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_qfUNL_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_qfUNL_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: 2.0936
|
| 22 |
+
- Rewards/chosen: -5.3144
|
| 23 |
+
- Rewards/rejected: -7.1419
|
| 24 |
+
- Rewards/accuracies: 0.7771
|
| 25 |
+
- Rewards/margins: 1.8275
|
| 26 |
+
- Logps/rejected: -0.7142
|
| 27 |
+
- Logps/chosen: -0.5314
|
| 28 |
+
- Logits/rejected: -1.3326
|
| 29 |
+
- Logits/chosen: -1.3731
|
| 30 |
+
- Semantic Entropy: 0.9976
|
| 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 |
+
| 2.0424 | 0.8743 | 400 | 2.0936 | -5.3144 | -7.1419 | 0.7771 | 1.8275 | -0.7142 | -0.5314 | -1.3326 | -1.3731 | 0.9976 |
|
| 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": 2.4655840506438875,
|
| 5 |
+
"train_runtime": 5955.1851,
|
| 6 |
+
"train_samples": 58558,
|
| 7 |
+
"train_samples_per_second": 9.833,
|
| 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": 2.4655840506438875,
|
| 5 |
+
"train_runtime": 5955.1851,
|
| 6 |
+
"train_samples": 58558,
|
| 7 |
+
"train_samples_per_second": 9.833,
|
| 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": 315.2451683922351,
|
| 14 |
+
"learning_rate": 1.0869565217391303e-07,
|
| 15 |
+
"logits/chosen": -1.0065257549285889,
|
| 16 |
+
"logits/rejected": -1.0008176565170288,
|
| 17 |
+
"logps/chosen": -0.28065255284309387,
|
| 18 |
+
"logps/rejected": -0.28539329767227173,
|
| 19 |
+
"loss": 3.4114,
|
| 20 |
+
"rewards/accuracies": 0.53125,
|
| 21 |
+
"rewards/chosen": -2.806525468826294,
|
| 22 |
+
"rewards/margins": 0.04740738496184349,
|
| 23 |
+
"rewards/rejected": -2.8539328575134277,
|
| 24 |
+
"semantic_entropy": 0.7513969540596008,
|
| 25 |
+
"step": 5
|
| 26 |
+
},
|
| 27 |
+
{
|
| 28 |
+
"epoch": 0.02185792349726776,
|
| 29 |
+
"grad_norm": 181.1484663719842,
|
| 30 |
+
"learning_rate": 2.1739130434782607e-07,
|
| 31 |
+
"logits/chosen": -1.0534369945526123,
|
| 32 |
+
"logits/rejected": -1.0029994249343872,
|
| 33 |
+
"logps/chosen": -0.2570807933807373,
|
| 34 |
+
"logps/rejected": -0.27113229036331177,
|
| 35 |
+
"loss": 3.3911,
|
| 36 |
+
"rewards/accuracies": 0.543749988079071,
|
| 37 |
+
"rewards/chosen": -2.570807933807373,
|
| 38 |
+
"rewards/margins": 0.14051488041877747,
|
| 39 |
+
"rewards/rejected": -2.711322784423828,
|
| 40 |
+
"semantic_entropy": 0.710273802280426,
|
| 41 |
+
"step": 10
|
| 42 |
+
},
|
| 43 |
+
{
|
| 44 |
+
"epoch": 0.03278688524590164,
|
| 45 |
+
"grad_norm": 178.65067568018998,
|
| 46 |
+
"learning_rate": 3.260869565217391e-07,
|
| 47 |
+
"logits/chosen": -1.0082308053970337,
|
| 48 |
+
"logits/rejected": -0.9609392285346985,
|
| 49 |
+
"logps/chosen": -0.26744094491004944,
|
| 50 |
+
"logps/rejected": -0.27332359552383423,
|
| 51 |
+
"loss": 3.3533,
|
| 52 |
+
"rewards/accuracies": 0.512499988079071,
|
| 53 |
+
"rewards/chosen": -2.6744089126586914,
|
| 54 |
+
"rewards/margins": 0.05882669612765312,
|
| 55 |
+
"rewards/rejected": -2.7332358360290527,
|
| 56 |
+
"semantic_entropy": 0.7273439168930054,
|
| 57 |
+
"step": 15
|
| 58 |
+
},
|
| 59 |
+
{
|
| 60 |
+
"epoch": 0.04371584699453552,
|
| 61 |
+
"grad_norm": 258.46792679447157,
|
| 62 |
+
"learning_rate": 4.3478260869565214e-07,
|
| 63 |
+
"logits/chosen": -0.9462105631828308,
|
| 64 |
+
"logits/rejected": -0.8957524299621582,
|
| 65 |
+
"logps/chosen": -0.27257752418518066,
|
| 66 |
+
"logps/rejected": -0.2848864197731018,
|
| 67 |
+
"loss": 3.3976,
|
| 68 |
+
"rewards/accuracies": 0.5687500238418579,
|
| 69 |
+
"rewards/chosen": -2.7257750034332275,
|
| 70 |
+
"rewards/margins": 0.12308906018733978,
|
| 71 |
+
"rewards/rejected": -2.8488640785217285,
|
| 72 |
+
"semantic_entropy": 0.7455072999000549,
|
| 73 |
+
"step": 20
|
| 74 |
+
},
|
| 75 |
+
{
|
| 76 |
+
"epoch": 0.0546448087431694,
|
| 77 |
+
"grad_norm": 273.0706988791117,
|
| 78 |
+
"learning_rate": 5.434782608695652e-07,
|
| 79 |
+
"logits/chosen": -0.9422909617424011,
|
| 80 |
+
"logits/rejected": -0.8697713613510132,
|
| 81 |
+
"logps/chosen": -0.2761459946632385,
|
| 82 |
+
"logps/rejected": -0.2941877543926239,
|
| 83 |
+
"loss": 3.346,
|
| 84 |
+
"rewards/accuracies": 0.581250011920929,
|
| 85 |
+
"rewards/chosen": -2.761460065841675,
|
| 86 |
+
