|
| 237 |
+
"learning_rate": 7.692307692307693e-07,
|
| 238 |
+
"loss": -0.0,
|
| 239 |
+
"reward": 0.3958333348855376,
|
| 240 |
+
"reward_std": 0.21344273164868355,
|
| 241 |
+
"rewards/accuracy_reward": 0.3854166679084301,
|
| 242 |
+
"rewards/format_reward": 0.010416666977107525,
|
| 243 |
+
"step": 18
|
| 244 |
+
},
|
| 245 |
+
{
|
| 246 |
+
"completion_length": 1783.3958587646484,
|
| 247 |
+
"epoch": 0.152,
|
| 248 |
+
"grad_norm": 0.0,
|
| 249 |
+
"kl": 0.0,
|
| 250 |
+
"learning_rate": 8.461538461538461e-07,
|
| 251 |
+
"loss": 0.0,
|
| 252 |
+
"reward": 0.13541666697710752,
|
| 253 |
+
"reward_std": 0.03608439117670059,
|
| 254 |
+
"rewards/accuracy_reward": 0.13541666697710752,
|
| 255 |
+
"rewards/format_reward": 0.0,
|
| 256 |
+
"step": 19
|
| 257 |
+
},
|
| 258 |
+
{
|
| 259 |
+
"completion_length": 1755.5208740234375,
|
| 260 |
+
"epoch": 0.16,
|
| 261 |
+
"grad_norm": 0.0,
|
| 262 |
+
"kl": 0.0,
|
| 263 |
+
"learning_rate": 8.461538461538461e-07,
|
| 264 |
+
"loss": 0.0,
|
| 265 |
+
"reward": 0.23958333674818277,
|
| 266 |
+
"reward_std": 0.16979892551898956,
|
| 267 |
+
"rewards/accuracy_reward": 0.22916666977107525,
|
| 268 |
+
"rewards/format_reward": 0.010416666977107525,
|
| 269 |
+
"step": 20
|
| 270 |
+
},
|
| 271 |
+
{
|
| 272 |
+
"completion_length": 1797.2083740234375,
|
| 273 |
+
"epoch": 0.168,
|
| 274 |
+
"grad_norm": 0.0,
|
| 275 |
+
"kl": 0.0,
|
| 276 |
+
"learning_rate": 8.461538461538461e-07,
|
| 277 |
+
"loss": -0.0,
|
| 278 |
+
"reward": 0.05208333395421505,
|
| 279 |
+
"reward_std": 0.10518955811858177,
|
| 280 |
+
"rewards/accuracy_reward": 0.05208333395421505,
|
| 281 |
+
"rewards/format_reward": 0.0,
|
| 282 |
+
"step": 21
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"completion_length": 1737.2916717529297,
|
| 286 |
+
"epoch": 0.176,
|
| 287 |
+
"grad_norm": 0.0,
|
| 288 |
+
"kl": 0.0,
|
| 289 |
+
"learning_rate": 8.461538461538461e-07,
|
| 290 |
+
"loss": 0.0,
|
| 291 |
+
"reward": 0.07291666883975267,
|
| 292 |
+
"reward_std": 0.14628632366657257,
|
| 293 |
+
"rewards/accuracy_reward": 0.06250000093132257,
|
| 294 |
+
"rewards/format_reward": 0.010416666977107525,
|
| 295 |
+
"step": 22
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"completion_length": 1865.281265258789,
|
| 299 |
+
"epoch": 0.184,
|
| 300 |
+
"grad_norm": 0.07227283716201782,
|
| 301 |
+
"kl": 0.0,
|
| 302 |
+
"learning_rate": 9.230769230769231e-07,
|
| 303 |
+
"loss": 0.0,
|
| 304 |
+
"reward": 0.052083334885537624,
|
| 305 |
+
"reward_std": 0.09763014316558838,
|
| 306 |
+
"rewards/accuracy_reward": 0.052083334885537624,
|
| 307 |
+
"rewards/format_reward": 0.0,
|
| 308 |
+
"step": 23
|
| 309 |
+
},
|
| 310 |
+
{
|
| 311 |
+
"completion_length": 1721.583351135254,
|
| 312 |
+
"epoch": 0.192,
|
| 313 |
+
"grad_norm": 3.1304349249694496e-05,
|
| 314 |
+
"kl": -7.547438144683838e-06,
|
| 315 |
+
"learning_rate": 1e-06,
|
| 316 |
+
"loss": -0.0,
|
| 317 |
+
"reward": 0.06250000279396772,
|
| 318 |
+
"reward_std": 0.10045047849416733,
|
| 319 |
+
"rewards/accuracy_reward": 0.0520833358168602,
|
| 320 |
+
"rewards/format_reward": 0.010416666977107525,
|
| 321 |
+
"step": 24
|
| 322 |
+
},
|
| 323 |
+
{
|
| 324 |
+
"completion_length": 1667.0521240234375,
|
| 325 |
+
"epoch": 0.2,
|
| 326 |
+
"grad_norm": 0.057344950735569,
|
| 327 |
+
"kl": -7.309019565582275e-06,
|
| 328 |
+
"learning_rate": 9.998229818723738e-07,
|
| 329 |
+
"loss": -0.0,
|
| 330 |
+
"reward": 0.1041666716337204,
|
| 331 |
+
"reward_std": 0.14010312780737877,
|
| 332 |
+
"rewards/accuracy_reward": 0.09375,
|
| 333 |
+
"rewards/format_reward": 0.010416666977107525,
|
| 334 |
+
"step": 25
|
| 335 |
+
},
|
| 336 |
+
{
|
| 337 |
+
"completion_length": 1752.8021240234375,
|
| 338 |
+
"epoch": 0.208,
|
| 339 |
+
"grad_norm": 0.11655361950397491,
|
| 340 |
+
"kl": -9.395182132720947e-06,
|
| 341 |
+
"learning_rate": 9.992920667580175e-07,
|
| 342 |
+
"loss": -0.0,
|
| 343 |
+
"reward": 0.26041667349636555,
|
| 344 |
+
"reward_std": 0.17456801980733871,
|
| 345 |
+
"rewards/accuracy_reward": 0.2291666716337204,
|
| 346 |
+
"rewards/format_reward": 0.031250000931322575,
|
| 347 |
+
"step": 26
|
| 348 |
+
},
|
| 349 |
+
{
|
| 350 |
+
"completion_length": 1865.9167175292969,
|
| 351 |
+
"epoch": 0.216,
|
| 352 |
+
"grad_norm": 0.0001482899097027257,
|
| 353 |
+
"kl": -1.2703239917755127e-05,
|
| 354 |
+
"learning_rate": 9.984076723529287e-07,
|
| 355 |
+
"loss": -0.0,
|
| 356 |
+
"reward": 0.13541666697710752,
|
| 357 |
+
"reward_std": 0.03608439117670059,
|
| 358 |
+
"rewards/accuracy_reward": 0.13541666697710752,
|
| 359 |
+
"rewards/format_reward": 0.0,
|
| 360 |
+
"step": 27
|
| 361 |
+
},
|
| 362 |
+
{
|
| 363 |
+
"completion_length": 1667.031265258789,
|
| 364 |
+
"epoch": 0.224,
|
| 365 |
+
"grad_norm": 0.00021164790086913854,
|
| 366 |
+
"kl": -1.093745231628418e-05,
|
| 367 |
+
"learning_rate": 9.971704944519593e-07,
|
| 368 |
+
"loss": -0.0,
|
| 369 |
+
"reward": 0.25,
|
| 370 |
+
"reward_std": 0.0,
|
| 371 |
+
"rewards/accuracy_reward": 0.25,
|
| 372 |
+
"rewards/format_reward": 0.0,
|
| 373 |
+
"step": 28
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"completion_length": 1666.7812805175781,
|
| 377 |
+
"epoch": 0.232,
|
| 378 |
+
"grad_norm": 0.000568240531720221,
|
| 379 |
+
"kl": -7.115304470062256e-06,
|
| 380 |
+
"learning_rate": 9.955815064014005e-07,
|
| 381 |
+
"loss": -0.0,
|
| 382 |
+
"reward": 0.031250000931322575,
|
| 383 |
+
"reward_std": 0.08474057912826538,
|
| 384 |
+
"rewards/accuracy_reward": 0.031250000931322575,
|
| 385 |
+
"rewards/format_reward": 0.0,
|
| 386 |
+
"step": 29
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"completion_length": 1486.2187881469727,
|
| 390 |
+
"epoch": 0.24,
|
| 391 |
+
"grad_norm": 0.1216636374592781,
|
| 392 |
+
"kl": -4.678964614868164e-06,
|
| 393 |
+
"learning_rate": 9.93641958333206e-07,
|
| 394 |
+
"loss": -0.0,
|
| 395 |
+
"reward": 0.2187500037252903,
|
| 396 |
+
"reward_std": 0.24490400031208992,
|
| 397 |
+
"rewards/accuracy_reward": 0.2187500037252903,
|
| 398 |
+
"rewards/format_reward": 0.0,
|
| 399 |
+
"step": 30
|
| 400 |
+
},
|
| 401 |
+
{
|
| 402 |
+
"completion_length": 1651.020866394043,
|
| 403 |
+
"epoch": 0.248,
|
| 404 |
+
"grad_norm": 0.09538763761520386,
|
| 405 |
+
"kl": -6.183981895446777e-07,
|
| 406 |
+
"learning_rate": 9.913533761814537e-07,
|
| 407 |
+
"loss": -0.0,
|
| 408 |
+
"reward": 0.041666666977107525,
|
| 409 |
+
"reward_std": 0.09261776879429817,
|
| 410 |
+
"rewards/accuracy_reward": 0.041666666977107525,
|
| 411 |
+
"rewards/format_reward": 0.0,
|
| 412 |
+
"step": 31
|
| 413 |
+
},
|
| 414 |
+
{
|
| 415 |
+
"completion_length": 1917.9375305175781,
|
| 416 |
+
"epoch": 0.256,
|
| 417 |
+
"grad_norm": 0.056192267686128616,
|
| 418 |
+
"kl": 3.120303153991699e-05,
|
| 419 |
+
"learning_rate": 9.887175604818206e-07,
|
| 420 |
+
"loss": 0.0,
