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_l5_best_entropy_1
|
| 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_l5_best_entropy_1
|
| 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: 1.9563
|
| 22 |
+
- Rewards/chosen: -8.6021
|
| 23 |
+
- Rewards/rejected: -13.9123
|
| 24 |
+
- Rewards/accuracies: 0.8434
|
| 25 |
+
- Rewards/margins: 5.3102
|
| 26 |
+
- Logps/rejected: -1.3912
|
| 27 |
+
- Logps/chosen: -0.8602
|
| 28 |
+
- Logits/rejected: -1.3261
|
| 29 |
+
- Logits/chosen: -1.3707
|
| 30 |
+
- Semantic Entropy: 0.9091
|
| 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 |
+
| 1.9268 | 0.8743 | 400 | 1.9563 | -8.6021 | -13.9123 | 0.8434 | 5.3102 | -1.3912 | -0.8602 | -1.3261 | -1.3707 | 0.9091 |
|
| 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": 2.3507381581112385,
|
| 5 |
+
"train_runtime": 7991.7933,
|
| 6 |
+
"train_samples": 58558,
|
| 7 |
+
"train_samples_per_second": 7.327,
|
| 8 |
+
"train_steps_per_second": 0.057
|
| 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": 2.3507381581112385,
|
| 5 |
+
"train_runtime": 7991.7933,
|
| 6 |
+
"train_samples": 58558,
|
| 7 |
+
"train_samples_per_second": 7.327,
|
| 8 |
+
"train_steps_per_second": 0.057
|
| 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": 48.58164102385424,
|
| 14 |
+
"learning_rate": 1.0869565217391303e-07,
|
| 15 |
+
"logits/chosen": -1.0108345746994019,
|
| 16 |
+
"logits/rejected": -1.005958914756775,
|
| 17 |
+
"logps/chosen": -0.2804548144340515,
|
| 18 |
+
"logps/rejected": -0.2860378921031952,
|
| 19 |
+
"loss": 3.1505,
|
| 20 |
+
"rewards/accuracies": 0.53125,
|
| 21 |
+
"rewards/chosen": -2.8045480251312256,
|
| 22 |
+
"rewards/margins": 0.055831074714660645,
|
| 23 |
+
"rewards/rejected": -2.8603789806365967,
|
| 24 |
+
"semantic_entropy": 0.7518873810768127,
|
| 25 |
+
"step": 5
|
| 26 |
+
},
|
| 27 |
+
{
|
| 28 |
+
"epoch": 0.02185792349726776,
|
| 29 |
+
"grad_norm": 60.09792641017642,
|
| 30 |
+
"learning_rate": 2.1739130434782607e-07,
|
| 31 |
+
"logits/chosen": -1.0501697063446045,
|
| 32 |
+
"logits/rejected": -0.9994386434555054,
|
| 33 |
+
"logps/chosen": -0.25685304403305054,
|
| 34 |
+
"logps/rejected": -0.27076202630996704,
|
| 35 |
+
"loss": 3.1218,
|
| 36 |
+
"rewards/accuracies": 0.5375000238418579,
|
| 37 |
+
"rewards/chosen": -2.568530559539795,
|
| 38 |
+
"rewards/margins": 0.1390899419784546,
|
| 39 |
+
"rewards/rejected": -2.707620143890381,
|
| 40 |
+
"semantic_entropy": 0.7095814347267151,
|
| 41 |
+
"step": 10
|
| 42 |
+
},
|
| 43 |
+
{
|
| 44 |
+
"epoch": 0.03278688524590164,
|
| 45 |
+
"grad_norm": 53.218153749761065,
|
| 46 |
+
"learning_rate": 3.260869565217391e-07,
|
| 47 |
+
"logits/chosen": -1.010613203048706,
|
| 48 |
+
"logits/rejected": -0.964097797870636,
|
| 49 |
+
"logps/chosen": -0.2674953043460846,
|
| 50 |
+
"logps/rejected": -0.2733650207519531,
|
| 51 |
+
"loss": 3.1131,
|
| 52 |
+
"rewards/accuracies": 0.512499988079071,
|
| 53 |
+
"rewards/chosen": -2.674952983856201,
|
| 54 |
+
"rewards/margins": 0.05869739130139351,
|
| 55 |
+
"rewards/rejected": -2.7336502075195312,
|
| 56 |
+
"semantic_entropy": 0.7275451421737671,
|
| 57 |
+
"step": 15
|
| 58 |
+
},
|
| 59 |
+
{
|
| 60 |
+
"epoch": 0.04371584699453552,
|
| 61 |
+
"grad_norm": 67.45687349058657,
|
| 62 |
+
"learning_rate": 4.3478260869565214e-07,
|
| 63 |
+
"logits/chosen": -0.9450489282608032,
|
| 64 |
+
"logits/rejected": -0.8954045176506042,
|
| 65 |
+
"logps/chosen": -0.27203652262687683,
|
| 66 |
+
"logps/rejected": -0.2849101722240448,
|
| 67 |
+
"loss": 3.1483,
|
| 68 |
+
"rewards/accuracies": 0.5687500238418579,
|
| 69 |
+
"rewards/chosen": -2.720365047454834,
|
| 70 |
+
"rewards/margins": 0.1287367343902588,
|
| 71 |
+
"rewards/rejected": -2.8491017818450928,
|
| 72 |
+
"semantic_entropy": 0.7445512413978577,
|
| 73 |
+
"step": 20
|
| 74 |
+
},
|
| 75 |
+
{
|
| 76 |
+
"epoch": 0.0546448087431694,
|
| 77 |
+
"grad_norm": 31.667955833823346,
|
| 78 |
+
"learning_rate": 5.434782608695652e-07,
|
| 79 |
+
"logits/chosen": -0.9527080655097961,
|
| 80 |
+
"logits/rejected": -0.8783634901046753,
|
| 81 |
+
"logps/chosen": -0.27410227060317993,
|
| 82 |
+
"logps/rejected": -0.29486891627311707,
|
| 83 |
+
"loss": 3.1117,
|
| 84 |
+
"rewards/accuracies": 0.59375,
|
| 85 |
+
"rewards/chosen": -2.741022825241089,
|
| 86 |
+
"rewards/margins": 0.2076665163040161,
|
| 87 |
+
"rewards/rejected": -2.9486892223358154,
|
| 88 |
+
"semantic_entropy": 0.754062831401825,
|
| 89 |
+
"step": 25
|
| 90 |
+
},
|
| 91 |
+
{
|
| 92 |
+
"epoch": 0.06557377049180328,
|
| 93 |
+
"grad_norm": 49.885531132124974,
|
| 94 |
+
"learning_rate": 6.521739130434782e-07,
|
| 95 |
+
"logits/chosen": -1.05032479763031,
|
| 96 |
+
"logits/rejected": -0.9858748316764832,
|
| 97 |
+
"logps/chosen": -0.2675837576389313,
|
| 98 |
+
"logps/rejected": -0.28654658794403076,
|
| 99 |
+
"loss": 3.1024,
|
| 100 |
+
"rewards/accuracies": 0.5375000238418579,
|
| 101 |
+
"rewards/chosen": -2.675837278366089,
|
| 102 |
+
"rewards/margins": 0.1896287053823471,
|
| 103 |
+
"rewards/rejected": -2.8654661178588867,
|
| 104 |
+
"semantic_entropy": 0.726620078086853,
|
| 105 |
+
"step": 30
|
| 106 |
+
},
|
| 107 |
+
{
|
| 108 |
+
"epoch": 0.07650273224043716,
|
| 109 |
+
"grad_norm": 55.4615534844222,
|
| 110 |
+
"learning_rate": 7.608695652173913e-07,
|
| 111 |
+
"logits/chosen": -1.0076689720153809,
|
| 112 |
+
"logits/rejected": -0.9416168332099915,
|
| 113 |
+
"logps/chosen": -0.25796476006507874,
|
| 114 |
+
"logps/rejected": -0.2931690514087677,
|
| 115 |
+
"loss": 3.058,
|
| 116 |
+
"rewards/accuracies": 0.512499988079071,
|
| 117 |
+
"rewards/chosen": -2.5796475410461426,
|
| 118 |
+
"rewards/margins": 0.3520428538322449,
|
| 119 |
+
"rewards/rejected": -2.931690216064453,
|
| 120 |
+
"semantic_entropy": 0.7216796875,
|
| 121 |
+
"step": 35
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 0.08743169398907104,
|
| 125 |
+
"grad_norm": 61.12682886329354,
|
| 126 |
+
"learning_rate": 8.695652173913043e-07,
|
| 127 |
+
"logits/chosen": -0.9594043493270874,
|
| 128 |
+
"logits/rejected": -0.8999654650688171,
|
| 129 |
+
"logps/chosen": -0.28498369455337524,
|
| 130 |
+
"logps/rejected": -0.31468039751052856,
|
| 131 |
+
"loss": 3.1476,
|
| 132 |
+
"rewards/accuracies": 0.512499988079071,
|
| 133 |
+
"rewards/chosen": -2.849836826324463,
|
| 134 |
+
"rewards/margins": 0.296967089176178,
|
| 135 |
+
"rewards/rejected": -3.146804094314575,
|
| 136 |
+
"semantic_entropy": 0.7657366394996643,
|
| 137 |
+
"step": 40
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 0.09836065573770492,
|
| 141 |
+
"grad_norm": 30.63183736560673,
|
| 142 |
+
"learning_rate": 9.782608695652173e-07,
|
| 143 |
+
"logits/chosen": -1.0224157571792603,
|
| 144 |
+
"logits/rejected": -0.9429336786270142,
|
| 145 |
+
"logps/chosen": -0.29369351267814636,
|
| 146 |
+
"logps/rejected": -0.3366171717643738,
|
| 147 |
+
"loss": 3.0689,
|
| 148 |
+
"rewards/accuracies": 0.518750011920929,
|
| 149 |
+
"rewards/chosen": -2.9369354248046875,
|
| 150 |
+
"rewards/margins": 0.42923641204833984,
|
| 151 |
+
"rewards/rejected": -3.3661720752716064,
|
| 152 |
+
"semantic_entropy": 0.7680700421333313,
|
| 153 |
+
"step": 45
|
| 154 |
+
},
|
| 155 |
+
{
|
| 156 |
+
"epoch": 0.1092896174863388,
|
| 157 |
+
"grad_norm": 66.51868162378878,
|
| 158 |
+
"learning_rate": 9.997663088532014e-07,
|
| 159 |
+
"logits/chosen": -0.9610111117362976,
|
| 160 |
+
"logits/rejected": -0.8836190104484558,
|
| 161 |
+
"logps/chosen": -0.3035297095775604,
|
| 162 |
+
"logps/rejected": -0.3295886814594269,
|
| 163 |
+
"loss": 3.0733,
|
| 164 |
+
"rewards/accuracies": 0.4937500059604645,
|
| 165 |
+
"rewards/chosen": -3.03529691696167,
|
| 166 |
+
"rewards/margins": 0.26059025526046753,
|
| 167 |
+
"rewards/rejected": -3.295886993408203,
|
| 168 |
+
"semantic_entropy": 0.7657495141029358,
|
| 169 |
+
"step": 50
|
| 170 |
+
},
|
| 171 |
+
{
|
| 172 |
+
"epoch": 0.12021857923497267,
|
| 173 |
+
"grad_norm": 36.53428450530809,
|
| 174 |
+
"learning_rate": 9.98817312944725e-07,
|
| 175 |