"rewards/margins": 0.18041765689849854,
|
| 87 |
+
"rewards/rejected": -2.9418773651123047,
|
| 88 |
+
"semantic_entropy": 0.7553174495697021,
|
| 89 |
+
"step": 25
|
| 90 |
+
},
|
| 91 |
+
{
|
| 92 |
+
"epoch": 0.06557377049180328,
|
| 93 |
+
"grad_norm": 260.6158887162735,
|
| 94 |
+
"learning_rate": 6.521739130434782e-07,
|
| 95 |
+
"logits/chosen": -1.0548616647720337,
|
| 96 |
+
"logits/rejected": -0.9892600774765015,
|
| 97 |
+
"logps/chosen": -0.2682558596134186,
|
| 98 |
+
"logps/rejected": -0.284037709236145,
|
| 99 |
+
"loss": 3.4058,
|
| 100 |
+
"rewards/accuracies": 0.5062500238418579,
|
| 101 |
+
"rewards/chosen": -2.682558536529541,
|
| 102 |
+
"rewards/margins": 0.1578185111284256,
|
| 103 |
+
"rewards/rejected": -2.8403773307800293,
|
| 104 |
+
"semantic_entropy": 0.7250551581382751,
|
| 105 |
+
"step": 30
|
| 106 |
+
},
|
| 107 |
+
{
|
| 108 |
+
"epoch": 0.07650273224043716,
|
| 109 |
+
"grad_norm": 124.4121289092538,
|
| 110 |
+
"learning_rate": 7.608695652173913e-07,
|
| 111 |
+
"logits/chosen": -1.0096337795257568,
|
| 112 |
+
"logits/rejected": -0.9423675537109375,
|
| 113 |
+
"logps/chosen": -0.2600244879722595,
|
| 114 |
+
"logps/rejected": -0.27900081872940063,
|
| 115 |
+
"loss": 3.313,
|
| 116 |
+
"rewards/accuracies": 0.4937500059604645,
|
| 117 |
+
"rewards/chosen": -2.6002449989318848,
|
| 118 |
+
"rewards/margins": 0.18976299464702606,
|
| 119 |
+
"rewards/rejected": -2.790008068084717,
|
| 120 |
+
"semantic_entropy": 0.7209498882293701,
|
| 121 |
+
"step": 35
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 0.08743169398907104,
|
| 125 |
+
"grad_norm": 147.2873491379567,
|
| 126 |
+
"learning_rate": 8.695652173913043e-07,
|
| 127 |
+
"logits/chosen": -0.9600120782852173,
|
| 128 |
+
"logits/rejected": -0.8983286619186401,
|
| 129 |
+
"logps/chosen": -0.2835314869880676,
|
| 130 |
+
"logps/rejected": -0.2980334460735321,
|
| 131 |
+
"loss": 3.454,
|
| 132 |
+
"rewards/accuracies": 0.512499988079071,
|
| 133 |
+
"rewards/chosen": -2.835314989089966,
|
| 134 |
+
"rewards/margins": 0.1450195610523224,
|
| 135 |
+
"rewards/rejected": -2.980334520339966,
|
| 136 |
+
"semantic_entropy": 0.7609063386917114,
|
| 137 |
+
"step": 40
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 0.09836065573770492,
|
| 141 |
+
"grad_norm": 93.94573860721647,
|
| 142 |
+
"learning_rate": 9.782608695652173e-07,
|
| 143 |
+
"logits/chosen": -1.0171349048614502,
|
| 144 |
+
"logits/rejected": -0.9333709478378296,
|
| 145 |
+
"logps/chosen": -0.28658169507980347,
|
| 146 |
+
"logps/rejected": -0.3058907389640808,
|
| 147 |
+
"loss": 3.1741,
|
| 148 |
+
"rewards/accuracies": 0.5062500238418579,
|
| 149 |
+
"rewards/chosen": -2.865816831588745,
|
| 150 |
+
"rewards/margins": 0.1930905282497406,
|
| 151 |
+
"rewards/rejected": -3.0589072704315186,
|
| 152 |
+
"semantic_entropy": 0.7610034346580505,
|
| 153 |
+
"step": 45
|
| 154 |
+
},
|
| 155 |
+
{
|
| 156 |
+
"epoch": 0.1092896174863388,
|
| 157 |
+
"grad_norm": 286.4609615603791,
|
| 158 |
+
"learning_rate": 9.997663088532014e-07,
|
| 159 |
+
"logits/chosen": -0.9543835520744324,
|
| 160 |
+
"logits/rejected": -0.8730956315994263,
|
| 161 |
+
"logps/chosen": -0.2823755145072937,
|
| 162 |
+
"logps/rejected": -0.2902544140815735,
|
| 163 |
+
"loss": 3.3278,
|
| 164 |
+
"rewards/accuracies": 0.48124998807907104,
|
| 165 |
+
"rewards/chosen": -2.8237555027008057,
|
| 166 |
+
"rewards/margins": 0.07878823578357697,
|
| 167 |
+
"rewards/rejected": -2.902543544769287,
|
| 168 |
+
"semantic_entropy": 0.7529318332672119,
|
| 169 |
+
"step": 50
|
| 170 |
+
},
|
| 171 |
+
{
|
| 172 |
+
"epoch": 0.12021857923497267,
|
| 173 |
+
"grad_norm": 122.74091082377944,
|
| 174 |
+
"learning_rate": 9.98817312944725e-07,
|
| 175 |
+
"logits/chosen": -0.9809161424636841,
|
| 176 |
+
"logits/rejected": -0.8649328947067261,
|
| 177 |
+
"logps/chosen": -0.2803560495376587,
|
| 178 |
+
"logps/rejected": -0.3149644732475281,
|
| 179 |
+
"loss": 3.1758,
|
| 180 |
+
"rewards/accuracies": 0.543749988079071,
|
| 181 |
+
"rewards/chosen": -2.803560256958008,
|
| 182 |
+
"rewards/margins": 0.3460845947265625,
|
| 183 |
+
"rewards/rejected": -3.1496450901031494,
|
| 184 |
+
"semantic_entropy": 0.7631497979164124,
|
| 185 |