|
| 421 |
+
"reward": 0.15625000093132257,
|
| 422 |
+
"reward_std": 0.10825317353010178,
|
| 423 |
+
"rewards/accuracy_reward": 0.14583333395421505,
|
| 424 |
+
"rewards/format_reward": 0.010416666977107525,
|
| 425 |
+
"step": 32
|
| 426 |
+
},
|
| 427 |
+
{
|
| 428 |
+
"completion_length": 1721.8646087646484,
|
| 429 |
+
"epoch": 0.264,
|
| 430 |
+
"grad_norm": 0.0013622839469462633,
|
| 431 |
+
"kl": 5.451589822769165e-05,
|
| 432 |
+
"learning_rate": 9.857365849550177e-07,
|
| 433 |
+
"loss": 0.0,
|
| 434 |
+
"reward": 0.13541666697710752,
|
| 435 |
+
"reward_std": 0.03608439117670059,
|
| 436 |
+
"rewards/accuracy_reward": 0.125,
|
| 437 |
+
"rewards/format_reward": 0.010416666977107525,
|
| 438 |
+
"step": 33
|
| 439 |
+
},
|
| 440 |
+
{
|
| 441 |
+
"completion_length": 1964.8750305175781,
|
| 442 |
+
"epoch": 0.272,
|
| 443 |
+
"grad_norm": 0.0009671057923696935,
|
| 444 |
+
"kl": 6.0245394706726074e-05,
|
| 445 |
+
"learning_rate": 9.824127948752948e-07,
|
| 446 |
+
"loss": 0.0,
|
| 447 |
+
"reward": 0.02083333395421505,
|
| 448 |
+
"reward_std": 0.07216878235340118,
|
| 449 |
+
"rewards/accuracy_reward": 0.02083333395421505,
|
| 450 |
+
"rewards/format_reward": 0.0,
|
| 451 |
+
"step": 34
|
| 452 |
+
},
|
| 453 |
+
{
|
| 454 |
+
"completion_length": 1607.7708587646484,
|
| 455 |
+
"epoch": 0.28,
|
| 456 |
+
"grad_norm": 0.11371063441038132,
|
| 457 |
+
"kl": 6.452202796936035e-05,
|
| 458 |
+
"learning_rate": 9.787488052253033e-07,
|
| 459 |
+
"loss": 0.0,
|
| 460 |
+
"reward": 0.041666666977107525,
|
| 461 |
+
"reward_std": 0.09261776879429817,
|
| 462 |
+
"rewards/accuracy_reward": 0.02083333395421505,
|
| 463 |
+
"rewards/format_reward": 0.02083333395421505,
|
| 464 |
+
"step": 35
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"completion_length": 1803.7500305175781,
|
| 468 |
+
"epoch": 0.288,
|
| 469 |
+
"grad_norm": 0.0009732124162837863,
|
| 470 |
+
"kl": 0.00013828277587890625,
|
| 471 |
+
"learning_rate": 9.747474986387654e-07,
|
| 472 |
+
"loss": 0.0,
|
| 473 |
+
"reward": 0.19791666697710752,
|
| 474 |
+
"reward_std": 0.10136350989341736,
|
| 475 |
+
"rewards/accuracy_reward": 0.19791666697710752,
|
| 476 |
+
"rewards/format_reward": 0.0,
|
| 477 |
+
"step": 36
|
| 478 |
+
},
|
| 479 |
+
{
|
| 480 |
+
"completion_length": 1660.7292175292969,
|
| 481 |
+
"epoch": 0.296,
|
| 482 |
+
"grad_norm": 0.054171618074178696,
|
| 483 |
+
"kl": 0.000163249671459198,
|
| 484 |
+
"learning_rate": 9.7041202313257e-07,
|
| 485 |
+
"loss": 0.0,
|
| 486 |
+
"reward": 0.19791666977107525,
|
| 487 |
+
"reward_std": 0.15001969039440155,
|
| 488 |
+
"rewards/accuracy_reward": 0.16666666977107525,
|
| 489 |
+
"rewards/format_reward": 0.03125,
|
| 490 |
+
"step": 37
|
| 491 |
+
},
|
| 492 |
+
{
|
| 493 |
+
"completion_length": 1631.750015258789,
|
| 494 |
+
"epoch": 0.304,
|
| 495 |
+
"grad_norm": 0.10330933332443237,
|
| 496 |
+
"kl": 0.0003886893391609192,
|
| 497 |
+
"learning_rate": 9.657457896300791e-07,
|
| 498 |
+
"loss": 0.0,
|
| 499 |
+
"reward": 0.25000000558793545,
|
| 500 |
+
"reward_std": 0.16188634932041168,
|
| 501 |
+
"rewards/accuracy_reward": 0.21875000279396772,
|
| 502 |
+
"rewards/format_reward": 0.031250000931322575,
|
| 503 |
+
"step": 38
|
| 504 |
+
},
|
| 505 |
+
{
|
| 506 |
+
"completion_length": 1778.6041946411133,
|
| 507 |
+
"epoch": 0.312,
|
| 508 |
+
"grad_norm": 0.0016962133813649416,
|
| 509 |
+
"kl": 0.00030538439750671387,
|
| 510 |
+
"learning_rate": 9.607524692775935e-07,
|
| 511 |
+
"loss": 0.0,
|
| 512 |
+
"reward": 0.02083333395421505,
|
| 513 |
+
"reward_std": 0.04865618795156479,
|
| 514 |
+
"rewards/accuracy_reward": 0.02083333395421505,
|
| 515 |
+
"rewards/format_reward": 0.0,
|
| 516 |
+
"step": 39
|
| 517 |
+
},
|
| 518 |
+
{
|
| 519 |
+
"completion_length": 1545.9167175292969,
|
| 520 |
+
"epoch": 0.32,
|
| 521 |
+
"grad_norm": 0.050718892365694046,
|
| 522 |
+
"kl": 0.0006757676601409912,
|
| 523 |
+
"learning_rate": 9.554359905560885e-07,
|
| 524 |
+
"loss": 0.0,
|
| 525 |
+
"reward": 0.33333334140479565,
|
| 526 |
+
"reward_std": 0.27968307584524155,
|
| 527 |
+
"rewards/accuracy_reward": 0.33333334140479565,
|
| 528 |
+
"rewards/format_reward": 0.0,
|
| 529 |
+
"step": 40
|
| 530 |
+
},
|
| 531 |
+
{
|
| 532 |
+
"completion_length": 1609.9271087646484,
|
| 533 |
+
"epoch": 0.328,
|
| 534 |
+
"grad_norm": 0.11008451133966446,
|
| 535 |
+
"kl": 0.0004702061414718628,
|
| 536 |
+
"learning_rate": 9.498005361904924e-07,
|
| 537 |
+
"loss": 0.0,
|
| 538 |
+
"reward": 0.1145833358168602,
|
| 539 |
+
"reward_std": 0.21539149060845375,
|
| 540 |
+
"rewards/accuracy_reward": 0.06250000186264515,
|
| 541 |
+
"rewards/format_reward": 0.05208333395421505,
|
| 542 |
+
"step": 41
|
| 543 |
+
},
|
| 544 |
+
{
|
| 545 |
+
"completion_length": 1728.614616394043,
|
| 546 |
+
"epoch": 0.336,
|
| 547 |
+
"grad_norm": 0.09986075758934021,
|
| 548 |
+
"kl": 0.0009259581565856934,
|
| 549 |
+
"learning_rate": 9.438505398589392e-07,
|
| 550 |
+
"loss": 0.0,
|
| 551 |
+
"reward": 0.08333333488553762,
|
| 552 |
+
"reward_std": 0.15416352078318596,
|
| 553 |
+
"rewards/accuracy_reward": 0.0416666679084301,
|
| 554 |
+
"rewards/format_reward": 0.041666666977107525,
|
| 555 |
+
"step": 42
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"completion_length": 1522.0312728881836,
|
| 559 |
+
"epoch": 0.344,
|
| 560 |
+
"grad_norm": 0.06469858437776566,
|
| 561 |
+
"kl": 0.0012334585189819336,
|
| 562 |
+
"learning_rate": 9.37590682704584e-07,
|
| 563 |
+
"loss": 0.0,
|
| 564 |
+
"reward": 0.21875000093132257,
|
| 565 |
+
"reward_std": 0.15001969039440155,
|
| 566 |
+
"rewards/accuracy_reward": 0.1875,
|
| 567 |
+
"rewards/format_reward": 0.031250000931322575,
|
| 568 |
+
"step": 43
|
| 569 |
+
},
|
| 570 |
+
{
|
| 571 |
+
"completion_length": 1845.1875305175781,
|
| 572 |
+
"epoch": 0.352,
|
| 573 |
+
"grad_norm": 0.05903354659676552,
|
| 574 |
+
"kl": 0.0015518367290496826,
|
| 575 |
+
"learning_rate": 9.310258896527278e-07,
|
| 576 |
+
"loss": 0.0001,
|
| 577 |
+
"reward": 0.11458333674818277,
|
| 578 |
+
"reward_std": 0.20783207565546036,
|
| 579 |
+
"rewards/accuracy_reward": 0.08333333395421505,
|
| 580 |
+
"rewards/format_reward": 0.031250000931322575,
|
| 581 |
+
"step": 44
|
| 582 |
+
},
|
| 583 |
+
{
|
| 584 |
+
"completion_length": 1626.8541793823242,
|
| 585 |
+
"epoch": 0.36,
|
| 586 |
+
"grad_norm": 0.002270436380058527,
|
| 587 |
+
"kl": 0.0011032521724700928,
|
| 588 |
+
"learning_rate": 9.241613255361454e-07,
|
| 589 |
+
"loss": 0.0,
|
| 590 |
+
"reward": 0.0833333358168602,
|
| 591 |
+
"reward_std": 0.12089946493506432,
|
| 592 |
+
"rewards/accuracy_reward": 0.0,
|
| 593 |
+
"rewards/format_reward": 0.0833333358168602,
|
| 594 |
+
"step": 45
|
| 595 |
+
},
|
| 596 |
+
{
|
| 597 |
+
"completion_length": 1350.5729446411133,
|
| 598 |
+
"epoch": 0.368,
|
| 599 |
+
"grad_norm": 0.08985026180744171,