+
"logits/chosen": -1.002439260482788,
|
| 176 |
+
"logits/rejected": -0.8919259905815125,
|
| 177 |
+
"logps/chosen": -0.2904648780822754,
|
| 178 |
+
"logps/rejected": -0.3420776426792145,
|
| 179 |
+
"loss": 2.9863,
|
| 180 |
+
"rewards/accuracies": 0.581250011920929,
|
| 181 |
+
"rewards/chosen": -2.904649019241333,
|
| 182 |
+
"rewards/margins": 0.5161274671554565,
|
| 183 |
+
"rewards/rejected": -3.4207763671875,
|
| 184 |
+
"semantic_entropy": 0.7696514129638672,
|
| 185 |
+
"step": 55
|
| 186 |
+
},
|
| 187 |
+
{
|
| 188 |
+
"epoch": 0.13114754098360656,
|
| 189 |
+
"grad_norm": 31.199928628063066,
|
| 190 |
+
"learning_rate": 9.971397915250336e-07,
|
| 191 |
+
"logits/chosen": -1.0388410091400146,
|
| 192 |
+
"logits/rejected": -0.9951351284980774,
|
| 193 |
+
"logps/chosen": -0.27618950605392456,
|
| 194 |
+
"logps/rejected": -0.3456563353538513,
|
| 195 |
+
"loss": 2.9429,
|
| 196 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 197 |
+
"rewards/chosen": -2.761894941329956,
|
| 198 |
+
"rewards/margins": 0.694668173789978,
|
| 199 |
+
"rewards/rejected": -3.4565632343292236,
|
| 200 |
+
"semantic_entropy": 0.7672961354255676,
|
| 201 |
+
"step": 60
|
| 202 |
+
},
|
| 203 |
+
{
|
| 204 |
+
"epoch": 0.14207650273224043,
|
| 205 |
+
"grad_norm": 43.04608890956751,
|
| 206 |
+
"learning_rate": 9.94736194623663e-07,
|
| 207 |
+
"logits/chosen": -1.0304734706878662,
|
| 208 |
+
"logits/rejected": -0.9656420946121216,
|
| 209 |
+
"logps/chosen": -0.3235929608345032,
|
| 210 |
+
"logps/rejected": -0.3579636216163635,
|
| 211 |
+
"loss": 3.063,
|
| 212 |
+
"rewards/accuracies": 0.53125,
|
| 213 |
+
"rewards/chosen": -3.2359299659729004,
|
| 214 |
+
"rewards/margins": 0.34370657801628113,
|
| 215 |
+
"rewards/rejected": -3.579636335372925,
|
| 216 |
+
"semantic_entropy": 0.8141781091690063,
|
| 217 |
+
"step": 65
|
| 218 |
+
},
|
| 219 |
+
{
|
| 220 |
+
"epoch": 0.15300546448087432,
|
| 221 |
+
"grad_norm": 210.46235590585465,
|
| 222 |
+
"learning_rate": 9.916100327075037e-07,
|
| 223 |
+
"logits/chosen": -0.9860013723373413,
|
| 224 |
+
"logits/rejected": -0.9686506986618042,
|
| 225 |
+
"logps/chosen": -0.3108309209346771,
|
| 226 |
+
"logps/rejected": -0.36498162150382996,
|
| 227 |
+
"loss": 2.9623,
|
| 228 |
+
"rewards/accuracies": 0.581250011920929,
|
| 229 |
+
"rewards/chosen": -3.108308792114258,
|
| 230 |
+
"rewards/margins": 0.5415072441101074,
|
| 231 |
+
"rewards/rejected": -3.6498160362243652,
|
| 232 |
+
"semantic_entropy": 0.7746008038520813,
|
| 233 |
+
"step": 70
|
| 234 |
+
},
|
| 235 |
+
{
|
| 236 |
+
"epoch": 0.16393442622950818,
|
| 237 |
+
"grad_norm": 35.07537533113234,
|
| 238 |
+
"learning_rate": 9.877658715537428e-07,
|
| 239 |
+
"logits/chosen": -0.97590172290802,
|
| 240 |
+
"logits/rejected": -0.9608650207519531,
|
| 241 |
+
"logps/chosen": -0.32075995206832886,
|
| 242 |
+
"logps/rejected": -0.39185312390327454,
|
| 243 |
+
"loss": 3.0014,
|
| 244 |
+
"rewards/accuracies": 0.637499988079071,
|
| 245 |
+
"rewards/chosen": -3.2076001167297363,
|
| 246 |
+
"rewards/margins": 0.7109313011169434,
|
| 247 |
+
"rewards/rejected": -3.9185314178466797,
|
| 248 |
+
"semantic_entropy": 0.806628406047821,
|
| 249 |
+
"step": 75
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 0.17486338797814208,
|
| 253 |
+
"grad_norm": 52.85437823609445,
|
| 254 |
+
"learning_rate": 9.832093255815216e-07,
|
| 255 |
+
"logits/chosen": -0.9724711179733276,
|
| 256 |
+
"logits/rejected": -0.9133187532424927,
|
| 257 |
+
"logps/chosen": -0.33075493574142456,
|
| 258 |
+
"logps/rejected": -0.3698170483112335,
|
| 259 |
+
"loss": 2.9391,
|
| 260 |
+
"rewards/accuracies": 0.637499988079071,
|
| 261 |
+
"rewards/chosen": -3.307549238204956,
|
| 262 |
+
"rewards/margins": 0.39062100648880005,
|
| 263 |
+
"rewards/rejected": -3.6981704235076904,
|
| 264 |
+
"semantic_entropy": 0.7876138091087341,
|
| 265 |
+
"step": 80
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.18579234972677597,
|
| 269 |
+
"grad_norm": 68.01590434786175,
|
| 270 |
+
"learning_rate": 9.779470496520441e-07,
|
| 271 |
+
"logits/chosen": -0.973779022693634,
|
| 272 |
+
"logits/rejected": -0.9283053278923035,
|
| 273 |
+
"logps/chosen": -0.3297511339187622,
|
| 274 |
+
"logps/rejected": -0.41490381956100464,
|
| 275 |
+
"loss": 2.9392,
|
| 276 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 277 |
+
"rewards/chosen": -3.297511339187622,
|
| 278 |
+
"rewards/margins": 0.8515273928642273,
|
| 279 |
+
"rewards/rejected": -4.149038791656494,
|
| 280 |
+
"semantic_entropy": 0.8133514523506165,
|
| 281 |
+
"step": 85
|
| 282 |
+
},
|
| 283 |
+
{
|
| 284 |
+
"epoch": 0.19672131147540983,
|
| 285 |
+
"grad_norm": 43.8207236035398,
|
| 286 |
+
"learning_rate": 9.719867293491144e-07,
|
| 287 |
+
"logits/chosen": -1.0416749715805054,
|
| 288 |
+
"logits/rejected": -0.9649559259414673,
|
| 289 |
+
"logps/chosen": -0.3701372444629669,
|
| 290 |
+
"logps/rejected": -0.42724722623825073,
|
| 291 |
+
"loss": 2.9596,
|
| 292 |
+
"rewards/accuracies": 0.59375,
|
| 293 |
+
"rewards/chosen": -3.7013721466064453,
|
| 294 |
+
"rewards/margins": 0.5711001753807068,
|
| 295 |
+
"rewards/rejected": -4.272472381591797,
|
| 296 |
+
"semantic_entropy": 0.8610862493515015,
|
| 297 |
+
"step": 90
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"epoch": 0.20765027322404372,
|
| 301 |
+
"grad_norm": 37.490640629213566,
|
| 302 |
+
"learning_rate": 9.653370697542987e-07,
|
| 303 |
+
"logits/chosen": -0.9690491557121277,
|
| 304 |
+
"logits/rejected": -0.9741093516349792,
|
| 305 |
+
"logps/chosen": -0.358903706073761,
|
| 306 |
+
"logps/rejected": -0.41135644912719727,
|
| 307 |
+
"loss": 2.8808,
|
| 308 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 309 |
+
"rewards/chosen": -3.5890374183654785,
|
| 310 |
+
"rewards/margins": 0.5245276689529419,
|
| 311 |
+
"rewards/rejected": -4.113564968109131,
|
| 312 |
+
"semantic_entropy": 0.8484892845153809,
|
| 313 |
+
"step": 95
|
| 314 |
+
},
|
| 315 |
+
{
|
| 316 |
+
"epoch": 0.2185792349726776,
|
| 317 |
+
"grad_norm": 72.4012883538114,
|
| 318 |
+
"learning_rate": 9.580077827331037e-07,
|
| 319 |
+
"logits/chosen": -0.9927606582641602,
|
| 320 |
+
"logits/rejected": -0.9488978385925293,
|
| 321 |
+
"logps/chosen": -0.4221861958503723,
|
| 322 |
+
"logps/rejected": -0.5099042654037476,
|
| 323 |
+
"loss": 2.8815,
|
| 324 |
+
"rewards/accuracies": 0.637499988079071,
|
| 325 |
+
"rewards/chosen": -4.221861839294434,
|
| 326 |
+
"rewards/margins": 0.8771804571151733,
|
| 327 |
+
"rewards/rejected": -5.0990424156188965,
|
| 328 |
+
"semantic_entropy": 0.891185462474823,
|
| 329 |
+
"step": 100
|
| 330 |
+
},
|
| 331 |
+
{
|
| 332 |
+
"epoch": 0.22950819672131148,
|
| 333 |
+
"grad_norm": 53.762702275255066,
|
| 334 |
+
"learning_rate": 9.500095727507419e-07,
|
| 335 |
+
"logits/chosen": -1.0526697635650635,
|
| 336 |
+
"logits/rejected": -1.025564193725586,
|
| 337 |
+
"logps/chosen": -0.37911754846572876,
|
| 338 |
+
"logps/rejected": -0.46409544348716736,
|
| 339 |
+
"loss": 2.8288,
|
| 340 |
+
"rewards/accuracies": 0.637499988079071,
|
| 341 |
+
"rewards/chosen": -3.7911758422851562,
|
| 342 |
+
"rewards/margins": 0.8497791290283203,
|
| 343 |
+
"rewards/rejected": -4.640954971313477,
|
| 344 |
+
"semantic_entropy": 0.8762981295585632,
|
| 345 |
+
"step": 105
|
| 346 |
+
},
|
| 347 |
+
{
|
| 348 |
+
"epoch": 0.24043715846994534,
|
| 349 |
+
"grad_norm": 45.04185647458636,
|
| 350 |
+
"learning_rate": 9.413541212382004e-07,
|
| 351 |
+
"logits/chosen": -1.0478137731552124,
|
| 352 |
+
"logits/rejected": -1.028585433959961,
|
| 353 |
+
"logps/chosen": -0.4319530427455902,
|
| 354 |
+
"logps/rejected": -0.5620400905609131,
|
| 355 |
+
"loss": 2.8058,
|
| 356 |
+
"rewards/accuracies": 0.6875,
|
| 357 |
+
"rewards/chosen": -4.319530487060547,
|
| 358 |
+
"rewards/margins": 1.3008701801300049,
|
| 359 |
+
"rewards/rejected": -5.620400428771973,
|
| 360 |
+
"semantic_entropy": 0.922514796257019,
|
| 361 |
+
"step": 110
|
| 362 |
+
},
|
| 363 |
+
{
|
| 364 |
+
"epoch": 0.25136612021857924,
|
