Commit
•
de1d13d
1
Parent(s):
f7e0437
End of training
Browse files- README.md +64 -0
- all_results.json +9 -0
- config.json +1 -1
- generation_config.json +6 -0
- train_results.json +9 -0
- trainer_state.json +1722 -0
README.md
ADDED
|
@@ -0,0 +1,64 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: transformers
|
| 3 |
+
license: other
|
| 4 |
+
base_model: deepseek-ai/deepseek-coder-7b-instruct-v1.5
|
| 5 |
+
tags:
|
| 6 |
+
- alignment-handbook
|
| 7 |
+
- generated_from_trainer
|
| 8 |
+
- trl
|
| 9 |
+
- dpo
|
| 10 |
+
datasets:
|
| 11 |
+
- self-generate/ds_coder6.7b_reflct_original_cn_mining_oj_iter0-binarized-reflection-scored
|
| 12 |
+
model-index:
|
| 13 |
+
- name: ds_coder6.7b_reflct_adamw_iter1
|
| 14 |
+
results: []
|
| 15 |
+
---
|
| 16 |
+
|
| 17 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 18 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 19 |
+
|
| 20 |
+
# ds_coder6.7b_reflct_adamw_iter1
|
| 21 |
+
|
| 22 |
+
This model is a fine-tuned version of [deepseek-ai/deepseek-coder-7b-instruct-v1.5](https://huggingface.co/deepseek-ai/deepseek-coder-7b-instruct-v1.5) on the self-generate/ds_coder6.7b_reflct_original_cn_mining_oj_iter0-binarized-reflection-scored dataset.
|
| 23 |
+
|
| 24 |
+
## Model description
|
| 25 |
+
|
| 26 |
+
More information needed
|
| 27 |
+
|
| 28 |
+
## Intended uses & limitations
|
| 29 |
+
|
| 30 |
+
More information needed
|
| 31 |
+
|
| 32 |
+
## Training and evaluation data
|
| 33 |
+
|
| 34 |
+
More information needed
|
| 35 |
+
|
| 36 |
+
## Training procedure
|
| 37 |
+
|
| 38 |
+
### Training hyperparameters
|
| 39 |
+
|
| 40 |
+
The following hyperparameters were used during training:
|
| 41 |
+
- learning_rate: 1e-06
|
| 42 |
+
- train_batch_size: 8
|
| 43 |
+
- eval_batch_size: 4
|
| 44 |
+
- seed: 42
|
| 45 |
+
- distributed_type: multi-GPU
|
| 46 |
+
- num_devices: 8
|
| 47 |
+
- total_train_batch_size: 64
|
| 48 |
+
- total_eval_batch_size: 32
|
| 49 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 50 |
+
- lr_scheduler_type: constant
|
| 51 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 52 |
+
- lr_scheduler_warmup_steps: 100
|
| 53 |
+
- num_epochs: 1.0
|
| 54 |
+
|
| 55 |
+
### Training results
|
| 56 |
+
|
| 57 |
+
|
| 58 |
+
|
| 59 |
+
### Framework versions
|
| 60 |
+
|
| 61 |
+
- Transformers 4.45.0
|
| 62 |
+
- Pytorch 2.4.0+cu121
|
| 63 |
+
- Datasets 2.14.6
|
| 64 |
+
- Tokenizers 0.20.3
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.3231356654316187,
|
| 5 |
+
"train_runtime": 754.2708,
|
| 6 |
+
"train_samples": 5120,
|
| 7 |
+
"train_samples_per_second": 6.788,
|
| 8 |
+
"train_steps_per_second": 0.106
|
| 9 |
+
}
|
config.json
CHANGED
|
@@ -25,6 +25,6 @@
|
|
| 25 |
"tie_word_embeddings": false,
|
| 26 |
"torch_dtype": "bfloat16",
|
| 27 |
"transformers_version": "4.45.0",
|
| 28 |
-
"use_cache":
|
| 29 |
"vocab_size": 102400
|
| 30 |
}
|
|
|
|
| 25 |
"tie_word_embeddings": false,
|
| 26 |
"torch_dtype": "bfloat16",
|
| 27 |
"transformers_version": "4.45.0",
|
| 28 |
+
"use_cache": true,
|
| 29 |
"vocab_size": 102400
|
| 30 |
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 100000,
|
| 4 |
+
"eos_token_id": 100015,
|
| 5 |
+
"transformers_version": "4.45.0"
|
| 6 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.3231356654316187,
|
| 5 |
+
"train_runtime": 754.2708,
|
| 6 |
+
"train_samples": 5120,
|
| 7 |
+
"train_samples_per_second": 6.788,
|
| 8 |
+
"train_steps_per_second": 0.106
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,1722 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 1.0,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 80,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"debug/policy_chosen_logits": 16.754077911376953,
|
| 13 |
+
"debug/policy_chosen_logps": -410.5740966796875,
|
| 14 |
+
"debug/policy_rejected_logits": 27.903383255004883,
|
| 15 |
+
"debug/policy_rejected_logps": -454.6379699707031,
|
| 16 |
+
"debug/reference_chosen_logps": -410.5740966796875,
|
| 17 |
+
"debug/reference_rejected_logps": -454.6379699707031,
|
| 18 |
+
"epoch": 0.0125,
|
| 19 |
+
"grad_norm": 8.96581893169365,
|
| 20 |
+
"learning_rate": 1e-06,
|
| 21 |
+
"logits/chosen": 16.754077911376953,
|
| 22 |
+
"logits/rejected": 27.903383255004883,
|
| 23 |
+
"logps/chosen": -410.5740966796875,
|
| 24 |
+
"logps/rejected": -454.6379699707031,
|
| 25 |
+
"loss": 0.5,
|
| 26 |
+
"rewards/accuracies": 0.0,
|
| 27 |
+
"rewards/chosen": 0.0,
|
| 28 |
+
"rewards/margins": 0.0,
|
| 29 |
+
"rewards/rejected": 0.0,
|
| 30 |
+
"step": 1
|
| 31 |
+
},
|
| 32 |
+
{
|
| 33 |
+
"debug/policy_chosen_logits": 21.365768432617188,
|
| 34 |
+
"debug/policy_chosen_logps": -414.0776062011719,
|
| 35 |
+
"debug/policy_rejected_logits": 24.07695770263672,
|
| 36 |
+
"debug/policy_rejected_logps": -463.8160095214844,
|
| 37 |
+
"debug/reference_chosen_logps": -414.4972839355469,
|
| 38 |
+
"debug/reference_rejected_logps": -464.1080322265625,
|
| 39 |
+
"epoch": 0.025,
|
| 40 |
+
"grad_norm": 8.150809103935691,
|
| 41 |
+
"learning_rate": 1e-06,
|
| 42 |
+
"logits/chosen": 21.365768432617188,
|
| 43 |
+
"logits/rejected": 24.07695770263672,
|
| 44 |
+
"logps/chosen": -414.0776062011719,
|
| 45 |
+
"logps/rejected": -463.8160095214844,
|
| 46 |
+
"loss": 0.4994,
|
| 47 |
+
"rewards/accuracies": 0.5,
|
| 48 |
+
"rewards/chosen": 0.004196891561150551,
|
| 49 |
+
"rewards/margins": 0.0012764740968123078,
|
| 50 |
+
"rewards/rejected": 0.002920417580753565,
|
| 51 |
+
"step": 2
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"debug/policy_chosen_logits": 25.1464786529541,
|
| 55 |
+
"debug/policy_chosen_logps": -451.97662353515625,
|
| 56 |
+
"debug/policy_rejected_logits": 20.219999313354492,
|
| 57 |
+
"debug/policy_rejected_logps": -430.0290222167969,
|
| 58 |
+
"debug/reference_chosen_logps": -451.04254150390625,
|
| 59 |
+
"debug/reference_rejected_logps": -429.7825927734375,
|
| 60 |
+
"epoch": 0.0375,
|
| 61 |
+
"grad_norm": 7.628476816083828,
|
| 62 |
+
"learning_rate": 1e-06,
|
| 63 |
+
"logits/chosen": 25.1464786529541,
|
| 64 |
+
"logits/rejected": 20.219999313354492,
|
| 65 |
+
"logps/chosen": -451.97662353515625,
|
| 66 |
+
"logps/rejected": -430.0290222167969,
|
| 67 |
+
"loss": 0.4977,
|
| 68 |
+
"rewards/accuracies": 0.25,
|
| 69 |
+
"rewards/chosen": -0.009341010823845863,
|
| 70 |
+
"rewards/margins": -0.006876640021800995,
|
| 71 |
+
"rewards/rejected": -0.002464370569214225,
|
| 72 |
+
"step": 3
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"debug/policy_chosen_logits": 19.24092674255371,
|
| 76 |
+
"debug/policy_chosen_logps": -438.38348388671875,
|
| 77 |
+
"debug/policy_rejected_logits": 26.255001068115234,
|
| 78 |
+
"debug/policy_rejected_logps": -463.3494567871094,
|
| 79 |
+
"debug/reference_chosen_logps": -437.87542724609375,
|
| 80 |
+
"debug/reference_rejected_logps": -461.87445068359375,
|
| 81 |
+
"epoch": 0.05,
|
| 82 |
+
"grad_norm": 8.325323905761818,
|
| 83 |
+
"learning_rate": 1e-06,
|
| 84 |
+
"logits/chosen": 19.24092674255371,
|
| 85 |
+
"logits/rejected": 26.255001068115234,
|
| 86 |
+
"logps/chosen": -438.38348388671875,
|
| 87 |
+
"logps/rejected": -463.3494567871094,
|
| 88 |
+
"loss": 0.4907,
|
| 89 |
+
"rewards/accuracies": 0.75,
|
| 90 |
+
"rewards/chosen": -0.0050805662758648396,
|
| 91 |
+
"rewards/margins": 0.009669303894042969,
|
| 92 |
+
"rewards/rejected": -0.014749869704246521,
|
| 93 |
+
"step": 4
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"debug/policy_chosen_logits": 17.008541107177734,
|
| 97 |
+
"debug/policy_chosen_logps": -407.684814453125,
|
| 98 |
+
"debug/policy_rejected_logits": 23.88060188293457,
|
| 99 |
+
"debug/policy_rejected_logps": -430.1419982910156,
|
| 100 |
+
"debug/reference_chosen_logps": -407.943115234375,
|
| 101 |
+
"debug/reference_rejected_logps": -429.351806640625,
|
| 102 |
+
"epoch": 0.0625,
|
| 103 |
+
"grad_norm": 8.68321847373836,
|
| 104 |
+
"learning_rate": 1e-06,
|
| 105 |
+
"logits/chosen": 17.008541107177734,
|
| 106 |
+
"logits/rejected": 23.88060188293457,
|
| 107 |
+
"logps/chosen": -407.684814453125,
|
| 108 |
+
"logps/rejected": -430.1419982910156,
|
| 109 |
+
"loss": 0.4865,
|
| 110 |
+
"rewards/accuracies": 0.5,
|
| 111 |
+
"rewards/chosen": 0.002583122346550226,
|
| 112 |
+
"rewards/margins": 0.010485000908374786,
|
| 113 |
+
"rewards/rejected": -0.007901879027485847,
|
| 114 |
+
"step": 5
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"debug/policy_chosen_logits": 23.972543716430664,
|
| 118 |
+
"debug/policy_chosen_logps": -452.3487548828125,
|
| 119 |
+
"debug/policy_rejected_logits": 23.104774475097656,
|
| 120 |
+
"debug/policy_rejected_logps": -456.2618408203125,
|
| 121 |
+
"debug/reference_chosen_logps": -451.52825927734375,
|
| 122 |
+
"debug/reference_rejected_logps": -454.55670166015625,
|
| 123 |
+
"epoch": 0.075,
|
| 124 |
+
"grad_norm": 6.132910409397078,
|
| 125 |
+
"learning_rate": 1e-06,
|
| 126 |
+
"logits/chosen": 23.972543716430664,
|
| 127 |
+
"logits/rejected": 23.104774475097656,
|
| 128 |
+
"logps/chosen": -452.3487548828125,
|
| 129 |
+
"logps/rejected": -456.2618408203125,
|
| 130 |
+
"loss": 0.4884,
|
| 131 |
+
"rewards/accuracies": 0.625,
|
| 132 |
+
"rewards/chosen": -0.00820488017052412,
|
| 133 |
+
"rewards/margins": 0.008846473880112171,
|
| 134 |
+
"rewards/rejected": -0.01705135405063629,
|
| 135 |
+
"step": 6
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"debug/policy_chosen_logits": 18.909212112426758,
|
| 139 |
+
"debug/policy_chosen_logps": -402.86126708984375,
|
| 140 |
+
"debug/policy_rejected_logits": 26.987449645996094,
|
| 141 |
+
"debug/policy_rejected_logps": -445.4053039550781,
|
| 142 |
+
"debug/reference_chosen_logps": -403.7894287109375,
|
| 143 |
+
"debug/reference_rejected_logps": -444.8185729980469,
|
| 144 |
+
"epoch": 0.0875,
|
| 145 |
+
"grad_norm": 8.214444111303614,
|
| 146 |
+
"learning_rate": 1e-06,
|
| 147 |
+
"logits/chosen": 18.909212112426758,
|
| 148 |
+
"logits/rejected": 26.987449645996094,
|
| 149 |
+
"logps/chosen": -402.86126708984375,
|
| 150 |
+
"logps/rejected": -445.4053039550781,
|
| 151 |
+
"loss": 0.4736,
|
| 152 |
+
"rewards/accuracies": 0.625,
|
| 153 |
+
"rewards/chosen": 0.00928169209510088,
|
| 154 |
+
"rewards/margins": 0.015148963779211044,
|
| 155 |
+
