Model save
Browse files- README.md +75 -0
- all_results.json +9 -0
- generation_config.json +12 -0
- train_results.json +9 -0
- trainer_state.json +1515 -0
README.md
ADDED
|
@@ -0,0 +1,75 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: transformers
|
| 3 |
+
license: llama3
|
| 4 |
+
base_model: meta-llama/Meta-Llama-3-8B-Instruct
|
| 5 |
+
tags:
|
| 6 |
+
- trl
|
| 7 |
+
- simpo
|
| 8 |
+
- generated_from_trainer
|
| 9 |
+
model-index:
|
| 10 |
+
- name: llama3_orpo_best_entropy
|
| 11 |
+
results: []
|
| 12 |
+
---
|
| 13 |
+
|
| 14 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 15 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 16 |
+
|
| 17 |
+
# llama3_orpo_best_entropy
|
| 18 |
+
|
| 19 |
+
This model is a fine-tuned version of [meta-llama/Meta-Llama-3-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct) on an unknown dataset.
|
| 20 |
+
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 2.5608
|
| 22 |
+
- Rewards/chosen: -12.8797
|
| 23 |
+
- Rewards/rejected: -17.3972
|
| 24 |
+
- Rewards/accuracies: 0.8072
|
| 25 |
+
- Rewards/margins: 4.5175
|
| 26 |
+
- Logps/rejected: -1.7397
|
| 27 |
+
- Logps/chosen: -1.2880
|
| 28 |
+
- Logits/rejected: -1.3473
|
| 29 |
+
- Logits/chosen: -1.3813
|
| 30 |
+
- Semantic Entropy: 0.7719
|
| 31 |
+
|
| 32 |
+
## Model description
|
| 33 |
+
|
| 34 |
+
More information needed
|
| 35 |
+
|
| 36 |
+
## Intended uses & limitations
|
| 37 |
+
|
| 38 |
+
More information needed
|
| 39 |
+
|
| 40 |
+
## Training and evaluation data
|
| 41 |
+
|
| 42 |
+
More information needed
|
| 43 |
+
|
| 44 |
+
## Training procedure
|
| 45 |
+
|
| 46 |
+
### Training hyperparameters
|
| 47 |
+
|
| 48 |
+
The following hyperparameters were used during training:
|
| 49 |
+
- learning_rate: 1e-06
|
| 50 |
+
- train_batch_size: 2
|
| 51 |
+
- eval_batch_size: 4
|
| 52 |
+
- seed: 42
|
| 53 |
+
- distributed_type: multi-GPU
|
| 54 |
+
- num_devices: 4
|
| 55 |
+
- gradient_accumulation_steps: 16
|
| 56 |
+
- total_train_batch_size: 128
|
| 57 |
+
- total_eval_batch_size: 16
|
| 58 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 59 |
+
- lr_scheduler_type: cosine
|
| 60 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 61 |
+
- num_epochs: 1.0
|
| 62 |
+
|
| 63 |
+
### Training results
|
| 64 |
+
|
| 65 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen | Semantic Entropy |
|
| 66 |
+
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|:----------------:|
|
| 67 |
+
| 2.3262 | 0.8743 | 400 | 2.5608 | -12.8797 | -17.3972 | 0.8072 | 4.5175 | -1.7397 | -1.2880 | -1.3473 | -1.3813 | 0.7719 |
|
| 68 |
+
|
| 69 |
+
|
| 70 |
+
### Framework versions
|
| 71 |
+
|
| 72 |
+
- Transformers 4.44.2
|
| 73 |
+
- Pytorch 2.2.2+cu121
|
| 74 |
+
- Datasets 2.18.0
|
| 75 |
+
- Tokenizers 0.19.1
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9989071038251366,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 3.4314852449513107,
|
| 5 |
+
"train_runtime": 5934.8281,
|
| 6 |
+
"train_samples": 58558,
|
| 7 |
+
"train_samples_per_second": 9.867,
|
| 8 |
+
"train_steps_per_second": 0.077
|
| 9 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token_id": 128000,
|
| 3 |
+
"do_sample": true,
|
| 4 |
+
"eos_token_id": [
|
| 5 |
+
128001,
|
| 6 |
+
128009
|
| 7 |
+
],
|
| 8 |
+
"max_length": 4096,
|
| 9 |
+
"temperature": 0.6,
|
| 10 |
+
"top_p": 0.9,
|
| 11 |
+
"transformers_version": "4.44.2"
|
| 12 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9989071038251366,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 3.4314852449513107,
|
| 5 |
+
"train_runtime": 5934.8281,
|
| 6 |
+
"train_samples": 58558,
|
| 7 |
+
"train_samples_per_second": 9.867,
|
| 8 |
+
"train_steps_per_second": 0.077
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,1515 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.9989071038251366,
|
| 5 |
+
"eval_steps": 400,
|
| 6 |
+
"global_step": 457,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.01092896174863388,
|
| 13 |
+
"grad_norm": 358.5620562461214,
|
| 14 |
+
"learning_rate": 1.0869565217391303e-07,
|
| 15 |
+
"logits/chosen": -1.0113575458526611,
|
| 16 |
+
"logits/rejected": -1.0064939260482788,
|
| 17 |
+
"logps/chosen": -0.2803983986377716,
|
| 18 |
+
"logps/rejected": -0.2860395908355713,
|
| 19 |
+
"loss": 5.2315,
|
| 20 |
+
"rewards/accuracies": 0.53125,
|
| 21 |
+
"rewards/chosen": -2.8039839267730713,
|
| 22 |
+
"rewards/margins": 0.056411754339933395,
|
| 23 |
+
"rewards/rejected": -2.860395669937134,
|
| 24 |
+
"semantic_entropy": 0.7518940567970276,
|
| 25 |
+
"step": 5
|
| 26 |
+
},
|
| 27 |
+
{
|
| 28 |
+
"epoch": 0.02185792349726776,
|
| 29 |
+
"grad_norm": 233.11198507849488,
|
| 30 |
+
"learning_rate": 2.1739130434782607e-07,
|
| 31 |
+
"logits/chosen": -1.056563138961792,
|
| 32 |
+
"logits/rejected": -1.0053507089614868,
|
| 33 |
+
"logps/chosen": -0.2568749487400055,
|
| 34 |
+
"logps/rejected": -0.27021342515945435,
|
| 35 |
+
"loss": 5.206,
|
| 36 |
+
"rewards/accuracies": 0.5375000238418579,
|
| 37 |
+
"rewards/chosen": -2.568748950958252,
|
| 38 |
+
"rewards/margins": 0.13338527083396912,
|
| 39 |
+
"rewards/rejected": -2.702134370803833,
|
| 40 |
+
"semantic_entropy": 0.7094504237174988,
|
| 41 |
+
"step": 10
|
| 42 |
+
},
|
| 43 |
+
{
|
| 44 |
+
"epoch": 0.03278688524590164,
|
| 45 |
+
"grad_norm": 210.79380420075688,
|
| 46 |
+
"learning_rate": 3.260869565217391e-07,
|
| 47 |
+
"logits/chosen": -1.0091139078140259,
|
| 48 |
+
"logits/rejected": -0.9631060361862183,
|
| 49 |
+
"logps/chosen": -0.2674282491207123,
|
| 50 |
+
"logps/rejected": -0.27336788177490234,
|
| 51 |
+
"loss": 5.1278,
|
| 52 |
+
"rewards/accuracies": 0.512499988079071,
|
| 53 |
+
"rewards/chosen": -2.6742825508117676,
|
| 54 |
+
"rewards/margins": 0.05939612537622452,
|
| 55 |
+
"rewards/rejected": -2.7336785793304443,
|
| 56 |
+
"semantic_entropy": 0.7274739742279053,
|
| 57 |
+
"step": 15
|
| 58 |
+
},
|
| 59 |
+
{
|
| 60 |
+
"epoch": 0.04371584699453552,
|
| 61 |
+
"grad_norm": 375.0299222500816,
|
| 62 |
+
"learning_rate": 4.3478260869565214e-07,
|
| 63 |
+
"logits/chosen": -0.9487798810005188,
|
| 64 |
+
"logits/rejected": -0.8998070955276489,
|
| 65 |
+
"logps/chosen": -0.2723819613456726,
|
| 66 |
+
"logps/rejected": -0.28497135639190674,
|
| 67 |
+
"loss": 5.222,
|
| 68 |
+
"rewards/accuracies": 0.5687500238418579,
|
| 69 |
+
"rewards/chosen": -2.7238194942474365,
|
| 70 |
+
"rewards/margins": 0.1258939504623413,
|
| 71 |
+
"rewards/rejected": -2.8497138023376465,
|
| 72 |
+
"semantic_entropy": 0.7452942132949829,
|
| 73 |
+
"step": 20
|
| 74 |
+
},
|
| 75 |
+
{
|
| 76 |
+
"epoch": 0.0546448087431694,
|
| 77 |
+
"grad_norm": 299.8105605992341,
|
| 78 |
+
"learning_rate": 5.434782608695652e-07,
|
| 79 |
+
"logits/chosen": -0.9505411386489868,
|
| 80 |
+
"logits/rejected": -0.8759678602218628,
|
| 81 |
+
"logps/chosen": -0.27557066082954407,
|
| 82 |
+
"logps/rejected": -0.29382389783859253,
|
| 83 |
+
"loss": 5.1223,
|
| 84 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 85 |
+
"rewards/chosen": -2.755706548690796,
|
| 86 |
+
"rewards/margins": 0.1825324147939682,
|
| 87 |
+
"rewards/rejected": -2.9382388591766357,
|
| 88 |
+
"semantic_entropy": 0.7546485662460327,
|
| 89 |
+
"step": 25
|
| 90 |
+
},
|
| 91 |
+
{
|
| 92 |
+
"epoch": 0.06557377049180328,
|
| 93 |
+
"grad_norm": 300.511189676041,
|
| 94 |
+
"learning_rate": 6.521739130434782e-07,
|
| 95 |
+
"logits/chosen": -1.0557540655136108,
|
| 96 |
+
"logits/rejected": -0.990186333656311,
|
| 97 |
+
"logps/chosen": -0.267598032951355,
|
| 98 |
+
"logps/rejected": -0.28400346636772156,
|
| 99 |
+
"loss": 5.1993,
|
| 100 |
+
"rewards/accuracies": 0.518750011920929,
|
| 101 |
+
"rewards/chosen": -2.6759800910949707,
|
| 102 |
+
"rewards/margins": 0.1640542596578598,
|
| 103 |
+
"rewards/rejected": -2.8400347232818604,
|
| 104 |
+
"semantic_entropy": 0.7248786091804504,
|
| 105 |
+
"step": 30
|
| 106 |
+
},
|
| 107 |
+
{
|
| 108 |
+
"epoch": 0.07650273224043716,
|
| 109 |
+
"grad_norm": 186.14457687250982,
|
| 110 |
+
"learning_rate": 7.608695652173913e-07,
|
| 111 |
+
"logits/chosen": -1.0076847076416016,
|
| 112 |
+
"logits/rejected": -0.9405019879341125,
|
| 113 |
+
"logps/chosen": -0.2580474615097046,
|
| 114 |
+
"logps/rejected": -0.279694139957428,
|
| 115 |
+
"loss": 5.0812,
|
| 116 |
+
"rewards/accuracies": 0.4937500059604645,
|
| 117 |
+
"rewards/chosen": -2.580474615097046,
|
| 118 |
+
"rewards/margins": 0.2164669930934906,
|
| 119 |
+