+
"step": 55
|
| 186 |
+
},
|
| 187 |
+
{
|
| 188 |
+
"epoch": 0.13114754098360656,
|
| 189 |
+
"grad_norm": 87.3444735452583,
|
| 190 |
+
"learning_rate": 9.971397915250336e-07,
|
| 191 |
+
"logits/chosen": -1.033070683479309,
|
| 192 |
+
"logits/rejected": -0.9856836199760437,
|
| 193 |
+
"logps/chosen": -0.2708079218864441,
|
| 194 |
+
"logps/rejected": -0.31113672256469727,
|
| 195 |
+
"loss": 3.0283,
|
| 196 |
+
"rewards/accuracies": 0.59375,
|
| 197 |
+
"rewards/chosen": -2.7080790996551514,
|
| 198 |
+
"rewards/margins": 0.40328770875930786,
|
| 199 |
+
"rewards/rejected": -3.1113669872283936,
|
| 200 |
+
"semantic_entropy": 0.7593907713890076,
|
| 201 |
+
"step": 60
|
| 202 |
+
},
|
| 203 |
+
{
|
| 204 |
+
"epoch": 0.14207650273224043,
|
| 205 |
+
"grad_norm": 162.53486819318002,
|
| 206 |
+
"learning_rate": 9.94736194623663e-07,
|
| 207 |
+
"logits/chosen": -1.0021493434906006,
|
| 208 |
+
"logits/rejected": -0.9318512082099915,
|
| 209 |
+
"logps/chosen": -0.31085288524627686,
|
| 210 |
+
"logps/rejected": -0.3405633866786957,
|
| 211 |
+
"loss": 3.2559,
|
| 212 |
+
"rewards/accuracies": 0.5375000238418579,
|
| 213 |
+
"rewards/chosen": -3.1085288524627686,
|
| 214 |
+
"rewards/margins": 0.29710477590560913,
|
| 215 |
+
"rewards/rejected": -3.4056334495544434,
|
| 216 |
+
"semantic_entropy": 0.8082467317581177,
|
| 217 |
+
"step": 65
|
| 218 |
+
},
|
| 219 |
+
{
|
| 220 |
+
"epoch": 0.15300546448087432,
|
| 221 |
+
"grad_norm": 348.65729079716573,
|
| 222 |
+
"learning_rate": 9.916100327075037e-07,
|
| 223 |
+
"logits/chosen": -0.9440506100654602,
|
| 224 |
+
"logits/rejected": -0.9250672459602356,
|
| 225 |
+
"logps/chosen": -0.29765281081199646,
|
| 226 |
+
"logps/rejected": -0.3226909935474396,
|
| 227 |
+
"loss": 3.0703,
|
| 228 |
+
"rewards/accuracies": 0.5687500238418579,
|
| 229 |
+
"rewards/chosen": -2.9765281677246094,
|
| 230 |
+
"rewards/margins": 0.2503815293312073,
|
| 231 |
+
"rewards/rejected": -3.22691011428833,
|
| 232 |
+
"semantic_entropy": 0.7739163041114807,
|
| 233 |
+
"step": 70
|
| 234 |
+
},
|
| 235 |
+
{
|
| 236 |
+
"epoch": 0.16393442622950818,
|
| 237 |
+
"grad_norm": 78.90080101900976,
|
| 238 |
+
"learning_rate": 9.877658715537428e-07,
|
| 239 |
+
"logits/chosen": -0.9282974004745483,
|
| 240 |
+
"logits/rejected": -0.912223219871521,
|
| 241 |
+
"logps/chosen": -0.3213742971420288,
|
| 242 |
+
"logps/rejected": -0.351571649312973,
|
| 243 |
+
"loss": 3.1657,
|
| 244 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 245 |
+
"rewards/chosen": -3.213742733001709,
|
| 246 |
+
"rewards/margins": 0.30197376012802124,
|
| 247 |
+
"rewards/rejected": -3.515717029571533,
|
| 248 |
+
"semantic_entropy": 0.8054312467575073,
|
| 249 |
+
"step": 75
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 0.17486338797814208,
|
| 253 |
+
"grad_norm": 86.95229439852206,
|
| 254 |
+
"learning_rate": 9.832093255815216e-07,
|
| 255 |
+
"logits/chosen": -0.9330040812492371,
|
| 256 |
+
"logits/rejected": -0.8699474334716797,
|
| 257 |
+
"logps/chosen": -0.3141597807407379,
|
| 258 |
+
"logps/rejected": -0.3350343406200409,
|
| 259 |
+
"loss": 3.1277,
|
| 260 |
+
"rewards/accuracies": 0.612500011920929,
|
| 261 |
+
"rewards/chosen": -3.1415977478027344,
|
| 262 |
+
"rewards/margins": 0.20874571800231934,
|
| 263 |
+
"rewards/rejected": -3.350343704223633,
|
| 264 |
+
"semantic_entropy": 0.7814024686813354,
|
| 265 |
+
"step": 80
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.18579234972677597,
|
| 269 |
+
"grad_norm": 129.31831333568232,
|
| 270 |
+
"learning_rate": 9.779470496520441e-07,
|
| 271 |
+
"logits/chosen": -0.9329907298088074,
|
| 272 |
+
"logits/rejected": -0.8826324343681335,
|
| 273 |
+
"logps/chosen": -0.31158381700515747,
|
| 274 |
+
"logps/rejected": -0.3667066693305969,
|
| 275 |
+
"loss": 2.9946,
|
| 276 |
+
"rewards/accuracies": 0.643750011920929,
|
| 277 |
+
"rewards/chosen": -3.1158382892608643,
|
| 278 |
+
"rewards/margins": 0.5512284636497498,
|
| 279 |
+
"rewards/rejected": -3.667067050933838,
|
| 280 |
+
"semantic_entropy": 0.802832305431366,
|
| 281 |
+
"step": 85
|
| 282 |
+
},
|
| 283 |
+
{
|
| 284 |
+
"epoch": 0.19672131147540983,
|
| 285 |
+
"grad_norm": 101.69393552819945,
|
| 286 |