|
| 600 |
+
"kl": 0.0012712180614471436,
|
| 601 |
+
"learning_rate": 9.17002391031667e-07,
|
| 602 |
+
"loss": 0.0001,
|
| 603 |
+
"reward": 0.3541666753590107,
|
| 604 |
+
"reward_std": 0.3032701872289181,
|
| 605 |
+
"rewards/accuracy_reward": 0.2812500046566129,
|
| 606 |
+
"rewards/format_reward": 0.07291666697710752,
|
| 607 |
+
"step": 46
|
| 608 |
+
},
|
| 609 |
+
{
|
| 610 |
+
"completion_length": 1617.0312728881836,
|
| 611 |
+
"epoch": 0.376,
|
| 612 |
+
"grad_norm": 0.0014874560292810202,
|
| 613 |
+
"kl": 0.0019048452377319336,
|
| 614 |
+
"learning_rate": 9.095547184112122e-07,
|
| 615 |
+
"loss": 0.0001,
|
| 616 |
+
"reward": 0.0729166679084301,
|
| 617 |
+
"reward_std": 0.17735834792256355,
|
| 618 |
+
"rewards/accuracy_reward": 0.02083333395421505,
|
| 619 |
+
"rewards/format_reward": 0.05208333395421505,
|
| 620 |
+
"step": 47
|
| 621 |
+
},
|
| 622 |
+
{
|
| 623 |
+
"completion_length": 1839.6146087646484,
|
| 624 |
+
"epoch": 0.384,
|
| 625 |
+
"grad_norm": 0.05115436390042305,
|
| 626 |
+
"kl": 0.00142669677734375,
|
| 627 |
+
"learning_rate": 9.018241671106134e-07,
|
| 628 |
+
"loss": 0.0001,
|
| 629 |
+
"reward": 0.1145833395421505,
|
| 630 |
+
"reward_std": 0.18088211119174957,
|
| 631 |
+
"rewards/accuracy_reward": 0.09375000093132257,
|
| 632 |
+
"rewards/format_reward": 0.02083333395421505,
|
| 633 |
+
"step": 48
|
| 634 |
+
},
|
| 635 |
+
{
|
| 636 |
+
"completion_length": 1270.8229370117188,
|
| 637 |
+
"epoch": 0.392,
|
| 638 |
+
"grad_norm": 0.09885463118553162,
|
| 639 |
+
"kl": 0.001969635486602783,
|
| 640 |
+
"learning_rate": 8.938168191197233e-07,
|
| 641 |
+
"loss": 0.0001,
|
| 642 |
+
"reward": 0.4583333423361182,
|
| 643 |
+
"reward_std": 0.2759315297007561,
|
| 644 |
+
"rewards/accuracy_reward": 0.3958333432674408,
|
| 645 |
+
"rewards/format_reward": 0.06250000186264515,
|
| 646 |
+
"step": 49
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"completion_length": 1280.7396087646484,
|
| 650 |
+
"epoch": 0.4,
|
| 651 |
+
"grad_norm": 0.10244077444076538,
|
| 652 |
+
"kl": 0.008331120014190674,
|
| 653 |
+
"learning_rate": 8.855389741974244e-07,
|
| 654 |
+
"loss": 0.0003,
|
| 655 |
+
"reward": 0.33333334513008595,
|
| 656 |
+
"reward_std": 0.2721981704235077,
|
| 657 |
+
"rewards/accuracy_reward": 0.21875000838190317,
|
| 658 |
+
"rewards/format_reward": 0.11458333674818277,
|
| 659 |
+
"step": 50
|
| 660 |
+
},
|
| 661 |
+
{
|
| 662 |
+
"completion_length": 1760.2396240234375,
|
| 663 |
+
"epoch": 0.408,
|
| 664 |
+
"grad_norm": 0.15419930219650269,
|
| 665 |
+
"kl": 0.001972787082195282,
|
| 666 |
+
"learning_rate": 8.769971449153122e-07,
|
| 667 |
+
"loss": 0.0001,
|
| 668 |
+
"reward": 0.22916667349636555,
|
| 669 |
+
"reward_std": 0.20281970128417015,
|
| 670 |
+
"rewards/accuracy_reward": 0.18750000186264515,
|
| 671 |
+
"rewards/format_reward": 0.0416666679084301,
|
| 672 |
+
"step": 51
|
| 673 |
+
},
|
| 674 |
+
{
|
| 675 |
+
"completion_length": 1589.5937728881836,
|
| 676 |
+
"epoch": 0.416,
|
| 677 |
+
"grad_norm": 0.08847062289714813,
|
| 678 |
+
"kl": 0.00470280647277832,
|
| 679 |
+
"learning_rate": 8.681980515339463e-07,
|
| 680 |
+
"loss": 0.0002,
|
| 681 |
+
"reward": 0.2083333395421505,
|
| 682 |
+
"reward_std": 0.15885811299085617,
|
| 683 |
+
"rewards/accuracy_reward": 0.125,
|
| 684 |
+
"rewards/format_reward": 0.0833333358168602,
|
| 685 |
+
"step": 52
|
| 686 |
+
},
|
| 687 |
+
{
|
| 688 |
+
"completion_length": 1216.8437843322754,
|
| 689 |
+
"epoch": 0.424,
|
| 690 |
+
"grad_norm": 0.09562092274427414,
|
| 691 |
+
"kl": 0.0038253068923950195,
|
| 692 |
+
"learning_rate": 8.591486167157057e-07,
|
| 693 |
+
"loss": 0.0002,
|
| 694 |
+
"reward": 0.32291668094694614,
|
| 695 |
+
"reward_std": 0.4110877886414528,
|
| 696 |
+
"rewards/accuracy_reward": 0.2187500074505806,
|
| 697 |
+
"rewards/format_reward": 0.10416666883975267,
|
| 698 |
+
"step": 53
|
| 699 |
+
},
|
| 700 |
+
{
|
| 701 |
+
"completion_length": 1607.8125381469727,
|
| 702 |
+
"epoch": 0.432,
|
| 703 |
+
"grad_norm": 0.09562092274427414,
|
| 704 |
+
"kl": 0.0024794340133666992,
|
| 705 |
+
"learning_rate": 8.591486167157057e-07,
|
| 706 |
+
"loss": 0.0001,
|
| 707 |
+
"reward": 0.40625000931322575,
|
| 708 |
+
"reward_std": 0.29860204458236694,
|
| 709 |
+
"rewards/accuracy_reward": 0.2916666716337204,
|
| 710 |
+
"rewards/format_reward": 0.11458333488553762,
|
| 711 |
+
"step": 54
|
| 712 |
+
},
|
| 713 |
+
{
|
| 714 |
+
"completion_length": 1466.2187728881836,
|
| 715 |
+
"epoch": 0.44,
|
| 716 |
+
"grad_norm": 0.12597741186618805,
|
| 717 |
+
"kl": 0.0032531023025512695,
|
| 718 |
+
"learning_rate": 8.498559600784018e-07,
|
| 719 |
+
"loss": 0.0001,
|
| 720 |
+
"reward": 0.10416666883975267,
|
| 721 |
+
"reward_std": 0.28561151400208473,
|
| 722 |
+
"rewards/accuracy_reward": 0.03125,
|
| 723 |
+
"rewards/format_reward": 0.07291666883975267,
|
| 724 |
+
"step": 55
|
| 725 |
+
},
|
| 726 |
+
{
|
| 727 |
+
"completion_length": 1158.5313034057617,
|
| 728 |
+
"epoch": 0.448,
|
| 729 |
+
"grad_norm": 0.1452835649251938,
|
| 730 |
+
"kl": 0.007259368896484375,
|
| 731 |
+
"learning_rate": 8.403273925939395e-07,
|
| 732 |
+
"loss": 0.0003,
|
| 733 |
+
"reward": 0.43750000838190317,
|
| 734 |
+
"reward_std": 0.32371916621923447,
|
| 735 |
+
"rewards/accuracy_reward": 0.2604166669771075,
|
| 736 |
+
"rewards/format_reward": 0.17708333767950535,
|
| 737 |
+
"step": 56
|
| 738 |
+
},
|
| 739 |
+
{
|
| 740 |
+
"completion_length": 1328.3958587646484,
|
| 741 |
+
"epoch": 0.456,
|
| 742 |
+
"grad_norm": 0.0959007516503334,
|
| 743 |
+
"kl": 0.012780427932739258,
|
| 744 |
+
"learning_rate": 8.305704108364301e-07,
|
| 745 |
+
"loss": 0.0005,
|
| 746 |
+
"reward": 0.4062500111758709,
|
| 747 |
+
"reward_std": 0.37512117996811867,
|
| 748 |
+
"rewards/accuracy_reward": 0.22916667442768812,
|
| 749 |
+
"rewards/format_reward": 0.17708333674818277,
|
| 750 |
+
"step": 57
|
| 751 |
+
},
|
| 752 |
+
{
|
| 753 |
+
"completion_length": 1199.7917022705078,
|
| 754 |
+
"epoch": 0.464,
|
| 755 |
+
"grad_norm": 0.053846701979637146,
|
| 756 |
+
"kl": 0.006056308746337891,
|
| 757 |
+
"learning_rate": 8.205926910842825e-07,
|
| 758 |
+
"loss": 0.0002,
|
| 759 |
+
"reward": 0.4791666716337204,
|
| 760 |
+
"reward_std": 0.30927008762955666,
|
| 761 |
+
"rewards/accuracy_reward": 0.19791667442768812,
|
| 762 |
+
"rewards/format_reward": 0.2812500037252903,
|
| 763 |
+
"step": 58
|
| 764 |
+
},
|
| 765 |
+
{
|
| 766 |
+
"completion_length": 1548.4062805175781,
|
| 767 |
+
"epoch": 0.472,
|
| 768 |
+
"grad_norm": 0.15665732324123383,
|
| 769 |
+
"kl": 0.0060198307037353516,
|
| 770 |
+
"learning_rate": 8.104020832809126e-07,
|
| 771 |
+
"loss": 0.0002,
|
| 772 |
+
"reward": 0.4166666865348816,
|
| 773 |
+
"reward_std": 0.37532175332307816,
|
| 774 |
+
"rewards/accuracy_reward": 0.3020833432674408,
|
| 775 |
+