"rewards/rejected": -0.005867272149771452,
|
| 156 |
+
"step": 7
|
| 157 |
+
},
|
| 158 |
+
{
|
| 159 |
+
"debug/policy_chosen_logits": 22.914825439453125,
|
| 160 |
+
"debug/policy_chosen_logps": -427.63922119140625,
|
| 161 |
+
"debug/policy_rejected_logits": 26.372671127319336,
|
| 162 |
+
"debug/policy_rejected_logps": -483.4498291015625,
|
| 163 |
+
"debug/reference_chosen_logps": -427.9068603515625,
|
| 164 |
+
"debug/reference_rejected_logps": -479.41705322265625,
|
| 165 |
+
"epoch": 0.1,
|
| 166 |
+
"grad_norm": 8.86676608632659,
|
| 167 |
+
"learning_rate": 1e-06,
|
| 168 |
+
"logits/chosen": 22.914825439453125,
|
| 169 |
+
"logits/rejected": 26.372671127319336,
|
| 170 |
+
"logps/chosen": -427.63922119140625,
|
| 171 |
+
"logps/rejected": -483.4498291015625,
|
| 172 |
+
"loss": 0.4702,
|
| 173 |
+
"rewards/accuracies": 0.875,
|
| 174 |
+
"rewards/chosen": 0.0026765051297843456,
|
| 175 |
+
"rewards/margins": 0.04300403222441673,
|
| 176 |
+
"rewards/rejected": -0.0403275266289711,
|
| 177 |
+
"step": 8
|
| 178 |
+
},
|
| 179 |
+
{
|
| 180 |
+
"debug/policy_chosen_logits": 20.00334930419922,
|
| 181 |
+
"debug/policy_chosen_logps": -432.1858215332031,
|
| 182 |
+
"debug/policy_rejected_logits": 26.359827041625977,
|
| 183 |
+
"debug/policy_rejected_logps": -458.9487609863281,
|
| 184 |
+
"debug/reference_chosen_logps": -433.03985595703125,
|
| 185 |
+
"debug/reference_rejected_logps": -455.154296875,
|
| 186 |
+
"epoch": 0.1125,
|
| 187 |
+
"grad_norm": 7.71461629791825,
|
| 188 |
+
"learning_rate": 1e-06,
|
| 189 |
+
"logits/chosen": 20.00334930419922,
|
| 190 |
+
"logits/rejected": 26.359827041625977,
|
| 191 |
+
"logps/chosen": -432.1858215332031,
|
| 192 |
+
"logps/rejected": -458.9487609863281,
|
| 193 |
+
"loss": 0.4593,
|
| 194 |
+
"rewards/accuracies": 0.875,
|
| 195 |
+
"rewards/chosen": 0.008540572598576546,
|
| 196 |
+
"rewards/margins": 0.04648509621620178,
|
| 197 |
+
"rewards/rejected": -0.037944525480270386,
|
| 198 |
+
"step": 9
|
| 199 |
+
},
|
| 200 |
+
{
|
| 201 |
+
"debug/policy_chosen_logits": 20.777141571044922,
|
| 202 |
+
"debug/policy_chosen_logps": -417.86517333984375,
|
| 203 |
+
"debug/policy_rejected_logits": 20.1833553314209,
|
| 204 |
+
"debug/policy_rejected_logps": -449.54583740234375,
|
| 205 |
+
"debug/reference_chosen_logps": -420.6201171875,
|
| 206 |
+
"debug/reference_rejected_logps": -447.97467041015625,
|
| 207 |
+
"epoch": 0.125,
|
| 208 |
+
"grad_norm": 7.863102292412372,
|
| 209 |
+
"learning_rate": 1e-06,
|
| 210 |
+
"logits/chosen": 20.777141571044922,
|
| 211 |
+
"logits/rejected": 20.1833553314209,
|
| 212 |
+
"logps/chosen": -417.86517333984375,
|
| 213 |
+
"logps/rejected": -449.54583740234375,
|
| 214 |
+
"loss": 0.4364,
|
| 215 |
+
"rewards/accuracies": 0.5,
|
| 216 |
+
"rewards/chosen": 0.027549132704734802,
|
| 217 |
+
"rewards/margins": 0.04326057434082031,
|
| 218 |
+
"rewards/rejected": -0.01571143977344036,
|
| 219 |
+
"step": 10
|
| 220 |
+
},
|
| 221 |
+
{
|
| 222 |
+
"debug/policy_chosen_logits": 20.944578170776367,
|
| 223 |
+
"debug/policy_chosen_logps": -417.01837158203125,
|
| 224 |
+
"debug/policy_rejected_logits": 25.846759796142578,
|
| 225 |
+
"debug/policy_rejected_logps": -450.23468017578125,
|
| 226 |
+
"debug/reference_chosen_logps": -422.13604736328125,
|
| 227 |
+
"debug/reference_rejected_logps": -443.507568359375,
|
| 228 |
+
"epoch": 0.1375,
|
| 229 |
+
"grad_norm": 6.8158984104886,
|
| 230 |
+
"learning_rate": 1e-06,
|
| 231 |
+
"logits/chosen": 20.944578170776367,
|
| 232 |
+
"logits/rejected": 25.846759796142578,
|
| 233 |
+
"logps/chosen": -417.01837158203125,
|
| 234 |
+
"logps/rejected": -450.23468017578125,
|
| 235 |
+
"loss": 0.4331,
|
| 236 |
+
"rewards/accuracies": 1.0,
|
| 237 |
+
"rewards/chosen": 0.05117690563201904,
|
| 238 |
+
"rewards/margins": 0.11844829469919205,
|
| 239 |
+
"rewards/rejected": -0.067271389067173,
|
| 240 |
+
"step": 11
|
| 241 |
+
},
|
| 242 |
+
{
|
| 243 |
+
"debug/policy_chosen_logits": 18.338014602661133,
|
| 244 |
+
"debug/policy_chosen_logps": -406.2593994140625,
|
| 245 |
+
"debug/policy_rejected_logits": 27.694448471069336,
|
| 246 |
+
"debug/policy_rejected_logps": -448.88671875,
|
| 247 |
+
"debug/reference_chosen_logps": -411.54815673828125,
|
| 248 |
+
"debug/reference_rejected_logps": -443.79315185546875,
|
| 249 |
+
"epoch": 0.15,
|
| 250 |
+
"grad_norm": 5.494973875316365,
|
| 251 |
+
"learning_rate": 1e-06,
|
| 252 |
+
"logits/chosen": 18.338014602661133,
|
| 253 |
+
"logits/rejected": 27.694448471069336,
|
| 254 |
+
"logps/chosen": -406.2593994140625,
|
| 255 |
+
"logps/rejected": -448.88671875,
|
| 256 |
+
"loss": 0.4392,
|
| 257 |
+
"rewards/accuracies": 0.75,
|
| 258 |
+
"rewards/chosen": 0.05288715288043022,
|
| 259 |
+
"rewards/margins": 0.10382324457168579,
|
| 260 |
+
"rewards/rejected": -0.05093608796596527,
|
| 261 |
+
"step": 12
|
| 262 |
+
},
|
| 263 |
+
{
|
| 264 |
+
"debug/policy_chosen_logits": 17.940452575683594,
|
| 265 |
+
"debug/policy_chosen_logps": -409.26666259765625,
|
| 266 |
+
"debug/policy_rejected_logits": 26.594871520996094,
|
| 267 |
+
"debug/policy_rejected_logps": -462.8791809082031,
|
| 268 |
+
"debug/reference_chosen_logps": -418.971923828125,
|
| 269 |
+
"debug/reference_rejected_logps": -455.71661376953125,
|
| 270 |
+
"epoch": 0.1625,
|
| 271 |
+
"grad_norm": 6.839480437899935,
|
| 272 |
+
"learning_rate": 1e-06,
|
| 273 |
+
"logits/chosen": 17.940452575683594,
|
| 274 |
+
"logits/rejected": 26.594871520996094,
|
| 275 |
+
"logps/chosen": -409.26666259765625,
|
| 276 |
+
"logps/rejected": -462.8791809082031,
|
| 277 |
+
"loss": 0.4146,
|
| 278 |
+
"rewards/accuracies": 0.875,
|
| 279 |
+
"rewards/chosen": 0.09705245494842529,
|
| 280 |
+
"rewards/margins": 0.1686779260635376,
|
| 281 |
+
"rewards/rejected": -0.0716254711151123,
|
| 282 |
+
"step": 13
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"debug/policy_chosen_logits": 21.89854621887207,
|
| 286 |
+
"debug/policy_chosen_logps": -419.85137939453125,
|
| 287 |
+
"debug/policy_rejected_logits": 28.44005012512207,
|
| 288 |
+
"debug/policy_rejected_logps": -485.2733154296875,
|
| 289 |
+
"debug/reference_chosen_logps": -422.12200927734375,
|
| 290 |
+
"debug/reference_rejected_logps": -476.89013671875,
|
| 291 |
+
"epoch": 0.175,
|
| 292 |
+
"grad_norm": 6.253412161075545,
|
| 293 |
+
"learning_rate": 1e-06,
|
| 294 |
+
"logits/chosen": 21.89854621887207,
|
| 295 |
+
"logits/rejected": 28.44005012512207,
|
| 296 |
+
"logps/chosen": -419.85137939453125,
|
| 297 |
+
"logps/rejected": -485.2733154296875,
|
| 298 |
+
"loss": 0.4071,
|
| 299 |
+
"rewards/accuracies": 0.75,
|
| 300 |
+
"rewards/chosen": 0.022706221789121628,
|
| 301 |
+
"rewards/margins": 0.10653793066740036,
|
| 302 |
+
"rewards/rejected": -0.08383171260356903,
|
| 303 |
+
"step": 14
|
| 304 |
+
},
|
| 305 |
+
{
|
| 306 |
+
"debug/policy_chosen_logits": 17.64580726623535,
|
| 307 |
+
"debug/policy_chosen_logps": -411.6397705078125,
|
| 308 |
+
"debug/policy_rejected_logits": 21.619295120239258,
|
| 309 |
+
"debug/policy_rejected_logps": -427.70172119140625,
|
| 310 |
+
"debug/reference_chosen_logps": -417.8538818359375,
|
| 311 |
+
"debug/reference_rejected_logps": -421.78173828125,
|
| 312 |
+
"epoch": 0.1875,
|
| 313 |
+
"grad_norm": 5.383646265219509,
|
| 314 |
+
"learning_rate": 1e-06,
|
| 315 |
+
"logits/chosen": 17.64580726623535,
|
| 316 |
+
"logits/rejected": 21.619295120239258,
|
| 317 |
+
"logps/chosen": -411.6397705078125,
|
| 318 |
+
"logps/rejected": -427.70172119140625,
|
| 319 |
+
"loss": 0.4213,
|
| 320 |
+
"rewards/accuracies": 0.75,
|
| 321 |
+
"rewards/chosen": 0.0621412992477417,
|
| 322 |
+
"rewards/margins": 0.12134108692407608,
|
| 323 |
+
"rewards/rejected": -0.05919978767633438,
|
| 324 |
+
"step": 15
|
| 325 |
+
},
|
| 326 |
+
{
|
| 327 |
+
"debug/policy_chosen_logits": 20.415315628051758,
|
| 328 |
+
"debug/policy_chosen_logps": -423.58721923828125,
|
| 329 |
+
"debug/policy_rejected_logits": 28.420116424560547,
|
| 330 |
+
"debug/policy_rejected_logps": -477.6827697753906,
|
| 331 |
+
"debug/reference_chosen_logps": -427.49700927734375,
|
| 332 |
+
"debug/reference_rejected_logps": -467.93035888671875,
|
| 333 |
+
"epoch": 0.2,
|
| 334 |
+
"grad_norm": 5.208847435610174,
|
| 335 |
+
"learning_rate": 1e-06,
|
| 336 |
+
"logits/chosen": 20.415315628051758,
|
| 337 |
+
"logits/rejected": 28.420116424560547,
|
| 338 |
+
"logps/chosen": -423.58721923828125,
|
| 339 |
+
"logps/rejected": -477.6827697753906,
|
| 340 |
+
"loss": 0.4097,
|
| 341 |
+
"rewards/accuracies": 0.875,
|
| 342 |
+
"rewards/chosen": 0.03909797593951225,
|
| 343 |
+
"rewards/margins": 0.1366221308708191,
|
| 344 |
+
"rewards/rejected": -0.09752415120601654,
|
| 345 |
+
"step": 16
|
| 346 |
+
},
|
| 347 |
+
{
|
| 348 |
+
"debug/policy_chosen_logits": 17.148258209228516,
|
| 349 |
+
"debug/policy_chosen_logps": -400.81634521484375,
|
| 350 |
+
"debug/policy_rejected_logits": 23.263225555419922,
|
| 351 |
+
"debug/policy_rejected_logps": -491.70166015625,
|
| 352 |
+
"debug/reference_chosen_logps": -413.73779296875,
|
| 353 |
+
"debug/reference_rejected_logps": -475.34991455078125,
|
| 354 |
+
"epoch": 0.2125,
|
| 355 |
+
"grad_norm": 6.371009448433531,
|
| 356 |
+
"learning_rate": 1e-06,
|
| 357 |
+
"logits/chosen": 17.148258209228516,
|
| 358 |
+
"logits/rejected": 23.263225555419922,
|
| 359 |
+
"logps/chosen": -400.81634521484375,
|
| 360 |
+
"logps/rejected": -491.70166015625,
|
| 361 |
+
"loss": 0.362,
|
| 362 |
+
"rewards/accuracies": 0.875,
|
| 363 |
+
"rewards/chosen": 0.12921428680419922,
|
| 364 |
+
"rewards/margins": 0.2927318513393402,
|
| 365 |
+
"rewards/rejected": -0.163517564535141,
|
| 366 |
+
"step": 17
|
| 367 |
+
},
|
| 368 |
+
{
|
| 369 |
+
"debug/policy_chosen_logits": 19.67523956298828,
|
| 370 |
+
"debug/policy_chosen_logps": -417.3398742675781,
|
| 371 |
+
"debug/policy_rejected_logits": 28.897300720214844,
|
| 372 |
+
"debug/policy_rejected_logps": -495.4722900390625,
|
| 373 |
+
"debug/reference_chosen_logps": -429.54852294921875,
|
| 374 |
+
"debug/reference_rejected_logps": -480.121337890625,
|
| 375 |
+
"epoch": 0.225,
|
| 376 |
+
"grad_norm": 4.6720603095118305,
|
| 377 |
+
"learning_rate": 1e-06,
|
| 378 |
+
"logits/chosen": 19.67523956298828,
|
| 379 |
+