"rewards/rejected": -2.7969415187835693,
|
| 120 |
+
"semantic_entropy": 0.7201561331748962,
|
| 121 |
+
"step": 35
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 0.08743169398907104,
|
| 125 |
+
"grad_norm": 223.67562662995005,
|
| 126 |
+
"learning_rate": 8.695652173913043e-07,
|
| 127 |
+
"logits/chosen": -0.9655061960220337,
|
| 128 |
+
"logits/rejected": -0.9045132398605347,
|
| 129 |
+
"logps/chosen": -0.2820321023464203,
|
| 130 |
+
"logps/rejected": -0.2990434765815735,
|
| 131 |
+
"loss": 5.3116,
|
| 132 |
+
"rewards/accuracies": 0.518750011920929,
|
| 133 |
+
"rewards/chosen": -2.820321559906006,
|
| 134 |
+
"rewards/margins": 0.17011362314224243,
|
| 135 |
+
"rewards/rejected": -2.9904348850250244,
|
| 136 |
+
"semantic_entropy": 0.7597802877426147,
|
| 137 |
+
"step": 40
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 0.09836065573770492,
|
| 141 |
+
"grad_norm": 131.00541589477967,
|
| 142 |
+
"learning_rate": 9.782608695652173e-07,
|
| 143 |
+
"logits/chosen": -1.0179851055145264,
|
| 144 |
+
"logits/rejected": -0.9359539151191711,
|
| 145 |
+
"logps/chosen": -0.2855134606361389,
|
| 146 |
+
"logps/rejected": -0.3078162968158722,
|
| 147 |
+
"loss": 4.919,
|
| 148 |
+
"rewards/accuracies": 0.4937500059604645,
|
| 149 |
+
"rewards/chosen": -2.8551342487335205,
|
| 150 |
+
"rewards/margins": 0.22302868962287903,
|
| 151 |
+
"rewards/rejected": -3.078162908554077,
|
| 152 |
+
"semantic_entropy": 0.7599790096282959,
|
| 153 |
+
"step": 45
|
| 154 |
+
},
|
| 155 |
+
{
|
| 156 |
+
"epoch": 0.1092896174863388,
|
| 157 |
+
"grad_norm": 360.5211336376954,
|
| 158 |
+
"learning_rate": 9.997663088532014e-07,
|
| 159 |
+
"logits/chosen": -0.9701619148254395,
|
| 160 |
+
"logits/rejected": -0.887865424156189,
|
| 161 |
+
"logps/chosen": -0.28138467669487,
|
| 162 |
+
"logps/rejected": -0.28913217782974243,
|
| 163 |
+
"loss": 5.1014,
|
| 164 |
+
"rewards/accuracies": 0.48750001192092896,
|
| 165 |
+
"rewards/chosen": -2.813847064971924,
|
| 166 |
+
"rewards/margins": 0.07747481018304825,
|
| 167 |
+
"rewards/rejected": -2.891321897506714,
|
| 168 |
+
"semantic_entropy": 0.7510823607444763,
|
| 169 |
+
"step": 50
|
| 170 |
+
},
|
| 171 |
+
{
|
| 172 |
+
"epoch": 0.12021857923497267,
|
| 173 |
+
"grad_norm": 162.42445549849919,
|
| 174 |
+
"learning_rate": 9.98817312944725e-07,
|
| 175 |
+
"logits/chosen": -1.0028311014175415,
|
| 176 |
+
"logits/rejected": -0.8850187063217163,
|
| 177 |
+
"logps/chosen": -0.27895885705947876,
|
| 178 |
+
"logps/rejected": -0.3143305778503418,
|
| 179 |
+
"loss": 4.8519,
|
| 180 |
+
"rewards/accuracies": 0.5562499761581421,
|
| 181 |
+
"rewards/chosen": -2.7895889282226562,
|
| 182 |
+
"rewards/margins": 0.3537173271179199,
|
| 183 |
+
"rewards/rejected": -3.143306255340576,
|
| 184 |
+
"semantic_entropy": 0.7608937621116638,
|
| 185 |
+
"step": 55
|
| 186 |
+
},
|
| 187 |
+
{
|
| 188 |
+
"epoch": 0.13114754098360656,
|
| 189 |
+
"grad_norm": 120.127890644715,
|
| 190 |
+
"learning_rate": 9.971397915250336e-07,
|
| 191 |
+
"logits/chosen": -1.0246554613113403,
|
| 192 |
+
"logits/rejected": -0.9786936640739441,
|
| 193 |
+
"logps/chosen": -0.26926669478416443,
|
| 194 |
+
"logps/rejected": -0.3113505244255066,
|
| 195 |
+
"loss": 4.6493,
|
| 196 |
+
"rewards/accuracies": 0.612500011920929,
|
| 197 |
+
"rewards/chosen": -2.692667245864868,
|
| 198 |
+
"rewards/margins": 0.42083826661109924,
|
| 199 |
+
"rewards/rejected": -3.1135053634643555,
|
| 200 |
+
"semantic_entropy": 0.7576217651367188,
|
| 201 |
+
"step": 60
|
| 202 |
+
},
|
| 203 |
+
{
|
| 204 |
+
"epoch": 0.14207650273224043,
|
| 205 |
+
"grad_norm": 214.34275301709462,
|
| 206 |
+
"learning_rate": 9.94736194623663e-07,
|
| 207 |
+
"logits/chosen": -0.9859918355941772,
|
| 208 |
+
"logits/rejected": -0.9181090593338013,
|
| 209 |
+
"logps/chosen": -0.3084973096847534,
|
| 210 |
+
"logps/rejected": -0.3399081826210022,
|
| 211 |
+
"loss": 4.9942,
|
| 212 |
+
"rewards/accuracies": 0.550000011920929,
|
| 213 |
+
"rewards/chosen": -3.084973096847534,
|
| 214 |
+
"rewards/margins": 0.31410855054855347,
|
| 215 |
+
"rewards/rejected": -3.3990814685821533,
|
| 216 |
+
"semantic_entropy": 0.8060176968574524,
|
| 217 |
+
"step": 65
|
| 218 |
+
},
|
| 219 |
+
{
|
| 220 |
+
"epoch": 0.15300546448087432,
|
| 221 |
+
"grad_norm": 544.9393207205336,
|
| 222 |
+
"learning_rate": 9.916100327075037e-07,
|
| 223 |
+
"logits/chosen": -0.9521551132202148,
|
| 224 |
+
"logits/rejected": -0.9334642291069031,
|
| 225 |
+
"logps/chosen": -0.29637694358825684,
|
| 226 |
+
"logps/rejected": -0.3240143656730652,
|
| 227 |
+
"loss": 4.7857,
|
| 228 |
+
"rewards/accuracies": 0.5687500238418579,
|
| 229 |
+
"rewards/chosen": -2.9637694358825684,
|
| 230 |
+
"rewards/margins": 0.27637460827827454,
|
| 231 |
+
"rewards/rejected": -3.2401435375213623,
|
| 232 |
+
"semantic_entropy": 0.7712014317512512,
|
| 233 |
+
"step": 70
|
| 234 |
+
},
|
| 235 |
+
{
|
| 236 |
+
"epoch": 0.16393442622950818,
|
| 237 |
+
"grad_norm": 123.2584567983735,
|
| 238 |
+
"learning_rate": 9.877658715537428e-07,
|
| 239 |
+
"logits/chosen": -0.9457874298095703,
|
| 240 |
+
"logits/rejected": -0.9289388656616211,
|
| 241 |
+
"logps/chosen": -0.31344500184059143,
|
| 242 |
+
"logps/rejected": -0.3473803997039795,
|
| 243 |
+
"loss": 4.8544,
|
| 244 |
+
"rewards/accuracies": 0.625,
|
| 245 |
+
"rewards/chosen": -3.1344501972198486,
|
| 246 |
+
"rewards/margins": 0.33935409784317017,
|
| 247 |
+
"rewards/rejected": -3.473804473876953,
|
| 248 |
+
"semantic_entropy": 0.8020299077033997,
|
| 249 |
+
"step": 75
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 0.17486338797814208,
|
| 253 |
+
"grad_norm": 136.55956527115663,
|
| 254 |
+
"learning_rate": 9.832093255815216e-07,
|
| 255 |
+
"logits/chosen": -0.9409273266792297,
|
| 256 |
+
"logits/rejected": -0.8797443509101868,
|
| 257 |
+
"logps/chosen": -0.3109641373157501,
|
| 258 |
+
"logps/rejected": -0.3346250057220459,
|
| 259 |
+
"loss": 4.7853,
|
| 260 |
+
"rewards/accuracies": 0.606249988079071,
|
| 261 |
+
"rewards/chosen": -3.1096413135528564,
|
| 262 |
+
"rewards/margins": 0.23660895228385925,
|
| 263 |
+
"rewards/rejected": -3.346250534057617,
|
| 264 |
+
"semantic_entropy": 0.7786868810653687,
|
| 265 |
+
"step": 80
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.18579234972677597,
|
| 269 |
+
"grad_norm": 202.99615488786125,
|
| 270 |
+
"learning_rate": 9.779470496520441e-07,
|
| 271 |
+
"logits/chosen": -0.9390329122543335,
|
| 272 |
+
"logits/rejected": -0.89063560962677,
|
| 273 |
+
"logps/chosen": -0.31133827567100525,
|
| 274 |
+
"logps/rejected": -0.3710102438926697,
|
| 275 |
+
"loss": 4.6394,
|
| 276 |
+
"rewards/accuracies": 0.643750011920929,
|
| 277 |
+
"rewards/chosen": -3.1133828163146973,
|
| 278 |
+
"rewards/margins": 0.5967196226119995,
|
| 279 |
+
"rewards/rejected": -3.7101027965545654,
|
| 280 |
+
"semantic_entropy": 0.8020746111869812,
|
| 281 |
+
"step": 85
|
| 282 |
+
},
|
| 283 |
+
{
|
| 284 |
+
"epoch": 0.19672131147540983,
|
| 285 |
+
"grad_norm": 147.8274369377715,
|
| 286 |
+
"learning_rate": 9.719867293491144e-07,
|
| 287 |
+
"logits/chosen": -0.9997242093086243,
|
| 288 |
+
"logits/rejected": -0.9218411445617676,
|
| 289 |
+
"logps/chosen": -0.33917468786239624,
|
| 290 |
+
"logps/rejected": -0.3771332800388336,
|
| 291 |
+
"loss": 4.7466,
|
| 292 |
+
"rewards/accuracies": 0.581250011920929,
|
| 293 |
+
"rewards/chosen": -3.391746997833252,
|
| 294 |
+
"rewards/margins": 0.37958595156669617,
|
| 295 |
+
"rewards/rejected": -3.7713329792022705,
|
| 296 |
+
"semantic_entropy": 0.8518031239509583,
|
| 297 |
+
"step": 90
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"epoch": 0.20765027322404372,
|
| 301 |
+
"grad_norm": 150.5965714109422,
|
| 302 |
+
"learning_rate": 9.653370697542987e-07,
|
| 303 |
+
"logits/chosen": -0.949033260345459,
|
| 304 |
+
"logits/rejected": -0.9504354596138,
|
| 305 |
+
"logps/chosen": -0.34006524085998535,
|
| 306 |
+
"logps/rejected": -0.3670746684074402,
|
| 307 |
+
"loss": 4.5023,
|
| 308 |
+
"rewards/accuracies": 0.574999988079071,
|
| 309 |
+
"rewards/chosen": -3.4006524085998535,
|
| 310 |
+
"rewards/margins": 0.27009397745132446,
|
| 311 |
+
"rewards/rejected": -3.6707465648651123,
|
| 312 |
+
"semantic_entropy": 0.8378095626831055,
|
| 313 |
+
"step": 95
|
| 314 |
+
},
|
| 315 |
+
{
|
| 316 |
+
"epoch": 0.2185792349726776,
|
| 317 |
+
"grad_norm": 205.55603229224332,
|
| 318 |
+
"learning_rate": 9.580077827331037e-07,
|
| 319 |
+