+
"learning_rate": 9.719867293491144e-07,
|
| 287 |
+
"logits/chosen": -1.0058772563934326,
|
| 288 |
+
"logits/rejected": -0.9232236742973328,
|
| 289 |
+
"logps/chosen": -0.33966144919395447,
|
| 290 |
+
"logps/rejected": -0.37590381503105164,
|
| 291 |
+
"loss": 3.0893,
|
| 292 |
+
"rewards/accuracies": 0.5625,
|
| 293 |
+
"rewards/chosen": -3.3966145515441895,
|
| 294 |
+
"rewards/margins": 0.3624236285686493,
|
| 295 |
+
"rewards/rejected": -3.759038209915161,
|
| 296 |
+
"semantic_entropy": 0.8507563471794128,
|
| 297 |
+
"step": 90
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"epoch": 0.20765027322404372,
|
| 301 |
+
"grad_norm": 91.30230852825771,
|
| 302 |
+
"learning_rate": 9.653370697542987e-07,
|
| 303 |
+
"logits/chosen": -0.9487398862838745,
|
| 304 |
+
"logits/rejected": -0.9483828544616699,
|
| 305 |
+
"logps/chosen": -0.3381520211696625,
|
| 306 |
+
"logps/rejected": -0.362968385219574,
|
| 307 |
+
"loss": 2.9005,
|
| 308 |
+
"rewards/accuracies": 0.574999988079071,
|
| 309 |
+
"rewards/chosen": -3.3815205097198486,
|
| 310 |
+
"rewards/margins": 0.24816343188285828,
|
| 311 |
+
"rewards/rejected": -3.62968373298645,
|
| 312 |
+
"semantic_entropy": 0.8359481692314148,
|
| 313 |
+
"step": 95
|
| 314 |
+
},
|
| 315 |
+
{
|
| 316 |
+
"epoch": 0.2185792349726776,
|
| 317 |
+
"grad_norm": 108.89056387914384,
|
| 318 |
+
"learning_rate": 9.580077827331037e-07,
|
| 319 |
+
"logits/chosen": -0.9571771621704102,
|
| 320 |
+
"logits/rejected": -0.9095252752304077,
|
| 321 |
+
"logps/chosen": -0.3725859522819519,
|
| 322 |
+
"logps/rejected": -0.4279399514198303,
|
| 323 |
+
"loss": 2.8907,
|
| 324 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 325 |
+
"rewards/chosen": -3.7258598804473877,
|
| 326 |
+
"rewards/margins": 0.553540050983429,
|
| 327 |
+
"rewards/rejected": -4.2793989181518555,
|
| 328 |
+
"semantic_entropy": 0.8787097930908203,
|
| 329 |
+
"step": 100
|
| 330 |
+
},
|
| 331 |
+
{
|
| 332 |
+
"epoch": 0.22950819672131148,
|
| 333 |
+
"grad_norm": 127.65723295420393,
|
| 334 |
+
"learning_rate": 9.500095727507419e-07,
|
| 335 |
+
"logits/chosen": -1.008998155593872,
|
| 336 |
+
"logits/rejected": -0.9757212400436401,
|
| 337 |
+
"logps/chosen": -0.3544849753379822,
|
| 338 |
+
"logps/rejected": -0.3977915942668915,
|
| 339 |
+
"loss": 2.9862,
|
| 340 |
+
"rewards/accuracies": 0.625,
|
| 341 |
+
"rewards/chosen": -3.5448498725891113,
|
| 342 |
+
"rewards/margins": 0.4330664277076721,
|
| 343 |
+
"rewards/rejected": -3.9779160022735596,
|
| 344 |
+
"semantic_entropy": 0.8548823595046997,
|
| 345 |
+
"step": 105
|
| 346 |
+
},
|
| 347 |
+
{
|
| 348 |
+
"epoch": 0.24043715846994534,
|
| 349 |
+
"grad_norm": 114.56272891566377,
|
| 350 |
+
"learning_rate": 9.413541212382004e-07,
|
| 351 |
+
"logits/chosen": -0.9939772486686707,
|
| 352 |
+
"logits/rejected": -0.9762369990348816,
|
| 353 |
+
"logps/chosen": -0.3617566227912903,
|
| 354 |
+
"logps/rejected": -0.4416491985321045,
|
| 355 |
+
"loss": 2.8457,
|
| 356 |
+
"rewards/accuracies": 0.637499988079071,
|
| 357 |
+
"rewards/chosen": -3.617565870285034,
|
| 358 |
+
"rewards/margins": 0.7989261746406555,
|
| 359 |
+
"rewards/rejected": -4.416492462158203,
|
| 360 |
+
"semantic_entropy": 0.8938226699829102,
|
| 361 |
+
"step": 110
|
| 362 |
+
},
|
| 363 |
+
{
|
| 364 |
+
"epoch": 0.25136612021857924,
|
| 365 |
+
"grad_norm": 93.3647181422965,
|
| 366 |
+
"learning_rate": 9.320540695314438e-07,
|
| 367 |
+
"logits/chosen": -1.007943868637085,
|
| 368 |
+
"logits/rejected": -0.9657400846481323,
|
| 369 |
+
"logps/chosen": -0.3646220564842224,
|
| 370 |
+
"logps/rejected": -0.46302324533462524,
|
| 371 |
+
"loss": 2.7542,
|
| 372 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 373 |
+
"rewards/chosen": -3.6462206840515137,
|
| 374 |
+
"rewards/margins": 0.9840116500854492,
|
| 375 |
+
"rewards/rejected": -4.630232334136963,
|
| 376 |
+
"semantic_entropy": 0.878681480884552,
|
| 377 |
+
"step": 115
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"epoch": 0.26229508196721313,
|
| 381 |
+
"grad_norm": 83.92874813803249,
|
| 382 |
+
"learning_rate": 9.221230004086721e-07,
|
| 383 |
+
"logits/chosen": -1.0344518423080444,
|
| 384 |
+
"logits/rejected": -0.9601195454597473,