"logits/rejected": 28.897300720214844,
|
| 380 |
+
"logps/chosen": -417.3398742675781,
|
| 381 |
+
"logps/rejected": -495.4722900390625,
|
| 382 |
+
"loss": 0.3625,
|
| 383 |
+
"rewards/accuracies": 1.0,
|
| 384 |
+
"rewards/chosen": 0.12208674848079681,
|
| 385 |
+
"rewards/margins": 0.2755962014198303,
|
| 386 |
+
"rewards/rejected": -0.1535094678401947,
|
| 387 |
+
"step": 18
|
| 388 |
+
},
|
| 389 |
+
{
|
| 390 |
+
"debug/policy_chosen_logits": 22.884260177612305,
|
| 391 |
+
"debug/policy_chosen_logps": -439.2691650390625,
|
| 392 |
+
"debug/policy_rejected_logits": 24.13343620300293,
|
| 393 |
+
"debug/policy_rejected_logps": -466.717041015625,
|
| 394 |
+
"debug/reference_chosen_logps": -442.8848876953125,
|
| 395 |
+
"debug/reference_rejected_logps": -443.7426452636719,
|
| 396 |
+
"epoch": 0.2375,
|
| 397 |
+
"grad_norm": 6.3874571813001,
|
| 398 |
+
"learning_rate": 1e-06,
|
| 399 |
+
"logits/chosen": 22.884260177612305,
|
| 400 |
+
"logits/rejected": 24.13343620300293,
|
| 401 |
+
"logps/chosen": -439.2691650390625,
|
| 402 |
+
"logps/rejected": -466.717041015625,
|
| 403 |
+
"loss": 0.2871,
|
| 404 |
+
"rewards/accuracies": 0.75,
|
| 405 |
+
"rewards/chosen": 0.03615710884332657,
|
| 406 |
+
"rewards/margins": 0.26590147614479065,
|
| 407 |
+
"rewards/rejected": -0.22974437475204468,
|
| 408 |
+
"step": 19
|
| 409 |
+
},
|
| 410 |
+
{
|
| 411 |
+
"debug/policy_chosen_logits": 18.085355758666992,
|
| 412 |
+
"debug/policy_chosen_logps": -405.03131103515625,
|
| 413 |
+
"debug/policy_rejected_logits": 25.65454864501953,
|
| 414 |
+
"debug/policy_rejected_logps": -455.36871337890625,
|
| 415 |
+
"debug/reference_chosen_logps": -411.85076904296875,
|
| 416 |
+
"debug/reference_rejected_logps": -429.8324890136719,
|
| 417 |
+
"epoch": 0.25,
|
| 418 |
+
"grad_norm": 4.779696568165662,
|
| 419 |
+
"learning_rate": 1e-06,
|
| 420 |
+
"logits/chosen": 18.085355758666992,
|
| 421 |
+
"logits/rejected": 25.65454864501953,
|
| 422 |
+
"logps/chosen": -405.03131103515625,
|
| 423 |
+
"logps/rejected": -455.36871337890625,
|
| 424 |
+
"loss": 0.3732,
|
| 425 |
+
"rewards/accuracies": 0.875,
|
| 426 |
+
"rewards/chosen": 0.06819470226764679,
|
| 427 |
+
"rewards/margins": 0.3235568404197693,
|
| 428 |
+
"rewards/rejected": -0.2553621232509613,
|
| 429 |
+
"step": 20
|
| 430 |
+
},
|
| 431 |
+
{
|
| 432 |
+
"debug/policy_chosen_logits": 20.691024780273438,
|
| 433 |
+
"debug/policy_chosen_logps": -395.2339782714844,
|
| 434 |
+
"debug/policy_rejected_logits": 28.38985824584961,
|
| 435 |
+
"debug/policy_rejected_logps": -447.9287109375,
|
| 436 |
+
"debug/reference_chosen_logps": -419.5431823730469,
|
| 437 |
+
"debug/reference_rejected_logps": -440.24517822265625,
|
| 438 |
+
"epoch": 0.2625,
|
| 439 |
+
"grad_norm": 4.138890165784992,
|
| 440 |
+
"learning_rate": 1e-06,
|
| 441 |
+
"logits/chosen": 20.691024780273438,
|
| 442 |
+
"logits/rejected": 28.38985824584961,
|
| 443 |
+
"logps/chosen": -395.2339782714844,
|
| 444 |
+
"logps/rejected": -447.9287109375,
|
| 445 |
+
"loss": 0.3061,
|
| 446 |
+
"rewards/accuracies": 0.75,
|
| 447 |
+
"rewards/chosen": 0.24309200048446655,
|
| 448 |
+
"rewards/margins": 0.3199271261692047,
|
| 449 |
+
"rewards/rejected": -0.07683513313531876,
|
| 450 |
+
"step": 21
|
| 451 |
+
},
|
| 452 |
+
{
|
| 453 |
+
"debug/policy_chosen_logits": 19.216005325317383,
|
| 454 |
+
"debug/policy_chosen_logps": -398.2820129394531,
|
| 455 |
+
"debug/policy_rejected_logits": 27.54266929626465,
|
| 456 |
+
"debug/policy_rejected_logps": -462.150146484375,
|
| 457 |
+
"debug/reference_chosen_logps": -406.9403076171875,
|
| 458 |
+
"debug/reference_rejected_logps": -450.525390625,
|
| 459 |
+
"epoch": 0.275,
|
| 460 |
+
"grad_norm": 4.456237735720388,
|
| 461 |
+
"learning_rate": 1e-06,
|
| 462 |
+
"logits/chosen": 19.216005325317383,
|
| 463 |
+
"logits/rejected": 27.54266929626465,
|
| 464 |
+
"logps/chosen": -398.2820129394531,
|
| 465 |
+
"logps/rejected": -462.150146484375,
|
| 466 |
+
"loss": 0.3342,
|
| 467 |
+
"rewards/accuracies": 0.875,
|
| 468 |
+
"rewards/chosen": 0.08658306300640106,
|
| 469 |
+
"rewards/margins": 0.20283077657222748,
|
| 470 |
+
"rewards/rejected": -0.11624770611524582,
|
| 471 |
+
"step": 22
|
| 472 |
+
},
|
| 473 |
+
{
|
| 474 |
+
"debug/policy_chosen_logits": 18.20200538635254,
|
| 475 |
+
"debug/policy_chosen_logps": -399.0750427246094,
|
| 476 |
+
"debug/policy_rejected_logits": 29.885364532470703,
|
| 477 |
+
"debug/policy_rejected_logps": -464.96728515625,
|
| 478 |
+
"debug/reference_chosen_logps": -416.9361877441406,
|
| 479 |
+
"debug/reference_rejected_logps": -453.67095947265625,
|
| 480 |
+
"epoch": 0.2875,
|
| 481 |
+
"grad_norm": 6.33340244824422,
|
| 482 |
+
"learning_rate": 1e-06,
|
| 483 |
+
"logits/chosen": 18.20200538635254,
|
| 484 |
+
"logits/rejected": 29.885364532470703,
|
| 485 |
+
"logps/chosen": -399.0750427246094,
|
| 486 |
+
"logps/rejected": -464.96728515625,
|
| 487 |
+
"loss": 0.3919,
|
| 488 |
+
"rewards/accuracies": 0.75,
|
| 489 |
+
"rewards/chosen": 0.17861154675483704,
|
| 490 |
+
"rewards/margins": 0.2915750741958618,
|
| 491 |
+
"rewards/rejected": -0.11296352744102478,
|
| 492 |
+
"step": 23
|
| 493 |
+
},
|
| 494 |
+
{
|
| 495 |
+
"debug/policy_chosen_logits": 20.623193740844727,
|
| 496 |
+
"debug/policy_chosen_logps": -405.6262512207031,
|
| 497 |
+
"debug/policy_rejected_logits": 30.310110092163086,
|
| 498 |
+
"debug/policy_rejected_logps": -461.8826599121094,
|
| 499 |
+
"debug/reference_chosen_logps": -436.78277587890625,
|
| 500 |
+
"debug/reference_rejected_logps": -433.28759765625,
|
| 501 |
+
"epoch": 0.3,
|
| 502 |
+
"grad_norm": 6.894047386241825,
|
| 503 |
+
"learning_rate": 1e-06,
|
| 504 |
+
"logits/chosen": 20.623193740844727,
|
| 505 |
+
"logits/rejected": 30.310110092163086,
|
| 506 |
+
"logps/chosen": -405.6262512207031,
|
| 507 |
+
"logps/rejected": -461.8826599121094,
|
| 508 |
+
"loss": 0.3025,
|
| 509 |
+
"rewards/accuracies": 1.0,
|
| 510 |
+
"rewards/chosen": 0.3115653395652771,
|
| 511 |
+
"rewards/margins": 0.5975160598754883,
|
| 512 |
+
"rewards/rejected": -0.2859506905078888,
|
| 513 |
+
"step": 24
|
| 514 |
+
},
|
| 515 |
+
{
|
| 516 |
+
"debug/policy_chosen_logits": 21.141868591308594,
|
| 517 |
+
"debug/policy_chosen_logps": -387.9573974609375,
|
| 518 |
+
"debug/policy_rejected_logits": 25.790800094604492,
|
| 519 |
+
"debug/policy_rejected_logps": -441.328369140625,
|
| 520 |
+
"debug/reference_chosen_logps": -414.5980529785156,
|
| 521 |
+
"debug/reference_rejected_logps": -434.82177734375,
|
| 522 |
+
"epoch": 0.3125,
|
| 523 |
+
"grad_norm": 4.7433049140532395,
|
| 524 |
+
"learning_rate": 1e-06,
|
| 525 |
+
"logits/chosen": 21.141868591308594,
|
| 526 |
+
"logits/rejected": 25.790800094604492,
|
| 527 |
+
"logps/chosen": -387.9573974609375,
|
| 528 |
+
"logps/rejected": -441.328369140625,
|
| 529 |
+
"loss": 0.3551,
|
| 530 |
+
"rewards/accuracies": 0.875,
|
| 531 |
+
"rewards/chosen": 0.26640671491622925,
|
| 532 |
+
"rewards/margins": 0.33147239685058594,
|
| 533 |
+
"rewards/rejected": -0.0650656521320343,
|
| 534 |
+
"step": 25
|
| 535 |
+
},
|
| 536 |
+
{
|
| 537 |
+
"debug/policy_chosen_logits": 18.596500396728516,
|
| 538 |
+
"debug/policy_chosen_logps": -389.6454162597656,
|
| 539 |
+
"debug/policy_rejected_logits": 27.930261611938477,
|
| 540 |
+
"debug/policy_rejected_logps": -470.5025329589844,
|
| 541 |
+
"debug/reference_chosen_logps": -414.76007080078125,
|
| 542 |
+
"debug/reference_rejected_logps": -440.7433166503906,
|
| 543 |
+
"epoch": 0.325,
|
| 544 |
+
"grad_norm": 4.306188553423187,
|
| 545 |
+
"learning_rate": 1e-06,
|
| 546 |
+
"logits/chosen": 18.596500396728516,
|
| 547 |
+
"logits/rejected": 27.930261611938477,
|
| 548 |
+
"logps/chosen": -389.6454162597656,
|
| 549 |
+
"logps/rejected": -470.5025329589844,
|
| 550 |
+
"loss": 0.2747,
|
| 551 |
+
"rewards/accuracies": 0.875,
|
| 552 |
+
"rewards/chosen": 0.25114673376083374,
|
| 553 |
+
"rewards/margins": 0.5487388968467712,
|
| 554 |
+
"rewards/rejected": -0.2975921928882599,
|
| 555 |
+
"step": 26
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"debug/policy_chosen_logits": 19.838258743286133,
|
| 559 |
+
"debug/policy_chosen_logps": -396.57708740234375,
|
| 560 |
+
"debug/policy_rejected_logits": 27.663278579711914,
|
| 561 |
+
"debug/policy_rejected_logps": -454.4193115234375,
|
| 562 |
+
"debug/reference_chosen_logps": -419.39898681640625,
|
| 563 |
+
"debug/reference_rejected_logps": -444.2556457519531,
|
| 564 |
+
"epoch": 0.3375,
|
| 565 |
+
"grad_norm": 5.503733763414167,
|
| 566 |
+
"learning_rate": 1e-06,
|
| 567 |
+
"logits/chosen": 19.838258743286133,
|
| 568 |
+
"logits/rejected": 27.663278579711914,
|
| 569 |
+
"logps/chosen": -396.57708740234375,
|
| 570 |
+
"logps/rejected": -454.4193115234375,
|
| 571 |
+
"loss": 0.3097,
|
| 572 |
+
"rewards/accuracies": 1.0,
|
| 573 |
+
"rewards/chosen": 0.22821906208992004,
|
| 574 |
+
"rewards/margins": 0.32985571026802063,
|
| 575 |
+
"rewards/rejected": -0.10163664817810059,
|
| 576 |
+
"step": 27
|
| 577 |
+
},
|
| 578 |
+
{
|
| 579 |
+
"debug/policy_chosen_logits": 19.494264602661133,
|
| 580 |
+
"debug/policy_chosen_logps": -386.3175048828125,
|
| 581 |
+
"debug/policy_rejected_logits": 25.22478675842285,
|
| 582 |
+
"debug/policy_rejected_logps": -454.5428466796875,
|
| 583 |
+
"debug/reference_chosen_logps": -409.989501953125,
|
| 584 |
+
"debug/reference_rejected_logps": -440.07342529296875,
|
| 585 |
+
"epoch": 0.35,
|
| 586 |
+
"grad_norm": 5.4619274191461775,
|
| 587 |
+
"learning_rate": 1e-06,
|
| 588 |
+
"logits/chosen": 19.494264602661133,
|
| 589 |
+
"logits/rejected": 25.22478675842285,
|
| 590 |
+
"logps/chosen": -386.3175048828125,
|
| 591 |
+
"logps/rejected": -454.5428466796875,
|
| 592 |
+
"loss": 0.3427,
|
| 593 |
+
"rewards/accuracies": 0.75,
|
| 594 |
+
"rewards/chosen": 0.23671993613243103,
|
| 595 |
+
"rewards/margins": 0.38141441345214844,
|
| 596 |
+
"rewards/rejected": -0.1446944773197174,
|
| 597 |
+
"step": 28
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"debug/policy_chosen_logits": 20.47563362121582,
|
| 601 |
+
"debug/policy_chosen_logps": -411.8681640625,
|
| 602 |
+
"debug/policy_rejected_logits": 29.555112838745117,
|
| 603 |
+
"debug/policy_rejected_logps": -495.034423828125,