| 365 |
+
"grad_norm": 56.505413378989054,
|
| 366 |
+
"learning_rate": 9.320540695314438e-07,
|
| 367 |
+
"logits/chosen": -1.06332266330719,
|
| 368 |
+
"logits/rejected": -1.0288236141204834,
|
| 369 |
+
"logps/chosen": -0.43162402510643005,
|
| 370 |
+
"logps/rejected": -0.5995403528213501,
|
| 371 |
+
"loss": 2.7772,
|
| 372 |
+
"rewards/accuracies": 0.706250011920929,
|
| 373 |
+
"rewards/chosen": -4.3162407875061035,
|
| 374 |
+
"rewards/margins": 1.6791629791259766,
|
| 375 |
+
"rewards/rejected": -5.995403289794922,
|
| 376 |
+
"semantic_entropy": 0.913605809211731,
|
| 377 |
+
"step": 115
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"epoch": 0.26229508196721313,
|
| 381 |
+
"grad_norm": 44.06382434081467,
|
| 382 |
+
"learning_rate": 9.221230004086721e-07,
|
| 383 |
+
"logits/chosen": -1.0795866250991821,
|
| 384 |
+
"logits/rejected": -1.0165742635726929,
|
| 385 |
+
"logps/chosen": -0.441800594329834,
|
| 386 |
+
"logps/rejected": -0.5526998043060303,
|
| 387 |
+
"loss": 2.736,
|
| 388 |
+
"rewards/accuracies": 0.6875,
|
| 389 |
+
"rewards/chosen": -4.41800594329834,
|
| 390 |
+
"rewards/margins": 1.1089919805526733,
|
| 391 |
+
"rewards/rejected": -5.526998519897461,
|
| 392 |
+
"semantic_entropy": 0.9433540105819702,
|
| 393 |
+
"step": 120
|
| 394 |
+
},
|
| 395 |
+
{
|
| 396 |
+
"epoch": 0.273224043715847,
|
| 397 |
+
"grad_norm": 115.05665224956637,
|
| 398 |
+
"learning_rate": 9.11575418252596e-07,
|
| 399 |
+
"logits/chosen": -0.9929405450820923,
|
| 400 |
+
"logits/rejected": -0.9629238247871399,
|
| 401 |
+
"logps/chosen": -0.4976869225502014,
|
| 402 |
+
"logps/rejected": -0.6475359201431274,
|
| 403 |
+
"loss": 2.7093,
|
| 404 |
+
"rewards/accuracies": 0.731249988079071,
|
| 405 |
+
"rewards/chosen": -4.976869106292725,
|
| 406 |
+
"rewards/margins": 1.4984896183013916,
|
| 407 |
+
"rewards/rejected": -6.4753594398498535,
|
| 408 |
+
"semantic_entropy": 0.9429339170455933,
|
| 409 |
+
"step": 125
|
| 410 |
+
},
|
| 411 |
+
{
|
| 412 |
+
"epoch": 0.28415300546448086,
|
| 413 |
+
"grad_norm": 45.61755920747003,
|
| 414 |
+
"learning_rate": 9.004267278667031e-07,
|
| 415 |
+
"logits/chosen": -1.0040867328643799,
|
| 416 |
+
"logits/rejected": -1.0052454471588135,
|
| 417 |
+
"logps/chosen": -0.5251004099845886,
|
| 418 |
+
"logps/rejected": -0.7621506452560425,
|
| 419 |
+
"loss": 2.7046,
|
| 420 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 421 |
+
"rewards/chosen": -5.251004219055176,
|
| 422 |
+
"rewards/margins": 2.3705029487609863,
|
| 423 |
+
"rewards/rejected": -7.621507167816162,
|
| 424 |
+
"semantic_entropy": 0.9312794804573059,
|
| 425 |
+
"step": 130
|
| 426 |
+
},
|
| 427 |
+
{
|
| 428 |
+
"epoch": 0.29508196721311475,
|
| 429 |
+
"grad_norm": 42.07329597975219,
|
| 430 |
+
"learning_rate": 8.886932119764565e-07,
|
| 431 |
+
"logits/chosen": -1.0416096448898315,
|
| 432 |
+
"logits/rejected": -0.9662833213806152,
|
| 433 |
+
"logps/chosen": -0.5473520755767822,
|
| 434 |
+
"logps/rejected": -0.7608135938644409,
|
| 435 |
+
"loss": 2.5919,
|
| 436 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 437 |
+
"rewards/chosen": -5.473520755767822,
|
| 438 |
+
"rewards/margins": 2.134615898132324,
|
| 439 |
+
"rewards/rejected": -7.608136177062988,
|
| 440 |
+
"semantic_entropy": 0.9771392941474915,
|
| 441 |
+
"step": 135
|
| 442 |
+
},
|
| 443 |
+
{
|
| 444 |
+
"epoch": 0.30601092896174864,
|
| 445 |
+
"grad_norm": 45.36734996455639,
|
| 446 |
+
"learning_rate": 8.763920074482809e-07,
|
| 447 |
+
"logits/chosen": -1.0434763431549072,
|
| 448 |
+
"logits/rejected": -0.9957412481307983,
|
| 449 |
+
"logps/chosen": -0.594224750995636,
|
| 450 |
+
"logps/rejected": -0.8605974316596985,
|
| 451 |
+
"loss": 2.3803,
|
| 452 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 453 |
+
"rewards/chosen": -5.94224739074707,
|
| 454 |
+
"rewards/margins": 2.663726806640625,
|
| 455 |
+
"rewards/rejected": -8.605974197387695,
|
| 456 |
+
"semantic_entropy": 0.9883469343185425,
|
| 457 |
+
"step": 140
|
| 458 |
+
},
|
| 459 |
+
{
|
| 460 |
+
"epoch": 0.31693989071038253,
|
| 461 |
+
"grad_norm": 57.35860265502167,
|
| 462 |
+
"learning_rate": 8.635410802610723e-07,
|
| 463 |
+
"logits/chosen": -1.0125925540924072,
|
| 464 |
+
"logits/rejected": -1.0003665685653687,
|
| 465 |
+
"logps/chosen": -0.6211115121841431,
|
| 466 |
+
"logps/rejected": -0.7806649804115295,
|
| 467 |
+
"loss": 2.5474,
|
| 468 |
+
"rewards/accuracies": 0.731249988079071,
|
| 469 |
+
"rewards/chosen": -6.211114406585693,
|
| 470 |
+
"rewards/margins": 1.5955346822738647,
|
| 471 |
+
"rewards/rejected": -7.806649208068848,
|
| 472 |
+
"semantic_entropy": 0.9819900393486023,
|
| 473 |
+
"step": 145
|
| 474 |
+
},
|
| 475 |
+
{
|
| 476 |
+
"epoch": 0.32786885245901637,
|
| 477 |
+
"grad_norm": 57.36843422854158,
|
| 478 |
+
"learning_rate": 8.501591992667849e-07,
|
| 479 |
+
"logits/chosen": -1.0663528442382812,
|
| 480 |
+
"logits/rejected": -1.0401932001113892,
|
| 481 |
+
"logps/chosen": -0.705702543258667,
|
| 482 |
+
"logps/rejected": -1.0317673683166504,
|
| 483 |
+
"loss": 2.4773,
|
| 484 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 485 |
+
"rewards/chosen": -7.057024955749512,
|
| 486 |
+
"rewards/margins": 3.2606492042541504,
|
| 487 |
+
"rewards/rejected": -10.317673683166504,
|
| 488 |
+
"semantic_entropy": 0.9794198870658875,
|
| 489 |
+
"step": 150
|
| 490 |
+
},
|
| 491 |
+
{
|
| 492 |
+
"epoch": 0.33879781420765026,
|
| 493 |
+
"grad_norm": 55.65247575131584,
|
| 494 |
+
"learning_rate": 8.362659087784152e-07,
|
| 495 |
+
"logits/chosen": -1.0362324714660645,
|
| 496 |
+
"logits/rejected": -0.9899821281433105,
|
| 497 |
+
"logps/chosen": -0.7101436257362366,
|
| 498 |
+
"logps/rejected": -0.9879515767097473,
|
| 499 |
+
"loss": 2.4284,
|
| 500 |
+
"rewards/accuracies": 0.731249988079071,
|
| 501 |
+
"rewards/chosen": -7.101436614990234,
|
| 502 |
+
"rewards/margins": 2.7780795097351074,
|
| 503 |
+
"rewards/rejected": -9.879515647888184,
|
| 504 |
+
"semantic_entropy": 0.9602919816970825,
|
| 505 |
+
"step": 155
|
| 506 |
+
},
|
| 507 |
+
{
|
| 508 |
+
"epoch": 0.34972677595628415,
|
| 509 |
+
"grad_norm": 57.58698714096881,
|
| 510 |
+
"learning_rate": 8.218815000254231e-07,
|
| 511 |
+
"logits/chosen": -1.0922863483428955,
|
| 512 |
+
"logits/rejected": -1.0440222024917603,
|
| 513 |
+
"logps/chosen": -0.8645855188369751,
|
| 514 |
+
"logps/rejected": -1.1068294048309326,
|
| 515 |
+
"loss": 2.4302,
|
| 516 |
+
"rewards/accuracies": 0.737500011920929,
|
| 517 |
+
"rewards/chosen": -8.645854949951172,
|
| 518 |
+
"rewards/margins": 2.4224390983581543,
|
| 519 |
+
"rewards/rejected": -11.068293571472168,
|
| 520 |
+
"semantic_entropy": 0.9356050491333008,
|
| 521 |
+
"step": 160
|
| 522 |
+
},
|
| 523 |
+
{
|
| 524 |
+
"epoch": 0.36065573770491804,
|
| 525 |
+
"grad_norm": 98.49758980833779,
|
| 526 |
+
"learning_rate": 8.07026981518276e-07,
|
| 527 |
+
"logits/chosen": -1.0802171230316162,
|
| 528 |
+
"logits/rejected": -1.0290331840515137,
|
| 529 |
+
"logps/chosen": -0.8299033045768738,
|
| 530 |
+
"logps/rejected": -1.0872141122817993,
|
| 531 |
+
"loss": 2.4163,
|
| 532 |
+
"rewards/accuracies": 0.75,
|
| 533 |
+
"rewards/chosen": -8.299032211303711,
|
| 534 |
+
"rewards/margins": 2.573108196258545,
|
| 535 |
+
"rewards/rejected": -10.872140884399414,
|
| 536 |
+
"semantic_entropy": 0.9756767153739929,
|
| 537 |
+
"step": 165
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 0.37158469945355194,
|
| 541 |
+
"grad_norm": 51.034857184899195,
|
| 542 |
+
"learning_rate": 7.917240483654e-07,
|
| 543 |
+
"logits/chosen": -1.0868977308273315,
|
| 544 |
+
"logits/rejected": -1.0327848196029663,
|
| 545 |
+
"logps/chosen": -0.8292908668518066,
|
| 546 |
+
"logps/rejected": -1.0890681743621826,
|
| 547 |
+
"loss": 2.4543,
|
| 548 |
+
"rewards/accuracies": 0.762499988079071,
|
| 549 |
+
"rewards/chosen": -8.29290771484375,
|
| 550 |
+
"rewards/margins": 2.597774028778076,
|
| 551 |
+
"rewards/rejected": -10.890682220458984,
|
| 552 |
+
"semantic_entropy": 0.949859619140625,
|