|
| 385 |
+
"logps/chosen": -0.37165606021881104,
|
| 386 |
+
"logps/rejected": -0.42726248502731323,
|
| 387 |
+
"loss": 2.8688,
|
| 388 |
+
"rewards/accuracies": 0.668749988079071,
|
| 389 |
+
"rewards/chosen": -3.7165608406066895,
|
| 390 |
+
"rewards/margins": 0.5560643076896667,
|
| 391 |
+
"rewards/rejected": -4.272624969482422,
|
| 392 |
+
"semantic_entropy": 0.8952873945236206,
|
| 393 |
+
"step": 120
|
| 394 |
+
},
|
| 395 |
+
{
|
| 396 |
+
"epoch": 0.273224043715847,
|
| 397 |
+
"grad_norm": 190.904470995411,
|
| 398 |
+
"learning_rate": 9.11575418252596e-07,
|
| 399 |
+
"logits/chosen": -0.9347244501113892,
|
| 400 |
+
"logits/rejected": -0.8975458145141602,
|
| 401 |
+
"logps/chosen": -0.38535335659980774,
|
| 402 |
+
"logps/rejected": -0.4541945457458496,
|
| 403 |
+
"loss": 2.7207,
|
| 404 |
+
"rewards/accuracies": 0.6875,
|
| 405 |
+
"rewards/chosen": -3.853533983230591,
|
| 406 |
+
"rewards/margins": 0.68841153383255,
|
| 407 |
+
"rewards/rejected": -4.541945457458496,
|
| 408 |
+
"semantic_entropy": 0.9004859924316406,
|
| 409 |
+
"step": 125
|
| 410 |
+
},
|
| 411 |
+
{
|
| 412 |
+
"epoch": 0.28415300546448086,
|
| 413 |
+
"grad_norm": 92.76045922249655,
|
| 414 |
+
"learning_rate": 9.004267278667031e-07,
|
| 415 |
+
"logits/chosen": -0.9624107480049133,
|
| 416 |
+
"logits/rejected": -0.9534618258476257,
|
| 417 |
+
"logps/chosen": -0.4056780934333801,
|
| 418 |
+
"logps/rejected": -0.5245551466941833,
|
| 419 |
+
"loss": 2.7139,
|
| 420 |
+
"rewards/accuracies": 0.706250011920929,
|
| 421 |
+
"rewards/chosen": -4.056780815124512,
|
| 422 |
+
"rewards/margins": 1.1887714862823486,
|
| 423 |
+
"rewards/rejected": -5.245552062988281,
|
| 424 |
+
"semantic_entropy": 0.8837997317314148,
|
| 425 |
+
"step": 130
|
| 426 |
+
},
|
| 427 |
+
{
|
| 428 |
+
"epoch": 0.29508196721311475,
|
| 429 |
+
"grad_norm": 73.9902335755549,
|
| 430 |
+
"learning_rate": 8.886932119764565e-07,
|
| 431 |
+
"logits/chosen": -1.0003821849822998,
|
| 432 |
+
"logits/rejected": -0.9125338792800903,
|
| 433 |
+
"logps/chosen": -0.3864729106426239,
|
| 434 |
+
"logps/rejected": -0.4857531189918518,
|
| 435 |
+
"loss": 2.6137,
|
| 436 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 437 |
+
"rewards/chosen": -3.8647284507751465,
|
| 438 |
+
"rewards/margins": 0.9928020238876343,
|
| 439 |
+
"rewards/rejected": -4.8575310707092285,
|
| 440 |
+
"semantic_entropy": 0.9038194417953491,
|
| 441 |
+
"step": 135
|
| 442 |
+
},
|
| 443 |
+
{
|
| 444 |
+
"epoch": 0.30601092896174864,
|
| 445 |
+
"grad_norm": 86.17327659708472,
|
| 446 |
+
"learning_rate": 8.763920074482809e-07,
|
| 447 |
+
"logits/chosen": -0.9963301420211792,
|
| 448 |
+
"logits/rejected": -0.9396141767501831,
|
| 449 |
+
"logps/chosen": -0.4139133095741272,
|
| 450 |
+
"logps/rejected": -0.5436104536056519,
|
| 451 |
+
"loss": 2.4158,
|
| 452 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 453 |
+
"rewards/chosen": -4.139132976531982,
|
| 454 |
+
"rewards/margins": 1.296971082687378,
|
| 455 |
+
"rewards/rejected": -5.4361042976379395,
|
| 456 |
+
"semantic_entropy": 0.9314233064651489,
|
| 457 |
+
"step": 140
|
| 458 |
+
},
|
| 459 |
+
{
|
| 460 |
+
"epoch": 0.31693989071038253,
|
| 461 |
+
"grad_norm": 95.91548705026597,
|
| 462 |
+
"learning_rate": 8.635410802610723e-07,
|
| 463 |
+
"logits/chosen": -0.9837471842765808,
|
| 464 |
+
"logits/rejected": -0.96197909116745,
|
| 465 |
+
"logps/chosen": -0.3892672657966614,
|
| 466 |
+
"logps/rejected": -0.4463191032409668,
|
| 467 |
+
"loss": 2.5469,
|
| 468 |
+
"rewards/accuracies": 0.71875,
|
| 469 |
+
"rewards/chosen": -3.8926727771759033,
|
| 470 |
+
"rewards/margins": 0.5705188512802124,
|
| 471 |
+
"rewards/rejected": -4.463191986083984,
|
| 472 |
+
"semantic_entropy": 0.9131715893745422,
|
| 473 |
+
"step": 145
|
| 474 |
+
},
|
| 475 |
+
{
|
| 476 |
+
"epoch": 0.32786885245901637,
|
| 477 |
+
"grad_norm": 62.831854072245996,
|
| 478 |
+
"learning_rate": 8.501591992667849e-07,
|
| 479 |
+
"logits/chosen": -1.0432965755462646,
|
| 480 |
+
"logits/rejected": -1.0063092708587646,
|
| 481 |
+
"logps/chosen": -0.42201298475265503,
|
| 482 |
+
"logps/rejected": -0.5824503302574158,
|
| 483 |
+
"loss": 2.4081,
|
| 484 |
+