|
| 604 |
+
"debug/reference_chosen_logps": -428.219970703125,
|
| 605 |
+
"debug/reference_rejected_logps": -462.74237060546875,
|
| 606 |
+
"epoch": 0.3625,
|
| 607 |
+
"grad_norm": 5.291532814248025,
|
| 608 |
+
"learning_rate": 1e-06,
|
| 609 |
+
"logits/chosen": 20.47563362121582,
|
| 610 |
+
"logits/rejected": 29.555112838745117,
|
| 611 |
+
"logps/chosen": -411.8681640625,
|
| 612 |
+
"logps/rejected": -495.034423828125,
|
| 613 |
+
"loss": 0.3448,
|
| 614 |
+
"rewards/accuracies": 1.0,
|
| 615 |
+
"rewards/chosen": 0.16351768374443054,
|
| 616 |
+
"rewards/margins": 0.486438125371933,
|
| 617 |
+
"rewards/rejected": -0.32292044162750244,
|
| 618 |
+
"step": 29
|
| 619 |
+
},
|
| 620 |
+
{
|
| 621 |
+
"debug/policy_chosen_logits": 20.484766006469727,
|
| 622 |
+
"debug/policy_chosen_logps": -403.1236572265625,
|
| 623 |
+
"debug/policy_rejected_logits": 28.859256744384766,
|
| 624 |
+
"debug/policy_rejected_logps": -473.91632080078125,
|
| 625 |
+
"debug/reference_chosen_logps": -417.843505859375,
|
| 626 |
+
"debug/reference_rejected_logps": -439.25738525390625,
|
| 627 |
+
"epoch": 0.375,
|
| 628 |
+
"grad_norm": 6.796682505768599,
|
| 629 |
+
"learning_rate": 1e-06,
|
| 630 |
+
"logits/chosen": 20.484766006469727,
|
| 631 |
+
"logits/rejected": 28.859256744384766,
|
| 632 |
+
"logps/chosen": -403.1236572265625,
|
| 633 |
+
"logps/rejected": -473.91632080078125,
|
| 634 |
+
"loss": 0.2546,
|
| 635 |
+
"rewards/accuracies": 0.75,
|
| 636 |
+
"rewards/chosen": 0.14719845354557037,
|
| 637 |
+
"rewards/margins": 0.4937880337238312,
|
| 638 |
+
"rewards/rejected": -0.3465895652770996,
|
| 639 |
+
"step": 30
|
| 640 |
+
},
|
| 641 |
+
{
|
| 642 |
+
"debug/policy_chosen_logits": 22.867298126220703,
|
| 643 |
+
"debug/policy_chosen_logps": -418.6160888671875,
|
| 644 |
+
"debug/policy_rejected_logits": 28.045475006103516,
|
| 645 |
+
"debug/policy_rejected_logps": -446.5901184082031,
|
| 646 |
+
"debug/reference_chosen_logps": -429.67620849609375,
|
| 647 |
+
"debug/reference_rejected_logps": -433.1312255859375,
|
| 648 |
+
"epoch": 0.3875,
|
| 649 |
+
"grad_norm": 7.032111537325665,
|
| 650 |
+
"learning_rate": 1e-06,
|
| 651 |
+
"logits/chosen": 22.867298126220703,
|
| 652 |
+
"logits/rejected": 28.045475006103516,
|
| 653 |
+
"logps/chosen": -418.6160888671875,
|
| 654 |
+
"logps/rejected": -446.5901184082031,
|
| 655 |
+
"loss": 0.3774,
|
| 656 |
+
"rewards/accuracies": 0.75,
|
| 657 |
+
"rewards/chosen": 0.11060100793838501,
|
| 658 |
+
"rewards/margins": 0.24519018828868866,
|
| 659 |
+
"rewards/rejected": -0.13458918035030365,
|
| 660 |
+
"step": 31
|
| 661 |
+
},
|
| 662 |
+
{
|
| 663 |
+
"debug/policy_chosen_logits": 21.08600425720215,
|
| 664 |
+
"debug/policy_chosen_logps": -425.2349853515625,
|
| 665 |
+
"debug/policy_rejected_logits": 28.975238800048828,
|
| 666 |
+
"debug/policy_rejected_logps": -470.2121887207031,
|
| 667 |
+
"debug/reference_chosen_logps": -433.0067443847656,
|
| 668 |
+
"debug/reference_rejected_logps": -453.880859375,
|
| 669 |
+
"epoch": 0.4,
|
| 670 |
+
"grad_norm": 6.839829841567521,
|
| 671 |
+
"learning_rate": 1e-06,
|
| 672 |
+
"logits/chosen": 21.08600425720215,
|
| 673 |
+
"logits/rejected": 28.975238800048828,
|
| 674 |
+
"logps/chosen": -425.2349853515625,
|
| 675 |
+
"logps/rejected": -470.2121887207031,
|
| 676 |
+
"loss": 0.2944,
|
| 677 |
+
"rewards/accuracies": 0.75,
|
| 678 |
+
"rewards/chosen": 0.07771759480237961,
|
| 679 |
+
"rewards/margins": 0.2410309612751007,
|
| 680 |
+
"rewards/rejected": -0.1633133739233017,
|
| 681 |
+
"step": 32
|
| 682 |
+
},
|
| 683 |
+
{
|
| 684 |
+
"debug/policy_chosen_logits": 25.196168899536133,
|
| 685 |
+
"debug/policy_chosen_logps": -411.496337890625,
|
| 686 |
+
"debug/policy_rejected_logits": 26.846052169799805,
|
| 687 |
+
"debug/policy_rejected_logps": -488.6064453125,
|
| 688 |
+
"debug/reference_chosen_logps": -432.0963439941406,
|
| 689 |
+
"debug/reference_rejected_logps": -467.6219787597656,
|
| 690 |
+
"epoch": 0.4125,
|
| 691 |
+
"grad_norm": 4.649169588980754,
|
| 692 |
+
"learning_rate": 1e-06,
|
| 693 |
+
"logits/chosen": 25.196168899536133,
|
| 694 |
+
"logits/rejected": 26.846052169799805,
|
| 695 |
+
"logps/chosen": -411.496337890625,
|
| 696 |
+
"logps/rejected": -488.6064453125,
|
| 697 |
+
"loss": 0.2653,
|
| 698 |
+
"rewards/accuracies": 0.875,
|
| 699 |
+
"rewards/chosen": 0.20599989593029022,
|
| 700 |
+
"rewards/margins": 0.41584473848342896,
|
| 701 |
+
"rewards/rejected": -0.20984479784965515,
|
| 702 |
+
"step": 33
|
| 703 |
+
},
|
| 704 |
+
{
|
| 705 |
+
"debug/policy_chosen_logits": 18.207822799682617,
|
| 706 |
+
"debug/policy_chosen_logps": -383.0249328613281,
|
| 707 |
+
"debug/policy_rejected_logits": 32.11717224121094,
|
| 708 |
+
"debug/policy_rejected_logps": -473.04638671875,
|
| 709 |
+
"debug/reference_chosen_logps": -408.9935302734375,
|
| 710 |
+
"debug/reference_rejected_logps": -424.98388671875,
|
| 711 |
+
"epoch": 0.425,
|
| 712 |
+
"grad_norm": 5.057572035835091,
|
| 713 |
+
"learning_rate": 1e-06,
|
| 714 |
+
"logits/chosen": 18.207822799682617,
|
| 715 |
+
"logits/rejected": 32.11717224121094,
|
| 716 |
+
"logps/chosen": -383.0249328613281,
|
| 717 |
+
"logps/rejected": -473.04638671875,
|
| 718 |
+
"loss": 0.3145,
|
| 719 |
+
"rewards/accuracies": 1.0,
|
| 720 |
+
"rewards/chosen": 0.2596858739852905,
|
| 721 |
+
"rewards/margins": 0.740310549736023,
|
| 722 |
+
"rewards/rejected": -0.4806246757507324,
|
| 723 |
+
"step": 34
|
| 724 |
+
},
|
| 725 |
+
{
|
| 726 |
+
"debug/policy_chosen_logits": 20.943693161010742,
|
| 727 |
+
"debug/policy_chosen_logps": -399.00933837890625,
|
| 728 |
+
"debug/policy_rejected_logits": 30.121450424194336,
|
| 729 |
+
"debug/policy_rejected_logps": -462.94781494140625,
|
| 730 |
+
"debug/reference_chosen_logps": -414.2557067871094,
|
| 731 |
+
"debug/reference_rejected_logps": -443.25909423828125,
|
| 732 |
+
"epoch": 0.4375,
|
| 733 |
+
"grad_norm": 3.7317792693885603,
|
| 734 |
+
"learning_rate": 1e-06,
|
| 735 |
+
"logits/chosen": 20.943693161010742,
|
| 736 |
+
"logits/rejected": 30.121450424194336,
|
| 737 |
+
"logps/chosen": -399.00933837890625,
|
| 738 |
+
"logps/rejected": -462.94781494140625,
|
| 739 |
+
"loss": 0.2713,
|
| 740 |
+
"rewards/accuracies": 0.875,
|
| 741 |
+
"rewards/chosen": 0.15246371924877167,
|
| 742 |
+
"rewards/margins": 0.34935063123703003,
|
| 743 |
+
"rewards/rejected": -0.19688692688941956,
|
| 744 |
+
"step": 35
|
| 745 |
+
},
|
| 746 |
+
{
|
| 747 |
+
"debug/policy_chosen_logits": 18.31939125061035,
|
| 748 |
+
"debug/policy_chosen_logps": -400.14715576171875,
|
| 749 |
+
"debug/policy_rejected_logits": 31.468852996826172,
|
| 750 |
+
"debug/policy_rejected_logps": -513.3980712890625,
|
| 751 |
+
"debug/reference_chosen_logps": -426.7197570800781,
|
| 752 |
+
"debug/reference_rejected_logps": -475.2760925292969,
|
| 753 |
+
"epoch": 0.45,
|
| 754 |
+
"grad_norm": 4.146366903828521,
|
| 755 |
+
"learning_rate": 1e-06,
|
| 756 |
+
"logits/chosen": 18.31939125061035,
|
| 757 |
+
"logits/rejected": 31.468852996826172,
|
| 758 |
+
"logps/chosen": -400.14715576171875,
|
| 759 |
+
"logps/rejected": -513.3980712890625,
|
| 760 |
+
"loss": 0.284,
|
| 761 |
+
"rewards/accuracies": 1.0,
|
| 762 |
+
"rewards/chosen": 0.2657262682914734,
|
| 763 |
+
"rewards/margins": 0.646946370601654,
|
| 764 |
+
"rewards/rejected": -0.3812200725078583,
|
| 765 |
+
"step": 36
|
| 766 |
+
},
|
| 767 |
+
{
|
| 768 |
+
"debug/policy_chosen_logits": 18.51481056213379,
|
| 769 |
+
"debug/policy_chosen_logps": -385.47406005859375,
|
| 770 |
+
"debug/policy_rejected_logits": 30.034343719482422,
|
| 771 |
+
"debug/policy_rejected_logps": -452.683837890625,
|
| 772 |
+
"debug/reference_chosen_logps": -420.1163635253906,
|
| 773 |
+
"debug/reference_rejected_logps": -444.416259765625,
|
| 774 |
+
"epoch": 0.4625,
|
| 775 |
+
"grad_norm": 4.656579513796896,
|
| 776 |
+
"learning_rate": 1e-06,
|
| 777 |
+
"logits/chosen": 18.51481056213379,
|
| 778 |
+
"logits/rejected": 30.034343719482422,
|
| 779 |
+
"logps/chosen": -385.47406005859375,
|
| 780 |
+
"logps/rejected": -452.683837890625,
|
| 781 |
+
"loss": 0.2506,
|
| 782 |
+
"rewards/accuracies": 1.0,
|
| 783 |
+
"rewards/chosen": 0.34642308950424194,
|
| 784 |
+
"rewards/margins": 0.4290991425514221,
|
| 785 |
+
"rewards/rejected": -0.08267608284950256,
|
| 786 |
+
"step": 37
|
| 787 |
+
},
|
| 788 |
+
{
|
| 789 |
+
"debug/policy_chosen_logits": 19.45956039428711,
|
| 790 |
+
"debug/policy_chosen_logps": -408.93182373046875,
|
| 791 |
+
"debug/policy_rejected_logits": 26.489351272583008,
|
| 792 |
+
"debug/policy_rejected_logps": -452.5537414550781,
|
| 793 |
+
"debug/reference_chosen_logps": -432.7354736328125,
|
| 794 |
+
"debug/reference_rejected_logps": -445.1405944824219,
|
| 795 |
+
"epoch": 0.475,
|
| 796 |
+
"grad_norm": 6.017458719962861,
|
| 797 |
+
"learning_rate": 1e-06,
|
| 798 |
+
"logits/chosen": 19.45956039428711,
|
| 799 |
+
"logits/rejected": 26.489351272583008,
|
| 800 |
+
"logps/chosen": -408.93182373046875,
|
| 801 |
+
"logps/rejected": -452.5537414550781,
|
| 802 |
+
"loss": 0.2954,
|
| 803 |
+
"rewards/accuracies": 1.0,
|
| 804 |
+
"rewards/chosen": 0.23803669214248657,
|
| 805 |
+
"rewards/margins": 0.3121681809425354,
|
| 806 |
+
"rewards/rejected": -0.07413151115179062,
|
| 807 |
+
"step": 38
|
| 808 |
+
},
|
| 809 |
+
{
|
| 810 |
+
"debug/policy_chosen_logits": 18.04475975036621,
|
| 811 |
+
"debug/policy_chosen_logps": -381.45330810546875,
|
| 812 |
+
"debug/policy_rejected_logits": 31.125469207763672,
|
| 813 |
+
"debug/policy_rejected_logps": -458.4730529785156,
|
| 814 |
+
"debug/reference_chosen_logps": -411.2501525878906,
|
| 815 |
+
"debug/reference_rejected_logps": -440.47845458984375,
|
| 816 |
+
"epoch": 0.4875,
|
| 817 |
+
"grad_norm": 7.182310738360888,
|
| 818 |
+
"learning_rate": 1e-06,
|
| 819 |
+
"logits/chosen": 18.04475975036621,
|
| 820 |
+
"logits/rejected": 31.125469207763672,
|
| 821 |
+
"logps/chosen": -381.45330810546875,
|
| 822 |
+
"logps/rejected": -458.4730529785156,
|
| 823 |
+
"loss": 0.2881,
|
| 824 |
+
"rewards/accuracies": 0.875,
|
| 825 |
+
"rewards/chosen": 0.29796817898750305,
|
| 826 |
+
"rewards/margins": 0.4779142439365387,
|
| 827 |
+
"rewards/rejected": -0.17994609475135803,
|
| 828 |
+