| 553 |
+
"step": 170
|
| 554 |
+
},
|
| 555 |
+
{
|
| 556 |
+
"epoch": 0.3825136612021858,
|
| 557 |
+
"grad_norm": 70.18286425215469,
|
| 558 |
+
"learning_rate": 7.759950505873521e-07,
|
| 559 |
+
"logits/chosen": -1.1532905101776123,
|
| 560 |
+
"logits/rejected": -1.1286249160766602,
|
| 561 |
+
"logps/chosen": -0.8882781863212585,
|
| 562 |
+
"logps/rejected": -1.0833543539047241,
|
| 563 |
+
"loss": 2.363,
|
| 564 |
+
"rewards/accuracies": 0.675000011920929,
|
| 565 |
+
"rewards/chosen": -8.882781982421875,
|
| 566 |
+
"rewards/margins": 1.9507627487182617,
|
| 567 |
+
"rewards/rejected": -10.833544731140137,
|
| 568 |
+
"semantic_entropy": 0.9320682287216187,
|
| 569 |
+
"step": 175
|
| 570 |
+
},
|
| 571 |
+
{
|
| 572 |
+
"epoch": 0.39344262295081966,
|
| 573 |
+
"grad_norm": 57.51037018197599,
|
| 574 |
+
"learning_rate": 7.598629604744872e-07,
|
| 575 |
+
"logits/chosen": -1.1318219900131226,
|
| 576 |
+
"logits/rejected": -1.1271886825561523,
|
| 577 |
+
"logps/chosen": -0.8509464263916016,
|
| 578 |
+
"logps/rejected": -1.2260388135910034,
|
| 579 |
+
"loss": 2.2789,
|
| 580 |
+
"rewards/accuracies": 0.768750011920929,
|
| 581 |
+
"rewards/chosen": -8.509464263916016,
|
| 582 |
+
"rewards/margins": 3.7509243488311768,
|
| 583 |
+
"rewards/rejected": -12.260388374328613,
|
| 584 |
+
"semantic_entropy": 0.9382311105728149,
|
| 585 |
+
"step": 180
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 0.40437158469945356,
|
| 589 |
+
"grad_norm": 52.68844433987736,
|
| 590 |
+
"learning_rate": 7.433513390357989e-07,
|
| 591 |
+
"logits/chosen": -1.147756814956665,
|
| 592 |
+
"logits/rejected": -1.167338490486145,
|
| 593 |
+
"logps/chosen": -0.8701988458633423,
|
| 594 |
+
"logps/rejected": -1.2299163341522217,
|
| 595 |
+
"loss": 2.2886,
|
| 596 |
+
"rewards/accuracies": 0.831250011920929,
|
| 597 |
+
"rewards/chosen": -8.701990127563477,
|
| 598 |
+
"rewards/margins": 3.5971744060516357,
|
| 599 |
+
"rewards/rejected": -12.299162864685059,
|
| 600 |
+
"semantic_entropy": 0.9483189582824707,
|
| 601 |
+
"step": 185
|
| 602 |
+
},
|
| 603 |
+
{
|
| 604 |
+
"epoch": 0.41530054644808745,
|
| 605 |
+
"grad_norm": 67.57716327384249,
|
| 606 |
+
"learning_rate": 7.264843015879321e-07,
|
| 607 |
+
"logits/chosen": -1.156224012374878,
|
| 608 |
+
"logits/rejected": -1.115622878074646,
|
| 609 |
+
"logps/chosen": -0.8300007581710815,
|
| 610 |
+
"logps/rejected": -1.1879067420959473,
|
| 611 |
+
"loss": 2.3281,
|
| 612 |
+
"rewards/accuracies": 0.824999988079071,
|
| 613 |
+
"rewards/chosen": -8.300007820129395,
|
| 614 |
+
"rewards/margins": 3.579059600830078,
|
| 615 |
+
"rewards/rejected": -11.879068374633789,
|
| 616 |
+
"semantic_entropy": 0.9479382634162903,
|
| 617 |
+
"step": 190
|
| 618 |
+
},
|
| 619 |
+
{
|
| 620 |
+
"epoch": 0.4262295081967213,
|
| 621 |
+
"grad_norm": 49.46269476072645,
|
| 622 |
+
"learning_rate": 7.092864825346266e-07,
|
| 623 |
+
"logits/chosen": -1.1936676502227783,
|
| 624 |
+
"logits/rejected": -1.173877477645874,
|
| 625 |
+
"logps/chosen": -0.9850988388061523,
|
| 626 |
+
"logps/rejected": -1.3840601444244385,
|
| 627 |
+
"loss": 2.3467,
|
| 628 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 629 |
+
"rewards/chosen": -9.850988388061523,
|
| 630 |
+
"rewards/margins": 3.9896130561828613,
|
| 631 |
+
"rewards/rejected": -13.840600967407227,
|
| 632 |
+
"semantic_entropy": 0.9018501043319702,
|
| 633 |
+
"step": 195
|
| 634 |
+
},
|
| 635 |
+
{
|
| 636 |
+
"epoch": 0.4371584699453552,
|
| 637 |
+
"grad_norm": 52.77170794873292,
|
| 638 |
+
"learning_rate": 6.917829993880302e-07,
|
| 639 |
+
"logits/chosen": -1.1570117473602295,
|
| 640 |
+
"logits/rejected": -1.0735323429107666,
|
| 641 |
+
"logps/chosen": -0.884504497051239,
|
| 642 |
+
"logps/rejected": -1.2432372570037842,
|
| 643 |
+
"loss": 2.2946,
|
| 644 |
+
"rewards/accuracies": 0.78125,
|
| 645 |
+
"rewards/chosen": -8.84504508972168,
|
| 646 |
+
"rewards/margins": 3.5873265266418457,
|
| 647 |
+
"rewards/rejected": -12.432373046875,
|
| 648 |
+
"semantic_entropy": 0.9433660507202148,
|
| 649 |
+
"step": 200
|
| 650 |
+
},
|
| 651 |
+
{
|
| 652 |
+
"epoch": 0.44808743169398907,
|
| 653 |
+
"grad_norm": 44.7279934671613,
|
| 654 |
+
"learning_rate": 6.739994160844309e-07,
|
| 655 |
+
"logits/chosen": -1.11527419090271,
|
| 656 |
+
"logits/rejected": -1.1322429180145264,
|
| 657 |
+
"logps/chosen": -0.8004055023193359,
|
| 658 |
+
"logps/rejected": -1.1100412607192993,
|
| 659 |
+
"loss": 2.2536,
|
| 660 |
+
"rewards/accuracies": 0.800000011920929,
|
| 661 |
+
"rewards/chosen": -8.004055976867676,
|
| 662 |
+
"rewards/margins": 3.0963568687438965,
|
| 663 |
+
"rewards/rejected": -11.100412368774414,
|
| 664 |
+
"semantic_entropy": 0.9715192914009094,
|
| 665 |
+
"step": 205
|
| 666 |
+
},
|
| 667 |
+
{
|
| 668 |
+
"epoch": 0.45901639344262296,
|
| 669 |
+
"grad_norm": 74.86910632781094,
|
| 670 |
+
"learning_rate": 6.559617056479827e-07,
|
| 671 |
+
"logits/chosen": -1.1598584651947021,
|
| 672 |
+
"logits/rejected": -1.1630977392196655,
|
| 673 |
+
"logps/chosen": -0.8840648531913757,
|
| 674 |
+
"logps/rejected": -1.3052994012832642,
|
| 675 |
+
"loss": 2.1909,
|
| 676 |
+
"rewards/accuracies": 0.824999988079071,
|
| 677 |
+
"rewards/chosen": -8.840649604797363,
|
| 678 |
+
"rewards/margins": 4.212344169616699,
|
| 679 |
+
"rewards/rejected": -13.052993774414062,
|
| 680 |
+
"semantic_entropy": 0.9134693145751953,
|
| 681 |
+
"step": 210
|
| 682 |
+
},
|
| 683 |
+
{
|
| 684 |
+
"epoch": 0.46994535519125685,
|
| 685 |
+
"grad_norm": 75.49182823367542,
|
| 686 |
+
"learning_rate": 6.376962122569567e-07,
|
| 687 |
+
"logits/chosen": -1.1508004665374756,
|
| 688 |
+
"logits/rejected": -1.0994528532028198,
|
| 689 |
+
"logps/chosen": -0.8758481740951538,
|
| 690 |
+
"logps/rejected": -1.3154475688934326,
|
| 691 |
+
"loss": 2.2578,
|
| 692 |
+
"rewards/accuracies": 0.831250011920929,
|
| 693 |
+
"rewards/chosen": -8.758482933044434,
|
| 694 |
+
"rewards/margins": 4.395995140075684,
|
| 695 |
+
"rewards/rejected": -13.154478073120117,
|
| 696 |
+
"semantic_entropy": 0.9499451518058777,
|
| 697 |
+
"step": 215
|
| 698 |
+
},
|
| 699 |
+
{
|
| 700 |
+
"epoch": 0.4808743169398907,
|
| 701 |
+
"grad_norm": 48.75460798689588,
|
| 702 |
+
"learning_rate": 6.192296127679192e-07,
|
| 703 |
+
"logits/chosen": -1.216799020767212,
|
| 704 |
+
"logits/rejected": -1.1946138143539429,
|
| 705 |
+
"logps/chosen": -0.9245316386222839,
|
| 706 |
+
"logps/rejected": -1.2553398609161377,
|
| 707 |
+
"loss": 2.1888,
|
| 708 |
+
"rewards/accuracies": 0.762499988079071,
|
| 709 |
+
"rewards/chosen": -9.245316505432129,
|
| 710 |
+
"rewards/margins": 3.3080811500549316,
|
| 711 |
+
"rewards/rejected": -12.553396224975586,
|
| 712 |
+
"semantic_entropy": 0.920149028301239,
|
| 713 |
+
"step": 220
|
| 714 |
+
},
|
| 715 |
+
{
|
| 716 |
+
"epoch": 0.4918032786885246,
|
| 717 |
+
"grad_norm": 51.690790536277476,
|
| 718 |
+
"learning_rate": 6.005888777540319e-07,
|
| 719 |
+
"logits/chosen": -1.2422759532928467,
|
| 720 |
+
"logits/rejected": -1.2019939422607422,
|
| 721 |
+
"logps/chosen": -0.8988568186759949,
|
| 722 |
+
"logps/rejected": -1.281859278678894,
|
| 723 |
+
"loss": 2.1988,
|
| 724 |
+
"rewards/accuracies": 0.824999988079071,
|
| 725 |
+
"rewards/chosen": -8.988569259643555,
|
| 726 |
+
"rewards/margins": 3.8300251960754395,
|
| 727 |
+
"rewards/rejected": -12.81859302520752,
|
| 728 |
+
"semantic_entropy": 0.9307478070259094,
|
| 729 |
+
"step": 225
|
| 730 |
+
},
|
| 731 |
+
{
|
| 732 |
+
"epoch": 0.5027322404371585,
|
| 733 |
+
"grad_norm": 54.336949187520304,
|
| 734 |
+
"learning_rate": 5.818012321143773e-07,
|
| 735 |
+
"logits/chosen": -1.1455602645874023,
|
| 736 |
+
"logits/rejected": -1.1488093137741089,
|
| 737 |
+
"logps/chosen": -0.8657233119010925,
|
| 738 |
+
"logps/rejected": -1.2924095392227173,
|
| 739 |
+
"loss": 2.2216,
|
| 740 |
+
"rewards/accuracies": 0.824999988079071,
|
| 741 |
+
"rewards/chosen": -8.657234191894531,
|
| 742 |
+
"rewards/margins": 4.266862392425537,
|
| 743 |
+