"logits/chosen": -0.9659525752067566,
|
| 320 |
+
"logits/rejected": -0.9214147329330444,
|
| 321 |
+
"logps/chosen": -0.37815189361572266,
|
| 322 |
+
"logps/rejected": -0.4408392012119293,
|
| 323 |
+
"loss": 4.4615,
|
| 324 |
+
"rewards/accuracies": 0.637499988079071,
|
| 325 |
+
"rewards/chosen": -3.7815189361572266,
|
| 326 |
+
"rewards/margins": 0.6268730163574219,
|
| 327 |
+
"rewards/rejected": -4.408391952514648,
|
| 328 |
+
"semantic_entropy": 0.8833344578742981,
|
| 329 |
+
"step": 100
|
| 330 |
+
},
|
| 331 |
+
{
|
| 332 |
+
"epoch": 0.22950819672131148,
|
| 333 |
+
"grad_norm": 168.46156689250674,
|
| 334 |
+
"learning_rate": 9.500095727507419e-07,
|
| 335 |
+
"logits/chosen": -1.0187479257583618,
|
| 336 |
+
"logits/rejected": -0.9876689910888672,
|
| 337 |
+
"logps/chosen": -0.3568420708179474,
|
| 338 |
+
"logps/rejected": -0.4102093577384949,
|
| 339 |
+
"loss": 4.4882,
|
| 340 |
+
"rewards/accuracies": 0.625,
|
| 341 |
+
"rewards/chosen": -3.568420886993408,
|
| 342 |
+
"rewards/margins": 0.5336726903915405,
|
| 343 |
+
"rewards/rejected": -4.102093696594238,
|
| 344 |
+
"semantic_entropy": 0.8596333265304565,
|
| 345 |
+
"step": 105
|
| 346 |
+
},
|
| 347 |
+
{
|
| 348 |
+
"epoch": 0.24043715846994534,
|
| 349 |
+
"grad_norm": 163.05795657074705,
|
| 350 |
+
"learning_rate": 9.413541212382004e-07,
|
| 351 |
+
"logits/chosen": -1.013091802597046,
|
| 352 |
+
"logits/rejected": -0.9950464367866516,
|
| 353 |
+
"logps/chosen": -0.37098902463912964,
|
| 354 |
+
"logps/rejected": -0.45781344175338745,
|
| 355 |
+
"loss": 4.3876,
|
| 356 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 357 |
+
"rewards/chosen": -3.709890842437744,
|
| 358 |
+
"rewards/margins": 0.8682438135147095,
|
| 359 |
+
"rewards/rejected": -4.578134059906006,
|
| 360 |
+
"semantic_entropy": 0.9028980135917664,
|
| 361 |
+
"step": 110
|
| 362 |
+
},
|
| 363 |
+
{
|
| 364 |
+
"epoch": 0.25136612021857924,
|
| 365 |
+
"grad_norm": 159.92608074418774,
|
| 366 |
+
"learning_rate": 9.320540695314438e-07,
|
| 367 |
+
"logits/chosen": -1.0206435918807983,
|
| 368 |
+
"logits/rejected": -0.9817934036254883,
|
| 369 |
+
"logps/chosen": -0.37152332067489624,
|
| 370 |
+
"logps/rejected": -0.4835759103298187,
|
| 371 |
+
"loss": 4.1862,
|
| 372 |
+
"rewards/accuracies": 0.699999988079071,
|
| 373 |
+
"rewards/chosen": -3.715233325958252,
|
| 374 |
+
"rewards/margins": 1.1205257177352905,
|
| 375 |
+
"rewards/rejected": -4.835759162902832,
|
| 376 |
+
"semantic_entropy": 0.8893026113510132,
|
| 377 |
+
"step": 115
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"epoch": 0.26229508196721313,
|
| 381 |
+
"grad_norm": 126.53869344949423,
|
| 382 |
+
"learning_rate": 9.221230004086721e-07,
|
| 383 |
+
"logits/chosen": -1.0430892705917358,
|
| 384 |
+
"logits/rejected": -0.9731811285018921,
|
| 385 |
+
"logps/chosen": -0.38018742203712463,
|
| 386 |
+
"logps/rejected": -0.44267600774765015,
|
| 387 |
+
"loss": 4.3042,
|
| 388 |
+
"rewards/accuracies": 0.706250011920929,
|
| 389 |
+
"rewards/chosen": -3.8018736839294434,
|
| 390 |
+
"rewards/margins": 0.6248863935470581,
|
| 391 |
+
"rewards/rejected": -4.426759719848633,
|
| 392 |
+
"semantic_entropy": 0.9051868319511414,
|
| 393 |
+
"step": 120
|
| 394 |
+
},
|
| 395 |
+
{
|
| 396 |
+
"epoch": 0.273224043715847,
|
| 397 |
+
"grad_norm": 330.21792860177044,
|
| 398 |
+
"learning_rate": 9.11575418252596e-07,
|
| 399 |
+
"logits/chosen": -0.9472485780715942,
|
| 400 |
+
"logits/rejected": -0.9132539629936218,
|
| 401 |
+
"logps/chosen": -0.396454393863678,
|
| 402 |
+
"logps/rejected": -0.4733741283416748,
|
| 403 |
+
"loss": 4.0912,
|
| 404 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 405 |
+
"rewards/chosen": -3.9645442962646484,
|
| 406 |
+
"rewards/margins": 0.7691973447799683,
|
| 407 |
+
"rewards/rejected": -4.733741283416748,
|
| 408 |
+
"semantic_entropy": 0.9098461866378784,
|
| 409 |
+
"step": 125
|
| 410 |
+
},
|
| 411 |
+
{
|
| 412 |
+
"epoch": 0.28415300546448086,
|
| 413 |
+
"grad_norm": 136.7218535491149,
|
| 414 |
+
"learning_rate": 9.004267278667031e-07,
|
| 415 |
+
"logits/chosen": -0.9847833514213562,
|
| 416 |
+
"logits/rejected": -0.9780336618423462,
|
| 417 |
+
"logps/chosen": -0.4162030816078186,
|
| 418 |
+
"logps/rejected": -0.5500742793083191,
|
| 419 |
+
"loss": 4.1012,
|
| 420 |
+
"rewards/accuracies": 0.731249988079071,
|
| 421 |
+
"rewards/chosen": -4.162030220031738,
|
| 422 |
+
"rewards/margins": 1.3387129306793213,
|
| 423 |
+
"rewards/rejected": -5.500743389129639,
|
| 424 |
+
"semantic_entropy": 0.9023580551147461,
|
| 425 |
+
"step": 130
|
| 426 |
+
},
|
| 427 |
+
{
|
| 428 |
+
"epoch": 0.29508196721311475,
|
| 429 |
+
"grad_norm": 114.59260330847327,
|
| 430 |
+
"learning_rate": 8.886932119764565e-07,
|
| 431 |
+
"logits/chosen": -1.018243670463562,
|
| 432 |
+
"logits/rejected": -0.934001624584198,
|
| 433 |
+
"logps/chosen": -0.40604203939437866,
|
| 434 |
+
"logps/rejected": -0.521537184715271,
|
| 435 |
+
"loss": 3.933,
|
| 436 |
+
"rewards/accuracies": 0.675000011920929,
|
| 437 |
+
"rewards/chosen": -4.060420036315918,
|
| 438 |
+
"rewards/margins": 1.154951810836792,
|
| 439 |
+
"rewards/rejected": -5.215372562408447,
|
| 440 |
+
"semantic_entropy": 0.9238536953926086,
|
| 441 |
+
"step": 135
|
| 442 |
+
},
|
| 443 |
+
{
|
| 444 |
+
"epoch": 0.30601092896174864,
|
| 445 |
+
"grad_norm": 151.22961611030988,
|
| 446 |
+
"learning_rate": 8.763920074482809e-07,
|
| 447 |
+
"logits/chosen": -1.02057683467865,
|
| 448 |
+
"logits/rejected": -0.9660438299179077,
|
| 449 |
+
"logps/chosen": -0.43255481123924255,
|
| 450 |
+
"logps/rejected": -0.5811036229133606,
|
| 451 |
+
"loss": 3.5202,
|
| 452 |
+
"rewards/accuracies": 0.71875,
|
| 453 |
+
"rewards/chosen": -4.32554817199707,
|
| 454 |
+
"rewards/margins": 1.485487699508667,
|
| 455 |
+
"rewards/rejected": -5.811036109924316,
|
| 456 |
+
"semantic_entropy": 0.9533751606941223,
|
| 457 |
+
"step": 140
|
| 458 |
+
},
|
| 459 |
+
{
|
| 460 |
+
"epoch": 0.31693989071038253,
|
| 461 |
+
"grad_norm": 171.9485188405326,
|
| 462 |
+
"learning_rate": 8.635410802610723e-07,
|
| 463 |
+
"logits/chosen": -1.0028008222579956,
|
| 464 |
+
"logits/rejected": -0.9838630557060242,
|
| 465 |
+
"logps/chosen": -0.4235480725765228,
|
| 466 |
+
"logps/rejected": -0.49707216024398804,
|
| 467 |
+
"loss": 3.7809,
|
| 468 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 469 |
+
"rewards/chosen": -4.235480785369873,
|
| 470 |
+
"rewards/margins": 0.7352409958839417,
|
| 471 |
+
"rewards/rejected": -4.970722198486328,
|
| 472 |
+
"semantic_entropy": 0.9417489767074585,
|
| 473 |
+
"step": 145
|
| 474 |
+
},
|
| 475 |
+
{
|
| 476 |
+
"epoch": 0.32786885245901637,
|
| 477 |
+
"grad_norm": 135.1994551251087,
|
| 478 |
+
"learning_rate": 8.501591992667849e-07,
|
| 479 |
+
"logits/chosen": -1.0660603046417236,
|
| 480 |
+
"logits/rejected": -1.0330188274383545,
|
| 481 |
+
"logps/chosen": -0.458204448223114,
|
| 482 |
+
"logps/rejected": -0.6369711756706238,
|
| 483 |
+
"loss": 3.6343,
|
| 484 |
+
"rewards/accuracies": 0.762499988079071,
|
| 485 |
+
"rewards/chosen": -4.58204460144043,
|
| 486 |
+
"rewards/margins": 1.7876676321029663,
|
| 487 |
+
"rewards/rejected": -6.369711875915527,
|
| 488 |
+
"semantic_entropy": 0.9522278904914856,
|
| 489 |
+
"step": 150
|
| 490 |
+
},
|
| 491 |
+
{
|
| 492 |
+
"epoch": 0.33879781420765026,
|
| 493 |
+
"grad_norm": 123.13149743090504,
|
| 494 |
+
"learning_rate": 8.362659087784152e-07,
|
| 495 |
+
"logits/chosen": -0.9992470741271973,
|
| 496 |
+
"logits/rejected": -0.946983814239502,
|
| 497 |
+
"logps/chosen": -0.45955339074134827,
|
| 498 |
+
"logps/rejected": -0.5732384324073792,
|
| 499 |
+
"loss": 3.6764,
|
| 500 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 501 |
+
"rewards/chosen": -4.5955328941345215,
|
| 502 |
+
"rewards/margins": 1.1368510723114014,
|
| 503 |
+
"rewards/rejected": -5.732384204864502,
|
| 504 |
+
"semantic_entropy": 0.9510468244552612,
|
| 505 |
+
"step": 155
|
| 506 |
+
},
|
| 507 |
+
{
|
| 508 |
+
"epoch": 0.34972677595628415,
|
| 509 |
+
"grad_norm": 245.75242739092204,
|
| 510 |
+
"learning_rate": 8.218815000254231e-07,
|
| 511 |
+
"logits/chosen": -1.0509494543075562,
|
| 512 |
+
"logits/rejected": -0.9942834973335266,
|
| 513 |
+
"logps/chosen": -0.5233258008956909,
|
| 514 |
+
"logps/rejected": -0.6250703930854797,
|
| 515 |
+
"loss": 3.7281,
|
| 516 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 517 |
+
"rewards/chosen": -5.233258247375488,
|