"rewards/accuracies": 0.731249988079071,
|
| 485 |
+
"rewards/chosen": -4.22012996673584,
|
| 486 |
+
"rewards/margins": 1.6043736934661865,
|
| 487 |
+
"rewards/rejected": -5.8245038986206055,
|
| 488 |
+
"semantic_entropy": 0.9204033613204956,
|
| 489 |
+
"step": 150
|
| 490 |
+
},
|
| 491 |
+
{
|
| 492 |
+
"epoch": 0.33879781420765026,
|
| 493 |
+
"grad_norm": 71.34642946039108,
|
| 494 |
+
"learning_rate": 8.362659087784152e-07,
|
| 495 |
+
"logits/chosen": -1.0033342838287354,
|
| 496 |
+
"logits/rejected": -0.943057656288147,
|
| 497 |
+
"logps/chosen": -0.4163185656070709,
|
| 498 |
+
"logps/rejected": -0.5151209831237793,
|
| 499 |
+
"loss": 2.5079,
|
| 500 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 501 |
+
"rewards/chosen": -4.163184642791748,
|
| 502 |
+
"rewards/margins": 0.9880247116088867,
|
| 503 |
+
"rewards/rejected": -5.151209831237793,
|
| 504 |
+
"semantic_entropy": 0.9188691973686218,
|
| 505 |
+
"step": 155
|
| 506 |
+
},
|
| 507 |
+
{
|
| 508 |
+
"epoch": 0.34972677595628415,
|
| 509 |
+
"grad_norm": 95.49736374278223,
|
| 510 |
+
"learning_rate": 8.218815000254231e-07,
|
| 511 |
+
"logits/chosen": -1.036727786064148,
|
| 512 |
+
"logits/rejected": -0.9749704599380493,
|
| 513 |
+
"logps/chosen": -0.46870869398117065,
|
| 514 |
+
"logps/rejected": -0.5485578775405884,
|
| 515 |
+
"loss": 2.5679,
|
| 516 |
+
"rewards/accuracies": 0.6875,
|
| 517 |
+
"rewards/chosen": -4.68708610534668,
|
| 518 |
+
"rewards/margins": 0.7984916567802429,
|
| 519 |
+
"rewards/rejected": -5.4855780601501465,
|
| 520 |
+
"semantic_entropy": 0.9369996786117554,
|
| 521 |
+
"step": 160
|
| 522 |
+
},
|
| 523 |
+
{
|
| 524 |
+
"epoch": 0.36065573770491804,
|
| 525 |
+
"grad_norm": 90.06304805222751,
|
| 526 |
+
"learning_rate": 8.07026981518276e-07,
|
| 527 |
+
"logits/chosen": -1.0219743251800537,
|
| 528 |
+
"logits/rejected": -0.9637954831123352,
|
| 529 |
+
"logps/chosen": -0.4483868181705475,
|
| 530 |
+
"logps/rejected": -0.5347827076911926,
|
| 531 |
+
"loss": 2.5189,
|
| 532 |
+
"rewards/accuracies": 0.731249988079071,
|
| 533 |
+
"rewards/chosen": -4.48386812210083,
|
| 534 |
+
"rewards/margins": 0.8639583587646484,
|
| 535 |
+
"rewards/rejected": -5.3478264808654785,
|
| 536 |
+
"semantic_entropy": 0.9511811137199402,
|
| 537 |
+
"step": 165
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 0.37158469945355194,
|
| 541 |
+
"grad_norm": 99.59126068255084,
|
| 542 |
+
"learning_rate": 7.917240483654e-07,
|
| 543 |
+
"logits/chosen": -1.01731276512146,
|
| 544 |
+
"logits/rejected": -0.9495924115180969,
|
| 545 |
+
"logps/chosen": -0.44470348954200745,
|
| 546 |
+
"logps/rejected": -0.5198506712913513,
|
| 547 |
+
"loss": 2.6168,
|
| 548 |
+
"rewards/accuracies": 0.6875,
|
| 549 |
+
"rewards/chosen": -4.447035312652588,
|
| 550 |
+
"rewards/margins": 0.7514716386795044,
|
| 551 |
+
"rewards/rejected": -5.198506832122803,
|
| 552 |
+
"semantic_entropy": 0.9501636624336243,
|
| 553 |
+
"step": 170
|
| 554 |
+
},
|
| 555 |
+
{
|
| 556 |
+
"epoch": 0.3825136612021858,
|
| 557 |
+
"grad_norm": 91.44599232569668,
|
| 558 |
+
"learning_rate": 7.759950505873521e-07,
|
| 559 |
+
"logits/chosen": -1.067455768585205,
|
| 560 |
+
"logits/rejected": -1.031198263168335,
|
| 561 |
+
"logps/chosen": -0.4639251232147217,
|
| 562 |
+
"logps/rejected": -0.5280762910842896,
|
| 563 |
+
"loss": 2.4665,
|
| 564 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 565 |
+
"rewards/chosen": -4.639250755310059,
|
| 566 |
+
"rewards/margins": 0.6415112614631653,
|
| 567 |
+
"rewards/rejected": -5.280762672424316,
|
| 568 |
+
"semantic_entropy": 0.9261935949325562,
|
| 569 |
+
"step": 175
|
| 570 |
+
},
|
| 571 |
+
{
|
| 572 |
+
"epoch": 0.39344262295081966,
|
| 573 |
+
"grad_norm": 90.82885753738844,
|
| 574 |
+
"learning_rate": 7.598629604744872e-07,
|
| 575 |
+
"logits/chosen": -1.0707954168319702,
|
| 576 |
+
"logits/rejected": -1.0595567226409912,
|
| 577 |
+
"logps/chosen": -0.43981847167015076,
|
| 578 |
+
"logps/rejected": -0.5758017897605896,
|
| 579 |
+
"loss": 2.3437,
|
| 580 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 581 |
+
"rewards/chosen": -4.398184776306152,
|
| 582 |
+
"rewards/margins": 1.3598332405090332,
|
| 583 |
+