"rewards/format_reward": 0.11458333674818277,
|
| 776 |
+
"step": 59
|
| 777 |
+
},
|
| 778 |
+
{
|
| 779 |
+
"completion_length": 1357.1250457763672,
|
| 780 |
+
"epoch": 0.48,
|
| 781 |
+
"grad_norm": 0.17250967025756836,
|
| 782 |
+
"kl": 0.005661368370056152,
|
| 783 |
+
"learning_rate": 8.00006604858821e-07,
|
| 784 |
+
"loss": 0.0002,
|
| 785 |
+
"reward": 0.42708334419876337,
|
| 786 |
+
"reward_std": 0.44033083319664,
|
| 787 |
+
"rewards/accuracy_reward": 0.1458333358168602,
|
| 788 |
+
"rewards/format_reward": 0.28125000838190317,
|
| 789 |
+
"step": 60
|
| 790 |
+
},
|
| 791 |
+
{
|
| 792 |
+
"completion_length": 1528.5104598999023,
|
| 793 |
+
"epoch": 0.488,
|
| 794 |
+
"grad_norm": 0.39410439133644104,
|
| 795 |
+
"kl": 0.012995481491088867,
|
| 796 |
+
"learning_rate": 7.894144344319013e-07,
|
| 797 |
+
"loss": 0.0005,
|
| 798 |
+
"reward": 0.33333334140479565,
|
| 799 |
+
"reward_std": 0.3153763897716999,
|
| 800 |
+
"rewards/accuracy_reward": 0.18750000279396772,
|
| 801 |
+
"rewards/format_reward": 0.14583333861082792,
|
| 802 |
+
"step": 61
|
| 803 |
+
},
|
| 804 |
+
{
|
| 805 |
+
"completion_length": 1634.0833740234375,
|
| 806 |
+
"epoch": 0.496,
|
| 807 |
+
"grad_norm": 0.003453546669334173,
|
| 808 |
+
"kl": 0.005750775337219238,
|
| 809 |
+
"learning_rate": 7.786339053609382e-07,
|
| 810 |
+
"loss": 0.0002,
|
| 811 |
+
"reward": 0.23958333861082792,
|
| 812 |
+
"reward_std": 0.3065379671752453,
|
| 813 |
+
"rewards/accuracy_reward": 0.010416666977107525,
|
| 814 |
+
"rewards/format_reward": 0.22916667442768812,
|
| 815 |
+
"step": 62
|
| 816 |
+
},
|
| 817 |
+
{
|
| 818 |
+
"completion_length": 1417.6042022705078,
|
| 819 |
+
"epoch": 0.504,
|
| 820 |
+
"grad_norm": 0.2036859691143036,
|
| 821 |
+
"kl": 0.0056678056716918945,
|
| 822 |
+
"learning_rate": 7.676734991973579e-07,
|
| 823 |
+
"loss": 0.0002,
|
| 824 |
+
"reward": 0.2916666744276881,
|
| 825 |
+
"reward_std": 0.35410021990537643,
|
| 826 |
+
"rewards/accuracy_reward": 0.010416666977107525,
|
| 827 |
+
"rewards/format_reward": 0.2812500074505806,
|
| 828 |
+
"step": 63
|
| 829 |
+
},
|
| 830 |
+
{
|
| 831 |
+
"completion_length": 1605.0521087646484,
|
| 832 |
+
"epoch": 0.512,
|
| 833 |
+
"grad_norm": 0.11567196249961853,
|
| 834 |
+
"kl": 0.005307435989379883,
|
| 835 |
+
"learning_rate": 7.56541839010392e-07,
|
| 836 |
+
"loss": 0.0002,
|
| 837 |
+
"reward": 0.32291667349636555,
|
| 838 |
+
"reward_std": 0.4613286033272743,
|
| 839 |
+
"rewards/accuracy_reward": 0.07291666883975267,
|
| 840 |
+
"rewards/format_reward": 0.2500000046566129,
|
| 841 |
+
"step": 64
|
| 842 |
+
},
|
| 843 |
+
{
|
| 844 |
+
"completion_length": 1362.8125076293945,
|
| 845 |
+
"epoch": 0.52,
|
| 846 |
+
"grad_norm": 0.14872920513153076,
|
| 847 |
+
"kl": 0.004861712455749512,
|
| 848 |
+
"learning_rate": 7.45247682602901e-07,
|
| 849 |
+
"loss": 0.0002,
|
| 850 |
+
"reward": 0.47916668839752674,
|
| 851 |
+
"reward_std": 0.47535645961761475,
|
| 852 |
+
"rewards/accuracy_reward": 0.14583333395421505,
|
| 853 |
+
"rewards/format_reward": 0.33333334513008595,
|
| 854 |
+
"step": 65
|
| 855 |
+
},
|
| 856 |
+
{
|
| 857 |
+
"completion_length": 1370.4167022705078,
|
| 858 |
+
"epoch": 0.528,
|
| 859 |
+
"grad_norm": 0.14111247658729553,
|
| 860 |
+
"kl": 0.005552053451538086,
|
| 861 |
+
"learning_rate": 7.337999156211983e-07,
|
| 862 |
+
"loss": 0.0002,
|
| 863 |
+
"reward": 0.5104166753590107,
|
| 864 |
+
"reward_std": 0.46016135439276695,
|
| 865 |
+
"rewards/accuracy_reward": 0.09375000279396772,
|
| 866 |
+
"rewards/format_reward": 0.416666672565043,
|
| 867 |
+
"step": 66
|
| 868 |
+
},
|
| 869 |
+
{
|
| 870 |
+
"completion_length": 1390.9062805175781,
|
| 871 |
+
"epoch": 0.536,
|
| 872 |
+
"grad_norm": 0.0858726054430008,
|
| 873 |
+
"kl": 0.010532855987548828,
|
| 874 |
+
"learning_rate": 7.222075445642904e-07,
|
| 875 |
+
"loss": 0.0004,
|
| 876 |
+
"reward": 0.5000000083819032,
|
| 877 |
+
"reward_std": 0.4547799788415432,
|
| 878 |
+
"rewards/accuracy_reward": 0.15625000093132257,
|
| 879 |
+
"rewards/format_reward": 0.34375000931322575,
|
| 880 |
+
"step": 67
|
| 881 |
+
},
|
| 882 |
+
{
|
| 883 |
+
"completion_length": 1446.6667098999023,
|
| 884 |
+
"epoch": 0.544,
|
| 885 |
+
"grad_norm": 0.12936249375343323,
|
| 886 |
+
"kl": 0.006559848785400391,
|
| 887 |
+
"learning_rate": 7.104796896980408e-07,
|
| 888 |
+
"loss": 0.0003,
|
| 889 |
+
"reward": 0.41666667722165585,
|
| 890 |
+
"reward_std": 0.45394138246774673,
|
| 891 |
+
"rewards/accuracy_reward": 0.03125,
|
| 892 |
+
"rewards/format_reward": 0.38541667722165585,
|
| 893 |
+
"step": 68
|
| 894 |
+
},
|
| 895 |
+
{
|
| 896 |
+
"completion_length": 1062.3229522705078,
|
| 897 |
+
"epoch": 0.552,
|
| 898 |
+
"grad_norm": 0.0991756021976471,
|
| 899 |
+
"kl": 0.028393268585205078,
|
| 900 |
+
"learning_rate": 6.986255778798252e-07,
|
| 901 |
+
"loss": 0.0011,
|
| 902 |
+
"reward": 0.9583333656191826,
|
| 903 |
+
"reward_std": 0.488203439861536,
|
| 904 |
+
"rewards/accuracy_reward": 0.31250000186264515,
|
| 905 |
+
"rewards/format_reward": 0.6458333544433117,
|
| 906 |
+
"step": 69
|
| 907 |
+
},
|
| 908 |
+
{
|
| 909 |
+
"completion_length": 1272.927116394043,
|
| 910 |
+
"epoch": 0.56,
|
| 911 |
+
"grad_norm": 0.12522751092910767,
|
| 912 |
+
"kl": 0.006715059280395508,
|
| 913 |
+
"learning_rate": 6.866545352993266e-07,
|
| 914 |
+
"loss": 0.0003,
|
| 915 |
+
"reward": 0.7187500223517418,
|
| 916 |
+
"reward_std": 0.4580589644610882,
|
| 917 |
+
"rewards/accuracy_reward": 0.27083333395421505,
|
| 918 |
+
"rewards/format_reward": 0.44791667722165585,
|
| 919 |
+
"step": 70
|
| 920 |
+
},
|
| 921 |
+
{
|
| 922 |
+
"completion_length": 1295.0104446411133,
|
| 923 |
+
"epoch": 0.568,
|
| 924 |
+
"grad_norm": 0.09719894826412201,
|
| 925 |
+
"kl": 0.006929874420166016,
|
| 926 |
+
"learning_rate": 6.745759801411822e-07,
|
| 927 |
+
"loss": 0.0003,
|
| 928 |
+
"reward": 0.6770833628252149,
|
| 929 |
+
"reward_std": 0.605831079185009,
|
| 930 |
+
"rewards/accuracy_reward": 0.22916667256504297,
|
| 931 |
+
"rewards/format_reward": 0.44791668839752674,
|
| 932 |
+
"step": 71
|
| 933 |
+
},
|
| 934 |
+
{
|
| 935 |
+
"completion_length": 1681.520866394043,
|
| 936 |
+
"epoch": 0.576,
|
| 937 |
+
"grad_norm": 0.12705527245998383,
|
| 938 |
+
"kl": 0.005926847457885742,
|
| 939 |
+
"learning_rate": 6.623994151752521e-07,
|
| 940 |
+
"loss": 0.0002,
|
| 941 |
+
"reward": 0.2187500037252903,
|
| 942 |
+
"reward_std": 0.25818225741386414,
|
| 943 |
+
"rewards/accuracy_reward": 0.010416666977107525,
|
| 944 |
+
"rewards/format_reward": 0.2083333358168602,
|
| 945 |
+
"step": 72
|
| 946 |
+
},
|
| 947 |
+
{
|
| 948 |
+
"completion_length": 972.6771011352539,
|
| 949 |
+
"epoch": 0.584,
|
| 950 |
+
"grad_norm": 0.09554693847894669,
|
| 951 |
+
"kl": 0.011373519897460938,
|
| 952 |
+
"learning_rate": 6.501344202803414e-07,