| 518 |
+
"rewards/margins": 1.0174460411071777,
|
| 519 |
+
"rewards/rejected": -6.250703811645508,
|
| 520 |
+
"semantic_entropy": 0.9671458005905151,
|
| 521 |
+
"step": 160
|
| 522 |
+
},
|
| 523 |
+
{
|
| 524 |
+
"epoch": 0.36065573770491804,
|
| 525 |
+
"grad_norm": 158.28325663577223,
|
| 526 |
+
"learning_rate": 8.07026981518276e-07,
|
| 527 |
+
"logits/chosen": -1.0312683582305908,
|
| 528 |
+
"logits/rejected": -0.9780334234237671,
|
| 529 |
+
"logps/chosen": -0.5141728520393372,
|
| 530 |
+
"logps/rejected": -0.6252259016036987,
|
| 531 |
+
"loss": 3.6091,
|
| 532 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 533 |
+
"rewards/chosen": -5.141728401184082,
|
| 534 |
+
"rewards/margins": 1.1105303764343262,
|
| 535 |
+
"rewards/rejected": -6.252258777618408,
|
| 536 |
+
"semantic_entropy": 0.9860242605209351,
|
| 537 |
+
"step": 165
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 0.37158469945355194,
|
| 541 |
+
"grad_norm": 136.70967385020978,
|
| 542 |
+
"learning_rate": 7.917240483654e-07,
|
| 543 |
+
"logits/chosen": -1.0271486043930054,
|
| 544 |
+
"logits/rejected": -0.9650506973266602,
|
| 545 |
+
"logps/chosen": -0.5105966329574585,
|
| 546 |
+
"logps/rejected": -0.6147508025169373,
|
| 547 |
+
"loss": 3.7493,
|
| 548 |
+
"rewards/accuracies": 0.71875,
|
| 549 |
+
"rewards/chosen": -5.105965614318848,
|
| 550 |
+
"rewards/margins": 1.0415422916412354,
|
| 551 |
+
"rewards/rejected": -6.147508144378662,
|
| 552 |
+
"semantic_entropy": 0.9853399991989136,
|
| 553 |
+
"step": 170
|
| 554 |
+
},
|
| 555 |
+
{
|
| 556 |
+
"epoch": 0.3825136612021858,
|
| 557 |
+
"grad_norm": 192.89943041726585,
|
| 558 |
+
"learning_rate": 7.759950505873521e-07,
|
| 559 |
+
"logits/chosen": -1.0819661617279053,
|
| 560 |
+
"logits/rejected": -1.050621747970581,
|
| 561 |
+
"logps/chosen": -0.5413715839385986,
|
| 562 |
+
"logps/rejected": -0.6315657496452332,
|
| 563 |
+
"loss": 3.5161,
|
| 564 |
+
"rewards/accuracies": 0.65625,
|
| 565 |
+
"rewards/chosen": -5.413715362548828,
|
| 566 |
+
"rewards/margins": 0.9019424319267273,
|
| 567 |
+
"rewards/rejected": -6.315657615661621,
|
| 568 |
+
"semantic_entropy": 0.9645811319351196,
|
| 569 |
+
"step": 175
|
| 570 |
+
},
|
| 571 |
+
{
|
| 572 |
+
"epoch": 0.39344262295081966,
|
| 573 |
+
"grad_norm": 152.89832261081025,
|
| 574 |
+
"learning_rate": 7.598629604744872e-07,
|
| 575 |
+
"logits/chosen": -1.0869873762130737,
|
| 576 |
+
"logits/rejected": -1.0785080194473267,
|
| 577 |
+
"logps/chosen": -0.5242325067520142,
|
| 578 |
+
"logps/rejected": -0.7012667655944824,
|
| 579 |
+
"loss": 3.3521,
|
| 580 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 581 |
+
"rewards/chosen": -5.242325305938721,
|
| 582 |
+
"rewards/margins": 1.7703425884246826,
|
| 583 |
+
"rewards/rejected": -7.012667655944824,
|
| 584 |
+
"semantic_entropy": 1.0011693239212036,
|
| 585 |
+
"step": 180
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 0.40437158469945356,
|
| 589 |
+
"grad_norm": 150.16456326171243,
|
| 590 |
+
"learning_rate": 7.433513390357989e-07,
|
| 591 |
+
"logits/chosen": -1.1106479167938232,
|
| 592 |
+
"logits/rejected": -1.1203057765960693,
|
| 593 |
+
"logps/chosen": -0.5613775253295898,
|
| 594 |
+
"logps/rejected": -0.7369329333305359,
|
| 595 |
+
"loss": 3.3325,
|
| 596 |
+
"rewards/accuracies": 0.793749988079071,
|
| 597 |
+
"rewards/chosen": -5.61377477645874,
|
| 598 |
+
"rewards/margins": 1.755553960800171,
|
| 599 |
+
"rewards/rejected": -7.36932897567749,
|
| 600 |
+
"semantic_entropy": 1.0098798274993896,
|
| 601 |
+
"step": 185
|
| 602 |
+
},
|
| 603 |
+
{
|
| 604 |
+
"epoch": 0.41530054644808745,
|
| 605 |
+
"grad_norm": 177.03981164325452,
|
| 606 |
+
"learning_rate": 7.264843015879321e-07,
|
| 607 |
+
"logits/chosen": -1.118817925453186,
|
| 608 |
+
"logits/rejected": -1.075656771659851,
|
| 609 |
+
"logps/chosen": -0.5620681047439575,
|
| 610 |
+
"logps/rejected": -0.7573049664497375,
|
| 611 |
+
"loss": 3.3889,
|
| 612 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 613 |
+
"rewards/chosen": -5.620680332183838,
|
| 614 |
+
"rewards/margins": 1.9523694515228271,
|
| 615 |
+
"rewards/rejected": -7.573049068450928,
|
| 616 |
+
"semantic_entropy": 0.998263955116272,
|
| 617 |
+
"step": 190
|
| 618 |
+
},
|
| 619 |
+
{
|
| 620 |
+
"epoch": 0.4262295081967213,
|
| 621 |
+
"grad_norm": 126.76170022708793,
|
| 622 |
+
"learning_rate": 7.092864825346266e-07,
|
| 623 |
+
"logits/chosen": -1.1385692358016968,
|
| 624 |
+
"logits/rejected": -1.1158863306045532,
|
| 625 |
+
"logps/chosen": -0.6738488078117371,
|
| 626 |
+
"logps/rejected": -0.8919061422348022,
|
| 627 |
+
"loss": 3.431,
|
| 628 |
+
"rewards/accuracies": 0.762499988079071,
|
| 629 |
+
"rewards/chosen": -6.73848819732666,
|
| 630 |
+
"rewards/margins": 2.180572986602783,
|
| 631 |
+
"rewards/rejected": -8.919061660766602,
|
| 632 |
+
"semantic_entropy": 0.9914792776107788,
|
| 633 |
+
"step": 195
|
| 634 |
+
},
|
| 635 |
+
{
|
| 636 |
+
"epoch": 0.4371584699453552,
|
| 637 |
+
"grad_norm": 146.12776815167663,
|
| 638 |
+
"learning_rate": 6.917829993880302e-07,
|
| 639 |
+
"logits/chosen": -1.1107165813446045,
|
| 640 |
+
"logits/rejected": -1.0282893180847168,
|
| 641 |
+
"logps/chosen": -0.6398170590400696,
|
| 642 |
+
"logps/rejected": -0.8206952810287476,
|
| 643 |
+
"loss": 3.2848,
|
| 644 |
+
"rewards/accuracies": 0.78125,
|
| 645 |
+
"rewards/chosen": -6.398170471191406,
|
| 646 |
+
"rewards/margins": 1.8087825775146484,
|
| 647 |
+
"rewards/rejected": -8.206953048706055,
|
| 648 |
+
"semantic_entropy": 1.0069334506988525,
|
| 649 |
+
"step": 200
|
| 650 |
+
},
|
| 651 |
+
{
|
| 652 |
+
"epoch": 0.44808743169398907,
|
| 653 |
+
"grad_norm": 114.70543290599623,
|
| 654 |
+
"learning_rate": 6.739994160844309e-07,
|
| 655 |
+
"logits/chosen": -1.0968120098114014,
|
| 656 |
+
"logits/rejected": -1.1100647449493408,
|
| 657 |
+
"logps/chosen": -0.6191304326057434,
|
| 658 |
+
"logps/rejected": -0.8075233697891235,
|
| 659 |
+
"loss": 3.113,
|
| 660 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 661 |
+
"rewards/chosen": -6.1913042068481445,
|
| 662 |
+
"rewards/margins": 1.8839294910430908,
|
| 663 |
+
"rewards/rejected": -8.075233459472656,
|
| 664 |
+
"semantic_entropy": 1.0141515731811523,
|
| 665 |
+
"step": 205
|
| 666 |
+
},
|
| 667 |
+
{
|
| 668 |
+
"epoch": 0.45901639344262296,
|
| 669 |
+
"grad_norm": 207.23320066468338,
|
| 670 |
+
"learning_rate": 6.559617056479827e-07,
|
| 671 |
+
"logits/chosen": -1.1229215860366821,
|
| 672 |
+
"logits/rejected": -1.1227028369903564,
|
| 673 |
+
"logps/chosen": -0.6795850396156311,
|
| 674 |
+
"logps/rejected": -0.9186260104179382,
|
| 675 |
+
"loss": 3.1442,
|
| 676 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 677 |
+
"rewards/chosen": -6.7958502769470215,
|
| 678 |
+
"rewards/margins": 2.3904099464416504,
|
| 679 |
+
"rewards/rejected": -9.186259269714355,
|
| 680 |
+
"semantic_entropy": 0.9720403552055359,
|
| 681 |
+
"step": 210
|
| 682 |
+
},
|
| 683 |
+
{
|
| 684 |
+
"epoch": 0.46994535519125685,
|
| 685 |
+
"grad_norm": 154.50502979258272,
|
| 686 |
+
"learning_rate": 6.376962122569567e-07,
|
| 687 |
+
"logits/chosen": -1.1284302473068237,
|
| 688 |
+
"logits/rejected": -1.0769437551498413,
|
| 689 |
+
"logps/chosen": -0.6974600553512573,
|
| 690 |
+
"logps/rejected": -0.9393995404243469,
|
| 691 |
+
"loss": 3.4054,
|
| 692 |
+
"rewards/accuracies": 0.800000011920929,
|
| 693 |
+
"rewards/chosen": -6.974600315093994,
|
| 694 |
+
"rewards/margins": 2.4193947315216064,
|
| 695 |
+
"rewards/rejected": -9.39399528503418,
|
| 696 |
+
"semantic_entropy": 1.0159441232681274,
|
| 697 |
+
"step": 215
|
| 698 |
+
},
|
| 699 |
+
{
|
| 700 |
+
"epoch": 0.4808743169398907,
|
| 701 |
+
"grad_norm": 138.6943703552703,
|
| 702 |
+
"learning_rate": 6.192296127679192e-07,
|
| 703 |
+
"logits/chosen": -1.1995110511779785,
|
| 704 |
+
"logits/rejected": -1.1735769510269165,
|
| 705 |
+
"logps/chosen": -0.7303147912025452,
|
| 706 |
+
"logps/rejected": -0.9362783432006836,
|
| 707 |
+
"loss": 3.1146,
|
| 708 |
+
"rewards/accuracies": 0.75,
|
| 709 |
+
"rewards/chosen": -7.303147792816162,
|
| 710 |
+
"rewards/margins": 2.0596346855163574,
|
| 711 |
+
"rewards/rejected": -9.362783432006836,
|
| 712 |
+
"semantic_entropy": 0.9887905120849609,
|
| 713 |
+
"step": 220
|
| 714 |
+
},
|
| 715 |
+
{
|
| 716 |
+
"epoch": 0.4918032786885246,
|
| 717 |
+
"grad_norm": 120.75185480754473,
|
| 718 |
+
"learning_rate": 6.005888777540319e-07,
|