"step": 39
|
| 829 |
+
},
|
| 830 |
+
{
|
| 831 |
+
"debug/policy_chosen_logits": 18.85544776916504,
|
| 832 |
+
"debug/policy_chosen_logps": -395.41802978515625,
|
| 833 |
+
"debug/policy_rejected_logits": 23.536022186279297,
|
| 834 |
+
"debug/policy_rejected_logps": -410.3462829589844,
|
| 835 |
+
"debug/reference_chosen_logps": -413.8913269042969,
|
| 836 |
+
"debug/reference_rejected_logps": -419.77679443359375,
|
| 837 |
+
"epoch": 0.5,
|
| 838 |
+
"grad_norm": 7.078963246488324,
|
| 839 |
+
"learning_rate": 1e-06,
|
| 840 |
+
"logits/chosen": 18.85544776916504,
|
| 841 |
+
"logits/rejected": 23.536022186279297,
|
| 842 |
+
"logps/chosen": -395.41802978515625,
|
| 843 |
+
"logps/rejected": -410.3462829589844,
|
| 844 |
+
"loss": 0.3245,
|
| 845 |
+
"rewards/accuracies": 0.625,
|
| 846 |
+
"rewards/chosen": 0.18473277986049652,
|
| 847 |
+
"rewards/margins": 0.09042750298976898,
|
| 848 |
+
"rewards/rejected": 0.09430526942014694,
|
| 849 |
+
"step": 40
|
| 850 |
+
},
|
| 851 |
+
{
|
| 852 |
+
"debug/policy_chosen_logits": 19.017606735229492,
|
| 853 |
+
"debug/policy_chosen_logps": -385.16949462890625,
|
| 854 |
+
"debug/policy_rejected_logits": 26.136432647705078,
|
| 855 |
+
"debug/policy_rejected_logps": -455.77178955078125,
|
| 856 |
+
"debug/reference_chosen_logps": -409.3904724121094,
|
| 857 |
+
"debug/reference_rejected_logps": -446.24786376953125,
|
| 858 |
+
"epoch": 0.5125,
|
| 859 |
+
"grad_norm": 4.348585522756699,
|
| 860 |
+
"learning_rate": 1e-06,
|
| 861 |
+
"logits/chosen": 19.017606735229492,
|
| 862 |
+
"logits/rejected": 26.136432647705078,
|
| 863 |
+
"logps/chosen": -385.16949462890625,
|
| 864 |
+
"logps/rejected": -455.77178955078125,
|
| 865 |
+
"loss": 0.3493,
|
| 866 |
+
"rewards/accuracies": 0.75,
|
| 867 |
+
"rewards/chosen": 0.24220973253250122,
|
| 868 |
+
"rewards/margins": 0.33744922280311584,
|
| 869 |
+
"rewards/rejected": -0.09523948282003403,
|
| 870 |
+
"step": 41
|
| 871 |
+
},
|
| 872 |
+
{
|
| 873 |
+
"debug/policy_chosen_logits": 18.70126724243164,
|
| 874 |
+
"debug/policy_chosen_logps": -393.8923645019531,
|
| 875 |
+
"debug/policy_rejected_logits": 28.3759765625,
|
| 876 |
+
"debug/policy_rejected_logps": -474.95550537109375,
|
| 877 |
+
"debug/reference_chosen_logps": -413.5538330078125,
|
| 878 |
+
"debug/reference_rejected_logps": -451.510498046875,
|
| 879 |
+
"epoch": 0.525,
|
| 880 |
+
"grad_norm": 6.403780077024945,
|
| 881 |
+
"learning_rate": 1e-06,
|
| 882 |
+
"logits/chosen": 18.70126724243164,
|
| 883 |
+
"logits/rejected": 28.3759765625,
|
| 884 |
+
"logps/chosen": -393.8923645019531,
|
| 885 |
+
"logps/rejected": -474.95550537109375,
|
| 886 |
+
"loss": 0.2725,
|
| 887 |
+
"rewards/accuracies": 0.625,
|
| 888 |
+
"rewards/chosen": 0.19661453366279602,
|
| 889 |
+
"rewards/margins": 0.4310648441314697,
|
| 890 |
+
"rewards/rejected": -0.2344503253698349,
|
| 891 |
+
"step": 42
|
| 892 |
+
},
|
| 893 |
+
{
|
| 894 |
+
"debug/policy_chosen_logits": 20.40342903137207,
|
| 895 |
+
"debug/policy_chosen_logps": -410.5350341796875,
|
| 896 |
+
"debug/policy_rejected_logits": 31.70136833190918,
|
| 897 |
+
"debug/policy_rejected_logps": -484.47381591796875,
|
| 898 |
+
"debug/reference_chosen_logps": -424.74542236328125,
|
| 899 |
+
"debug/reference_rejected_logps": -446.88623046875,
|
| 900 |
+
"epoch": 0.5375,
|
| 901 |
+
"grad_norm": 9.04429983740781,
|
| 902 |
+
"learning_rate": 1e-06,
|
| 903 |
+
"logits/chosen": 20.40342903137207,
|
| 904 |
+
"logits/rejected": 31.70136833190918,
|
| 905 |
+
"logps/chosen": -410.5350341796875,
|
| 906 |
+
"logps/rejected": -484.47381591796875,
|
| 907 |
+
"loss": 0.2588,
|
| 908 |
+
"rewards/accuracies": 0.875,
|
| 909 |
+
"rewards/chosen": 0.14210368692874908,
|
| 910 |
+
"rewards/margins": 0.5179791450500488,
|
| 911 |
+
"rewards/rejected": -0.37587541341781616,
|
| 912 |
+
"step": 43
|
| 913 |
+
},
|
| 914 |
+
{
|
| 915 |
+
"debug/policy_chosen_logits": 19.714599609375,
|
| 916 |
+
"debug/policy_chosen_logps": -411.0771484375,
|
| 917 |
+
"debug/policy_rejected_logits": 22.63053321838379,
|
| 918 |
+
"debug/policy_rejected_logps": -460.9835510253906,
|
| 919 |
+
"debug/reference_chosen_logps": -425.8843994140625,
|
| 920 |
+
"debug/reference_rejected_logps": -441.40985107421875,
|
| 921 |
+
"epoch": 0.55,
|
| 922 |
+
"grad_norm": 5.462116354601322,
|
| 923 |
+
"learning_rate": 1e-06,
|
| 924 |
+
"logits/chosen": 19.714599609375,
|
| 925 |
+
"logits/rejected": 22.63053321838379,
|
| 926 |
+
"logps/chosen": -411.0771484375,
|
| 927 |
+
"logps/rejected": -460.9835510253906,
|
| 928 |
+
"loss": 0.3296,
|
| 929 |
+
"rewards/accuracies": 1.0,
|
| 930 |
+
"rewards/chosen": 0.14807261526584625,
|
| 931 |
+
"rewards/margins": 0.34380990266799927,
|
| 932 |
+
"rewards/rejected": -0.19573725759983063,
|
| 933 |
+
"step": 44
|
| 934 |
+
},
|
| 935 |
+
{
|
| 936 |
+
"debug/policy_chosen_logits": 23.49720001220703,
|
| 937 |
+
"debug/policy_chosen_logps": -402.09844970703125,
|
| 938 |
+
"debug/policy_rejected_logits": 29.42737579345703,
|
| 939 |
+
"debug/policy_rejected_logps": -443.2966003417969,
|
| 940 |
+
"debug/reference_chosen_logps": -417.8492431640625,
|
| 941 |
+
"debug/reference_rejected_logps": -440.06658935546875,
|
| 942 |
+
"epoch": 0.5625,
|
| 943 |
+
"grad_norm": 5.887141950804354,
|
| 944 |
+
"learning_rate": 1e-06,
|
| 945 |
+
"logits/chosen": 23.49720001220703,
|
| 946 |
+
"logits/rejected": 29.42737579345703,
|
| 947 |
+
"logps/chosen": -402.09844970703125,
|
| 948 |
+
"logps/rejected": -443.2966003417969,
|
| 949 |
+
"loss": 0.3101,
|
| 950 |
+
"rewards/accuracies": 0.75,
|
| 951 |
+
"rewards/chosen": 0.15750765800476074,
|
| 952 |
+
"rewards/margins": 0.1898074597120285,
|
| 953 |
+
"rewards/rejected": -0.03229980170726776,
|
| 954 |
+
"step": 45
|
| 955 |
+
},
|
| 956 |
+
{
|
| 957 |
+
"debug/policy_chosen_logits": 18.707294464111328,
|
| 958 |
+
"debug/policy_chosen_logps": -381.8944396972656,
|
| 959 |
+
"debug/policy_rejected_logits": 26.21320343017578,
|
| 960 |
+
"debug/policy_rejected_logps": -449.5614318847656,
|
| 961 |
+
"debug/reference_chosen_logps": -402.29058837890625,
|
| 962 |
+
"debug/reference_rejected_logps": -439.96856689453125,
|
| 963 |
+
"epoch": 0.575,
|
| 964 |
+
"grad_norm": 7.109794252989897,
|
| 965 |
+
"learning_rate": 1e-06,
|
| 966 |
+
"logits/chosen": 18.707294464111328,
|
| 967 |
+
"logits/rejected": 26.21320343017578,
|
| 968 |
+
"logps/chosen": -381.8944396972656,
|
| 969 |
+
"logps/rejected": -449.5614318847656,
|
| 970 |
+
"loss": 0.2665,
|
| 971 |
+
"rewards/accuracies": 0.875,
|
| 972 |
+
"rewards/chosen": 0.20396174490451813,
|
| 973 |
+
"rewards/margins": 0.2998904585838318,
|
| 974 |
+
"rewards/rejected": -0.09592871367931366,
|
| 975 |
+
"step": 46
|
| 976 |
+
},
|
| 977 |
+
{
|
| 978 |
+
"debug/policy_chosen_logits": 25.76456069946289,
|
| 979 |
+
"debug/policy_chosen_logps": -413.05450439453125,
|
| 980 |
+
"debug/policy_rejected_logits": 29.15418815612793,
|
| 981 |
+
"debug/policy_rejected_logps": -446.2088623046875,
|
| 982 |
+
"debug/reference_chosen_logps": -437.90911865234375,
|
| 983 |
+
"debug/reference_rejected_logps": -437.53997802734375,
|
| 984 |
+
"epoch": 0.5875,
|
| 985 |
+
"grad_norm": 4.107110423899632,
|
| 986 |
+
"learning_rate": 1e-06,
|
| 987 |
+
"logits/chosen": 25.76456069946289,
|
| 988 |
+
"logits/rejected": 29.15418815612793,
|
| 989 |
+
"logps/chosen": -413.05450439453125,
|
| 990 |
+
"logps/rejected": -446.2088623046875,
|
| 991 |
+
"loss": 0.2756,
|
| 992 |
+
"rewards/accuracies": 0.75,
|
| 993 |
+
"rewards/chosen": 0.24854573607444763,
|
| 994 |
+
"rewards/margins": 0.3352348506450653,
|
| 995 |
+
"rewards/rejected": -0.08668914437294006,
|
| 996 |
+
"step": 47
|
| 997 |
+
},
|
| 998 |
+
{
|
| 999 |
+
"debug/policy_chosen_logits": 18.897085189819336,
|
| 1000 |
+
"debug/policy_chosen_logps": -395.95501708984375,
|
| 1001 |
+
"debug/policy_rejected_logits": 24.991113662719727,
|
| 1002 |
+
"debug/policy_rejected_logps": -443.54205322265625,
|
| 1003 |
+
"debug/reference_chosen_logps": -417.3414611816406,
|
| 1004 |
+
"debug/reference_rejected_logps": -438.97320556640625,
|
| 1005 |
+
"epoch": 0.6,
|
| 1006 |
+
"grad_norm": 3.7538316619274052,
|
| 1007 |
+
"learning_rate": 1e-06,
|
| 1008 |
+
"logits/chosen": 18.897085189819336,
|
| 1009 |
+
"logits/rejected": 24.991113662719727,
|
| 1010 |
+
"logps/chosen": -395.95501708984375,
|
| 1011 |
+
"logps/rejected": -443.54205322265625,
|
| 1012 |
+
"loss": 0.3109,
|
| 1013 |
+
"rewards/accuracies": 0.625,
|
| 1014 |
+
"rewards/chosen": 0.21386432647705078,
|
| 1015 |
+
"rewards/margins": 0.25955283641815186,
|
| 1016 |
+
"rewards/rejected": -0.04568850249052048,
|
| 1017 |
+
"step": 48
|
| 1018 |
+
},
|
| 1019 |
+
{
|
| 1020 |
+
"debug/policy_chosen_logits": 17.76668357849121,
|
| 1021 |
+
"debug/policy_chosen_logps": -381.77435302734375,
|
| 1022 |
+
"debug/policy_rejected_logits": 28.45110511779785,
|
| 1023 |
+
"debug/policy_rejected_logps": -451.37335205078125,
|
| 1024 |
+
"debug/reference_chosen_logps": -409.0759582519531,
|
| 1025 |
+
"debug/reference_rejected_logps": -429.466552734375,
|
| 1026 |
+
"epoch": 0.6125,
|
| 1027 |
+
"grad_norm": 4.0217662693111516,
|
| 1028 |
+
"learning_rate": 1e-06,
|
| 1029 |
+
"logits/chosen": 17.76668357849121,
|
| 1030 |
+
"logits/rejected": 28.45110511779785,
|
| 1031 |
+
"logps/chosen": -381.77435302734375,
|
| 1032 |
+
"logps/rejected": -451.37335205078125,
|
| 1033 |
+
"loss": 0.2499,
|
| 1034 |
+
"rewards/accuracies": 0.875,
|
| 1035 |
+
"rewards/chosen": 0.27301597595214844,
|
| 1036 |
+
"rewards/margins": 0.49208390712738037,
|
| 1037 |
+
"rewards/rejected": -0.21906791627407074,
|
| 1038 |
+
"step": 49
|
| 1039 |
+
},
|
| 1040 |
+
{
|
| 1041 |
+
"debug/policy_chosen_logits": 23.378732681274414,
|
| 1042 |
+
"debug/policy_chosen_logps": -419.51385498046875,
|
| 1043 |
+
"debug/policy_rejected_logits": 25.671220779418945,
|
| 1044 |
+
"debug/policy_rejected_logps": -420.00140380859375,
|
| 1045 |
+
"debug/reference_chosen_logps": -430.8975524902344,
|
| 1046 |
+
"debug/reference_rejected_logps": -418.96746826171875,
|
| 1047 |
+
"epoch": 0.625,
|
| 1048 |
+
"grad_norm": 4.909954105878226,
|
| 1049 |
+
"learning_rate": 1e-06,
|
| 1050 |
+
"logits/chosen": 23.378732681274414,
|
| 1051 |
+