"rewards/rejected": -5.7580180168151855,
|
| 584 |
+
"semantic_entropy": 0.966151237487793,
|
| 585 |
+
"step": 180
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 0.40437158469945356,
|
| 589 |
+
"grad_norm": 67.92403731887116,
|
| 590 |
+
"learning_rate": 7.433513390357989e-07,
|
| 591 |
+
"logits/chosen": -1.108884572982788,
|
| 592 |
+
"logits/rejected": -1.1143901348114014,
|
| 593 |
+
"logps/chosen": -0.46474918723106384,
|
| 594 |
+
"logps/rejected": -0.5912537574768066,
|
| 595 |
+
"loss": 2.3628,
|
| 596 |
+
"rewards/accuracies": 0.793749988079071,
|
| 597 |
+
"rewards/chosen": -4.647492408752441,
|
| 598 |
+
"rewards/margins": 1.265044927597046,
|
| 599 |
+
"rewards/rejected": -5.912537097930908,
|
| 600 |
+
"semantic_entropy": 0.9750612378120422,
|
| 601 |
+
"step": 185
|
| 602 |
+
},
|
| 603 |
+
{
|
| 604 |
+
"epoch": 0.41530054644808745,
|
| 605 |
+
"grad_norm": 108.44092763175198,
|
| 606 |
+
"learning_rate": 7.264843015879321e-07,
|
| 607 |
+
"logits/chosen": -1.1020928621292114,
|
| 608 |
+
"logits/rejected": -1.0545780658721924,
|
| 609 |
+
"logps/chosen": -0.4519892632961273,
|
| 610 |
+
"logps/rejected": -0.6003154516220093,
|
| 611 |
+
"loss": 2.4166,
|
| 612 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 613 |
+
"rewards/chosen": -4.51989221572876,
|
| 614 |
+
"rewards/margins": 1.483262062072754,
|
| 615 |
+
"rewards/rejected": -6.003154754638672,
|
| 616 |
+
"semantic_entropy": 0.9666361808776855,
|
| 617 |
+
"step": 190
|
| 618 |
+
},
|
| 619 |
+
{
|
| 620 |
+
"epoch": 0.4262295081967213,
|
| 621 |
+
"grad_norm": 83.4733675057699,
|
| 622 |
+
"learning_rate": 7.092864825346266e-07,
|
| 623 |
+
"logits/chosen": -1.129482626914978,
|
| 624 |
+
"logits/rejected": -1.0993843078613281,
|
| 625 |
+
"logps/chosen": -0.5358282327651978,
|
| 626 |
+
"logps/rejected": -0.7053772211074829,
|
| 627 |
+
"loss": 2.467,
|
| 628 |
+
"rewards/accuracies": 0.737500011920929,
|
| 629 |
+
"rewards/chosen": -5.358282566070557,
|
| 630 |
+
"rewards/margins": 1.6954904794692993,
|
| 631 |
+
"rewards/rejected": -7.053772926330566,
|
| 632 |
+
"semantic_entropy": 0.9725033044815063,
|
| 633 |
+
"step": 195
|
| 634 |
+
},
|
| 635 |
+
{
|
| 636 |
+
"epoch": 0.4371584699453552,
|
| 637 |
+
"grad_norm": 84.36629950168863,
|
| 638 |
+
"learning_rate": 6.917829993880302e-07,
|
| 639 |
+
"logits/chosen": -1.1204617023468018,
|
| 640 |
+
"logits/rejected": -1.0279228687286377,
|
| 641 |
+
"logps/chosen": -0.5025330185890198,
|
| 642 |
+
"logps/rejected": -0.6288330554962158,
|
| 643 |
+
"loss": 2.3371,
|
| 644 |
+
"rewards/accuracies": 0.762499988079071,
|
| 645 |
+
"rewards/chosen": -5.025330543518066,
|
| 646 |
+
"rewards/margins": 1.263000249862671,
|
| 647 |
+
"rewards/rejected": -6.288330078125,
|
| 648 |
+
"semantic_entropy": 0.9786258935928345,
|
| 649 |
+
"step": 200
|
| 650 |
+
},
|
| 651 |
+
{
|
| 652 |
+
"epoch": 0.44808743169398907,
|
| 653 |
+
"grad_norm": 69.08292746057649,
|
| 654 |
+
"learning_rate": 6.739994160844309e-07,
|
| 655 |
+
"logits/chosen": -1.0733792781829834,
|
| 656 |
+
"logits/rejected": -1.0833173990249634,
|
| 657 |
+
"logps/chosen": -0.4674602448940277,
|
| 658 |
+
"logps/rejected": -0.6061697006225586,
|
| 659 |
+
"loss": 2.2298,
|
| 660 |
+
"rewards/accuracies": 0.800000011920929,
|
| 661 |
+
"rewards/chosen": -4.67460298538208,
|
| 662 |
+
"rewards/margins": 1.3870941400527954,
|
| 663 |
+
"rewards/rejected": -6.061697006225586,
|
| 664 |
+
"semantic_entropy": 0.9808717966079712,
|
| 665 |
+
"step": 205
|
| 666 |
+
},
|
| 667 |
+
{
|
| 668 |
+
"epoch": 0.45901639344262296,
|
| 669 |
+
"grad_norm": 173.88772431903314,
|
| 670 |
+
"learning_rate": 6.559617056479827e-07,
|
| 671 |
+
"logits/chosen": -1.1001962423324585,
|
| 672 |
+
"logits/rejected": -1.0926573276519775,
|
| 673 |
+
"logps/chosen": -0.5027323961257935,
|
| 674 |
+
"logps/rejected": -0.6716328263282776,
|
| 675 |
+
"loss": 2.2974,
|
| 676 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 677 |
+
"rewards/chosen": -5.027324676513672,
|
| 678 |
+
"rewards/margins": 1.6890045404434204,
|
| 679 |
+
"rewards/rejected": -6.7163286209106445,
|
| 680 |
+
"semantic_entropy": 0.9590319395065308,
|
| 681 |
+
"step": 210
|
| 682 |
+
},
|
| 683 |
+
{
|
| 684 |
+