| 719 |
+
"logits/chosen": -1.2185839414596558,
|
| 720 |
+
"logits/rejected": -1.1783679723739624,
|
| 721 |
+
"logps/chosen": -0.7471610307693481,
|
| 722 |
+
"logps/rejected": -0.9852075576782227,
|
| 723 |
+
"loss": 3.1221,
|
| 724 |
+
"rewards/accuracies": 0.8125,
|
| 725 |
+
"rewards/chosen": -7.471610069274902,
|
| 726 |
+
"rewards/margins": 2.380465030670166,
|
| 727 |
+
"rewards/rejected": -9.852075576782227,
|
| 728 |
+
"semantic_entropy": 0.9979068040847778,
|
| 729 |
+
"step": 225
|
| 730 |
+
},
|
| 731 |
+
{
|
| 732 |
+
"epoch": 0.5027322404371585,
|
| 733 |
+
"grad_norm": 122.09354415954165,
|
| 734 |
+
"learning_rate": 5.818012321143773e-07,
|
| 735 |
+
"logits/chosen": -1.1167972087860107,
|
| 736 |
+
"logits/rejected": -1.1158543825149536,
|
| 737 |
+
"logps/chosen": -0.7506524920463562,
|
| 738 |
+
"logps/rejected": -1.0082188844680786,
|
| 739 |
+
"loss": 3.1675,
|
| 740 |
+
"rewards/accuracies": 0.8125,
|
| 741 |
+
"rewards/chosen": -7.50652551651001,
|
| 742 |
+
"rewards/margins": 2.5756633281707764,
|
| 743 |
+
"rewards/rejected": -10.082188606262207,
|
| 744 |
+
"semantic_entropy": 0.993320643901825,
|
| 745 |
+
"step": 230
|
| 746 |
+
},
|
| 747 |
+
{
|
| 748 |
+
"epoch": 0.5136612021857924,
|
| 749 |
+
"grad_norm": 147.5670346142303,
|
| 750 |
+
"learning_rate": 5.628941153118388e-07,
|
| 751 |
+
"logits/chosen": -1.1291579008102417,
|
| 752 |
+
"logits/rejected": -1.0918009281158447,
|
| 753 |
+
"logps/chosen": -0.775924563407898,
|
| 754 |
+
"logps/rejected": -1.0093990564346313,
|
| 755 |
+
"loss": 2.9612,
|
| 756 |
+
"rewards/accuracies": 0.824999988079071,
|
| 757 |
+
"rewards/chosen": -7.759246826171875,
|
| 758 |
+
"rewards/margins": 2.3347439765930176,
|
| 759 |
+
"rewards/rejected": -10.093989372253418,
|
| 760 |
+
"semantic_entropy": 0.981053352355957,
|
| 761 |
+
"step": 235
|
| 762 |
+
},
|
| 763 |
+
{
|
| 764 |
+
"epoch": 0.5245901639344263,
|
| 765 |
+
"grad_norm": 133.14046441252583,
|
| 766 |
+
"learning_rate": 5.438951412976098e-07,
|
| 767 |
+
"logits/chosen": -1.1797640323638916,
|
| 768 |
+
"logits/rejected": -1.1900447607040405,
|
| 769 |
+
"logps/chosen": -0.7702494859695435,
|
| 770 |
+
"logps/rejected": -1.056715488433838,
|
| 771 |
+
"loss": 2.7098,
|
| 772 |
+
"rewards/accuracies": 0.8187500238418579,
|
| 773 |
+
"rewards/chosen": -7.702495574951172,
|
| 774 |
+
"rewards/margins": 2.864659547805786,
|
| 775 |
+
"rewards/rejected": -10.567155838012695,
|
| 776 |
+
"semantic_entropy": 0.991602897644043,
|
| 777 |
+
"step": 240
|
| 778 |
+
},
|
| 779 |
+
{
|
| 780 |
+
"epoch": 0.5355191256830601,
|
| 781 |
+
"grad_norm": 133.0444183672,
|
| 782 |
+
"learning_rate": 5.248320581808619e-07,
|
| 783 |
+
"logits/chosen": -1.1087061166763306,
|
| 784 |
+
"logits/rejected": -1.0653207302093506,
|
| 785 |
+
"logps/chosen": -0.792129397392273,
|
| 786 |
+
"logps/rejected": -1.075703501701355,
|
| 787 |
+
"loss": 2.9413,
|
| 788 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 789 |
+
"rewards/chosen": -7.921294212341309,
|
| 790 |
+
"rewards/margins": 2.835740566253662,
|
| 791 |
+
"rewards/rejected": -10.757036209106445,
|
| 792 |
+
"semantic_entropy": 0.9732475280761719,
|
| 793 |
+
"step": 245
|
| 794 |
+
},
|
| 795 |
+
{
|
| 796 |
+
"epoch": 0.546448087431694,
|
| 797 |
+
"grad_norm": 122.1994564086186,
|
| 798 |
+
"learning_rate": 5.057327077024744e-07,
|
| 799 |
+
"logits/chosen": -1.1738090515136719,
|
| 800 |
+
"logits/rejected": -1.1409178972244263,
|
| 801 |
+
"logps/chosen": -0.806613564491272,
|
| 802 |
+
"logps/rejected": -1.0335161685943604,
|
| 803 |
+
"loss": 3.0864,
|
| 804 |
+
"rewards/accuracies": 0.737500011920929,
|
| 805 |
+
"rewards/chosen": -8.06613540649414,
|
| 806 |
+
"rewards/margins": 2.2690269947052,
|
| 807 |
+
"rewards/rejected": -10.335162162780762,
|
| 808 |
+
"semantic_entropy": 0.9622586965560913,
|
| 809 |
+
"step": 250
|
| 810 |
+
},
|
| 811 |
+
{
|
| 812 |
+
"epoch": 0.5573770491803278,
|
| 813 |
+
"grad_norm": 232.10642636329976,
|
| 814 |
+
"learning_rate": 4.866249845720132e-07,
|
| 815 |
+
"logits/chosen": -1.1730918884277344,
|
| 816 |
+
"logits/rejected": -1.1448460817337036,
|
| 817 |
+
"logps/chosen": -0.8790150880813599,
|
| 818 |
+
"logps/rejected": -1.1891063451766968,
|
| 819 |
+
"loss": 2.8134,
|
| 820 |
+
"rewards/accuracies": 0.800000011920929,
|
| 821 |
+
"rewards/chosen": -8.790151596069336,
|
| 822 |
+
"rewards/margins": 3.1009116172790527,
|
| 823 |
+
"rewards/rejected": -11.891061782836914,
|
| 824 |
+
"semantic_entropy": 0.9595390558242798,
|
| 825 |
+
"step": 255
|
| 826 |
+
},
|
| 827 |
+
{
|
| 828 |
+
"epoch": 0.5683060109289617,
|
| 829 |
+
"grad_norm": 131.2755348822229,
|
| 830 |
+
"learning_rate": 4.675367957273505e-07,
|
| 831 |
+
"logits/chosen": -1.141157865524292,
|
| 832 |
+
"logits/rejected": -1.1335737705230713,
|
| 833 |
+
"logps/chosen": -0.8586977124214172,
|
| 834 |
+
"logps/rejected": -1.1454424858093262,
|
| 835 |
+
"loss": 2.8892,
|
| 836 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 837 |
+
"rewards/chosen": -8.586977005004883,
|
| 838 |
+
"rewards/margins": 2.8674476146698,
|
| 839 |
+
"rewards/rejected": -11.454424858093262,
|
| 840 |
+
"semantic_entropy": 0.9425627589225769,
|
| 841 |
+
"step": 260
|
| 842 |
+
},
|
| 843 |
+
{
|
| 844 |
+
"epoch": 0.5792349726775956,
|
| 845 |
+
"grad_norm": 140.64129624293926,
|
| 846 |
+
"learning_rate": 4.4849601957642285e-07,
|
| 847 |
+
"logits/chosen": -1.1671048402786255,
|
| 848 |
+
"logits/rejected": -1.1342850923538208,
|
| 849 |
+
"logps/chosen": -0.884295642375946,
|
| 850 |
+
"logps/rejected": -1.1824935674667358,
|
| 851 |
+
"loss": 2.8754,
|
| 852 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 853 |
+
"rewards/chosen": -8.84295654296875,
|
| 854 |
+
"rewards/margins": 2.9819793701171875,
|
| 855 |
+
"rewards/rejected": -11.824935913085938,
|
| 856 |
+
"semantic_entropy": 0.9442464709281921,
|
| 857 |
+
"step": 265
|
| 858 |
+
},
|
| 859 |
+
{
|
| 860 |
+
"epoch": 0.5901639344262295,
|
| 861 |
+
"grad_norm": 147.4714622434341,
|
| 862 |
+
"learning_rate": 4.295304652806592e-07,
|
| 863 |
+
"logits/chosen": -1.161628246307373,
|
| 864 |
+
"logits/rejected": -1.1403006315231323,
|
| 865 |
+
"logps/chosen": -0.9028175473213196,
|
| 866 |
+
"logps/rejected": -1.2475910186767578,
|
| 867 |
+
"loss": 2.6617,
|
| 868 |
+
"rewards/accuracies": 0.8125,
|
| 869 |
+
"rewards/chosen": -9.028175354003906,
|
| 870 |
+
"rewards/margins": 3.447734832763672,
|
| 871 |
+
"rewards/rejected": -12.475909233093262,
|
| 872 |
+
"semantic_entropy": 0.9366561770439148,
|
| 873 |
+
"step": 270
|
| 874 |
+
},
|
| 875 |
+
{
|
| 876 |
+
"epoch": 0.6010928961748634,
|
| 877 |
+
"grad_norm": 103.487924065158,
|
| 878 |
+
"learning_rate": 4.106678321395433e-07,
|
| 879 |
+
"logits/chosen": -1.149213194847107,
|
| 880 |
+
"logits/rejected": -1.091584324836731,
|
| 881 |
+
"logps/chosen": -0.944200873374939,
|
| 882 |
+
"logps/rejected": -1.1400898694992065,
|
| 883 |
+
"loss": 2.9059,
|
| 884 |
+
"rewards/accuracies": 0.668749988079071,
|
| 885 |
+
"rewards/chosen": -9.442008972167969,
|
| 886 |
+
"rewards/margins": 1.9588892459869385,
|
| 887 |
+
"rewards/rejected": -11.400897979736328,
|
| 888 |
+
"semantic_entropy": 0.9353906512260437,
|
| 889 |
+
"step": 275
|
| 890 |
+
},
|
| 891 |
+
{
|
| 892 |
+
"epoch": 0.6120218579234973,
|
| 893 |
+
"grad_norm": 134.66475933979152,
|
| 894 |
+
"learning_rate": 3.9193566913562915e-07,
|
| 895 |
+
"logits/chosen": -1.112555742263794,
|
| 896 |
+
"logits/rejected": -1.1183173656463623,
|
| 897 |
+
"logps/chosen": -0.9673782587051392,
|
| 898 |
+
"logps/rejected": -1.330810308456421,
|
| 899 |
+
"loss": 2.8642,
|
| 900 |
+
"rewards/accuracies": 0.731249988079071,
|
| 901 |
+
"rewards/chosen": -9.673782348632812,
|
| 902 |
+
"rewards/margins": 3.6343204975128174,
|
| 903 |
+
"rewards/rejected": -13.30810260772705,
|
| 904 |
+
"semantic_entropy": 0.8982528448104858,
|
| 905 |
+
"step": 280
|
| 906 |
+
},
|
| 907 |
+
{
|
| 908 |
+
"epoch": 0.6229508196721312,
|
| 909 |
+
"grad_norm": 122.90153644042246,
|
| 910 |
+
"learning_rate": 3.7336133469909623e-07,
|
| 911 |
+
"logits/chosen": -1.2511584758758545,
|
| 912 |
+
"logits/rejected": -1.2238370180130005,
|
| 913 |
+
"logps/chosen": -0.9596298933029175,
|
| 914 |
+
"logps/rejected": -1.3390361070632935,
|
| 915 |
+
"loss": 2.5195,
|
| 916 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 917 |
+
"rewards/chosen": -9.59630012512207,
|
| 918 |
+
"rewards/margins": 3.794060468673706,