"logits/rejected": 25.671220779418945,
|
| 1052 |
+
"logps/chosen": -419.51385498046875,
|
| 1053 |
+
"logps/rejected": -420.00140380859375,
|
| 1054 |
+
"loss": 0.315,
|
| 1055 |
+
"rewards/accuracies": 0.75,
|
| 1056 |
+
"rewards/chosen": 0.11383689939975739,
|
| 1057 |
+
"rewards/margins": 0.12417624145746231,
|
| 1058 |
+
"rewards/rejected": -0.010339349508285522,
|
| 1059 |
+
"step": 50
|
| 1060 |
+
},
|
| 1061 |
+
{
|
| 1062 |
+
"debug/policy_chosen_logits": 23.499061584472656,
|
| 1063 |
+
"debug/policy_chosen_logps": -395.154541015625,
|
| 1064 |
+
"debug/policy_rejected_logits": 26.088699340820312,
|
| 1065 |
+
"debug/policy_rejected_logps": -437.3359375,
|
| 1066 |
+
"debug/reference_chosen_logps": -418.2652587890625,
|
| 1067 |
+
"debug/reference_rejected_logps": -433.26739501953125,
|
| 1068 |
+
"epoch": 0.6375,
|
| 1069 |
+
"grad_norm": 5.626401478627072,
|
| 1070 |
+
"learning_rate": 1e-06,
|
| 1071 |
+
"logits/chosen": 23.499061584472656,
|
| 1072 |
+
"logits/rejected": 26.088699340820312,
|
| 1073 |
+
"logps/chosen": -395.154541015625,
|
| 1074 |
+
"logps/rejected": -437.3359375,
|
| 1075 |
+
"loss": 0.2901,
|
| 1076 |
+
"rewards/accuracies": 0.75,
|
| 1077 |
+
"rewards/chosen": 0.23110699653625488,
|
| 1078 |
+
"rewards/margins": 0.27179256081581116,
|
| 1079 |
+
"rewards/rejected": -0.04068557173013687,
|
| 1080 |
+
"step": 51
|
| 1081 |
+
},
|
| 1082 |
+
{
|
| 1083 |
+
"debug/policy_chosen_logits": 21.639236450195312,
|
| 1084 |
+
"debug/policy_chosen_logps": -402.419189453125,
|
| 1085 |
+
"debug/policy_rejected_logits": 27.019371032714844,
|
| 1086 |
+
"debug/policy_rejected_logps": -426.9820251464844,
|
| 1087 |
+
"debug/reference_chosen_logps": -423.36669921875,
|
| 1088 |
+
"debug/reference_rejected_logps": -438.14837646484375,
|
| 1089 |
+
"epoch": 0.65,
|
| 1090 |
+
"grad_norm": 5.769823478560454,
|
| 1091 |
+
"learning_rate": 1e-06,
|
| 1092 |
+
"logits/chosen": 21.639236450195312,
|
| 1093 |
+
"logits/rejected": 27.019371032714844,
|
| 1094 |
+
"logps/chosen": -402.419189453125,
|
| 1095 |
+
"logps/rejected": -426.9820251464844,
|
| 1096 |
+
"loss": 0.2988,
|
| 1097 |
+
"rewards/accuracies": 0.375,
|
| 1098 |
+
"rewards/chosen": 0.20947498083114624,
|
| 1099 |
+
"rewards/margins": 0.09781143069267273,
|
| 1100 |
+
"rewards/rejected": 0.11166355758905411,
|
| 1101 |
+
"step": 52
|
| 1102 |
+
},
|
| 1103 |
+
{
|
| 1104 |
+
"debug/policy_chosen_logits": 23.30719757080078,
|
| 1105 |
+
"debug/policy_chosen_logps": -396.96221923828125,
|
| 1106 |
+
"debug/policy_rejected_logits": 32.520904541015625,
|
| 1107 |
+
"debug/policy_rejected_logps": -487.1607666015625,
|
| 1108 |
+
"debug/reference_chosen_logps": -425.7676696777344,
|
| 1109 |
+
"debug/reference_rejected_logps": -464.8005676269531,
|
| 1110 |
+
"epoch": 0.6625,
|
| 1111 |
+
"grad_norm": 3.958996117827781,
|
| 1112 |
+
"learning_rate": 1e-06,
|
| 1113 |
+
"logits/chosen": 23.30719757080078,
|
| 1114 |
+
"logits/rejected": 32.520904541015625,
|
| 1115 |
+
"logps/chosen": -396.96221923828125,
|
| 1116 |
+
"logps/rejected": -487.1607666015625,
|
| 1117 |
+
"loss": 0.2418,
|
| 1118 |
+
"rewards/accuracies": 1.0,
|
| 1119 |
+
"rewards/chosen": 0.2880541682243347,
|
| 1120 |
+
"rewards/margins": 0.5116561651229858,
|
| 1121 |
+
"rewards/rejected": -0.2236020267009735,
|
| 1122 |
+
"step": 53
|
| 1123 |
+
},
|
| 1124 |
+
{
|
| 1125 |
+
"debug/policy_chosen_logits": 19.69259262084961,
|
| 1126 |
+
"debug/policy_chosen_logps": -380.706298828125,
|
| 1127 |
+
"debug/policy_rejected_logits": 34.78654098510742,
|
| 1128 |
+
"debug/policy_rejected_logps": -464.5356140136719,
|
| 1129 |
+
"debug/reference_chosen_logps": -408.298583984375,
|
| 1130 |
+
"debug/reference_rejected_logps": -437.8012390136719,
|
| 1131 |
+
"epoch": 0.675,
|
| 1132 |
+
"grad_norm": 5.599167260869624,
|
| 1133 |
+
"learning_rate": 1e-06,
|
| 1134 |
+
"logits/chosen": 19.69259262084961,
|
| 1135 |
+
"logits/rejected": 34.78654098510742,
|
| 1136 |
+
"logps/chosen": -380.706298828125,
|
| 1137 |
+
"logps/rejected": -464.5356140136719,
|
| 1138 |
+
"loss": 0.2732,
|
| 1139 |
+
"rewards/accuracies": 0.875,
|
| 1140 |
+
"rewards/chosen": 0.27592289447784424,
|
| 1141 |
+
"rewards/margins": 0.5432666540145874,
|
| 1142 |
+
"rewards/rejected": -0.26734381914138794,
|
| 1143 |
+
"step": 54
|
| 1144 |
+
},
|
| 1145 |
+
{
|
| 1146 |
+
"debug/policy_chosen_logits": 19.60610580444336,
|
| 1147 |
+
"debug/policy_chosen_logps": -388.4259948730469,
|
| 1148 |
+
"debug/policy_rejected_logits": 28.09248161315918,
|
| 1149 |
+
"debug/policy_rejected_logps": -442.44500732421875,
|
| 1150 |
+
"debug/reference_chosen_logps": -414.5915832519531,
|
| 1151 |
+
"debug/reference_rejected_logps": -430.89666748046875,
|
| 1152 |
+
"epoch": 0.6875,
|
| 1153 |
+
"grad_norm": 3.7882949756257345,
|
| 1154 |
+
"learning_rate": 1e-06,
|
| 1155 |
+
"logits/chosen": 19.60610580444336,
|
| 1156 |
+
"logits/rejected": 28.09248161315918,
|
| 1157 |
+
"logps/chosen": -388.4259948730469,
|
| 1158 |
+
"logps/rejected": -442.44500732421875,
|
| 1159 |
+
"loss": 0.3028,
|
| 1160 |
+
"rewards/accuracies": 1.0,
|
| 1161 |
+
"rewards/chosen": 0.2616558372974396,
|
| 1162 |
+
"rewards/margins": 0.3771390914916992,
|
| 1163 |
+
"rewards/rejected": -0.11548325419425964,
|
| 1164 |
+
"step": 55
|
| 1165 |
+
},
|
| 1166 |
+
{
|
| 1167 |
+
"debug/policy_chosen_logits": 19.7696533203125,
|
| 1168 |
+
"debug/policy_chosen_logps": -400.4974365234375,
|
| 1169 |
+
"debug/policy_rejected_logits": 30.27635383605957,
|
| 1170 |
+
"debug/policy_rejected_logps": -492.6361999511719,
|
| 1171 |
+
"debug/reference_chosen_logps": -423.21453857421875,
|
| 1172 |
+
"debug/reference_rejected_logps": -456.6308288574219,
|
| 1173 |
+
"epoch": 0.7,
|
| 1174 |
+
"grad_norm": 3.5927401238775176,
|
| 1175 |
+
"learning_rate": 1e-06,
|
| 1176 |
+
"logits/chosen": 19.7696533203125,
|
| 1177 |
+
"logits/rejected": 30.27635383605957,
|
| 1178 |
+
"logps/chosen": -400.4974365234375,
|
| 1179 |
+
"logps/rejected": -492.6361999511719,
|
| 1180 |
+
"loss": 0.2819,
|
| 1181 |
+
"rewards/accuracies": 0.875,
|
| 1182 |
+
"rewards/chosen": 0.22717097401618958,
|
| 1183 |
+
"rewards/margins": 0.5872249007225037,
|
| 1184 |
+
"rewards/rejected": -0.3600539267063141,
|
| 1185 |
+
"step": 56
|
| 1186 |
+
},
|
| 1187 |
+
{
|
| 1188 |
+
"debug/policy_chosen_logits": 18.955514907836914,
|
| 1189 |
+
"debug/policy_chosen_logps": -407.9693603515625,
|
| 1190 |
+
"debug/policy_rejected_logits": 29.411741256713867,
|
| 1191 |
+
"debug/policy_rejected_logps": -474.837646484375,
|
| 1192 |
+
"debug/reference_chosen_logps": -418.5709228515625,
|
| 1193 |
+
"debug/reference_rejected_logps": -455.1912841796875,
|
| 1194 |
+
"epoch": 0.7125,
|
| 1195 |
+
"grad_norm": 5.251957841550457,
|
| 1196 |
+
"learning_rate": 1e-06,
|
| 1197 |
+
"logits/chosen": 18.955514907836914,
|
| 1198 |
+
"logits/rejected": 29.411741256713867,
|
| 1199 |
+
"logps/chosen": -407.9693603515625,
|
| 1200 |
+
"logps/rejected": -474.837646484375,
|
| 1201 |
+
"loss": 0.2251,
|
| 1202 |
+
"rewards/accuracies": 0.875,
|
| 1203 |
+
"rewards/chosen": 0.10601577907800674,
|
| 1204 |
+
"rewards/margins": 0.3024793565273285,
|
| 1205 |
+
"rewards/rejected": -0.19646358489990234,
|
| 1206 |
+
"step": 57
|
| 1207 |
+
},
|
| 1208 |
+
{
|
| 1209 |
+
"debug/policy_chosen_logits": 19.137195587158203,
|
| 1210 |
+
"debug/policy_chosen_logps": -394.4432678222656,
|
| 1211 |
+
"debug/policy_rejected_logits": 25.144182205200195,
|
| 1212 |
+
"debug/policy_rejected_logps": -469.2935791015625,
|
| 1213 |
+
"debug/reference_chosen_logps": -414.51580810546875,
|
| 1214 |
+
"debug/reference_rejected_logps": -443.5001220703125,
|
| 1215 |
+
"epoch": 0.725,
|
| 1216 |
+
"grad_norm": 4.675802502460992,
|
| 1217 |
+
"learning_rate": 1e-06,
|
| 1218 |
+
"logits/chosen": 19.137195587158203,
|
| 1219 |
+
"logits/rejected": 25.144182205200195,
|
| 1220 |
+
"logps/chosen": -394.4432678222656,
|
| 1221 |
+
"logps/rejected": -469.2935791015625,
|
| 1222 |
+
"loss": 0.2485,
|
| 1223 |
+
"rewards/accuracies": 0.75,
|
| 1224 |
+
"rewards/chosen": 0.20072558522224426,
|
| 1225 |
+
"rewards/margins": 0.4586600065231323,
|
| 1226 |
+
"rewards/rejected": -0.25793442130088806,
|
| 1227 |
+
"step": 58
|
| 1228 |
+
},
|
| 1229 |
+
{
|
| 1230 |
+
"debug/policy_chosen_logits": 28.265625,
|
| 1231 |
+
"debug/policy_chosen_logps": -439.6889953613281,
|
| 1232 |
+
"debug/policy_rejected_logits": 26.68956756591797,
|
| 1233 |
+
"debug/policy_rejected_logps": -480.45562744140625,
|
| 1234 |
+
"debug/reference_chosen_logps": -452.0250244140625,
|
| 1235 |
+
"debug/reference_rejected_logps": -461.8011474609375,
|
| 1236 |
+
"epoch": 0.7375,
|
| 1237 |
+
"grad_norm": 4.874826559571265,
|
| 1238 |
+
"learning_rate": 1e-06,
|
| 1239 |
+
"logits/chosen": 28.265625,
|
| 1240 |
+
"logits/rejected": 26.68956756591797,
|
| 1241 |
+
"logps/chosen": -439.6889953613281,
|
| 1242 |
+
"logps/rejected": -480.45562744140625,
|
| 1243 |
+
"loss": 0.3103,
|
| 1244 |
+
"rewards/accuracies": 0.75,
|
| 1245 |
+
"rewards/chosen": 0.12336035817861557,
|
| 1246 |
+
"rewards/margins": 0.30990517139434814,
|
| 1247 |
+
"rewards/rejected": -0.18654480576515198,
|
| 1248 |
+
"step": 59
|
| 1249 |
+
},
|
| 1250 |
+
{
|
| 1251 |
+
"debug/policy_chosen_logits": 17.890281677246094,
|
| 1252 |
+
"debug/policy_chosen_logps": -388.18328857421875,
|
| 1253 |
+
"debug/policy_rejected_logits": 29.060598373413086,
|
| 1254 |
+
"debug/policy_rejected_logps": -480.3656921386719,
|
| 1255 |
+
"debug/reference_chosen_logps": -411.6337890625,
|
| 1256 |
+
"debug/reference_rejected_logps": -446.0444641113281,
|
| 1257 |
+
"epoch": 0.75,
|
| 1258 |
+
"grad_norm": 3.8092727948167764,
|
| 1259 |
+
"learning_rate": 1e-06,
|
| 1260 |
+
"logits/chosen": 17.890281677246094,
|
| 1261 |
+
"logits/rejected": 29.060598373413086,
|
| 1262 |
+
"logps/chosen": -388.18328857421875,
|
| 1263 |
+
"logps/rejected": -480.3656921386719,
|
| 1264 |
+
"loss": 0.3023,
|
| 1265 |
+
"rewards/accuracies": 1.0,
|
| 1266 |
+
"rewards/chosen": 0.23450516164302826,
|
| 1267 |
+
"rewards/margins": 0.5777173638343811,
|
| 1268 |
+
"rewards/rejected": -0.3432122468948364,
|
| 1269 |
+
"step": 60
|
| 1270 |
+
},
|
| 1271 |
+
{
|
| 1272 |
+
"debug/policy_chosen_logits": 21.306364059448242,
|
| 1273 |
+