|
| 953 |
+
"loss": 0.0005,
|
| 954 |
+
"reward": 0.6666666716337204,
|
| 955 |
+
"reward_std": 0.39136262610554695,
|
| 956 |
+
"rewards/accuracy_reward": 0.06250000186264515,
|
| 957 |
+
"rewards/format_reward": 0.604166679084301,
|
| 958 |
+
"step": 73
|
| 959 |
+
},
|
| 960 |
+
{
|
| 961 |
+
"completion_length": 1333.552116394043,
|
| 962 |
+
"epoch": 0.592,
|
| 963 |
+
"grad_norm": 0.1676543802022934,
|
| 964 |
+
"kl": 0.006979465484619141,
|
| 965 |
+
"learning_rate": 6.377906449072577e-07,
|
| 966 |
+
"loss": 0.0003,
|
| 967 |
+
"reward": 0.5312500149011612,
|
| 968 |
+
"reward_std": 0.4429461173713207,
|
| 969 |
+
"rewards/accuracy_reward": 0.1250000037252903,
|
| 970 |
+
"rewards/format_reward": 0.4062500149011612,
|
| 971 |
+
"step": 74
|
| 972 |
+
},
|
| 973 |
+
{
|
| 974 |
+
"completion_length": 1172.7604446411133,
|
| 975 |
+
"epoch": 0.6,
|
| 976 |
+
"grad_norm": 0.1972622275352478,
|
| 977 |
+
"kl": 0.00921487808227539,
|
| 978 |
+
"learning_rate": 6.253778004871314e-07,
|
| 979 |
+
"loss": 0.0004,
|
| 980 |
+
"reward": 0.6041666902601719,
|
| 981 |
+
"reward_std": 0.505381915718317,
|
| 982 |
+
"rewards/accuracy_reward": 0.11458333861082792,
|
| 983 |
+
"rewards/format_reward": 0.4895833432674408,
|
| 984 |
+
"step": 75
|
| 985 |
+
},
|
| 986 |
+
{
|
| 987 |
+
"completion_length": 1327.0833740234375,
|
| 988 |
+
"epoch": 0.608,
|
| 989 |
+
"grad_norm": 0.16929921507835388,
|
| 990 |
+
"kl": 0.00800466537475586,
|
| 991 |
+
"learning_rate": 6.129056527909748e-07,
|
| 992 |
+
"loss": 0.0003,
|
| 993 |
+
"reward": 0.46875000558793545,
|
| 994 |
+
"reward_std": 0.5199594870209694,
|
| 995 |
+
"rewards/accuracy_reward": 0.010416666977107525,
|
| 996 |
+
"rewards/format_reward": 0.45833334140479565,
|
| 997 |
+
"step": 76
|
| 998 |
+
},
|
| 999 |
+
{
|
| 1000 |
+
"completion_length": 1060.5833587646484,
|
| 1001 |
+
"epoch": 0.616,
|
| 1002 |
+
"grad_norm": 0.2086302489042282,
|
| 1003 |
+
"kl": 0.012459754943847656,
|
| 1004 |
+
"learning_rate": 6.003840142464885e-07,
|
| 1005 |
+
"loss": 0.0005,
|
| 1006 |
+
"reward": 0.8020833656191826,
|
| 1007 |
+
"reward_std": 0.5204068273305893,
|
| 1008 |
+
"rewards/accuracy_reward": 0.15625,
|
| 1009 |
+
"rewards/format_reward": 0.6458333507180214,
|
| 1010 |
+
"step": 77
|
| 1011 |
+
},
|
| 1012 |
+
{
|
| 1013 |
+
"completion_length": 1362.739616394043,
|
| 1014 |
+
"epoch": 0.624,
|
| 1015 |
+
"grad_norm": 0.07928597182035446,
|
| 1016 |
+
"kl": 0.006580352783203125,
|
| 1017 |
+
"learning_rate": 5.878227362181614e-07,
|
| 1018 |
+
"loss": 0.0003,
|
| 1019 |
+
"reward": 0.43750000931322575,
|
| 1020 |
+
"reward_std": 0.25147588178515434,
|
| 1021 |
+
"rewards/accuracy_reward": 0.0,
|
| 1022 |
+
"rewards/format_reward": 0.43750000931322575,
|
| 1023 |
+
"step": 78
|
| 1024 |
+
},
|
| 1025 |
+
{
|
| 1026 |
+
"completion_length": 1356.2604446411133,
|
| 1027 |
+
"epoch": 0.632,
|
| 1028 |
+
"grad_norm": 0.11168374121189117,
|
| 1029 |
+
"kl": 0.00591731071472168,
|
| 1030 |
+
"learning_rate": 5.752317012567362e-07,
|
| 1031 |
+
"loss": 0.0002,
|
| 1032 |
+
"reward": 0.6458333488553762,
|
| 1033 |
+
"reward_std": 0.49627581238746643,
|
| 1034 |
+
"rewards/accuracy_reward": 0.22916667722165585,
|
| 1035 |
+
"rewards/format_reward": 0.41666668094694614,
|
| 1036 |
+
"step": 79
|
| 1037 |
+
},
|
| 1038 |
+
{
|
| 1039 |
+
"completion_length": 960.5625381469727,
|
| 1040 |
+
"epoch": 0.64,
|
| 1041 |
+
"grad_norm": 0.11304420977830887,
|
| 1042 |
+
"kl": 0.011513233184814453,
|
| 1043 |
+
"learning_rate": 5.626208153241411e-07,
|
| 1044 |
+
"loss": 0.0005,
|
| 1045 |
+
"reward": 0.9062500074505806,
|
| 1046 |
+
"reward_std": 0.44416263699531555,
|
| 1047 |
+
"rewards/accuracy_reward": 0.18750001024454832,
|
| 1048 |
+
"rewards/format_reward": 0.718750013038516,
|
| 1049 |
+
"step": 80
|
| 1050 |
+
},
|
| 1051 |
+
{
|
| 1052 |
+
"completion_length": 1400.8750305175781,
|
| 1053 |
+
"epoch": 0.648,
|
| 1054 |
+
"grad_norm": 0.0740160122513771,
|
| 1055 |
+
"kl": 0.0074901580810546875,
|
| 1056 |
+
"learning_rate": 5.5e-07,
|
| 1057 |
+
"loss": 0.0003,
|
| 1058 |
+
"reward": 0.5729166753590107,
|
| 1059 |
+
"reward_std": 0.3763520009815693,
|
| 1060 |
+
"rewards/accuracy_reward": 0.0416666679084301,
|
| 1061 |
+
"rewards/format_reward": 0.5312500111758709,
|
| 1062 |
+
"step": 81
|
| 1063 |
+
},
|
| 1064 |
+
{
|
| 1065 |
+
"completion_length": 754.583366394043,
|
| 1066 |
+
"epoch": 0.656,
|
| 1067 |
+
"grad_norm": 0.22753088176250458,
|
| 1068 |
+
"kl": 0.015882015228271484,
|
| 1069 |
+
"learning_rate": 5.373791846758589e-07,
|
| 1070 |
+
"loss": 0.0006,
|
| 1071 |
+
"reward": 1.041666692122817,
|
| 1072 |
+
"reward_std": 0.43276379629969597,
|
| 1073 |
+
"rewards/accuracy_reward": 0.3437500074505806,
|
| 1074 |
+
"rewards/format_reward": 0.6979166772216558,
|
| 1075 |
+
"step": 82
|
| 1076 |
+
},
|
| 1077 |
+
{
|
| 1078 |
+
"completion_length": 1032.552116394043,
|
| 1079 |
+
"epoch": 0.664,
|
| 1080 |
+
"grad_norm": 0.10078462958335876,
|
| 1081 |
+
"kl": 0.012661933898925781,
|
| 1082 |
+
"learning_rate": 5.247682987432637e-07,
|
| 1083 |
+
"loss": 0.0005,
|
| 1084 |
+
"reward": 0.7500000111758709,
|
| 1085 |
+
"reward_std": 0.5134871490299702,
|
| 1086 |
+
"rewards/accuracy_reward": 0.15625,
|
| 1087 |
+
"rewards/format_reward": 0.5937500111758709,
|
| 1088 |
+
"step": 83
|
| 1089 |
+
},
|
| 1090 |
+
{
|
| 1091 |
+
"completion_length": 1074.0000267028809,
|
| 1092 |
+
"epoch": 0.672,
|
| 1093 |
+
"grad_norm": 0.06993328779935837,
|
| 1094 |
+
"kl": 0.011388182640075684,
|
| 1095 |
+
"learning_rate": 5.121772637818387e-07,
|
| 1096 |
+
"loss": 0.0005,
|
| 1097 |
+
"reward": 0.7291666846722364,
|
| 1098 |
+
"reward_std": 0.3966032788157463,
|
| 1099 |
+
"rewards/accuracy_reward": 0.0416666679084301,
|
| 1100 |
+
"rewards/format_reward": 0.6875000223517418,
|
| 1101 |
+
"step": 84
|
| 1102 |
+
},
|
| 1103 |
+
{
|
| 1104 |
+
"completion_length": 1188.4062728881836,
|
| 1105 |
+
"epoch": 0.68,
|
| 1106 |
+
"grad_norm": 0.11544036120176315,
|
| 1107 |
+
"kl": 0.009247779846191406,
|
| 1108 |
+
"learning_rate": 4.996159857535115e-07,
|
| 1109 |
+
"loss": 0.0004,
|
| 1110 |
+
"reward": 0.8437500149011612,
|
| 1111 |
+
"reward_std": 0.5094321705400944,
|
| 1112 |
+
"rewards/accuracy_reward": 0.2500000074505806,
|
| 1113 |
+
"rewards/format_reward": 0.5937500074505806,
|
| 1114 |
+
"step": 85
|
| 1115 |
+
},
|
| 1116 |
+
{
|
| 1117 |
+
"completion_length": 988.0000457763672,
|
| 1118 |
+
"epoch": 0.688,
|
| 1119 |
+
"grad_norm": 0.13851231336593628,
|
| 1120 |
+
"kl": 0.0120697021484375,
|
| 1121 |
+
"learning_rate": 4.870943472090254e-07,
|
| 1122 |
+
"loss": 0.0005,
|
| 1123 |
+
"reward": 0.739583358168602,
|
| 1124 |
+
"reward_std": 0.42295433580875397,
|
| 1125 |
+
"rewards/accuracy_reward": 0.041666666977107525,
|
| 1126 |
+
"rewards/format_reward": 0.6979166865348816,
|
| 1127 |