"rewards/rejected": -12.924097061157227,
|
| 744 |
+
"semantic_entropy": 0.9354921579360962,
|
| 745 |
+
"step": 230
|
| 746 |
+
},
|
| 747 |
+
{
|
| 748 |
+
"epoch": 0.5136612021857924,
|
| 749 |
+
"grad_norm": 65.64434177548276,
|
| 750 |
+
"learning_rate": 5.628941153118388e-07,
|
| 751 |
+
"logits/chosen": -1.158027172088623,
|
| 752 |
+
"logits/rejected": -1.1202760934829712,
|
| 753 |
+
"logps/chosen": -0.8916556239128113,
|
| 754 |
+
"logps/rejected": -1.255821704864502,
|
| 755 |
+
"loss": 2.147,
|
| 756 |
+
"rewards/accuracies": 0.831250011920929,
|
| 757 |
+
"rewards/chosen": -8.916555404663086,
|
| 758 |
+
"rewards/margins": 3.641660690307617,
|
| 759 |
+
"rewards/rejected": -12.55821704864502,
|
| 760 |
+
"semantic_entropy": 0.9296435117721558,
|
| 761 |
+
"step": 235
|
| 762 |
+
},
|
| 763 |
+
{
|
| 764 |
+
"epoch": 0.5245901639344263,
|
| 765 |
+
"grad_norm": 50.4169580491656,
|
| 766 |
+
"learning_rate": 5.438951412976098e-07,
|
| 767 |
+
"logits/chosen": -1.230419397354126,
|
| 768 |
+
"logits/rejected": -1.24024498462677,
|
| 769 |
+
"logps/chosen": -0.8406194448471069,
|
| 770 |
+
"logps/rejected": -1.2802150249481201,
|
| 771 |
+
"loss": 2.048,
|
| 772 |
+
"rewards/accuracies": 0.831250011920929,
|
| 773 |
+
"rewards/chosen": -8.406194686889648,
|
| 774 |
+
"rewards/margins": 4.395954608917236,
|
| 775 |
+
"rewards/rejected": -12.802148818969727,
|
| 776 |
+
"semantic_entropy": 0.9510146379470825,
|
| 777 |
+
"step": 240
|
| 778 |
+
},
|
| 779 |
+
{
|
| 780 |
+
"epoch": 0.5355191256830601,
|
| 781 |
+
"grad_norm": 49.4313854721959,
|
| 782 |
+
"learning_rate": 5.248320581808619e-07,
|
| 783 |
+
"logits/chosen": -1.1573108434677124,
|
| 784 |
+
"logits/rejected": -1.1132813692092896,
|
| 785 |
+
"logps/chosen": -0.8796091079711914,
|
| 786 |
+
"logps/rejected": -1.3321001529693604,
|
| 787 |
+
"loss": 2.1276,
|
| 788 |
+
"rewards/accuracies": 0.8125,
|
| 789 |
+
"rewards/chosen": -8.796091079711914,
|
| 790 |
+
"rewards/margins": 4.524909973144531,
|
| 791 |
+
"rewards/rejected": -13.321001052856445,
|
| 792 |
+
"semantic_entropy": 0.9308155179023743,
|
| 793 |
+
"step": 245
|
| 794 |
+
},
|
| 795 |
+
{
|
| 796 |
+
"epoch": 0.546448087431694,
|
| 797 |
+
"grad_norm": 48.57429116569769,
|
| 798 |
+
"learning_rate": 5.057327077024744e-07,
|
| 799 |
+
"logits/chosen": -1.2085868120193481,
|
| 800 |
+
"logits/rejected": -1.1734802722930908,
|
| 801 |
+
"logps/chosen": -0.8411601185798645,
|
| 802 |
+
"logps/rejected": -1.2031333446502686,
|
| 803 |
+
"loss": 2.1896,
|
| 804 |
+
"rewards/accuracies": 0.793749988079071,
|
| 805 |
+
"rewards/chosen": -8.411601066589355,
|
| 806 |
+
"rewards/margins": 3.619731903076172,
|
| 807 |
+
"rewards/rejected": -12.031333923339844,
|
| 808 |
+
"semantic_entropy": 0.9265835881233215,
|
| 809 |
+
"step": 250
|
| 810 |
+
},
|
| 811 |
+
{
|
| 812 |
+
"epoch": 0.5573770491803278,
|
| 813 |
+
"grad_norm": 66.94044733232491,
|
| 814 |
+
"learning_rate": 4.866249845720132e-07,
|
| 815 |
+
"logits/chosen": -1.1839255094528198,
|
| 816 |
+
"logits/rejected": -1.1519018411636353,
|
| 817 |
+
"logps/chosen": -0.9079425930976868,
|
| 818 |
+
"logps/rejected": -1.3595575094223022,
|
| 819 |
+
"loss": 2.1027,
|
| 820 |
+
"rewards/accuracies": 0.800000011920929,
|
| 821 |
+
"rewards/chosen": -9.079425811767578,
|
| 822 |
+
"rewards/margins": 4.516149044036865,
|
| 823 |
+
"rewards/rejected": -13.595575332641602,
|
| 824 |
+
"semantic_entropy": 0.9337202310562134,
|
| 825 |
+
"step": 255
|
| 826 |
+
},
|
| 827 |
+
{
|
| 828 |
+
"epoch": 0.5683060109289617,
|
| 829 |
+
"grad_norm": 52.69841931787451,
|
| 830 |
+
"learning_rate": 4.675367957273505e-07,
|
| 831 |
+
"logits/chosen": -1.1522729396820068,
|
| 832 |
+
"logits/rejected": -1.1428637504577637,
|
| 833 |
+
"logps/chosen": -0.8813174366950989,
|
| 834 |
+
"logps/rejected": -1.3057996034622192,
|
| 835 |
+
"loss": 2.093,
|
| 836 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 837 |
+
"rewards/chosen": -8.813173294067383,
|
| 838 |
+
"rewards/margins": 4.2448225021362305,
|
| 839 |
+
"rewards/rejected": -13.057995796203613,
|
| 840 |
+
"semantic_entropy": 0.9142274856567383,
|
| 841 |
+
"step": 260
|
| 842 |
+
},
|
| 843 |
+
{
|
| 844 |
+
"epoch": 0.5792349726775956,
|
| 845 |
+
"grad_norm": 43.72494095619232,
|
| 846 |
+
"learning_rate": 4.4849601957642285e-07,
|
| 847 |
+
"logits/chosen": -1.1581261157989502,
|
| 848 |
+
"logits/rejected": -1.1219886541366577,
|
| 849 |
+
"logps/chosen": -0.851395308971405,
|
| 850 |
+
"logps/rejected": -1.285305380821228,
|
| 851 |
+
"loss": 2.0682,
|
| 852 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 853 |
+
"rewards/chosen": -8.51395320892334,
|
| 854 |
+
"rewards/margins": 4.339099407196045,
|
| 855 |
+
"rewards/rejected": -12.853053092956543,
|
| 856 |
+
"semantic_entropy": 0.9402335286140442,
|
| 857 |
+
"step": 265
|
| 858 |
+
},
|
| 859 |
+
{
|
| 860 |
+
"epoch": 0.5901639344262295,
|
| 861 |
+
"grad_norm": 74.3004033684429,
|
| 862 |
+
"learning_rate": 4.295304652806592e-07,
|
| 863 |
+
"logits/chosen": -1.1579556465148926,
|
| 864 |
+
"logits/rejected": -1.1343352794647217,
|
| 865 |
+
"logps/chosen": -0.8544819951057434,
|
| 866 |
+
"logps/rejected": -1.3365771770477295,
|
| 867 |
+
"loss": 2.03,
|
| 868 |
+
"rewards/accuracies": 0.800000011920929,
|
| 869 |
+
"rewards/chosen": -8.544819831848145,
|
| 870 |
+
"rewards/margins": 4.820953369140625,
|
| 871 |
+
"rewards/rejected": -13.36577320098877,
|
| 872 |
+
"semantic_entropy": 0.9244714975357056,
|
| 873 |
+
"step": 270
|
| 874 |
+
},
|
| 875 |
+
{
|
| 876 |
+
"epoch": 0.6010928961748634,
|
| 877 |
+
"grad_norm": 47.25525711691707,
|
| 878 |
+
"learning_rate": 4.106678321395433e-07,
|
| 879 |
+
"logits/chosen": -1.1198965311050415,
|
| 880 |
+
"logits/rejected": -1.0580785274505615,
|
| 881 |
+
"logps/chosen": -0.8487468957901001,
|
| 882 |
+
"logps/rejected": -1.1780356168746948,
|
| 883 |
+
"loss": 2.0436,
|
| 884 |
+
"rewards/accuracies": 0.768750011920929,
|
| 885 |
+
"rewards/chosen": -8.487468719482422,
|
| 886 |
+
"rewards/margins": 3.2928879261016846,
|
| 887 |
+
"rewards/rejected": -11.780357360839844,
|
| 888 |
+
"semantic_entropy": 0.9419177174568176,
|
| 889 |
+
"step": 275
|
| 890 |
+
},
|
| 891 |
+
{
|
| 892 |
+
"epoch": 0.6120218579234973,
|
| 893 |
+
"grad_norm": 50.516107327592174,
|
| 894 |
+
"learning_rate": 3.9193566913562915e-07,
|
| 895 |
+
"logits/chosen": -1.0839662551879883,
|
| 896 |
+
"logits/rejected": -1.088769555091858,
|
| 897 |
+
"logps/chosen": -0.8575767278671265,
|
| 898 |
+
"logps/rejected": -1.3727636337280273,
|
| 899 |
+
"loss": 2.0447,
|
| 900 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 901 |
+
"rewards/chosen": -8.57576847076416,
|
| 902 |
+
"rewards/margins": 5.151867866516113,
|
| 903 |
+
"rewards/rejected": -13.727636337280273,
|
| 904 |
+
"semantic_entropy": 0.9023324251174927,
|
| 905 |
+
"step": 280
|
| 906 |
+
},
|
| 907 |
+
{
|
| 908 |
+
"epoch": 0.6229508196721312,
|
| 909 |
+
"grad_norm": 53.0473998547049,
|
| 910 |
+
"learning_rate": 3.7336133469909623e-07,
|
| 911 |
+
"logits/chosen": -1.2294294834136963,
|
| 912 |
+
"logits/rejected": -1.1984007358551025,
|
| 913 |
+
"logps/chosen": -0.8300548791885376,
|
| 914 |
+
"logps/rejected": -1.328940987586975,
|
| 915 |
+
"loss": 1.9766,
|
| 916 |
+
"rewards/accuracies": 0.8812500238418579,
|
| 917 |
+
"rewards/chosen": -8.300548553466797,
|
| 918 |
+
"rewards/margins": 4.988862037658691,
|
| 919 |
+
"rewards/rejected": -13.289410591125488,
|
| 920 |
+
"semantic_entropy": 0.929417610168457,
|
| 921 |
+
"step": 285
|
| 922 |
+
},
|
| 923 |
+
{
|
| 924 |
+
"epoch": 0.6338797814207651,
|
| 925 |
+
"grad_norm": 90.39553328719175,
|
| 926 |
+
"learning_rate": 3.549719567506076e-07,
|
| 927 |
+
"logits/chosen": -1.153247594833374,
|
| 928 |
+
"logits/rejected": -1.1141259670257568,
|
| 929 |
+
"logps/chosen": -0.865193247795105,
|
| 930 |
+
"logps/rejected": -1.3289250135421753,
|
| 931 |
+
"loss": 2.0316,
|
| 932 |
+
"rewards/accuracies": 0.8187500238418579,
|
| 933 |
+
"rewards/chosen": -8.651933670043945,