"debug/policy_chosen_logps": -417.54681396484375,
|
| 1274 |
+
"debug/policy_rejected_logits": 23.916120529174805,
|
| 1275 |
+
"debug/policy_rejected_logps": -446.38153076171875,
|
| 1276 |
+
"debug/reference_chosen_logps": -437.6956787109375,
|
| 1277 |
+
"debug/reference_rejected_logps": -429.464599609375,
|
| 1278 |
+
"epoch": 0.7625,
|
| 1279 |
+
"grad_norm": 3.732343889219544,
|
| 1280 |
+
"learning_rate": 1e-06,
|
| 1281 |
+
"logits/chosen": 21.306364059448242,
|
| 1282 |
+
"logits/rejected": 23.916120529174805,
|
| 1283 |
+
"logps/chosen": -417.54681396484375,
|
| 1284 |
+
"logps/rejected": -446.38153076171875,
|
| 1285 |
+
"loss": 0.2784,
|
| 1286 |
+
"rewards/accuracies": 0.75,
|
| 1287 |
+
"rewards/chosen": 0.2014881819486618,
|
| 1288 |
+
"rewards/margins": 0.3706575632095337,
|
| 1289 |
+
"rewards/rejected": -0.16916939616203308,
|
| 1290 |
+
"step": 61
|
| 1291 |
+
},
|
| 1292 |
+
{
|
| 1293 |
+
"debug/policy_chosen_logits": 19.258764266967773,
|
| 1294 |
+
"debug/policy_chosen_logps": -400.505126953125,
|
| 1295 |
+
"debug/policy_rejected_logits": 29.247554779052734,
|
| 1296 |
+
"debug/policy_rejected_logps": -488.6958923339844,
|
| 1297 |
+
"debug/reference_chosen_logps": -424.3946533203125,
|
| 1298 |
+
"debug/reference_rejected_logps": -456.62408447265625,
|
| 1299 |
+
"epoch": 0.775,
|
| 1300 |
+
"grad_norm": 3.9288590692913727,
|
| 1301 |
+
"learning_rate": 1e-06,
|
| 1302 |
+
"logits/chosen": 19.258764266967773,
|
| 1303 |
+
"logits/rejected": 29.247554779052734,
|
| 1304 |
+
"logps/chosen": -400.505126953125,
|
| 1305 |
+
"logps/rejected": -488.6958923339844,
|
| 1306 |
+
"loss": 0.2174,
|
| 1307 |
+
"rewards/accuracies": 1.0,
|
| 1308 |
+
"rewards/chosen": 0.23889553546905518,
|
| 1309 |
+
"rewards/margins": 0.5596135854721069,
|
| 1310 |
+
"rewards/rejected": -0.32071805000305176,
|
| 1311 |
+
"step": 62
|
| 1312 |
+
},
|
| 1313 |
+
{
|
| 1314 |
+
"debug/policy_chosen_logits": 25.917665481567383,
|
| 1315 |
+
"debug/policy_chosen_logps": -414.66156005859375,
|
| 1316 |
+
"debug/policy_rejected_logits": 27.687381744384766,
|
| 1317 |
+
"debug/policy_rejected_logps": -445.0428771972656,
|
| 1318 |
+
"debug/reference_chosen_logps": -433.19635009765625,
|
| 1319 |
+
"debug/reference_rejected_logps": -456.6318664550781,
|
| 1320 |
+
"epoch": 0.7875,
|
| 1321 |
+
"grad_norm": 5.374299437309603,
|
| 1322 |
+
"learning_rate": 1e-06,
|
| 1323 |
+
"logits/chosen": 25.917665481567383,
|
| 1324 |
+
"logits/rejected": 27.687381744384766,
|
| 1325 |
+
"logps/chosen": -414.66156005859375,
|
| 1326 |
+
"logps/rejected": -445.0428771972656,
|
| 1327 |
+
"loss": 0.3391,
|
| 1328 |
+
"rewards/accuracies": 0.25,
|
| 1329 |
+
"rewards/chosen": 0.18534794449806213,
|
| 1330 |
+
"rewards/margins": 0.06945788860321045,
|
| 1331 |
+
"rewards/rejected": 0.11589004844427109,
|
| 1332 |
+
"step": 63
|
| 1333 |
+
},
|
| 1334 |
+
{
|
| 1335 |
+
"debug/policy_chosen_logits": 23.454923629760742,
|
| 1336 |
+
"debug/policy_chosen_logps": -392.32305908203125,
|
| 1337 |
+
"debug/policy_rejected_logits": 28.59387969970703,
|
| 1338 |
+
"debug/policy_rejected_logps": -434.9425048828125,
|
| 1339 |
+
"debug/reference_chosen_logps": -421.68670654296875,
|
| 1340 |
+
"debug/reference_rejected_logps": -419.7337646484375,
|
| 1341 |
+
"epoch": 0.8,
|
| 1342 |
+
"grad_norm": 6.44340155168501,
|
| 1343 |
+
"learning_rate": 1e-06,
|
| 1344 |
+
"logits/chosen": 23.454923629760742,
|
| 1345 |
+
"logits/rejected": 28.59387969970703,
|
| 1346 |
+
"logps/chosen": -392.32305908203125,
|
| 1347 |
+
"logps/rejected": -434.9425048828125,
|
| 1348 |
+
"loss": 0.286,
|
| 1349 |
+
"rewards/accuracies": 0.875,
|
| 1350 |
+
"rewards/chosen": 0.29363659024238586,
|
| 1351 |
+
"rewards/margins": 0.44572439789772034,
|
| 1352 |
+
"rewards/rejected": -0.15208780765533447,
|
| 1353 |
+
"step": 64
|
| 1354 |
+
},
|
| 1355 |
+
{
|
| 1356 |
+
"debug/policy_chosen_logits": 24.071760177612305,
|
| 1357 |
+
"debug/policy_chosen_logps": -397.78515625,
|
| 1358 |
+
"debug/policy_rejected_logits": 24.690345764160156,
|
| 1359 |
+
"debug/policy_rejected_logps": -474.0267028808594,
|
| 1360 |
+
"debug/reference_chosen_logps": -413.3990783691406,
|
| 1361 |
+
"debug/reference_rejected_logps": -447.173095703125,
|
| 1362 |
+
"epoch": 0.8125,
|
| 1363 |
+
"grad_norm": 4.265641795749149,
|
| 1364 |
+
"learning_rate": 1e-06,
|
| 1365 |
+
"logits/chosen": 24.071760177612305,
|
| 1366 |
+
"logits/rejected": 24.690345764160156,
|
| 1367 |
+
"logps/chosen": -397.78515625,
|
| 1368 |
+
"logps/rejected": -474.0267028808594,
|
| 1369 |
+
"loss": 0.2996,
|
| 1370 |
+
"rewards/accuracies": 0.625,
|
| 1371 |
+
"rewards/chosen": 0.15613901615142822,
|
| 1372 |
+
"rewards/margins": 0.42467522621154785,
|
| 1373 |
+
"rewards/rejected": -0.26853621006011963,
|
| 1374 |
+
"step": 65
|
| 1375 |
+
},
|
| 1376 |
+
{
|
| 1377 |
+
"debug/policy_chosen_logits": 27.323055267333984,
|
| 1378 |
+
"debug/policy_chosen_logps": -423.77105712890625,
|
| 1379 |
+
"debug/policy_rejected_logits": 32.6217041015625,
|
| 1380 |
+
"debug/policy_rejected_logps": -454.37091064453125,
|
| 1381 |
+
"debug/reference_chosen_logps": -444.37615966796875,
|
| 1382 |
+
"debug/reference_rejected_logps": -440.6434326171875,
|
| 1383 |
+
"epoch": 0.825,
|
| 1384 |
+
"grad_norm": 4.106006750125632,
|
| 1385 |
+
"learning_rate": 1e-06,
|
| 1386 |
+
"logits/chosen": 27.323055267333984,
|
| 1387 |
+
"logits/rejected": 32.6217041015625,
|
| 1388 |
+
"logps/chosen": -423.77105712890625,
|
| 1389 |
+
"logps/rejected": -454.37091064453125,
|
| 1390 |
+
"loss": 0.3016,
|
| 1391 |
+
"rewards/accuracies": 1.0,
|
| 1392 |
+
"rewards/chosen": 0.2060512900352478,
|
| 1393 |
+
"rewards/margins": 0.3433259427547455,
|
| 1394 |
+
"rewards/rejected": -0.13727465271949768,
|
| 1395 |
+
"step": 66
|
| 1396 |
+
},
|
| 1397 |
+
{
|
| 1398 |
+
"debug/policy_chosen_logits": 20.215694427490234,
|
| 1399 |
+
"debug/policy_chosen_logps": -392.8240966796875,
|
| 1400 |
+
"debug/policy_rejected_logits": 27.513879776000977,
|
| 1401 |
+
"debug/policy_rejected_logps": -464.026611328125,
|
| 1402 |
+
"debug/reference_chosen_logps": -423.304931640625,
|
| 1403 |
+
"debug/reference_rejected_logps": -444.7491455078125,
|
| 1404 |
+
"epoch": 0.8375,
|
| 1405 |
+
"grad_norm": 3.4052279815687894,
|
| 1406 |
+
"learning_rate": 1e-06,
|
| 1407 |
+
"logits/chosen": 20.215694427490234,
|
| 1408 |
+
"logits/rejected": 27.513879776000977,
|
| 1409 |
+
"logps/chosen": -392.8240966796875,
|
| 1410 |
+
"logps/rejected": -464.026611328125,
|
| 1411 |
+
"loss": 0.2463,
|
| 1412 |
+
"rewards/accuracies": 1.0,
|
| 1413 |
+
"rewards/chosen": 0.3048083484172821,
|
| 1414 |
+
"rewards/margins": 0.49758297204971313,
|
| 1415 |
+
"rewards/rejected": -0.19277465343475342,
|
| 1416 |
+
"step": 67
|
| 1417 |
+
},
|
| 1418 |
+
{
|
| 1419 |
+
"debug/policy_chosen_logits": 20.277034759521484,
|
| 1420 |
+
"debug/policy_chosen_logps": -390.63189697265625,
|
| 1421 |
+
"debug/policy_rejected_logits": 26.955398559570312,
|
| 1422 |
+
"debug/policy_rejected_logps": -494.57330322265625,
|
| 1423 |
+
"debug/reference_chosen_logps": -411.6377868652344,
|
| 1424 |
+
"debug/reference_rejected_logps": -470.29150390625,
|
| 1425 |
+
"epoch": 0.85,
|
| 1426 |
+
"grad_norm": 4.314646861788122,
|
| 1427 |
+
"learning_rate": 1e-06,
|
| 1428 |
+
"logits/chosen": 20.277034759521484,
|
| 1429 |
+
"logits/rejected": 26.955398559570312,
|
| 1430 |
+
"logps/chosen": -390.63189697265625,
|
| 1431 |
+
"logps/rejected": -494.57330322265625,
|
| 1432 |
+
"loss": 0.2748,
|
| 1433 |
+
"rewards/accuracies": 0.875,
|
| 1434 |
+
"rewards/chosen": 0.21005921065807343,
|
| 1435 |
+
"rewards/margins": 0.4528772830963135,
|
| 1436 |
+
"rewards/rejected": -0.24281807243824005,
|
| 1437 |
+
"step": 68
|
| 1438 |
+
},
|
| 1439 |
+
{
|
| 1440 |
+
"debug/policy_chosen_logits": 17.920215606689453,
|
| 1441 |
+
"debug/policy_chosen_logps": -383.5913391113281,
|
| 1442 |
+
"debug/policy_rejected_logits": 22.582468032836914,
|
| 1443 |
+
"debug/policy_rejected_logps": -483.46142578125,
|
| 1444 |
+
"debug/reference_chosen_logps": -406.38525390625,
|
| 1445 |
+
"debug/reference_rejected_logps": -442.0745544433594,
|
| 1446 |
+
"epoch": 0.8625,
|
| 1447 |
+
"grad_norm": 7.009734793629414,
|
| 1448 |
+
"learning_rate": 1e-06,
|
| 1449 |
+
"logits/chosen": 17.920215606689453,
|
| 1450 |
+
"logits/rejected": 22.582468032836914,
|
| 1451 |
+
"logps/chosen": -383.5913391113281,
|
| 1452 |
+
"logps/rejected": -483.46142578125,
|
| 1453 |
+
"loss": 0.2491,
|
| 1454 |
+
"rewards/accuracies": 1.0,
|
| 1455 |
+
"rewards/chosen": 0.22793909907341003,
|
| 1456 |
+
"rewards/margins": 0.641808032989502,
|
| 1457 |
+
"rewards/rejected": -0.4138689637184143,
|
| 1458 |
+
"step": 69
|
| 1459 |
+
},
|
| 1460 |
+
{
|
| 1461 |
+
"debug/policy_chosen_logits": 21.91058349609375,
|
| 1462 |
+
"debug/policy_chosen_logps": -391.1068115234375,
|
| 1463 |
+
"debug/policy_rejected_logits": 28.602935791015625,
|
| 1464 |
+
"debug/policy_rejected_logps": -462.3726501464844,
|
| 1465 |
+
"debug/reference_chosen_logps": -408.93109130859375,
|
| 1466 |
+
"debug/reference_rejected_logps": -436.0927429199219,
|
| 1467 |
+
"epoch": 0.875,
|
| 1468 |
+
"grad_norm": 6.986157067049592,
|
| 1469 |
+
"learning_rate": 1e-06,
|
| 1470 |
+
"logits/chosen": 21.91058349609375,
|
| 1471 |
+
"logits/rejected": 28.602935791015625,
|
| 1472 |
+
"logps/chosen": -391.1068115234375,
|
| 1473 |
+
"logps/rejected": -462.3726501464844,
|
| 1474 |
+
"loss": 0.2593,
|
| 1475 |
+
"rewards/accuracies": 0.875,
|
| 1476 |
+
"rewards/chosen": 0.1782427579164505,
|
| 1477 |
+
"rewards/margins": 0.4410419762134552,
|
| 1478 |
+
"rewards/rejected": -0.2627992033958435,
|
| 1479 |
+
"step": 70
|
| 1480 |
+
},
|
| 1481 |
+
{
|
| 1482 |
+
"debug/policy_chosen_logits": 19.776662826538086,
|
| 1483 |
+
"debug/policy_chosen_logps": -398.9197082519531,
|
| 1484 |
+
"debug/policy_rejected_logits": 32.09967803955078,
|
| 1485 |
+
"debug/policy_rejected_logps": -487.4455871582031,
|
| 1486 |
+
"debug/reference_chosen_logps": -420.3005676269531,
|
| 1487 |
+
"debug/reference_rejected_logps": -452.2755126953125,
|
| 1488 |
+
"epoch": 0.8875,
|
| 1489 |
+
"grad_norm": 5.014513725137227,
|
| 1490 |
+
"learning_rate": 1e-06,
|
| 1491 |
+
"logits/chosen": 19.776662826538086,