+
"step": 86
|
| 1128 |
+
},
|
| 1129 |
+
{
|
| 1130 |
+
"completion_length": 853.0833473205566,
|
| 1131 |
+
"epoch": 0.696,
|
| 1132 |
+
"grad_norm": 0.11069481074810028,
|
| 1133 |
+
"kl": 0.014725208282470703,
|
| 1134 |
+
"learning_rate": 4.7462219951286864e-07,
|
| 1135 |
+
"loss": 0.0006,
|
| 1136 |
+
"reward": 0.6979166716337204,
|
| 1137 |
+
"reward_std": 0.4276985712349415,
|
| 1138 |
+
"rewards/accuracy_reward": 0.010416666977107525,
|
| 1139 |
+
"rewards/format_reward": 0.6875000074505806,
|
| 1140 |
+
"step": 87
|
| 1141 |
+
},
|
| 1142 |
+
{
|
| 1143 |
+
"completion_length": 1298.302116394043,
|
| 1144 |
+
"epoch": 0.704,
|
| 1145 |
+
"grad_norm": 0.127773255109787,
|
| 1146 |
+
"kl": 0.01068568229675293,
|
| 1147 |
+
"learning_rate": 4.6220935509274227e-07,
|
| 1148 |
+
"loss": 0.0004,
|
| 1149 |
+
"reward": 0.5729166893288493,
|
| 1150 |
+
"reward_std": 0.4483075179159641,
|
| 1151 |
+
"rewards/accuracy_reward": 0.0,
|
| 1152 |
+
"rewards/format_reward": 0.5729166893288493,
|
| 1153 |
+
"step": 88
|
| 1154 |
+
},
|
| 1155 |
+
{
|
| 1156 |
+
"completion_length": 806.1354351043701,
|
| 1157 |
+
"epoch": 0.712,
|
| 1158 |
+
"grad_norm": 0.09778036177158356,
|
| 1159 |
+
"kl": 0.018845558166503906,
|
| 1160 |
+
"learning_rate": 4.4986557971965856e-07,
|
| 1161 |
+
"loss": 0.0008,
|
| 1162 |
+
"reward": 1.0937500223517418,
|
| 1163 |
+
"reward_std": 0.2953929826617241,
|
| 1164 |
+
"rewards/accuracy_reward": 0.27083333395421505,
|
| 1165 |
+
"rewards/format_reward": 0.8229166865348816,
|
| 1166 |
+
"step": 89
|
| 1167 |
+
},
|
| 1168 |
+
{
|
| 1169 |
+
"completion_length": 687.0000152587891,
|
| 1170 |
+
"epoch": 0.72,
|
| 1171 |
+
"grad_norm": 0.17915241420269012,
|
| 1172 |
+
"kl": 0.017522811889648438,
|
| 1173 |
+
"learning_rate": 4.3760058482474783e-07,
|
| 1174 |
+
"loss": 0.0007,
|
| 1175 |
+
"reward": 1.041666679084301,
|
| 1176 |
+
"reward_std": 0.38721640035510063,
|
| 1177 |
+
"rewards/accuracy_reward": 0.16666667349636555,
|
| 1178 |
+
"rewards/format_reward": 0.8750000074505806,
|
| 1179 |
+
"step": 90
|
| 1180 |
+
},
|
| 1181 |
+
{
|
| 1182 |
+
"completion_length": 881.3437614440918,
|
| 1183 |
+
"epoch": 0.728,
|
| 1184 |
+
"grad_norm": 0.14008083939552307,
|
| 1185 |
+
"kl": 0.013111591339111328,
|
| 1186 |
+
"learning_rate": 4.254240198588178e-07,
|
| 1187 |
+
"loss": 0.0005,
|
| 1188 |
+
"reward": 0.8854166865348816,
|
| 1189 |
+
"reward_std": 0.31409377604722977,
|
| 1190 |
+
"rewards/accuracy_reward": 0.05208333395421505,
|
| 1191 |
+
"rewards/format_reward": 0.8333333507180214,
|
| 1192 |
+
"step": 91
|
| 1193 |
+
},
|
| 1194 |
+
{
|
| 1195 |
+
"completion_length": 927.1354446411133,
|
| 1196 |
+
"epoch": 0.736,
|
| 1197 |
+
"grad_norm": 0.04579927399754524,
|
| 1198 |
+
"kl": 0.015148162841796875,
|
| 1199 |
+
"learning_rate": 4.133454647006733e-07,
|
| 1200 |
+
"loss": 0.0006,
|
| 1201 |
+
"reward": 1.0520833656191826,
|
| 1202 |
+
"reward_std": 0.3989434242248535,
|
| 1203 |
+
"rewards/accuracy_reward": 0.3645833395421505,
|
| 1204 |
+
"rewards/format_reward": 0.6875000149011612,
|
| 1205 |
+
"step": 92
|
| 1206 |
+
},
|
| 1207 |
+
{
|
| 1208 |
+
"completion_length": 1096.2708702087402,
|
| 1209 |
+
"epoch": 0.744,
|
| 1210 |
+
"grad_norm": 0.11566469073295593,
|
| 1211 |
+
"kl": 0.012713432312011719,
|
| 1212 |
+
"learning_rate": 4.013744221201749e-07,
|
| 1213 |
+
"loss": 0.0005,
|
| 1214 |
+
"reward": 0.6770833414047956,
|
| 1215 |
+
"reward_std": 0.4284285344183445,
|
| 1216 |
+
"rewards/accuracy_reward": 0.0416666679084301,
|
| 1217 |
+
"rewards/format_reward": 0.6354166772216558,
|
| 1218 |
+
"step": 93
|
| 1219 |
+
},
|
| 1220 |
+
{
|
| 1221 |
+
"completion_length": 690.1041774749756,
|
| 1222 |
+
"epoch": 0.752,
|
| 1223 |
+
"grad_norm": 0.15324266254901886,
|
| 1224 |
+
"kl": 0.016754150390625,
|
| 1225 |
+
"learning_rate": 3.895203103019592e-07,
|
| 1226 |
+
"loss": 0.0007,
|
| 1227 |
+
"reward": 1.0208333507180214,
|
| 1228 |
+
"reward_std": 0.4104958660900593,
|
| 1229 |
+
"rewards/accuracy_reward": 0.15625000838190317,
|
| 1230 |
+
"rewards/format_reward": 0.8645833507180214,
|
| 1231 |
+
"step": 94
|
| 1232 |
+
},
|
| 1233 |
+
{
|
| 1234 |
+
"completion_length": 814.8437843322754,
|
| 1235 |
+
"epoch": 0.76,
|
| 1236 |
+
"grad_norm": 0.11410548537969589,
|
| 1237 |
+
"kl": 0.018395423889160156,
|
| 1238 |
+
"learning_rate": 3.777924554357096e-07,
|
| 1239 |
+
"loss": 0.0007,
|
| 1240 |
+
"reward": 0.9479166865348816,
|
| 1241 |
+
"reward_std": 0.4380828067660332,
|
| 1242 |
+
"rewards/accuracy_reward": 0.1145833358168602,
|
| 1243 |
+
"rewards/format_reward": 0.8333333507180214,
|
| 1244 |
+
"step": 95
|
| 1245 |
+
},
|
| 1246 |
+
{
|
| 1247 |
+
"completion_length": 918.2187805175781,
|
| 1248 |
+
"epoch": 0.768,
|
| 1249 |
+
"grad_norm": 0.16425909101963043,
|
| 1250 |
+
"kl": 0.012192249298095703,
|
| 1251 |
+
"learning_rate": 3.662000843788018e-07,
|
| 1252 |
+
"loss": 0.0005,
|
| 1253 |
+
"reward": 0.8125000149011612,
|
| 1254 |
+
"reward_std": 0.3481748141348362,
|
| 1255 |
+
"rewards/accuracy_reward": 0.07291666977107525,
|
| 1256 |
+
"rewards/format_reward": 0.7395833432674408,
|
| 1257 |
+
"step": 96
|
| 1258 |
+
},
|
| 1259 |
+
{
|
| 1260 |
+
"completion_length": 844.8541870117188,
|
| 1261 |
+
"epoch": 0.776,
|
| 1262 |
+
"grad_norm": 0.1520727574825287,
|
| 1263 |
+
"kl": 0.0149383544921875,
|
| 1264 |
+
"learning_rate": 3.547523173970989e-07,
|
| 1265 |
+
"loss": 0.0006,
|
| 1266 |
+
"reward": 0.9791666939854622,
|
| 1267 |
+
"reward_std": 0.42518816888332367,
|
| 1268 |
+
"rewards/accuracy_reward": 0.17708333488553762,
|
| 1269 |
+
"rewards/format_reward": 0.802083358168602,
|
| 1270 |
+
"step": 97
|
| 1271 |
+
},
|
| 1272 |
+
{
|
| 1273 |
+
"completion_length": 678.458366394043,
|
| 1274 |
+
"epoch": 0.784,
|
| 1275 |
+
"grad_norm": 0.11712879687547684,
|
| 1276 |
+
"kl": 0.018598556518554688,
|
| 1277 |
+
"learning_rate": 3.4345816098960794e-07,
|
| 1278 |
+
"loss": 0.0007,
|
| 1279 |
+
"reward": 1.270833358168602,
|
| 1280 |
+
"reward_std": 0.4059586226940155,
|
| 1281 |
+
"rewards/accuracy_reward": 0.4166666753590107,
|
| 1282 |
+
"rewards/format_reward": 0.854166679084301,
|
| 1283 |
+
"step": 98
|
| 1284 |
+
},
|
| 1285 |
+
{
|
| 1286 |
+
"completion_length": 707.4479446411133,
|
| 1287 |
+
"epoch": 0.792,
|
| 1288 |
+
"grad_norm": 0.00943591445684433,
|
| 1289 |
+
"kl": 0.018640995025634766,
|
| 1290 |
+
"learning_rate": 3.323265008026421e-07,
|
| 1291 |
+
"loss": 0.0007,
|
| 1292 |
+
"reward": 1.020833346992731,
|
| 1293 |
+
"reward_std": 0.3158864565193653,
|
| 1294 |
+
"rewards/accuracy_reward": 0.1666666679084301,
|
| 1295 |
+
"rewards/format_reward": 0.8541666753590107,
|
| 1296 |
+
"step": 99
|
| 1297 |
+
},
|
| 1298 |
+
{
|
| 1299 |
+
"completion_length": 900.7083511352539,
|
| 1300 |
"epoch": 0.8,
|
| 1301 |
+
"grad_norm": 0.1339350938796997,
|
| 1302 |