|
| 919 |
+
"rewards/rejected": -13.390359878540039,
|
| 920 |
+
"semantic_entropy": 0.9061362147331238,
|
| 921 |
+
"step": 285
|
| 922 |
+
},
|
| 923 |
+
{
|
| 924 |
+
"epoch": 0.6338797814207651,
|
| 925 |
+
"grad_norm": 155.0095981855771,
|
| 926 |
+
"learning_rate": 3.549719567506076e-07,
|
| 927 |
+
"logits/chosen": -1.190308690071106,
|
| 928 |
+
"logits/rejected": -1.1568708419799805,
|
| 929 |
+
"logps/chosen": -1.0390634536743164,
|
| 930 |
+
"logps/rejected": -1.3806968927383423,
|
| 931 |
+
"loss": 2.8348,
|
| 932 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 933 |
+
"rewards/chosen": -10.390633583068848,
|
| 934 |
+
"rewards/margins": 3.4163336753845215,
|
| 935 |
+
"rewards/rejected": -13.806968688964844,
|
| 936 |
+
"semantic_entropy": 0.8884953260421753,
|
| 937 |
+
"step": 290
|
| 938 |
+
},
|
| 939 |
+
{
|
| 940 |
+
"epoch": 0.644808743169399,
|
| 941 |
+
"grad_norm": 128.49285257186182,
|
| 942 |
+
"learning_rate": 3.3679439308082774e-07,
|
| 943 |
+
"logits/chosen": -1.1692029237747192,
|
| 944 |
+
"logits/rejected": -1.172719120979309,
|
| 945 |
+
"logps/chosen": -1.0511713027954102,
|
| 946 |
+
"logps/rejected": -1.4550367593765259,
|
| 947 |
+
"loss": 2.2826,
|
| 948 |
+
"rewards/accuracies": 0.856249988079071,
|
| 949 |
+
"rewards/chosen": -10.511713027954102,
|
| 950 |
+
"rewards/margins": 4.0386552810668945,
|
| 951 |
+
"rewards/rejected": -14.55036735534668,
|
| 952 |
+
"semantic_entropy": 0.878921389579773,
|
| 953 |
+
"step": 295
|
| 954 |
+
},
|
| 955 |
+
{
|
| 956 |
+
"epoch": 0.6557377049180327,
|
| 957 |
+
"grad_norm": 113.97800775884723,
|
| 958 |
+
"learning_rate": 3.1885519212446716e-07,
|
| 959 |
+
"logits/chosen": -1.2200191020965576,
|
| 960 |
+
"logits/rejected": -1.209221601486206,
|
| 961 |
+
"logps/chosen": -1.0968992710113525,
|
| 962 |
+
"logps/rejected": -1.4734210968017578,
|
| 963 |
+
"loss": 2.5534,
|
| 964 |
+
"rewards/accuracies": 0.793749988079071,
|
| 965 |
+
"rewards/chosen": -10.968993186950684,
|
| 966 |
+
"rewards/margins": 3.7652194499969482,
|
| 967 |
+
"rewards/rejected": -14.734212875366211,
|
| 968 |
+
"semantic_entropy": 0.8491713404655457,
|
| 969 |
+
"step": 300
|
| 970 |
+
},
|
| 971 |
+
{
|
| 972 |
+
"epoch": 0.6666666666666666,
|
| 973 |
+
"grad_norm": 153.40901841678263,
|
| 974 |
+
"learning_rate": 3.0118055418614295e-07,
|
| 975 |
+
"logits/chosen": -1.210296392440796,
|
| 976 |
+
"logits/rejected": -1.166100263595581,
|
| 977 |
+
"logps/chosen": -1.0915873050689697,
|
| 978 |
+
"logps/rejected": -1.5003201961517334,
|
| 979 |
+
"loss": 2.7077,
|
| 980 |
+
"rewards/accuracies": 0.8187500238418579,
|
| 981 |
+
"rewards/chosen": -10.915873527526855,
|
| 982 |
+
"rewards/margins": 4.087328910827637,
|
| 983 |
+
"rewards/rejected": -15.003199577331543,
|
| 984 |
+
"semantic_entropy": 0.851865291595459,
|
| 985 |
+
"step": 305
|
| 986 |
+
},
|
| 987 |
+
{
|
| 988 |
+
"epoch": 0.6775956284153005,
|
| 989 |
+
"grad_norm": 188.92294277142176,
|
| 990 |
+
"learning_rate": 2.83796293174686e-07,
|
| 991 |
+
"logits/chosen": -1.144047498703003,
|
| 992 |
+
"logits/rejected": -1.1552207469940186,
|
| 993 |
+
"logps/chosen": -1.1141725778579712,
|
| 994 |
+
"logps/rejected": -1.5596559047698975,
|
| 995 |
+
"loss": 2.9239,
|
| 996 |
+
"rewards/accuracies": 0.800000011920929,
|
| 997 |
+
"rewards/chosen": -11.141725540161133,
|
| 998 |
+
"rewards/margins": 4.454832077026367,
|
| 999 |
+
"rewards/rejected": -15.5965576171875,
|
| 1000 |
+
"semantic_entropy": 0.8461610078811646,
|
| 1001 |
+
"step": 310
|
| 1002 |
+
},
|
| 1003 |
+
{
|
| 1004 |
+
"epoch": 0.6885245901639344,
|
| 1005 |
+
"grad_norm": 158.1416950642293,
|
| 1006 |
+
"learning_rate": 2.6672779890178046e-07,
|
| 1007 |
+
"logits/chosen": -1.1891381740570068,
|
| 1008 |
+
"logits/rejected": -1.1934657096862793,
|
| 1009 |
+
"logps/chosen": -1.1909846067428589,
|
| 1010 |
+
"logps/rejected": -1.4780324697494507,
|
| 1011 |
+
"loss": 2.6254,
|
| 1012 |
+
"rewards/accuracies": 0.75,
|
| 1013 |
+
"rewards/chosen": -11.9098482131958,
|
| 1014 |
+
"rewards/margins": 2.8704779148101807,
|
| 1015 |
+
"rewards/rejected": -14.780324935913086,
|
| 1016 |
+
"semantic_entropy": 0.8265172243118286,
|
| 1017 |
+
"step": 315
|
| 1018 |
+
},
|
| 1019 |
+
{
|
| 1020 |
+
"epoch": 0.6994535519125683,
|
| 1021 |
+
"grad_norm": 117.90491663822478,
|
| 1022 |
+
"learning_rate": 2.500000000000001e-07,
|
| 1023 |
+
"logits/chosen": -1.2632883787155151,
|
| 1024 |
+
"logits/rejected": -1.2213891744613647,
|
| 1025 |
+
"logps/chosen": -1.2055654525756836,
|
| 1026 |
+
"logps/rejected": -1.6077735424041748,
|
| 1027 |
+
"loss": 2.6783,
|
| 1028 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1029 |
+
"rewards/chosen": -12.05565357208252,
|
| 1030 |
+
"rewards/margins": 4.022082805633545,
|
| 1031 |
+
"rewards/rejected": -16.077735900878906,
|
| 1032 |
+
"semantic_entropy": 0.8192012906074524,
|
| 1033 |
+
"step": 320
|
| 1034 |
+
},
|
| 1035 |
+
{
|
| 1036 |
+
"epoch": 0.7103825136612022,
|
| 1037 |
+
"grad_norm": 164.40962706267163,
|
| 1038 |
+
"learning_rate": 2.3363732751439923e-07,
|
| 1039 |
+
"logits/chosen": -1.2160775661468506,
|
| 1040 |
+
"logits/rejected": -1.205322027206421,
|
| 1041 |
+
"logps/chosen": -1.1661893129348755,
|
| 1042 |
+
"logps/rejected": -1.5277663469314575,
|
| 1043 |
+
"loss": 2.6624,
|
| 1044 |
+
"rewards/accuracies": 0.793749988079071,
|
| 1045 |
+
"rewards/chosen": -11.661892890930176,
|
| 1046 |
+
"rewards/margins": 3.615769863128662,
|
| 1047 |
+
"rewards/rejected": -15.27766227722168,
|
| 1048 |
+
"semantic_entropy": 0.823623538017273,
|
| 1049 |
+
"step": 325
|
| 1050 |
+
},
|
| 1051 |
+
{
|
| 1052 |
+
"epoch": 0.7213114754098361,
|
| 1053 |
+
"grad_norm": 107.16070244814924,
|
| 1054 |
+
"learning_rate": 2.1766367922083283e-07,
|
| 1055 |
+
"logits/chosen": -1.1548130512237549,
|
| 1056 |
+
"logits/rejected": -1.1359449625015259,
|
| 1057 |
+
"logps/chosen": -1.1202858686447144,
|
| 1058 |
+
"logps/rejected": -1.626604676246643,
|
| 1059 |
+
"loss": 2.5554,
|
| 1060 |
+
"rewards/accuracies": 0.84375,
|
| 1061 |
+
"rewards/chosen": -11.20285701751709,
|
| 1062 |
+
"rewards/margins": 5.063187599182129,
|
| 1063 |
+
"rewards/rejected": -16.26604461669922,
|
| 1064 |
+
"semantic_entropy": 0.8284898996353149,
|
| 1065 |
+
"step": 330
|
| 1066 |
+
},
|
| 1067 |
+
{
|
| 1068 |
+
"epoch": 0.73224043715847,
|
| 1069 |
+
"grad_norm": 143.54805923440733,
|
| 1070 |
+
"learning_rate": 2.021023847231202e-07,
|
| 1071 |
+
"logits/chosen": -1.1354625225067139,
|
| 1072 |
+
"logits/rejected": -1.1055997610092163,
|
| 1073 |
+
"logps/chosen": -1.2420397996902466,
|
| 1074 |
+
"logps/rejected": -1.6342121362686157,
|
| 1075 |
+
"loss": 2.5194,
|
| 1076 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 1077 |
+
"rewards/chosen": -12.42039680480957,
|
| 1078 |
+
"rewards/margins": 3.9217231273651123,
|
| 1079 |
+
"rewards/rejected": -16.342121124267578,
|
| 1080 |
+
"semantic_entropy": 0.8055655360221863,
|
| 1081 |
+
"step": 335
|
| 1082 |
+
},
|
| 1083 |
+
{
|
| 1084 |
+
"epoch": 0.7431693989071039,
|
| 1085 |
+
"grad_norm": 143.90601843179704,
|
| 1086 |
+
"learning_rate": 1.869761713800254e-07,
|
| 1087 |
+
"logits/chosen": -1.159055471420288,
|
| 1088 |
+
"logits/rejected": -1.1217933893203735,
|
| 1089 |
+
"logps/chosen": -1.216088056564331,
|
| 1090 |
+
"logps/rejected": -1.6224826574325562,
|
| 1091 |
+
"loss": 2.6871,
|
| 1092 |
+
"rewards/accuracies": 0.768750011920929,
|
| 1093 |
+
"rewards/chosen": -12.160881996154785,
|
| 1094 |
+
"rewards/margins": 4.0639448165893555,
|
| 1095 |
+
"rewards/rejected": -16.22482681274414,
|
| 1096 |
+
"semantic_entropy": 0.7940818667411804,
|
| 1097 |
+
"step": 340
|
| 1098 |
+
},
|
| 1099 |
+
{
|
| 1100 |
+
"epoch": 0.7540983606557377,
|
| 1101 |
+
"grad_norm": 146.77126754239146,
|
| 1102 |
+
"learning_rate": 1.7230713111182164e-07,
|
| 1103 |
+
"logits/chosen": -1.2234665155410767,
|
| 1104 |
+
"logits/rejected": -1.2274234294891357,
|
| 1105 |
+
"logps/chosen": -1.2689043283462524,
|
| 1106 |
+
"logps/rejected": -1.6999661922454834,
|
| 1107 |
+
"loss": 2.6683,
|
| 1108 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 1109 |
+
"rewards/chosen": -12.689043998718262,
|
| 1110 |
+
"rewards/margins": 4.310617923736572,
|
| 1111 |
+
"rewards/rejected": -16.99966049194336,
|
| 1112 |
+
"semantic_entropy": 0.7804916501045227,
|
| 1113 |
+
"step": 345
|
| 1114 |
+
},
|
| 1115 |
+
{
|
| 1116 |
+
"epoch": 0.7650273224043715,
|
| 1117 |