|
| 934 |
+
"rewards/margins": 4.6373186111450195,
|
| 935 |
+
"rewards/rejected": -13.289251327514648,
|
| 936 |
+
"semantic_entropy": 0.9263349771499634,
|
| 937 |
+
"step": 290
|
| 938 |
+
},
|
| 939 |
+
{
|
| 940 |
+
"epoch": 0.644808743169399,
|
| 941 |
+
"grad_norm": 55.4386947782287,
|
| 942 |
+
"learning_rate": 3.3679439308082774e-07,
|
| 943 |
+
"logits/chosen": -1.1450589895248413,
|
| 944 |
+
"logits/rejected": -1.1411672830581665,
|
| 945 |
+
"logps/chosen": -0.8747571706771851,
|
| 946 |
+
"logps/rejected": -1.3667054176330566,
|
| 947 |
+
"loss": 1.9546,
|
| 948 |
+
"rewards/accuracies": 0.856249988079071,
|
| 949 |
+
"rewards/chosen": -8.747570991516113,
|
| 950 |
+
"rewards/margins": 4.919483661651611,
|
| 951 |
+
"rewards/rejected": -13.66705322265625,
|
| 952 |
+
"semantic_entropy": 0.9194013476371765,
|
| 953 |
+
"step": 295
|
| 954 |
+
},
|
| 955 |
+
{
|
| 956 |
+
"epoch": 0.6557377049180327,
|
| 957 |
+
"grad_norm": 45.252881799709705,
|
| 958 |
+
"learning_rate": 3.1885519212446716e-07,
|
| 959 |
+
"logits/chosen": -1.1887695789337158,
|
| 960 |
+
"logits/rejected": -1.1733474731445312,
|
| 961 |
+
"logps/chosen": -0.9111166000366211,
|
| 962 |
+
"logps/rejected": -1.3687089681625366,
|
| 963 |
+
"loss": 1.9751,
|
| 964 |
+
"rewards/accuracies": 0.762499988079071,
|
| 965 |
+
"rewards/chosen": -9.111166000366211,
|
| 966 |
+
"rewards/margins": 4.575922966003418,
|
| 967 |
+
"rewards/rejected": -13.687089920043945,
|
| 968 |
+
"semantic_entropy": 0.8905431032180786,
|
| 969 |
+
"step": 300
|
| 970 |
+
},
|
| 971 |
+
{
|
| 972 |
+
"epoch": 0.6666666666666666,
|
| 973 |
+
"grad_norm": 63.443681351494455,
|
| 974 |
+
"learning_rate": 3.0118055418614295e-07,
|
| 975 |
+
"logits/chosen": -1.1734164953231812,
|
| 976 |
+
"logits/rejected": -1.1228505373001099,
|
| 977 |
+
"logps/chosen": -0.8473879098892212,
|
| 978 |
+
"logps/rejected": -1.3491770029067993,
|
| 979 |
+
"loss": 2.0108,
|
| 980 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 981 |
+
"rewards/chosen": -8.473878860473633,
|
| 982 |
+
"rewards/margins": 5.017890453338623,
|
| 983 |
+
"rewards/rejected": -13.491769790649414,
|
| 984 |
+
"semantic_entropy": 0.9206064343452454,
|
| 985 |
+
"step": 305
|
| 986 |
+
},
|
| 987 |
+
{
|
| 988 |
+
"epoch": 0.6775956284153005,
|
| 989 |
+
"grad_norm": 48.352495630764054,
|
| 990 |
+
"learning_rate": 2.83796293174686e-07,
|
| 991 |
+
"logits/chosen": -1.117762565612793,
|
| 992 |
+
"logits/rejected": -1.1260699033737183,
|
| 993 |
+
"logps/chosen": -0.901767909526825,
|
| 994 |
+
"logps/rejected": -1.443933129310608,
|
| 995 |
+
"loss": 2.0817,
|
| 996 |
+
"rewards/accuracies": 0.824999988079071,
|
| 997 |
+
"rewards/chosen": -9.017679214477539,
|
| 998 |
+
"rewards/margins": 5.421651840209961,
|
| 999 |
+
"rewards/rejected": -14.4393310546875,
|
| 1000 |
+
"semantic_entropy": 0.9092252850532532,
|
| 1001 |
+
"step": 310
|
| 1002 |
+
},
|
| 1003 |
+
{
|
| 1004 |
+
"epoch": 0.6885245901639344,
|
| 1005 |
+
"grad_norm": 48.159593233413936,
|
| 1006 |
+
"learning_rate": 2.6672779890178046e-07,
|
| 1007 |
+
"logits/chosen": -1.1639653444290161,
|
| 1008 |
+
"logits/rejected": -1.1631653308868408,
|
| 1009 |
+
"logps/chosen": -0.9492608904838562,
|
| 1010 |
+
"logps/rejected": -1.3062164783477783,
|
| 1011 |
+
"loss": 2.0243,
|
| 1012 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 1013 |
+
"rewards/chosen": -9.492609977722168,
|
| 1014 |
+
"rewards/margins": 3.5695548057556152,
|
| 1015 |
+
"rewards/rejected": -13.062166213989258,
|
| 1016 |
+
"semantic_entropy": 0.9068315625190735,
|
| 1017 |
+
"step": 315
|
| 1018 |
+
},
|
| 1019 |
+
{
|
| 1020 |
+
"epoch": 0.6994535519125683,
|
| 1021 |
+
"grad_norm": 44.49675726254248,
|
| 1022 |
+
"learning_rate": 2.500000000000001e-07,
|
| 1023 |
+
"logits/chosen": -1.2249935865402222,
|
| 1024 |
+
"logits/rejected": -1.1779038906097412,
|
| 1025 |
+
"logps/chosen": -0.9137300252914429,
|
| 1026 |
+
"logps/rejected": -1.3746672868728638,
|
| 1027 |
+
"loss": 1.9926,
|
| 1028 |
+
"rewards/accuracies": 0.84375,
|
| 1029 |
+
"rewards/chosen": -9.137300491333008,
|
| 1030 |
+
"rewards/margins": 4.609372138977051,
|
| 1031 |
+
"rewards/rejected": -13.746671676635742,
|
| 1032 |
+
"semantic_entropy": 0.9218708872795105,
|
| 1033 |
+
"step": 320
|
| 1034 |
+
},
|
| 1035 |
+
{
|
| 1036 |
+
"epoch": 0.7103825136612022,
|
| 1037 |
+
"grad_norm": 62.06584454249009,
|
| 1038 |
+
"learning_rate": 2.3363732751439923e-07,
|
| 1039 |
+
"logits/chosen": -1.1833363771438599,
|
| 1040 |
+
"logits/rejected": -1.1670992374420166,
|
| 1041 |
+
"logps/chosen": -0.8321934938430786,
|
| 1042 |
+
"logps/rejected": -1.2726361751556396,
|
| 1043 |
+
"loss": 2.0087,
|
| 1044 |
+
"rewards/accuracies": 0.84375,
|
| 1045 |
+
"rewards/chosen": -8.32193374633789,
|
| 1046 |
+
"rewards/margins": 4.404426097869873,
|
| 1047 |
+
"rewards/rejected": -12.726360321044922,
|
| 1048 |
+
"semantic_entropy": 0.9296501278877258,
|
| 1049 |
+
"step": 325
|
| 1050 |
+
},
|
| 1051 |
+
{
|
| 1052 |
+
"epoch": 0.7213114754098361,
|
| 1053 |
+
"grad_norm": 47.582405569826236,
|
| 1054 |
+
"learning_rate": 2.1766367922083283e-07,
|
| 1055 |
+
"logits/chosen": -1.1103484630584717,
|
| 1056 |
+
"logits/rejected": -1.0834376811981201,
|
| 1057 |
+
"logps/chosen": -0.7902881503105164,
|
| 1058 |
+
"logps/rejected": -1.379204511642456,
|
| 1059 |
+
"loss": 1.9477,
|
| 1060 |
+
"rewards/accuracies": 0.824999988079071,
|
| 1061 |
+
"rewards/chosen": -7.9028825759887695,
|
| 1062 |
+
"rewards/margins": 5.889164447784424,
|
| 1063 |
+
"rewards/rejected": -13.792045593261719,
|
| 1064 |
+
"semantic_entropy": 0.9163787961006165,
|
| 1065 |
+
"step": 330
|
| 1066 |
+
},
|
| 1067 |
+
{
|
| 1068 |
+
"epoch": 0.73224043715847,
|
| 1069 |
+
"grad_norm": 51.055602602949044,
|
| 1070 |
+
"learning_rate": 2.021023847231202e-07,
|
| 1071 |
+
"logits/chosen": -1.0874385833740234,
|
| 1072 |
+
"logits/rejected": -1.0472285747528076,
|
| 1073 |
+
"logps/chosen": -0.9132916331291199,
|
| 1074 |
+
"logps/rejected": -1.3749035596847534,
|
| 1075 |
+
"loss": 1.9593,
|
| 1076 |
+
"rewards/accuracies": 0.856249988079071,
|
| 1077 |
+
"rewards/chosen": -9.132916450500488,
|
| 1078 |
+
"rewards/margins": 4.616118907928467,
|
| 1079 |
+
"rewards/rejected": -13.749035835266113,
|
| 1080 |
+
"semantic_entropy": 0.8924514651298523,
|
| 1081 |
+
"step": 335
|
| 1082 |
+
},
|
| 1083 |
+
{
|
| 1084 |
+
"epoch": 0.7431693989071039,
|
| 1085 |
+
"grad_norm": 53.38188724969566,
|
| 1086 |
+
"learning_rate": 1.869761713800254e-07,
|
| 1087 |
+
"logits/chosen": -1.1097468137741089,
|
| 1088 |
+
"logits/rejected": -1.064247965812683,
|
| 1089 |
+
"logps/chosen": -0.8977781534194946,
|
| 1090 |
+
"logps/rejected": -1.4049198627471924,
|
| 1091 |
+
"loss": 2.0313,
|
| 1092 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1093 |
+
"rewards/chosen": -8.977781295776367,
|
| 1094 |
+
"rewards/margins": 5.071417808532715,
|
| 1095 |
+
"rewards/rejected": -14.049200057983398,
|
| 1096 |
+
"semantic_entropy": 0.896017849445343,
|
| 1097 |
+
"step": 340
|
| 1098 |
+
},
|
| 1099 |
+
{
|
| 1100 |
+
"epoch": 0.7540983606557377,
|
| 1101 |
+
"grad_norm": 57.62309009254558,
|
| 1102 |
+
"learning_rate": 1.7230713111182164e-07,
|
| 1103 |
+
"logits/chosen": -1.1679933071136475,
|
| 1104 |
+
"logits/rejected": -1.167457103729248,
|
| 1105 |
+
"logps/chosen": -0.919908344745636,
|
| 1106 |
+
"logps/rejected": -1.4295835494995117,
|
| 1107 |
+
"loss": 2.0598,
|
| 1108 |
+
"rewards/accuracies": 0.824999988079071,
|
| 1109 |
+
"rewards/chosen": -9.19908332824707,
|
| 1110 |
+
"rewards/margins": 5.096750736236572,
|
| 1111 |
+
"rewards/rejected": -14.2958345413208,
|
| 1112 |
+
"semantic_entropy": 0.8877577781677246,
|
| 1113 |
+
"step": 345
|
| 1114 |
+
},
|
| 1115 |
+
{
|
| 1116 |
+
"epoch": 0.7650273224043715,
|
| 1117 |
+
"grad_norm": 56.46416368551912,
|
| 1118 |
+
"learning_rate": 1.5811668813491696e-07,
|
| 1119 |
+
"logits/chosen": -1.164217233657837,
|
| 1120 |
+
"logits/rejected": -1.140291452407837,
|
| 1121 |
+
"logps/chosen": -0.8424995541572571,