+
"kl": 0.014582633972167969,
|
| 1303 |
+
"learning_rate": 3.2136609463906184e-07,
|
| 1304 |
+
"loss": 0.0006,
|
| 1305 |
+
"reward": 0.7604166865348816,
|
| 1306 |
+
"reward_std": 0.4508432447910309,
|
| 1307 |
+
"rewards/accuracy_reward": 0.0416666679084301,
|
| 1308 |
+
"rewards/format_reward": 0.7187500223517418,
|
| 1309 |
+
"step": 100
|
| 1310 |
+
},
|
| 1311 |
+
{
|
| 1312 |
+
"completion_length": 753.3750286102295,
|
| 1313 |
+
"epoch": 0.808,
|
| 1314 |
+
"grad_norm": 0.09386853128671646,
|
| 1315 |
+
"kl": 0.01984691619873047,
|
| 1316 |
+
"learning_rate": 3.105855655680986e-07,
|
| 1317 |
+
"loss": 0.0008,
|
| 1318 |
+
"reward": 1.0312500149011612,
|
| 1319 |
+
"reward_std": 0.3750321790575981,
|
| 1320 |
+
"rewards/accuracy_reward": 0.16666666883975267,
|
| 1321 |
+
"rewards/format_reward": 0.8645833432674408,
|
| 1322 |
+
"step": 101
|
| 1323 |
+
},
|
| 1324 |
+
{
|
| 1325 |
+
"completion_length": 962.6666831970215,
|
| 1326 |
+
"epoch": 0.816,
|
| 1327 |
+
"grad_norm": 0.17427518963813782,
|
| 1328 |
+
"kl": 0.017024517059326172,
|
| 1329 |
+
"learning_rate": 2.999933951411791e-07,
|
| 1330 |
+
"loss": 0.0007,
|
| 1331 |
+
"reward": 0.9895833656191826,
|
| 1332 |
+
"reward_std": 0.3274080455303192,
|
| 1333 |
+
"rewards/accuracy_reward": 0.2812500074505806,
|
| 1334 |
+
"rewards/format_reward": 0.7083333432674408,
|
| 1335 |
+
"step": 102
|
| 1336 |
+
},
|
| 1337 |
+
{
|
| 1338 |
+
"completion_length": 812.7500228881836,
|
| 1339 |
+
"epoch": 0.824,
|
| 1340 |
+
"grad_norm": 0.08086265623569489,
|
| 1341 |
+
"kl": 0.01704549789428711,
|
| 1342 |
+
"learning_rate": 2.895979167190874e-07,
|
| 1343 |
+
"loss": 0.0007,
|
| 1344 |
+
"reward": 0.864583358168602,
|
| 1345 |
+
"reward_std": 0.4472285062074661,
|
| 1346 |
+
"rewards/accuracy_reward": 0.10416666977107525,
|
| 1347 |
+
"rewards/format_reward": 0.7604166939854622,
|
| 1348 |
+
"step": 103
|
| 1349 |
+
},
|
| 1350 |
+
{
|
| 1351 |
+
"completion_length": 592.4583549499512,
|
| 1352 |
+
"epoch": 0.832,
|
| 1353 |
+
"grad_norm": 0.1811506450176239,
|
| 1354 |
+
"kl": 0.02013111114501953,
|
| 1355 |
+
"learning_rate": 2.794073089157173e-07,
|
| 1356 |
+
"loss": 0.0008,
|
| 1357 |
+
"reward": 1.2187500298023224,
|
| 1358 |
+
"reward_std": 0.39934028312563896,
|
| 1359 |
+
"rewards/accuracy_reward": 0.3020833367481828,
|
| 1360 |
+
"rewards/format_reward": 0.916666679084301,
|
| 1361 |
+
"step": 104
|
| 1362 |
+
},
|
| 1363 |
+
{
|
| 1364 |
+
"completion_length": 906.2500267028809,
|
| 1365 |
+
"epoch": 0.84,
|
| 1366 |
+
"grad_norm": 0.15297923982143402,
|
| 1367 |
+
"kl": 0.018891334533691406,
|
| 1368 |
+
"learning_rate": 2.6942958916356994e-07,
|
| 1369 |
+
"loss": 0.0008,
|
| 1370 |
+
"reward": 0.8125000186264515,
|
| 1371 |
+
"reward_std": 0.44843045622110367,
|
| 1372 |
+
"rewards/accuracy_reward": 0.10416666697710752,
|
| 1373 |
+
"rewards/format_reward": 0.7083333544433117,
|
| 1374 |
+
"step": 105
|
| 1375 |
+
},
|
| 1376 |
+
{
|
| 1377 |
+
"completion_length": 995.7604522705078,
|
| 1378 |
+
"epoch": 0.848,
|
| 1379 |
+
"grad_norm": 0.12237068265676498,
|
| 1380 |
+
"kl": 0.014804840087890625,
|
| 1381 |
+
"learning_rate": 2.596726074060607e-07,
|
| 1382 |
+
"loss": 0.0006,
|
| 1383 |
+
"reward": 0.7604166865348816,
|
| 1384 |
+
"reward_std": 0.5074402429163456,
|
| 1385 |
+
"rewards/accuracy_reward": 0.05208333395421505,
|
| 1386 |
+
"rewards/format_reward": 0.7083333507180214,
|
| 1387 |
+
"step": 106
|
| 1388 |
+
},
|
| 1389 |
+
{
|
| 1390 |
+
"completion_length": 878.5937728881836,
|
| 1391 |
+
"epoch": 0.856,
|
| 1392 |
+
"grad_norm": 0.13649912178516388,
|
| 1393 |
+
"kl": 0.026726722717285156,
|
| 1394 |
+
"learning_rate": 2.501440399215983e-07,
|
| 1395 |
+
"loss": 0.0011,
|
| 1396 |
+
"reward": 0.7916666939854622,
|
| 1397 |
+
"reward_std": 0.4398350641131401,
|
| 1398 |
+
"rewards/accuracy_reward": 0.0416666679084301,
|
| 1399 |
+
"rewards/format_reward": 0.7500000223517418,
|
| 1400 |
+
"step": 107
|
| 1401 |
+
},
|
| 1402 |
+
{
|
| 1403 |
+
"completion_length": 822.510440826416,
|
| 1404 |
+
"epoch": 0.864,
|
| 1405 |
+
"grad_norm": 0.17916011810302734,
|
| 1406 |
+
"kl": 0.018551349639892578,
|
| 1407 |
+
"learning_rate": 2.4085138328429425e-07,
|
| 1408 |
+
"loss": 0.0007,
|
| 1409 |
+
"reward": 1.0520833656191826,
|
| 1410 |
+
"reward_std": 0.4777773655951023,
|
| 1411 |
+
"rewards/accuracy_reward": 0.21875000465661287,
|
| 1412 |
+
"rewards/format_reward": 0.8333333507180214,
|
| 1413 |
+
"step": 108
|
| 1414 |
+
},
|
| 1415 |
+
{
|
| 1416 |
+
"completion_length": 900.5312728881836,
|
| 1417 |
+
"epoch": 0.872,
|
| 1418 |
+
"grad_norm": 0.12796513736248016,
|
| 1419 |
+
"kl": 0.01479339599609375,
|
| 1420 |
+
"learning_rate": 2.3180194846605364e-07,
|
| 1421 |
+
"loss": 0.0006,
|
| 1422 |
+
"reward": 1.0625000223517418,
|
| 1423 |
+
"reward_std": 0.3528694063425064,
|
| 1424 |
+
"rewards/accuracy_reward": 0.2395833432674408,
|
| 1425 |
+
"rewards/format_reward": 0.822916679084301,
|
| 1426 |
+
"step": 109
|
| 1427 |
+
},
|
| 1428 |
+
{
|
| 1429 |
+
"completion_length": 847.4896087646484,
|
| 1430 |
+
"epoch": 0.88,
|
| 1431 |
+
"grad_norm": 0.14577722549438477,
|
| 1432 |
+
"kl": 0.015839576721191406,
|
| 1433 |
+
"learning_rate": 2.2300285508468792e-07,
|
| 1434 |
+
"loss": 0.0006,
|
| 1435 |
+
"reward": 1.0208333656191826,
|
| 1436 |
+
"reward_std": 0.3138932101428509,
|
| 1437 |
+
"rewards/accuracy_reward": 0.1458333358168602,
|
| 1438 |
+
"rewards/format_reward": 0.8750000149011612,
|
| 1439 |
+
"step": 110
|
| 1440 |
+
},
|
| 1441 |
+
{
|
| 1442 |
+
"completion_length": 684.7187690734863,
|
| 1443 |
+
"epoch": 0.888,
|
| 1444 |
+
"grad_norm": 0.08863761276006699,
|
| 1445 |
+
"kl": 0.02144336700439453,
|
| 1446 |
+
"learning_rate": 2.1446102580257546e-07,
|
| 1447 |
+
"loss": 0.0009,
|
| 1448 |
+
"reward": 1.0208333507180214,
|
| 1449 |
+
"reward_std": 0.32314179465174675,
|
| 1450 |
+
"rewards/accuracy_reward": 0.18750000279396772,
|
| 1451 |
+
"rewards/format_reward": 0.8333333507180214,
|
| 1452 |
+
"step": 111
|
| 1453 |
+
},
|
| 1454 |
+
{
|
| 1455 |
+
"completion_length": 648.2708473205566,
|
| 1456 |
+
"epoch": 0.896,
|
| 1457 |
+
"grad_norm": 0.1076781153678894,
|
| 1458 |
+
"kl": 0.018851280212402344,
|
| 1459 |
+
"learning_rate": 2.0618318088027664e-07,
|
| 1460 |
+
"loss": 0.0008,
|
| 1461 |
+
"reward": 0.9895833507180214,
|
| 1462 |
+
"reward_std": 0.34308793768286705,
|
| 1463 |
+
"rewards/accuracy_reward": 0.08333333395421505,
|
| 1464 |
+
"rewards/format_reward": 0.9062500149011612,
|
| 1465 |
+
"step": 112
|
| 1466 |
+
},
|
| 1467 |
+
{
|
| 1468 |
+
"completion_length": 848.979175567627,
|
| 1469 |
+
"epoch": 0.904,
|
| 1470 |
+
"grad_norm": 0.1795138269662857,
|
| 1471 |
+
"kl": 0.028219223022460938,
|
| 1472 |
+
"learning_rate": 1.9817583288938662e-07,
|
| 1473 |
+
"loss": 0.0011,
|
| 1474 |
+
"reward": 0.9687500298023224,