+
"grad_norm": 156.59132030322806,
|
| 1118 |
+
"learning_rate": 1.5811668813491696e-07,
|
| 1119 |
+
"logits/chosen": -1.2142775058746338,
|
| 1120 |
+
"logits/rejected": -1.1987954378128052,
|
| 1121 |
+
"logps/chosen": -1.1977766752243042,
|
| 1122 |
+
"logps/rejected": -1.5605593919754028,
|
| 1123 |
+
"loss": 2.6148,
|
| 1124 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1125 |
+
"rewards/chosen": -11.977766036987305,
|
| 1126 |
+
"rewards/margins": 3.62782621383667,
|
| 1127 |
+
"rewards/rejected": -15.605592727661133,
|
| 1128 |
+
"semantic_entropy": 0.8086638450622559,
|
| 1129 |
+
"step": 350
|
| 1130 |
+
},
|
| 1131 |
+
{
|
| 1132 |
+
"epoch": 0.7759562841530054,
|
| 1133 |
+
"grad_norm": 144.3372503884412,
|
| 1134 |
+
"learning_rate": 1.4442556767166369e-07,
|
| 1135 |
+
"logits/chosen": -1.1762679815292358,
|
| 1136 |
+
"logits/rejected": -1.1479089260101318,
|
| 1137 |
+
"logps/chosen": -1.193447470664978,
|
| 1138 |
+
"logps/rejected": -1.6105226278305054,
|
| 1139 |
+
"loss": 2.5697,
|
| 1140 |
+
"rewards/accuracies": 0.84375,
|
| 1141 |
+
"rewards/chosen": -11.934475898742676,
|
| 1142 |
+
"rewards/margins": 4.170751094818115,
|
| 1143 |
+
"rewards/rejected": -16.105228424072266,
|
| 1144 |
+
"semantic_entropy": 0.8080552220344543,
|
| 1145 |
+
"step": 355
|
| 1146 |
+
},
|
| 1147 |
+
{
|
| 1148 |
+
"epoch": 0.7868852459016393,
|
| 1149 |
+
"grad_norm": 148.5272314733488,
|
| 1150 |
+
"learning_rate": 1.312537656810549e-07,
|
| 1151 |
+
"logits/chosen": -1.135874629020691,
|
| 1152 |
+
"logits/rejected": -1.1405253410339355,
|
| 1153 |
+
"logps/chosen": -1.242232322692871,
|
| 1154 |
+
"logps/rejected": -1.6316314935684204,
|
| 1155 |
+
"loss": 2.629,
|
| 1156 |
+
"rewards/accuracies": 0.75,
|
| 1157 |
+
"rewards/chosen": -12.422323226928711,
|
| 1158 |
+
"rewards/margins": 3.8939926624298096,
|
| 1159 |
+
"rewards/rejected": -16.316316604614258,
|
| 1160 |
+
"semantic_entropy": 0.8038153648376465,
|
| 1161 |
+
"step": 360
|
| 1162 |
+
},
|
| 1163 |
+
{
|
| 1164 |
+
"epoch": 0.7978142076502732,
|
| 1165 |
+
"grad_norm": 189.12080572010774,
|
| 1166 |
+
"learning_rate": 1.1862051965451214e-07,
|
| 1167 |
+
"logits/chosen": -1.216966152191162,
|
| 1168 |
+
"logits/rejected": -1.2227869033813477,
|
| 1169 |
+
"logps/chosen": -1.2868871688842773,
|
| 1170 |
+
"logps/rejected": -1.718698263168335,
|
| 1171 |
+
"loss": 2.5421,
|
| 1172 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1173 |
+
"rewards/chosen": -12.868871688842773,
|
| 1174 |
+
"rewards/margins": 4.318110466003418,
|
| 1175 |
+
"rewards/rejected": -17.18698501586914,
|
| 1176 |
+
"semantic_entropy": 0.7714166045188904,
|
| 1177 |
+
"step": 365
|
| 1178 |
+
},
|
| 1179 |
+
{
|
| 1180 |
+
"epoch": 0.8087431693989071,
|
| 1181 |
+
"grad_norm": 130.34992194933218,
|
| 1182 |
+
"learning_rate": 1.0654428051942138e-07,
|
| 1183 |
+
"logits/chosen": -1.2347556352615356,
|
| 1184 |
+
"logits/rejected": -1.2042248249053955,
|
| 1185 |
+
"logps/chosen": -1.3064343929290771,
|
| 1186 |
+
"logps/rejected": -1.777515172958374,
|
| 1187 |
+
"loss": 2.6716,
|
| 1188 |
+
"rewards/accuracies": 0.78125,
|
| 1189 |
+
"rewards/chosen": -13.06434440612793,
|
| 1190 |
+
"rewards/margins": 4.710807800292969,
|
| 1191 |
+
"rewards/rejected": -17.7751522064209,
|
| 1192 |
+
"semantic_entropy": 0.7663524150848389,
|
| 1193 |
+
"step": 370
|
| 1194 |
+
},
|
| 1195 |
+
{
|
| 1196 |
+
"epoch": 0.819672131147541,
|
| 1197 |
+
"grad_norm": 120.52547478396558,
|
| 1198 |
+
"learning_rate": 9.504268569144763e-08,
|
| 1199 |
+
"logits/chosen": -1.2310011386871338,
|
| 1200 |
+
"logits/rejected": -1.179221749305725,
|
| 1201 |
+
"logps/chosen": -1.2675514221191406,
|
| 1202 |
+
"logps/rejected": -1.7146999835968018,
|
| 1203 |
+
"loss": 2.5273,
|
| 1204 |
+
"rewards/accuracies": 0.8187500238418579,
|
| 1205 |
+
"rewards/chosen": -12.675516128540039,
|
| 1206 |
+
"rewards/margins": 4.4714837074279785,
|
| 1207 |
+
"rewards/rejected": -17.14699935913086,
|
| 1208 |
+
"semantic_entropy": 0.7719418406486511,
|
| 1209 |
+
"step": 375
|
| 1210 |
+
},
|
| 1211 |
+
{
|
| 1212 |
+
"epoch": 0.8306010928961749,
|
| 1213 |
+
"grad_norm": 134.85584452412903,
|
| 1214 |
+
"learning_rate": 8.413253331499049e-08,
|
| 1215 |
+
"logits/chosen": -1.1171958446502686,
|
| 1216 |
+
"logits/rejected": -1.1349446773529053,
|
| 1217 |
+
"logps/chosen": -1.2904092073440552,
|
| 1218 |
+
"logps/rejected": -1.6963565349578857,
|
| 1219 |
+
"loss": 2.5116,
|
| 1220 |
+
"rewards/accuracies": 0.84375,
|
| 1221 |
+
"rewards/chosen": -12.904090881347656,
|
| 1222 |
+
"rewards/margins": 4.059475898742676,
|
| 1223 |
+
"rewards/rejected": -16.96356773376465,
|
| 1224 |
+
"semantic_entropy": 0.7810186147689819,
|
| 1225 |
+
"step": 380
|
| 1226 |
+
},
|
| 1227 |
+
{
|
| 1228 |
+
"epoch": 0.8415300546448088,
|
| 1229 |
+
"grad_norm": 137.9880749898867,
|
| 1230 |
+
"learning_rate": 7.382975772939865e-08,
|
| 1231 |
+
"logits/chosen": -1.219100832939148,
|
| 1232 |
+
"logits/rejected": -1.2109915018081665,
|
| 1233 |
+
"logps/chosen": -1.3701751232147217,
|
| 1234 |
+
"logps/rejected": -1.809171438217163,
|
| 1235 |
+
"loss": 2.8231,
|
| 1236 |
+
"rewards/accuracies": 0.831250011920929,
|
| 1237 |
+
"rewards/chosen": -13.701749801635742,
|
| 1238 |
+
"rewards/margins": 4.389963626861572,
|
| 1239 |
+
"rewards/rejected": -18.09171485900879,
|
| 1240 |
+
"semantic_entropy": 0.7575483918190002,
|
| 1241 |
+
"step": 385
|
| 1242 |
+
},
|
| 1243 |
+
{
|
| 1244 |
+
"epoch": 0.8524590163934426,
|
| 1245 |
+
"grad_norm": 192.7458723286622,
|
| 1246 |
+
"learning_rate": 6.414940619677734e-08,
|
| 1247 |
+
"logits/chosen": -1.2024834156036377,
|
| 1248 |
+
"logits/rejected": -1.1897705793380737,
|
| 1249 |
+
"logps/chosen": -1.2859004735946655,
|
| 1250 |
+
"logps/rejected": -1.8171262741088867,
|
| 1251 |
+
"loss": 2.4732,
|
| 1252 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 1253 |
+
"rewards/chosen": -12.85900592803955,
|
| 1254 |
+
"rewards/margins": 5.312258243560791,
|
| 1255 |
+
"rewards/rejected": -18.171262741088867,
|
| 1256 |
+
"semantic_entropy": 0.77564936876297,
|
| 1257 |
+
"step": 390
|
| 1258 |
+
},
|
| 1259 |
+
{
|
| 1260 |
+
"epoch": 0.8633879781420765,
|
| 1261 |
+
"grad_norm": 182.36527261020933,
|
| 1262 |
+
"learning_rate": 5.5105616925376296e-08,
|
| 1263 |
+
"logits/chosen": -1.1923558712005615,
|
| 1264 |
+
"logits/rejected": -1.1765515804290771,
|
| 1265 |
+
"logps/chosen": -1.3574202060699463,
|
| 1266 |
+
"logps/rejected": -1.7072795629501343,
|
| 1267 |
+
"loss": 2.4702,
|
| 1268 |
+
"rewards/accuracies": 0.768750011920929,
|
| 1269 |
+
"rewards/chosen": -13.574200630187988,
|
| 1270 |
+
"rewards/margins": 3.4985928535461426,
|
| 1271 |
+
"rewards/rejected": -17.07279396057129,
|
| 1272 |
+
"semantic_entropy": 0.7606626749038696,
|
| 1273 |
+
"step": 395
|
| 1274 |
+
},
|
| 1275 |
+
{
|
| 1276 |
+
"epoch": 0.8743169398907104,
|
| 1277 |
+
"grad_norm": 156.32689835666156,
|
| 1278 |
+
"learning_rate": 4.6711598420656976e-08,
|
| 1279 |
+
"logits/chosen": -1.144141435623169,
|
| 1280 |
+
"logits/rejected": -1.1195942163467407,
|
| 1281 |
+
"logps/chosen": -1.3408299684524536,
|
| 1282 |
+
"logps/rejected": -1.8310232162475586,
|
| 1283 |
+
"loss": 2.3262,
|
| 1284 |
+
"rewards/accuracies": 0.8687499761581421,
|
| 1285 |
+
"rewards/chosen": -13.408299446105957,
|
| 1286 |
+
"rewards/margins": 4.901931285858154,
|
| 1287 |
+
"rewards/rejected": -18.310230255126953,
|
| 1288 |
+
"semantic_entropy": 0.7442251443862915,
|
| 1289 |
+
"step": 400
|
| 1290 |
+
},
|
| 1291 |
+
{
|
| 1292 |
+
"epoch": 0.8743169398907104,
|
| 1293 |
+
"eval_logits/chosen": -1.3813101053237915,
|
| 1294 |
+
"eval_logits/rejected": -1.3473328351974487,
|
| 1295 |
+
"eval_logps/chosen": -1.2879669666290283,
|
| 1296 |
+
"eval_logps/rejected": -1.739721655845642,
|
| 1297 |
+
"eval_loss": 2.5607941150665283,
|
| 1298 |
+
"eval_rewards/accuracies": 0.8072289228439331,
|
| 1299 |
+
"eval_rewards/chosen": -12.879671096801758,
|
| 1300 |
+
"eval_rewards/margins": 4.517546653747559,
|
| 1301 |
+
"eval_rewards/rejected": -17.397218704223633,
|
| 1302 |
+
"eval_runtime": 36.6792,
|
| 1303 |
+
"eval_samples_per_second": 35.933,
|
| 1304 |
+
"eval_semantic_entropy": 0.7719007730484009,
|
| 1305 |
+
"eval_steps_per_second": 2.263,
|
| 1306 |
+
"step": 400
|
| 1307 |
+
},
|
| 1308 |
+
{
|
| 1309 |
+
"epoch": 0.8852459016393442,
|
| 1310 |
+
"grad_norm": 148.89570610138972,
|
| 1311 |
+
"learning_rate": 3.897961019419516e-08,