|
| 1122 |
+
"logps/rejected": -1.2386906147003174,
|
| 1123 |
+
"loss": 1.9725,
|
| 1124 |
+
"rewards/accuracies": 0.831250011920929,
|
| 1125 |
+
"rewards/chosen": -8.424995422363281,
|
| 1126 |
+
"rewards/margins": 3.96191143989563,
|
| 1127 |
+
"rewards/rejected": -12.386906623840332,
|
| 1128 |
+
"semantic_entropy": 0.9234525561332703,
|
| 1129 |
+
"step": 350
|
| 1130 |
+
},
|
| 1131 |
+
{
|
| 1132 |
+
"epoch": 0.7759562841530054,
|
| 1133 |
+
"grad_norm": 47.45687983835232,
|
| 1134 |
+
"learning_rate": 1.4442556767166369e-07,
|
| 1135 |
+
"logits/chosen": -1.1149598360061646,
|
| 1136 |
+
"logits/rejected": -1.0830504894256592,
|
| 1137 |
+
"logps/chosen": -0.8345939517021179,
|
| 1138 |
+
"logps/rejected": -1.2633658647537231,
|
| 1139 |
+
"loss": 1.9792,
|
| 1140 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 1141 |
+
"rewards/chosen": -8.345940589904785,
|
| 1142 |
+
"rewards/margins": 4.287718296051025,
|
| 1143 |
+
"rewards/rejected": -12.633659362792969,
|
| 1144 |
+
"semantic_entropy": 0.9200286865234375,
|
| 1145 |
+
"step": 355
|
| 1146 |
+
},
|
| 1147 |
+
{
|
| 1148 |
+
"epoch": 0.7868852459016393,
|
| 1149 |
+
"grad_norm": 69.3977290180621,
|
| 1150 |
+
"learning_rate": 1.312537656810549e-07,
|
| 1151 |
+
"logits/chosen": -1.0737590789794922,
|
| 1152 |
+
"logits/rejected": -1.0747511386871338,
|
| 1153 |
+
"logps/chosen": -0.9100669622421265,
|
| 1154 |
+
"logps/rejected": -1.410347580909729,
|
| 1155 |
+
"loss": 1.9621,
|
| 1156 |
+
"rewards/accuracies": 0.762499988079071,
|
| 1157 |
+
"rewards/chosen": -9.100671768188477,
|
| 1158 |
+
"rewards/margins": 5.002806186676025,
|
| 1159 |
+
"rewards/rejected": -14.103477478027344,
|
| 1160 |
+
"semantic_entropy": 0.8941181898117065,
|
| 1161 |
+
"step": 360
|
| 1162 |
+
},
|
| 1163 |
+
{
|
| 1164 |
+
"epoch": 0.7978142076502732,
|
| 1165 |
+
"grad_norm": 62.629527673002094,
|
| 1166 |
+
"learning_rate": 1.1862051965451214e-07,
|
| 1167 |
+
"logits/chosen": -1.1590697765350342,
|
| 1168 |
+
"logits/rejected": -1.1595137119293213,
|
| 1169 |
+
"logps/chosen": -0.8836385011672974,
|
| 1170 |
+
"logps/rejected": -1.3890182971954346,
|
| 1171 |
+
"loss": 1.9788,
|
| 1172 |
+
"rewards/accuracies": 0.793749988079071,
|
| 1173 |
+
"rewards/chosen": -8.836385726928711,
|
| 1174 |
+
"rewards/margins": 5.053797245025635,
|
| 1175 |
+
"rewards/rejected": -13.890182495117188,
|
| 1176 |
+
"semantic_entropy": 0.9085249900817871,
|
| 1177 |
+
"step": 365
|
| 1178 |
+
},
|
| 1179 |
+
{
|
| 1180 |
+
"epoch": 0.8087431693989071,
|
| 1181 |
+
"grad_norm": 65.41767650747809,
|
| 1182 |
+
"learning_rate": 1.0654428051942138e-07,
|
| 1183 |
+
"logits/chosen": -1.1633613109588623,
|
| 1184 |
+
"logits/rejected": -1.1319725513458252,
|
| 1185 |
+
"logps/chosen": -0.9073840975761414,
|
| 1186 |
+
"logps/rejected": -1.4780160188674927,
|
| 1187 |
+
"loss": 2.0047,
|
| 1188 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 1189 |
+
"rewards/chosen": -9.07384204864502,
|
| 1190 |
+
"rewards/margins": 5.7063188552856445,
|
| 1191 |
+
"rewards/rejected": -14.780160903930664,
|
| 1192 |
+
"semantic_entropy": 0.8961697816848755,
|
| 1193 |
+
"step": 370
|
| 1194 |
+
},
|
| 1195 |
+
{
|
| 1196 |
+
"epoch": 0.819672131147541,
|
| 1197 |
+
"grad_norm": 45.906819053741515,
|
| 1198 |
+
"learning_rate": 9.504268569144763e-08,
|
| 1199 |
+
"logits/chosen": -1.1833055019378662,
|
| 1200 |
+
"logits/rejected": -1.1207568645477295,
|
| 1201 |
+
"logps/chosen": -0.8244643211364746,
|
| 1202 |
+
"logps/rejected": -1.3258384466171265,
|
| 1203 |
+
"loss": 1.9881,
|
| 1204 |
+
"rewards/accuracies": 0.84375,
|
| 1205 |
+
"rewards/chosen": -8.244643211364746,
|
| 1206 |
+
"rewards/margins": 5.013741970062256,
|
| 1207 |
+
"rewards/rejected": -13.258384704589844,
|
| 1208 |
+
"semantic_entropy": 0.9239455461502075,
|
| 1209 |
+
"step": 375
|
| 1210 |
+
},
|
| 1211 |
+
{
|
| 1212 |
+
"epoch": 0.8306010928961749,
|
| 1213 |
+
"grad_norm": 41.61552479544314,
|
| 1214 |
+
"learning_rate": 8.413253331499049e-08,
|
| 1215 |
+
"logits/chosen": -1.067228078842163,
|
| 1216 |
+
"logits/rejected": -1.083414077758789,
|
| 1217 |
+
"logps/chosen": -0.8691232800483704,
|
| 1218 |
+
"logps/rejected": -1.315148949623108,
|
| 1219 |
+
"loss": 1.9039,
|
| 1220 |
+
"rewards/accuracies": 0.8187500238418579,
|
| 1221 |
+
"rewards/chosen": -8.69123363494873,
|
| 1222 |
+
"rewards/margins": 4.4602556228637695,
|
| 1223 |
+
"rewards/rejected": -13.151487350463867,
|
| 1224 |
+
"semantic_entropy": 0.9291045069694519,
|
| 1225 |
+
"step": 380
|
| 1226 |
+
},
|
| 1227 |
+
{
|
| 1228 |
+
"epoch": 0.8415300546448088,
|
| 1229 |
+
"grad_norm": 53.25699744382843,
|
| 1230 |
+
"learning_rate": 7.382975772939865e-08,
|
| 1231 |
+
"logits/chosen": -1.1655666828155518,
|
| 1232 |
+
"logits/rejected": -1.1497318744659424,
|
| 1233 |
+
"logps/chosen": -0.9494159817695618,
|
| 1234 |
+
"logps/rejected": -1.448509931564331,
|
| 1235 |
+
"loss": 2.061,
|
| 1236 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 1237 |
+
"rewards/chosen": -9.494159698486328,
|
| 1238 |
+
"rewards/margins": 4.990939617156982,
|
| 1239 |
+
"rewards/rejected": -14.485099792480469,
|
| 1240 |
+
"semantic_entropy": 0.9152476191520691,
|
| 1241 |
+
"step": 385
|
| 1242 |
+
},
|
| 1243 |
+
{
|
| 1244 |
+
"epoch": 0.8524590163934426,
|
| 1245 |
+
"grad_norm": 70.12618215621923,
|
| 1246 |
+
"learning_rate": 6.414940619677734e-08,
|
| 1247 |
+
"logits/chosen": -1.157553791999817,
|
| 1248 |
+
"logits/rejected": -1.1328723430633545,
|
| 1249 |
+
"logps/chosen": -0.8517138361930847,
|
| 1250 |
+
"logps/rejected": -1.4659974575042725,
|
| 1251 |
+
"loss": 1.9241,
|
| 1252 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1253 |
+
"rewards/chosen": -8.51713752746582,
|
| 1254 |
+
"rewards/margins": 6.14283561706543,
|
| 1255 |
+
"rewards/rejected": -14.659975051879883,
|
| 1256 |
+
"semantic_entropy": 0.9240752458572388,
|
| 1257 |
+
"step": 390
|
| 1258 |
+
},
|
| 1259 |
+
{
|
| 1260 |
+
"epoch": 0.8633879781420765,
|
| 1261 |
+
"grad_norm": 55.71292222513987,
|
| 1262 |
+
"learning_rate": 5.5105616925376296e-08,
|
| 1263 |
+
"logits/chosen": -1.134119987487793,
|
| 1264 |
+
"logits/rejected": -1.1163135766983032,
|
| 1265 |
+
"logps/chosen": -0.9063531160354614,
|
| 1266 |
+
"logps/rejected": -1.3211853504180908,
|
| 1267 |
+
"loss": 1.9119,
|
| 1268 |
+
"rewards/accuracies": 0.8125,
|
| 1269 |
+
"rewards/chosen": -9.063530921936035,
|
| 1270 |
+
"rewards/margins": 4.148322105407715,
|
| 1271 |
+
"rewards/rejected": -13.21185302734375,
|
| 1272 |
+
"semantic_entropy": 0.9249491691589355,
|
| 1273 |
+
"step": 395
|
| 1274 |
+
},
|
| 1275 |
+
{
|
| 1276 |
+
"epoch": 0.8743169398907104,
|
| 1277 |
+
"grad_norm": 55.00548916953222,
|
| 1278 |
+
"learning_rate": 4.6711598420656976e-08,
|
| 1279 |
+
"logits/chosen": -1.0777003765106201,
|
| 1280 |
+
"logits/rejected": -1.0427886247634888,
|
| 1281 |
+
"logps/chosen": -0.8970823287963867,
|
| 1282 |
+
"logps/rejected": -1.4476853609085083,
|
| 1283 |
+
"loss": 1.9268,
|
| 1284 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1285 |
+
"rewards/chosen": -8.970823287963867,
|
| 1286 |
+
"rewards/margins": 5.506030082702637,
|
| 1287 |
+
"rewards/rejected": -14.476852416992188,
|
| 1288 |
+
"semantic_entropy": 0.8907070159912109,
|
| 1289 |
+
"step": 400
|
| 1290 |
+
},
|
| 1291 |
+
{
|
| 1292 |
+
"epoch": 0.8743169398907104,
|
| 1293 |
+
"eval_logits/chosen": -1.3707057237625122,
|
| 1294 |
+
"eval_logits/rejected": -1.3261345624923706,
|
| 1295 |
+
"eval_logps/chosen": -0.8602119088172913,
|
| 1296 |
+
"eval_logps/rejected": -1.391231656074524,
|
| 1297 |
+
"eval_loss": 1.9562536478042603,
|
| 1298 |
+
"eval_rewards/accuracies": 0.8433734774589539,
|
| 1299 |
+
"eval_rewards/chosen": -8.602119445800781,
|
| 1300 |
+
"eval_rewards/margins": 5.31019926071167,
|
| 1301 |
+
"eval_rewards/rejected": -13.912318229675293,
|
| 1302 |
+
"eval_runtime": 46.5338,
|
| 1303 |
+
"eval_samples_per_second": 28.323,
|
| 1304 |
+
"eval_semantic_entropy": 0.9091164469718933,
|
| 1305 |