|
| 1475 |
+
"reward_std": 0.48388223350048065,
|
| 1476 |
+
"rewards/accuracy_reward": 0.20833334140479565,
|
| 1477 |
+
"rewards/format_reward": 0.7604166828095913,
|
| 1478 |
+
"step": 113
|
| 1479 |
+
},
|
| 1480 |
+
{
|
| 1481 |
+
"completion_length": 807.0416793823242,
|
| 1482 |
+
"epoch": 0.912,
|
| 1483 |
+
"grad_norm": 0.1863919496536255,
|
| 1484 |
+
"kl": 0.016091346740722656,
|
| 1485 |
+
"learning_rate": 1.9044528158878803e-07,
|
| 1486 |
+
"loss": 0.0006,
|
| 1487 |
+
"reward": 1.0625000149011612,
|
| 1488 |
+
"reward_std": 0.36854929849505424,
|
| 1489 |
+
"rewards/accuracy_reward": 0.16666666697710752,
|
| 1490 |
+
"rewards/format_reward": 0.8958333507180214,
|
| 1491 |
+
"step": 114
|
| 1492 |
+
},
|
| 1493 |
+
{
|
| 1494 |
+
"completion_length": 730.8125228881836,
|
| 1495 |
+
"epoch": 0.92,
|
| 1496 |
+
"grad_norm": 0.008086685091257095,
|
| 1497 |
+
"kl": 0.019659996032714844,
|
| 1498 |
+
"learning_rate": 1.8299760896833295e-07,
|
| 1499 |
+
"loss": 0.0008,
|
| 1500 |
+
"reward": 1.0312500149011612,
|
| 1501 |
+
"reward_std": 0.1851910501718521,
|
| 1502 |
+
"rewards/accuracy_reward": 0.125,
|
| 1503 |
+
"rewards/format_reward": 0.9062500149011612,
|
| 1504 |
+
"step": 115
|
| 1505 |
+
},
|
| 1506 |
+
{
|
| 1507 |
+
"completion_length": 1080.8958587646484,
|
| 1508 |
+
"epoch": 0.928,
|
| 1509 |
+
"grad_norm": 0.13679863512516022,
|
| 1510 |
+
"kl": 0.0175095796585083,
|
| 1511 |
+
"learning_rate": 1.758386744638546e-07,
|
| 1512 |
+
"loss": 0.0007,
|
| 1513 |
+
"reward": 0.6875000223517418,
|
| 1514 |
+
"reward_std": 0.4191872850060463,
|
| 1515 |
+
"rewards/accuracy_reward": 0.0416666679084301,
|
| 1516 |
+
"rewards/format_reward": 0.6458333507180214,
|
| 1517 |
+
"step": 116
|
| 1518 |
+
},
|
| 1519 |
+
{
|
| 1520 |
+
"completion_length": 924.6250343322754,
|
| 1521 |
+
"epoch": 0.936,
|
| 1522 |
+
"grad_norm": 0.1184714287519455,
|
| 1523 |
+
"kl": 0.015748023986816406,
|
| 1524 |
+
"learning_rate": 1.6897411034727217e-07,
|
| 1525 |
+
"loss": 0.0006,
|
| 1526 |
+
"reward": 0.989583358168602,
|
| 1527 |
+
"reward_std": 0.4329289048910141,
|
| 1528 |
+
"rewards/accuracy_reward": 0.16666667722165585,
|
| 1529 |
+
"rewards/format_reward": 0.8229166865348816,
|
| 1530 |
+
"step": 117
|
| 1531 |
+
},
|
| 1532 |
+
{
|
| 1533 |
+
"completion_length": 780.7083511352539,
|
| 1534 |
+
"epoch": 0.944,
|
| 1535 |
+
"grad_norm": 0.15466775000095367,
|
| 1536 |
+
"kl": 0.025674819946289062,
|
| 1537 |
+
"learning_rate": 1.6240931729541597e-07,
|
| 1538 |
+
"loss": 0.001,
|
| 1539 |
+
"reward": 0.9687500149011612,
|
| 1540 |
+
"reward_std": 0.34918052703142166,
|
| 1541 |
+
"rewards/accuracy_reward": 0.12500000093132257,
|
| 1542 |
+
"rewards/format_reward": 0.8437500149011612,
|
| 1543 |
+
"step": 118
|
| 1544 |
+
},
|
| 1545 |
+
{
|
| 1546 |
+
"completion_length": 614.0625038146973,
|
| 1547 |
+
"epoch": 0.952,
|
| 1548 |
+
"grad_norm": 0.21478953957557678,
|
| 1549 |
+
"kl": 0.022258758544921875,
|
| 1550 |
+
"learning_rate": 1.5614946014106085e-07,
|
| 1551 |
+
"loss": 0.0009,
|
| 1552 |
+
"reward": 0.9270833507180214,
|
| 1553 |
+
"reward_std": 0.35268617793917656,
|
| 1554 |
+
"rewards/accuracy_reward": 0.052083334885537624,
|
| 1555 |
+
"rewards/format_reward": 0.8750000149011612,
|
| 1556 |
+
"step": 119
|
| 1557 |
+
},
|
| 1558 |
+
{
|
| 1559 |
+
"completion_length": 785.1666870117188,
|
| 1560 |
+
"epoch": 0.96,
|
| 1561 |
+
"grad_norm": 0.053828902542591095,
|
| 1562 |
+
"kl": 0.01752948760986328,
|
| 1563 |
+
"learning_rate": 1.5019946380950755e-07,
|
| 1564 |
+
"loss": 0.0007,
|
| 1565 |
+
"reward": 1.010416679084301,
|
| 1566 |
+
"reward_std": 0.328002754598856,
|
| 1567 |
+
"rewards/accuracy_reward": 0.1458333358168602,
|
| 1568 |
+
"rewards/format_reward": 0.8645833507180214,
|
| 1569 |
+
"step": 120
|
| 1570 |
+
},
|
| 1571 |
+
{
|
| 1572 |
+
"completion_length": 675.9166831970215,
|
| 1573 |
+
"epoch": 0.968,
|
| 1574 |
+
"grad_norm": 0.31227222084999084,
|
| 1575 |
+
"kl": 0.02220916748046875,
|
| 1576 |
+
"learning_rate": 1.4456400944391144e-07,
|
| 1577 |
+
"loss": 0.0009,
|
| 1578 |
+
"reward": 0.9791666865348816,
|
| 1579 |
+
"reward_std": 0.3436807915568352,
|
| 1580 |
+
"rewards/accuracy_reward": 0.11458333861082792,
|
| 1581 |
+
"rewards/format_reward": 0.8645833507180214,
|
| 1582 |
+
"step": 121
|
| 1583 |
+
},
|
| 1584 |
+
{
|
| 1585 |
+
"completion_length": 901.2083511352539,
|
| 1586 |
+
"epoch": 0.976,
|
| 1587 |
+
"grad_norm": 0.13310351967811584,
|
| 1588 |
+
"kl": 0.020013809204101562,
|
| 1589 |
+
"learning_rate": 1.392475307224065e-07,
|
| 1590 |
+
"loss": 0.0008,
|
| 1591 |
+
"reward": 0.7395833507180214,
|
| 1592 |
+
"reward_std": 0.2778088189661503,
|
| 1593 |
+
"rewards/accuracy_reward": 0.0,
|
| 1594 |
+
"rewards/format_reward": 0.7395833507180214,
|
| 1595 |
+
"step": 122
|
| 1596 |
+
},
|
| 1597 |
+
{
|
| 1598 |
+
"completion_length": 674.9791946411133,
|
| 1599 |
+
"epoch": 0.984,
|
| 1600 |
+
"grad_norm": 0.09870073944330215,
|
| 1601 |
+
"kl": 0.02259063720703125,
|
| 1602 |
+
"learning_rate": 1.3425421036992097e-07,
|
| 1603 |
+
"loss": 0.0009,
|
| 1604 |
+
"reward": 1.0104166939854622,
|
| 1605 |
+
"reward_std": 0.32367467880249023,
|
| 1606 |
+
"rewards/accuracy_reward": 0.15625000093132257,
|
| 1607 |
+
"rewards/format_reward": 0.8541666865348816,
|
| 1608 |
+
"step": 123
|
| 1609 |
+
},
|
| 1610 |
+
{
|
| 1611 |
+
"completion_length": 878.7083625793457,
|
| 1612 |
+
"epoch": 0.992,
|
| 1613 |
+
"grad_norm": 0.2026570737361908,
|
| 1614 |
+
"kl": 0.018926620483398438,
|
| 1615 |
+
"learning_rate": 1.2958797686743014e-07,
|
| 1616 |
+
"loss": 0.0008,
|
| 1617 |
+
"reward": 0.8437500223517418,
|
| 1618 |
+
"reward_std": 0.5021120570600033,
|
| 1619 |
+
"rewards/accuracy_reward": 0.10416666977107525,
|
| 1620 |
+
"rewards/format_reward": 0.7395833507180214,
|
| 1621 |
+
"step": 124
|
| 1622 |
+
},
|
| 1623 |
+
{
|
| 1624 |
+
"completion_length": 611.5,
|
| 1625 |
+
"epoch": 1.0,
|
| 1626 |
+
"grad_norm": 0.2026570737361908,
|
| 1627 |
+
"kl": 0.03096485137939453,
|
| 1628 |
+
"learning_rate": 1.2958797686743014e-07,
|
| 1629 |
+
"loss": 0.0012,
|
| 1630 |
+
"reward": 0.9687500298023224,
|
| 1631 |
+
"reward_std": 0.3619955964386463,
|
| 1632 |
+
"rewards/accuracy_reward": 0.11458333395421505,
|
| 1633 |
+
"rewards/format_reward": 0.854166679084301,
|
| 1634 |
+
"step": 125
|
| 1635 |
+
},
|
| 1636 |
+
{
|
| 1637 |
+
"epoch": 1.0,
|
| 1638 |
+
"step": 125,
|
| 1639 |
"total_flos": 0.0,
|
| 1640 |
+
"train_loss": 0.0003328805177226313,
|
| 1641 |
+
"train_runtime": 21588.6366,
|
| 1642 |
+
"train_samples_per_second": 0.046,
|
| 1643 |
+
"train_steps_per_second": 0.006
|
| 1644 |
}
|
| 1645 |
],
|
| 1646 |
"logging_steps": 1,
|
| 1647 |
+
"max_steps": 125,
|
| 1648 |
"num_input_tokens_seen": 0,
|
| 1649 |
"num_train_epochs": 1,
|
| 1650 |
"save_steps": 500,
|