|
| 1312 |
+
"logits/chosen": -1.1572606563568115,
|
| 1313 |
+
"logits/rejected": -1.1004865169525146,
|
| 1314 |
+
"logps/chosen": -1.2273566722869873,
|
| 1315 |
+
"logps/rejected": -1.640681266784668,
|
| 1316 |
+
"loss": 2.4505,
|
| 1317 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1318 |
+
"rewards/chosen": -12.273566246032715,
|
| 1319 |
+
"rewards/margins": 4.13324499130249,
|
| 1320 |
+
"rewards/rejected": -16.406810760498047,
|
| 1321 |
+
"semantic_entropy": 0.7908967137336731,
|
| 1322 |
+
"step": 405
|
| 1323 |
+
},
|
| 1324 |
+
{
|
| 1325 |
+
"epoch": 0.8961748633879781,
|
| 1326 |
+
"grad_norm": 135.13513698899922,
|
| 1327 |
+
"learning_rate": 3.192094485859526e-08,
|
| 1328 |
+
"logits/chosen": -1.164189100265503,
|
| 1329 |
+
"logits/rejected": -1.1957439184188843,
|
| 1330 |
+
"logps/chosen": -1.340441346168518,
|
| 1331 |
+
"logps/rejected": -1.8204329013824463,
|
| 1332 |
+
"loss": 2.5453,
|
| 1333 |
+
"rewards/accuracies": 0.78125,
|
| 1334 |
+
"rewards/chosen": -13.404413223266602,
|
| 1335 |
+
"rewards/margins": 4.799916744232178,
|
| 1336 |
+
"rewards/rejected": -18.204330444335938,
|
| 1337 |
+
"semantic_entropy": 0.7424927949905396,
|
| 1338 |
+
"step": 410
|
| 1339 |
+
},
|
| 1340 |
+
{
|
| 1341 |
+
"epoch": 0.907103825136612,
|
| 1342 |
+
"grad_norm": 138.1252447502212,
|
| 1343 |
+
"learning_rate": 2.5545911634565265e-08,
|
| 1344 |
+
"logits/chosen": -1.2133328914642334,
|
| 1345 |
+
"logits/rejected": -1.217905879020691,
|
| 1346 |
+
"logps/chosen": -1.3356597423553467,
|
| 1347 |
+
"logps/rejected": -1.8359012603759766,
|
| 1348 |
+
"loss": 2.719,
|
| 1349 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1350 |
+
"rewards/chosen": -13.356595993041992,
|
| 1351 |
+
"rewards/margins": 5.002416133880615,
|
| 1352 |
+
"rewards/rejected": -18.359012603759766,
|
| 1353 |
+
"semantic_entropy": 0.7578593492507935,
|
| 1354 |
+
"step": 415
|
| 1355 |
+
},
|
| 1356 |
+
{
|
| 1357 |
+
"epoch": 0.9180327868852459,
|
| 1358 |
+
"grad_norm": 133.36843556957237,
|
| 1359 |
+
"learning_rate": 1.9863821294241522e-08,
|
| 1360 |
+
"logits/chosen": -1.2045748233795166,
|
| 1361 |
+
"logits/rejected": -1.188718557357788,
|
| 1362 |
+
"logps/chosen": -1.3029712438583374,
|
| 1363 |
+
"logps/rejected": -1.7860324382781982,
|
| 1364 |
+
"loss": 2.3706,
|
| 1365 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 1366 |
+
"rewards/chosen": -13.029711723327637,
|
| 1367 |
+
"rewards/margins": 4.830612659454346,
|
| 1368 |
+
"rewards/rejected": -17.86032485961914,
|
| 1369 |
+
"semantic_entropy": 0.7500916123390198,
|
| 1370 |
+
"step": 420
|
| 1371 |
+
},
|
| 1372 |
+
{
|
| 1373 |
+
"epoch": 0.9289617486338798,
|
| 1374 |
+
"grad_norm": 139.42311056578998,
|
| 1375 |
+
"learning_rate": 1.4882972562753615e-08,
|
| 1376 |
+
"logits/chosen": -1.2088757753372192,
|
| 1377 |
+
"logits/rejected": -1.2028796672821045,
|
| 1378 |
+
"logps/chosen": -1.4182978868484497,
|
| 1379 |
+
"logps/rejected": -1.899735450744629,
|
| 1380 |
+
"loss": 2.6805,
|
| 1381 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1382 |
+
"rewards/chosen": -14.182978630065918,
|
| 1383 |
+
"rewards/margins": 4.814376354217529,
|
| 1384 |
+
"rewards/rejected": -18.997356414794922,
|
| 1385 |
+
"semantic_entropy": 0.7167907953262329,
|
| 1386 |
+
"step": 425
|
| 1387 |
+
},
|
| 1388 |
+
{
|
| 1389 |
+
"epoch": 0.9398907103825137,
|
| 1390 |
+
"grad_norm": 138.05690419269808,
|
| 1391 |
+
"learning_rate": 1.0610639997888915e-08,
|
| 1392 |
+
"logits/chosen": -1.1360652446746826,
|
| 1393 |
+
"logits/rejected": -1.1394312381744385,
|
| 1394 |
+
"logps/chosen": -1.265080451965332,
|
| 1395 |
+
"logps/rejected": -1.7739555835723877,
|
| 1396 |
+
"loss": 2.2297,
|
| 1397 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1398 |
+
"rewards/chosen": -12.65080451965332,
|
| 1399 |
+
"rewards/margins": 5.0887532234191895,
|
| 1400 |
+
"rewards/rejected": -17.73955535888672,
|
| 1401 |
+
"semantic_entropy": 0.7776240110397339,
|
| 1402 |
+
"step": 430
|
| 1403 |
+
},
|
| 1404 |
+
{
|
| 1405 |
+
"epoch": 0.9508196721311475,
|
| 1406 |
+
"grad_norm": 156.31606318677163,
|
| 1407 |
+
"learning_rate": 7.053063365559997e-09,
|
| 1408 |
+
"logits/chosen": -1.1962147951126099,
|
| 1409 |
+
"logits/rejected": -1.2261369228363037,
|
| 1410 |
+
"logps/chosen": -1.330643892288208,
|
| 1411 |
+
"logps/rejected": -1.8400895595550537,
|
| 1412 |
+
"loss": 2.397,
|
| 1413 |
+
"rewards/accuracies": 0.8687499761581421,
|
| 1414 |
+
"rewards/chosen": -13.306437492370605,
|
| 1415 |
+
"rewards/margins": 5.094459056854248,
|
| 1416 |
+
"rewards/rejected": -18.400897979736328,
|
| 1417 |
+
"semantic_entropy": 0.7367173433303833,
|
| 1418 |
+
"step": 435
|
| 1419 |
+
},
|
| 1420 |
+
{
|
| 1421 |
+
"epoch": 0.9617486338797814,
|
| 1422 |
+
"grad_norm": 137.08276067025474,
|
| 1423 |
+
"learning_rate": 4.215438526591064e-09,
|
| 1424 |
+
"logits/chosen": -1.1730583906173706,
|
| 1425 |
+
"logits/rejected": -1.140211582183838,
|
| 1426 |
+
"logps/chosen": -1.3757076263427734,
|
| 1427 |
+
"logps/rejected": -1.7452484369277954,
|
| 1428 |
+
"loss": 2.5885,
|
| 1429 |
+
"rewards/accuracies": 0.78125,
|
| 1430 |
+
"rewards/chosen": -13.757074356079102,
|
| 1431 |
+
"rewards/margins": 3.6954073905944824,
|
| 1432 |
+
"rewards/rejected": -17.452484130859375,
|
| 1433 |
+
"semantic_entropy": 0.7384070754051208,
|
| 1434 |
+
"step": 440
|
| 1435 |
+
},
|
| 1436 |
+
{
|
| 1437 |
+
"epoch": 0.9726775956284153,
|
| 1438 |
+
"grad_norm": 153.84420514847153,
|
| 1439 |
+
"learning_rate": 2.1019098481337426e-09,
|
| 1440 |
+
"logits/chosen": -1.2059695720672607,
|
| 1441 |
+
"logits/rejected": -1.1932657957077026,
|
| 1442 |
+
"logps/chosen": -1.2917059659957886,
|
| 1443 |
+
"logps/rejected": -1.7843118906021118,
|
| 1444 |
+
"loss": 2.4083,
|
| 1445 |
+
"rewards/accuracies": 0.8125,
|
| 1446 |
+
"rewards/chosen": -12.917058944702148,
|
| 1447 |
+
"rewards/margins": 4.926059246063232,
|
| 1448 |
+
"rewards/rejected": -17.843120574951172,
|
| 1449 |
+
"semantic_entropy": 0.7641780376434326,
|
| 1450 |
+
"step": 445
|
| 1451 |
+
},
|
| 1452 |
+
{
|
| 1453 |
+
"epoch": 0.9836065573770492,
|
| 1454 |
+
"grad_norm": 169.02746393168292,
|
| 1455 |
+
"learning_rate": 7.155641507955445e-10,
|
| 1456 |
+
"logits/chosen": -1.122040033340454,
|
| 1457 |
+
"logits/rejected": -1.1226763725280762,
|
| 1458 |
+
"logps/chosen": -1.3791415691375732,
|
| 1459 |
+
"logps/rejected": -1.798413872718811,
|
| 1460 |
+
"loss": 2.6799,
|
| 1461 |
+
"rewards/accuracies": 0.8125,
|
| 1462 |
+
"rewards/chosen": -13.791415214538574,
|
| 1463 |
+
"rewards/margins": 4.192723274230957,
|
| 1464 |
+
"rewards/rejected": -17.98413848876953,
|
| 1465 |
+
"semantic_entropy": 0.7384847402572632,
|
| 1466 |
+
"step": 450
|
| 1467 |
+
},
|
| 1468 |
+
{
|
| 1469 |
+
"epoch": 0.994535519125683,
|
| 1470 |
+
"grad_norm": 160.91531446057363,
|
| 1471 |
+
"learning_rate": 5.842620032053824e-11,
|
| 1472 |
+
"logits/chosen": -1.1409598588943481,
|
| 1473 |
+
"logits/rejected": -1.139664649963379,
|
| 1474 |
+
"logps/chosen": -1.3967525959014893,
|
| 1475 |
+
"logps/rejected": -1.73647940158844,
|
| 1476 |
+
"loss": 2.866,
|
| 1477 |
+
"rewards/accuracies": 0.75,
|
| 1478 |
+
"rewards/chosen": -13.96752643585205,
|
| 1479 |
+
"rewards/margins": 3.3972675800323486,
|
| 1480 |
+
"rewards/rejected": -17.364791870117188,
|
| 1481 |
+
"semantic_entropy": 0.7418167591094971,
|
| 1482 |
+
"step": 455
|
| 1483 |
+
},
|
| 1484 |
+
{
|
| 1485 |
+
"epoch": 0.9989071038251366,
|
| 1486 |
+
"step": 457,
|
| 1487 |
+
"total_flos": 0.0,
|
| 1488 |
+
"train_loss": 3.4314852449513107,
|
| 1489 |
+
"train_runtime": 5934.8281,
|
| 1490 |
+
"train_samples_per_second": 9.867,
|
| 1491 |
+
"train_steps_per_second": 0.077
|
| 1492 |
+
}
|
| 1493 |
+
],
|
| 1494 |
+
"logging_steps": 5,
|
| 1495 |
+
"max_steps": 457,
|
| 1496 |
+
"num_input_tokens_seen": 0,
|
| 1497 |
+
"num_train_epochs": 1,
|
| 1498 |
+
"save_steps": 1000000,
|
| 1499 |
+
"stateful_callbacks": {
|
| 1500 |
+
"TrainerControl": {
|
| 1501 |
+
"args": {
|
| 1502 |
+
"should_epoch_stop": false,
|
| 1503 |
+
"should_evaluate": false,
|
| 1504 |
+
"should_log": false,
|
| 1505 |
+
"should_save": true,
|
| 1506 |
+
"should_training_stop": true
|
| 1507 |
+
},
|
| 1508 |
+
"attributes": {}
|
| 1509 |
+
}
|
| 1510 |
+
},
|
| 1511 |
+
"total_flos": 0.0,
|
| 1512 |
+
"train_batch_size": 2,
|
| 1513 |
+
"trial_name": null,
|
| 1514 |
+
"trial_params": null
|
| 1515 |
+
}
|