+
"eval_steps_per_second": 1.784,
|
| 1306 |
+
"step": 400
|
| 1307 |
+
},
|
| 1308 |
+
{
|
| 1309 |
+
"epoch": 0.8852459016393442,
|
| 1310 |
+
"grad_norm": 56.49464800000407,
|
| 1311 |
+
"learning_rate": 3.897961019419516e-08,
|
| 1312 |
+
"logits/chosen": -1.1149052381515503,
|
| 1313 |
+
"logits/rejected": -1.0388867855072021,
|
| 1314 |
+
"logps/chosen": -0.8156052827835083,
|
| 1315 |
+
"logps/rejected": -1.2854751348495483,
|
| 1316 |
+
"loss": 1.9256,
|
| 1317 |
+
"rewards/accuracies": 0.824999988079071,
|
| 1318 |
+
"rewards/chosen": -8.156051635742188,
|
| 1319 |
+
"rewards/margins": 4.698698997497559,
|
| 1320 |
+
"rewards/rejected": -12.854751586914062,
|
| 1321 |
+
"semantic_entropy": 0.9291986227035522,
|
| 1322 |
+
"step": 405
|
| 1323 |
+
},
|
| 1324 |
+
{
|
| 1325 |
+
"epoch": 0.8961748633879781,
|
| 1326 |
+
"grad_norm": 50.962161952757214,
|
| 1327 |
+
"learning_rate": 3.192094485859526e-08,
|
| 1328 |
+
"logits/chosen": -1.117832064628601,
|
| 1329 |
+
"logits/rejected": -1.1457985639572144,
|
| 1330 |
+
"logps/chosen": -0.8877307772636414,
|
| 1331 |
+
"logps/rejected": -1.4623702764511108,
|
| 1332 |
+
"loss": 1.9062,
|
| 1333 |
+
"rewards/accuracies": 0.856249988079071,
|
| 1334 |
+
"rewards/chosen": -8.877307891845703,
|
| 1335 |
+
"rewards/margins": 5.746395587921143,
|
| 1336 |
+
"rewards/rejected": -14.62370491027832,
|
| 1337 |
+
"semantic_entropy": 0.8888769149780273,
|
| 1338 |
+
"step": 410
|
| 1339 |
+
},
|
| 1340 |
+
{
|
| 1341 |
+
"epoch": 0.907103825136612,
|
| 1342 |
+
"grad_norm": 54.776769647326134,
|
| 1343 |
+
"learning_rate": 2.5545911634565265e-08,
|
| 1344 |
+
"logits/chosen": -1.1549928188323975,
|
| 1345 |
+
"logits/rejected": -1.1519380807876587,
|
| 1346 |
+
"logps/chosen": -0.8584114909172058,
|
| 1347 |
+
"logps/rejected": -1.409808874130249,
|
| 1348 |
+
"loss": 1.9933,
|
| 1349 |
+
"rewards/accuracies": 0.856249988079071,
|
| 1350 |
+
"rewards/chosen": -8.584114074707031,
|
| 1351 |
+
"rewards/margins": 5.513972282409668,
|
| 1352 |
+
"rewards/rejected": -14.0980863571167,
|
| 1353 |
+
"semantic_entropy": 0.9221046566963196,
|
| 1354 |
+
"step": 415
|
| 1355 |
+
},
|
| 1356 |
+
{
|
| 1357 |
+
"epoch": 0.9180327868852459,
|
| 1358 |
+
"grad_norm": 57.98778672833022,
|
| 1359 |
+
"learning_rate": 1.9863821294241522e-08,
|
| 1360 |
+
"logits/chosen": -1.149745225906372,
|
| 1361 |
+
"logits/rejected": -1.125451683998108,
|
| 1362 |
+
"logps/chosen": -0.8198378682136536,
|
| 1363 |
+
"logps/rejected": -1.4263782501220703,
|
| 1364 |
+
"loss": 1.8964,
|
| 1365 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1366 |
+
"rewards/chosen": -8.198378562927246,
|
| 1367 |
+
"rewards/margins": 6.065403938293457,
|
| 1368 |
+
"rewards/rejected": -14.26378345489502,
|
| 1369 |
+
"semantic_entropy": 0.9195442199707031,
|
| 1370 |
+
"step": 420
|
| 1371 |
+
},
|
| 1372 |
+
{
|
| 1373 |
+
"epoch": 0.9289617486338798,
|
| 1374 |
+
"grad_norm": 52.73787429786629,
|
| 1375 |
+
"learning_rate": 1.4882972562753615e-08,
|
| 1376 |
+
"logits/chosen": -1.1456595659255981,
|
| 1377 |
+
"logits/rejected": -1.1313962936401367,
|
| 1378 |
+
"logps/chosen": -0.9214321374893188,
|
| 1379 |
+
"logps/rejected": -1.5123398303985596,
|
| 1380 |
+
"loss": 1.99,
|
| 1381 |
+
"rewards/accuracies": 0.8187500238418579,
|
| 1382 |
+
"rewards/chosen": -9.21432113647461,
|
| 1383 |
+
"rewards/margins": 5.90907621383667,
|
| 1384 |
+
"rewards/rejected": -15.123395919799805,
|
| 1385 |
+
"semantic_entropy": 0.8927605748176575,
|
| 1386 |
+
"step": 425
|
| 1387 |
+
},
|
| 1388 |
+
{
|
| 1389 |
+
"epoch": 0.9398907103825137,
|
| 1390 |
+
"grad_norm": 44.1912015437529,
|
| 1391 |
+
"learning_rate": 1.0610639997888915e-08,
|
| 1392 |
+
"logits/chosen": -1.081146001815796,
|
| 1393 |
+
"logits/rejected": -1.075714111328125,
|
| 1394 |
+
"logps/chosen": -0.7971862554550171,
|
| 1395 |
+
"logps/rejected": -1.3901129961013794,
|
| 1396 |
+
"loss": 1.7868,
|
| 1397 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1398 |
+
"rewards/chosen": -7.971863746643066,
|
| 1399 |
+
"rewards/margins": 5.92926549911499,
|
| 1400 |
+
"rewards/rejected": -13.901128768920898,
|
| 1401 |
+
"semantic_entropy": 0.9434274435043335,
|
| 1402 |
+
"step": 430
|
| 1403 |
+
},
|
| 1404 |
+
{
|
| 1405 |
+
"epoch": 0.9508196721311475,
|
| 1406 |
+
"grad_norm": 61.224577711409715,
|
| 1407 |
+
"learning_rate": 7.053063365559997e-09,
|
| 1408 |
+
"logits/chosen": -1.1544030904769897,
|
| 1409 |
+
"logits/rejected": -1.1826975345611572,
|
| 1410 |
+
"logps/chosen": -0.8376800417900085,
|
| 1411 |
+
"logps/rejected": -1.3729816675186157,
|
| 1412 |
+
"loss": 1.9375,
|
| 1413 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1414 |
+
"rewards/chosen": -8.376801490783691,
|
| 1415 |
+
"rewards/margins": 5.353014945983887,
|
| 1416 |
+
"rewards/rejected": -13.729815483093262,
|
| 1417 |
+
"semantic_entropy": 0.9147855639457703,
|
| 1418 |
+
"step": 435
|
| 1419 |
+
},
|
| 1420 |
+
{
|
| 1421 |
+
"epoch": 0.9617486338797814,
|
| 1422 |
+
"grad_norm": 53.9815806254924,
|
| 1423 |
+
"learning_rate": 4.215438526591064e-09,
|
| 1424 |
+
"logits/chosen": -1.110844612121582,
|
| 1425 |
+
"logits/rejected": -1.0655571222305298,
|
| 1426 |
+
"logps/chosen": -0.9438391923904419,
|
| 1427 |
+
"logps/rejected": -1.4074232578277588,
|
| 1428 |
+
"loss": 1.8879,
|
| 1429 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1430 |
+
"rewards/chosen": -9.438390731811523,
|
| 1431 |
+
"rewards/margins": 4.635839939117432,
|
| 1432 |
+
"rewards/rejected": -14.074231147766113,
|
| 1433 |
+
"semantic_entropy": 0.879165530204773,
|
| 1434 |
+
"step": 440
|
| 1435 |
+
},
|
| 1436 |
+
{
|
| 1437 |
+
"epoch": 0.9726775956284153,
|
| 1438 |
+
"grad_norm": 57.824670175057065,
|
| 1439 |
+
"learning_rate": 2.1019098481337426e-09,
|
| 1440 |
+
"logits/chosen": -1.1616451740264893,
|
| 1441 |
+
"logits/rejected": -1.1365407705307007,
|
| 1442 |
+
"logps/chosen": -0.8202412724494934,
|
| 1443 |
+
"logps/rejected": -1.3376966714859009,
|
| 1444 |
+
"loss": 1.9265,
|
| 1445 |
+
"rewards/accuracies": 0.8125,
|
| 1446 |
+
"rewards/chosen": -8.202413558959961,
|
| 1447 |
+
"rewards/margins": 5.174552917480469,
|
| 1448 |
+
"rewards/rejected": -13.376965522766113,
|
| 1449 |
+
"semantic_entropy": 0.9334556460380554,
|
| 1450 |
+
"step": 445
|
| 1451 |
+
},
|
| 1452 |
+
{
|
| 1453 |
+
"epoch": 0.9836065573770492,
|
| 1454 |
+
"grad_norm": 55.05572574163813,
|
| 1455 |
+
"learning_rate": 7.155641507955445e-10,
|
| 1456 |
+
"logits/chosen": -1.076629877090454,
|
| 1457 |
+
"logits/rejected": -1.0707811117172241,
|
| 1458 |
+
"logps/chosen": -0.882774829864502,
|
| 1459 |
+
"logps/rejected": -1.3867313861846924,
|
| 1460 |
+
"loss": 1.9798,
|
| 1461 |
+
"rewards/accuracies": 0.8187500238418579,
|
| 1462 |
+
"rewards/chosen": -8.82774829864502,
|
| 1463 |
+
"rewards/margins": 5.039565563201904,
|
| 1464 |
+
"rewards/rejected": -13.86731243133545,
|
| 1465 |
+
"semantic_entropy": 0.910732626914978,
|
| 1466 |
+
"step": 450
|
| 1467 |
+
},
|
| 1468 |
+
{
|
| 1469 |
+
"epoch": 0.994535519125683,
|
| 1470 |
+
"grad_norm": 64.69250426850014,
|
| 1471 |
+
"learning_rate": 5.842620032053824e-11,
|
| 1472 |
+
"logits/chosen": -1.0871715545654297,
|
| 1473 |
+
"logits/rejected": -1.0820033550262451,
|
| 1474 |
+
"logps/chosen": -0.9286049604415894,
|
| 1475 |
+
"logps/rejected": -1.3075941801071167,
|
| 1476 |
+
"loss": 2.053,
|
| 1477 |
+
"rewards/accuracies": 0.78125,
|
| 1478 |
+
"rewards/chosen": -9.286049842834473,
|
| 1479 |
+
"rewards/margins": 3.7898917198181152,
|
| 1480 |
+
"rewards/rejected": -13.07594108581543,
|
| 1481 |
+
"semantic_entropy": 0.9091150164604187,
|
| 1482 |
+
"step": 455
|
| 1483 |
+
},
|
| 1484 |
+
{
|
| 1485 |
+
"epoch": 0.9989071038251366,
|
| 1486 |
+
"step": 457,
|
| 1487 |
+
"total_flos": 0.0,
|
| 1488 |
+
"train_loss": 2.3507381581112385,
|
| 1489 |
+
"train_runtime": 7991.7933,
|
| 1490 |
+
"train_samples_per_second": 7.327,
|
| 1491 |
+
"train_steps_per_second": 0.057
|
| 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 |
+
}
|