|
| 1492 |
+
"logits/rejected": 32.09967803955078,
|
| 1493 |
+
"logps/chosen": -398.9197082519531,
|
| 1494 |
+
"logps/rejected": -487.4455871582031,
|
| 1495 |
+
"loss": 0.2544,
|
| 1496 |
+
"rewards/accuracies": 1.0,
|
| 1497 |
+
"rewards/chosen": 0.2138085961341858,
|
| 1498 |
+
"rewards/margins": 0.5655093193054199,
|
| 1499 |
+
"rewards/rejected": -0.3517007827758789,
|
| 1500 |
+
"step": 71
|
| 1501 |
+
},
|
| 1502 |
+
{
|
| 1503 |
+
"debug/policy_chosen_logits": 21.157052993774414,
|
| 1504 |
+
"debug/policy_chosen_logps": -391.25897216796875,
|
| 1505 |
+
"debug/policy_rejected_logits": 26.412094116210938,
|
| 1506 |
+
"debug/policy_rejected_logps": -490.5589294433594,
|
| 1507 |
+
"debug/reference_chosen_logps": -413.3541259765625,
|
| 1508 |
+
"debug/reference_rejected_logps": -471.00921630859375,
|
| 1509 |
+
"epoch": 0.9,
|
| 1510 |
+
"grad_norm": 5.328562975855576,
|
| 1511 |
+
"learning_rate": 1e-06,
|
| 1512 |
+
"logits/chosen": 21.157052993774414,
|
| 1513 |
+
"logits/rejected": 26.412094116210938,
|
| 1514 |
+
"logps/chosen": -391.25897216796875,
|
| 1515 |
+
"logps/rejected": -490.5589294433594,
|
| 1516 |
+
"loss": 0.2412,
|
| 1517 |
+
"rewards/accuracies": 0.875,
|
| 1518 |
+
"rewards/chosen": 0.2209514081478119,
|
| 1519 |
+
"rewards/margins": 0.4164482653141022,
|
| 1520 |
+
"rewards/rejected": -0.19549685716629028,
|
| 1521 |
+
"step": 72
|
| 1522 |
+
},
|
| 1523 |
+
{
|
| 1524 |
+
"debug/policy_chosen_logits": 24.695026397705078,
|
| 1525 |
+
"debug/policy_chosen_logps": -397.16497802734375,
|
| 1526 |
+
"debug/policy_rejected_logits": 26.946775436401367,
|
| 1527 |
+
"debug/policy_rejected_logps": -438.91900634765625,
|
| 1528 |
+
"debug/reference_chosen_logps": -418.1241455078125,
|
| 1529 |
+
"debug/reference_rejected_logps": -434.0196228027344,
|
| 1530 |
+
"epoch": 0.9125,
|
| 1531 |
+
"grad_norm": 5.591526461654656,
|
| 1532 |
+
"learning_rate": 1e-06,
|
| 1533 |
+
"logits/chosen": 24.695026397705078,
|
| 1534 |
+
"logits/rejected": 26.946775436401367,
|
| 1535 |
+
"logps/chosen": -397.16497802734375,
|
| 1536 |
+
"logps/rejected": -438.91900634765625,
|
| 1537 |
+
"loss": 0.3228,
|
| 1538 |
+
"rewards/accuracies": 0.625,
|
| 1539 |
+
"rewards/chosen": 0.20959174633026123,
|
| 1540 |
+
"rewards/margins": 0.25858569145202637,
|
| 1541 |
+
"rewards/rejected": -0.04899394512176514,
|
| 1542 |
+
"step": 73
|
| 1543 |
+
},
|
| 1544 |
+
{
|
| 1545 |
+
"debug/policy_chosen_logits": 22.788793563842773,
|
| 1546 |
+
"debug/policy_chosen_logps": -402.5521240234375,
|
| 1547 |
+
"debug/policy_rejected_logits": 27.343595504760742,
|
| 1548 |
+
"debug/policy_rejected_logps": -457.029052734375,
|
| 1549 |
+
"debug/reference_chosen_logps": -427.92510986328125,
|
| 1550 |
+
"debug/reference_rejected_logps": -443.33807373046875,
|
| 1551 |
+
"epoch": 0.925,
|
| 1552 |
+
"grad_norm": 4.13982614264715,
|
| 1553 |
+
"learning_rate": 1e-06,
|
| 1554 |
+
"logits/chosen": 22.788793563842773,
|
| 1555 |
+
"logits/rejected": 27.343595504760742,
|
| 1556 |
+
"logps/chosen": -402.5521240234375,
|
| 1557 |
+
"logps/rejected": -457.029052734375,
|
| 1558 |
+
"loss": 0.2539,
|
| 1559 |
+
"rewards/accuracies": 0.75,
|
| 1560 |
+
"rewards/chosen": 0.2537294626235962,
|
| 1561 |
+
"rewards/margins": 0.39063963294029236,
|
| 1562 |
+
"rewards/rejected": -0.13691017031669617,
|
| 1563 |
+
"step": 74
|
| 1564 |
+
},
|
| 1565 |
+
{
|
| 1566 |
+
"debug/policy_chosen_logits": 17.950855255126953,
|
| 1567 |
+
"debug/policy_chosen_logps": -386.80126953125,
|
| 1568 |
+
"debug/policy_rejected_logits": 28.871490478515625,
|
| 1569 |
+
"debug/policy_rejected_logps": -489.54559326171875,
|
| 1570 |
+
"debug/reference_chosen_logps": -418.87652587890625,
|
| 1571 |
+
"debug/reference_rejected_logps": -437.8173828125,
|
| 1572 |
+
"epoch": 0.9375,
|
| 1573 |
+
"grad_norm": 3.573253258125765,
|
| 1574 |
+
"learning_rate": 1e-06,
|
| 1575 |
+
"logits/chosen": 17.950855255126953,
|
| 1576 |
+
"logits/rejected": 28.871490478515625,
|
| 1577 |
+
"logps/chosen": -386.80126953125,
|
| 1578 |
+
"logps/rejected": -489.54559326171875,
|
| 1579 |
+
"loss": 0.235,
|
| 1580 |
+
"rewards/accuracies": 1.0,
|
| 1581 |
+
"rewards/chosen": 0.3207521438598633,
|
| 1582 |
+
"rewards/margins": 0.8380340933799744,
|
| 1583 |
+
"rewards/rejected": -0.5172819495201111,
|
| 1584 |
+
"step": 75
|
| 1585 |
+
},
|
| 1586 |
+
{
|
| 1587 |
+
"debug/policy_chosen_logits": 20.622098922729492,
|
| 1588 |
+
"debug/policy_chosen_logps": -384.69195556640625,
|
| 1589 |
+
"debug/policy_rejected_logits": 31.19890594482422,
|
| 1590 |
+
"debug/policy_rejected_logps": -452.1475524902344,
|
| 1591 |
+
"debug/reference_chosen_logps": -407.402099609375,
|
| 1592 |
+
"debug/reference_rejected_logps": -432.25592041015625,
|
| 1593 |
+
"epoch": 0.95,
|
| 1594 |
+
"grad_norm": 6.4154538717209695,
|
| 1595 |
+
"learning_rate": 1e-06,
|
| 1596 |
+
"logits/chosen": 20.622098922729492,
|
| 1597 |
+
"logits/rejected": 31.19890594482422,
|
| 1598 |
+
"logps/chosen": -384.69195556640625,
|
| 1599 |
+
"logps/rejected": -452.1475524902344,
|
| 1600 |
+
"loss": 0.3166,
|
| 1601 |
+
"rewards/accuracies": 0.875,
|
| 1602 |
+
"rewards/chosen": 0.22710174322128296,
|
| 1603 |
+
"rewards/margins": 0.42601796984672546,
|
| 1604 |
+
"rewards/rejected": -0.1989162415266037,
|
| 1605 |
+
"step": 76
|
| 1606 |
+
},
|
| 1607 |
+
{
|
| 1608 |
+
"debug/policy_chosen_logits": 21.178997039794922,
|
| 1609 |
+
"debug/policy_chosen_logps": -384.0088806152344,
|
| 1610 |
+
"debug/policy_rejected_logits": 32.716346740722656,
|
| 1611 |
+
"debug/policy_rejected_logps": -477.344482421875,
|
| 1612 |
+
"debug/reference_chosen_logps": -413.40850830078125,
|
| 1613 |
+
"debug/reference_rejected_logps": -443.79180908203125,
|
| 1614 |
+
"epoch": 0.9625,
|
| 1615 |
+
"grad_norm": 4.479488616499859,
|
| 1616 |
+
"learning_rate": 1e-06,
|
| 1617 |
+
"logits/chosen": 21.178997039794922,
|
| 1618 |
+
"logits/rejected": 32.716346740722656,
|
| 1619 |
+
"logps/chosen": -384.0088806152344,
|
| 1620 |
+
"logps/rejected": -477.344482421875,
|
| 1621 |
+
"loss": 0.2545,
|
| 1622 |
+
"rewards/accuracies": 0.875,
|
| 1623 |
+
"rewards/chosen": 0.2939964532852173,
|
| 1624 |
+
"rewards/margins": 0.6295228600502014,
|
| 1625 |
+
"rewards/rejected": -0.33552637696266174,
|
| 1626 |
+
"step": 77
|
| 1627 |
+
},
|
| 1628 |
+
{
|
| 1629 |
+
"debug/policy_chosen_logits": 25.3104190826416,
|
| 1630 |
+
"debug/policy_chosen_logps": -409.6578063964844,
|
| 1631 |
+
"debug/policy_rejected_logits": 31.45233917236328,
|
| 1632 |
+
"debug/policy_rejected_logps": -462.6581115722656,
|
| 1633 |
+
"debug/reference_chosen_logps": -426.9599609375,
|
| 1634 |
+
"debug/reference_rejected_logps": -430.9308166503906,
|
| 1635 |
+
"epoch": 0.975,
|
| 1636 |
+
"grad_norm": 3.3409543416332235,
|
| 1637 |
+
"learning_rate": 1e-06,
|
| 1638 |
+
"logits/chosen": 25.3104190826416,
|
| 1639 |
+
"logits/rejected": 31.45233917236328,
|
| 1640 |
+
"logps/chosen": -409.6578063964844,
|
| 1641 |
+
"logps/rejected": -462.6581115722656,
|
| 1642 |
+
"loss": 0.2304,
|
| 1643 |
+
"rewards/accuracies": 0.875,
|
| 1644 |
+
"rewards/chosen": 0.17302149534225464,
|
| 1645 |
+
"rewards/margins": 0.49029436707496643,
|
| 1646 |
+
"rewards/rejected": -0.3172728717327118,
|
| 1647 |
+
"step": 78
|
| 1648 |
+
},
|
| 1649 |
+
{
|
| 1650 |
+
"debug/policy_chosen_logits": 20.39267349243164,
|
| 1651 |
+
"debug/policy_chosen_logps": -391.23089599609375,
|
| 1652 |
+
"debug/policy_rejected_logits": 29.421716690063477,
|
| 1653 |
+
"debug/policy_rejected_logps": -489.40728759765625,
|
| 1654 |
+
"debug/reference_chosen_logps": -423.12115478515625,
|
| 1655 |
+
"debug/reference_rejected_logps": -447.7901611328125,
|
| 1656 |
+
"epoch": 0.9875,
|
| 1657 |
+
"grad_norm": 5.428664997337287,
|
| 1658 |
+
"learning_rate": 1e-06,
|
| 1659 |
+
"logits/chosen": 20.39267349243164,
|
| 1660 |
+
"logits/rejected": 29.421716690063477,
|
| 1661 |
+
"logps/chosen": -391.23089599609375,
|
| 1662 |
+
"logps/rejected": -489.40728759765625,
|
| 1663 |
+
"loss": 0.2057,
|
| 1664 |
+
"rewards/accuracies": 1.0,
|
| 1665 |
+
"rewards/chosen": 0.3189027011394501,
|
| 1666 |
+
"rewards/margins": 0.7350739240646362,
|
| 1667 |
+
"rewards/rejected": -0.41617119312286377,
|
| 1668 |
+
"step": 79
|
| 1669 |
+
},
|
| 1670 |
+
{
|
| 1671 |
+
"debug/policy_chosen_logits": 18.865188598632812,
|
| 1672 |
+
"debug/policy_chosen_logps": -384.6566162109375,
|
| 1673 |
+
"debug/policy_rejected_logits": 29.98931884765625,
|
| 1674 |
+
"debug/policy_rejected_logps": -469.95892333984375,
|
| 1675 |
+
"debug/reference_chosen_logps": -407.10784912109375,
|
| 1676 |
+
"debug/reference_rejected_logps": -433.1329345703125,
|
| 1677 |
+
"epoch": 1.0,
|
| 1678 |
+
"grad_norm": 4.751087690542253,
|
| 1679 |
+
"learning_rate": 1e-06,
|
| 1680 |
+
"logits/chosen": 18.865188598632812,
|
| 1681 |
+
"logits/rejected": 29.98931884765625,
|
| 1682 |
+
"logps/chosen": -384.6566162109375,
|
| 1683 |
+
"logps/rejected": -469.95892333984375,
|
| 1684 |
+
"loss": 0.1778,
|
| 1685 |
+
"rewards/accuracies": 0.875,
|
| 1686 |
+
"rewards/chosen": 0.22451260685920715,
|
| 1687 |
+
"rewards/margins": 0.5927724838256836,
|
| 1688 |
+
"rewards/rejected": -0.36825984716415405,
|
| 1689 |
+
"step": 80
|
| 1690 |
+
},
|
| 1691 |
+
{
|
| 1692 |
+
"epoch": 1.0,
|
| 1693 |
+
"step": 80,
|
| 1694 |
+
"total_flos": 0.0,
|
| 1695 |
+
"train_loss": 0.3231356654316187,
|
| 1696 |
+
"train_runtime": 754.2708,
|
| 1697 |
+
"train_samples_per_second": 6.788,
|
| 1698 |
+
"train_steps_per_second": 0.106
|
| 1699 |
+
}
|
| 1700 |
+
],
|
| 1701 |
+
"logging_steps": 1,
|
| 1702 |
+
"max_steps": 80,
|
| 1703 |
+
"num_input_tokens_seen": 0,
|
| 1704 |
+
"num_train_epochs": 1,
|
| 1705 |
+
"save_steps": 500,
|
| 1706 |
+
"stateful_callbacks": {
|
| 1707 |
+
"TrainerControl": {
|
| 1708 |
+
"args": {
|
| 1709 |
+
"should_epoch_stop": false,
|
| 1710 |
+
"should_evaluate": false,
|
| 1711 |
+
"should_log": false,
|
| 1712 |
+
"should_save": true,
|
| 1713 |
+
"should_training_stop": true
|
| 1714 |
+
},
|
| 1715 |
+
"attributes": {}
|
| 1716 |
+
}
|
| 1717 |
+
},
|
| 1718 |
+
"total_flos": 0.0,
|
| 1719 |
+
"train_batch_size": 8,
|
| 1720 |
+
"trial_name": null,
|
| 1721 |
+
"trial_params": null
|
| 1722 |
+
}
|