"epoch": 0.46994535519125685,
|
| 685 |
+
"grad_norm": 94.73792447265116,
|
| 686 |
+
"learning_rate": 6.376962122569567e-07,
|
| 687 |
+
"logits/chosen": -1.1140978336334229,
|
| 688 |
+
"logits/rejected": -1.0545861721038818,
|
| 689 |
+
"logps/chosen": -0.5114679336547852,
|
| 690 |
+
"logps/rejected": -0.6812509298324585,
|
| 691 |
+
"loss": 2.4425,
|
| 692 |
+
"rewards/accuracies": 0.793749988079071,
|
| 693 |
+
"rewards/chosen": -5.114679336547852,
|
| 694 |
+
"rewards/margins": 1.6978304386138916,
|
| 695 |
+
"rewards/rejected": -6.812510013580322,
|
| 696 |
+
"semantic_entropy": 1.0018761157989502,
|
| 697 |
+
"step": 215
|
| 698 |
+
},
|
| 699 |
+
{
|
| 700 |
+
"epoch": 0.4808743169398907,
|
| 701 |
+
"grad_norm": 92.5570818396132,
|
| 702 |
+
"learning_rate": 6.192296127679192e-07,
|
| 703 |
+
"logits/chosen": -1.1659886837005615,
|
| 704 |
+
"logits/rejected": -1.1348073482513428,
|
| 705 |
+
"logps/chosen": -0.5338795781135559,
|
| 706 |
+
"logps/rejected": -0.6664601564407349,
|
| 707 |
+
"loss": 2.2908,
|
| 708 |
+
"rewards/accuracies": 0.71875,
|
| 709 |
+
"rewards/chosen": -5.338796138763428,
|
| 710 |
+
"rewards/margins": 1.325805425643921,
|
| 711 |
+
"rewards/rejected": -6.664601802825928,
|
| 712 |
+
"semantic_entropy": 0.9806681871414185,
|
| 713 |
+
"step": 220
|
| 714 |
+
},
|
| 715 |
+
{
|
| 716 |
+
"epoch": 0.4918032786885246,
|
| 717 |
+
"grad_norm": 69.72669762595787,
|
| 718 |
+
"learning_rate": 6.005888777540319e-07,
|
| 719 |
+
"logits/chosen": -1.1951611042022705,
|
| 720 |
+
"logits/rejected": -1.1498881578445435,
|
| 721 |
+
"logps/chosen": -0.5304981470108032,
|
| 722 |
+
"logps/rejected": -0.6832343935966492,
|
| 723 |
+
"loss": 2.335,
|
| 724 |
+
"rewards/accuracies": 0.762499988079071,
|
| 725 |
+
"rewards/chosen": -5.304981708526611,
|
| 726 |
+
"rewards/margins": 1.5273630619049072,
|
| 727 |
+
"rewards/rejected": -6.832344055175781,
|
| 728 |
+
"semantic_entropy": 1.0007470846176147,
|
| 729 |
+
"step": 225
|
| 730 |
+
},
|
| 731 |
+
{
|
| 732 |
+
"epoch": 0.5027322404371585,
|
| 733 |
+
"grad_norm": 75.21829123168051,
|
| 734 |
+
"learning_rate": 5.818012321143773e-07,
|
| 735 |
+
"logits/chosen": -1.0969598293304443,
|
| 736 |
+
"logits/rejected": -1.0912028551101685,
|
| 737 |
+
"logps/chosen": -0.5201154351234436,
|
| 738 |
+
"logps/rejected": -0.7016697525978088,
|
| 739 |
+
"loss": 2.2528,
|
| 740 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 741 |
+
"rewards/chosen": -5.2011542320251465,
|
| 742 |
+
"rewards/margins": 1.8155431747436523,
|
| 743 |
+
"rewards/rejected": -7.016697883605957,
|
| 744 |
+
"semantic_entropy": 0.9929972887039185,
|
| 745 |
+
"step": 230
|
| 746 |
+
},
|
| 747 |
+
{
|
| 748 |
+
"epoch": 0.5136612021857924,
|
| 749 |
+
"grad_norm": 83.71823549926938,
|
| 750 |
+
"learning_rate": 5.628941153118388e-07,
|
| 751 |
+
"logits/chosen": -1.1018104553222656,
|
| 752 |
+
"logits/rejected": -1.0575555562973022,
|
| 753 |
+
"logps/chosen": -0.5212110280990601,
|
| 754 |
+
"logps/rejected": -0.664139986038208,
|
| 755 |
+
"loss": 2.2375,
|
| 756 |
+
"rewards/accuracies": 0.78125,
|
| 757 |
+
"rewards/chosen": -5.21211051940918,
|
| 758 |
+
"rewards/margins": 1.4292891025543213,
|
| 759 |
+
"rewards/rejected": -6.641399383544922,
|
| 760 |
+
"semantic_entropy": 0.9852234125137329,
|
| 761 |
+
"step": 235
|
| 762 |
+
},
|
| 763 |
+
{
|
| 764 |
+
"epoch": 0.5245901639344263,
|
| 765 |
+
"grad_norm": 69.35272919599275,
|
| 766 |
+
"learning_rate": 5.438951412976098e-07,
|
| 767 |
+
"logits/chosen": -1.1364176273345947,
|
| 768 |
+
"logits/rejected": -1.141788125038147,
|
| 769 |
+
"logps/chosen": -0.49681615829467773,
|
| 770 |
+
"logps/rejected": -0.6832265853881836,
|
| 771 |
+
"loss": 2.0805,
|
| 772 |
+
"rewards/accuracies": 0.800000011920929,
|
| 773 |
+
"rewards/chosen": -4.9681620597839355,
|
| 774 |
+
"rewards/margins": 1.8641045093536377,
|
| 775 |
+
"rewards/rejected": -6.832266330718994,
|
| 776 |
+
"semantic_entropy": 1.0116462707519531,
|
| 777 |
+
"step": 240
|
| 778 |
+
},
|
| 779 |
+
{
|
| 780 |
+
"epoch": 0.5355191256830601,
|
| 781 |
+
"grad_norm": 71.19254935234711,
|
| 782 |
+
"learning_rate": 5.248320581808619e-07,
|
| 783 |
+
"logits/chosen": -1.0613957643508911,
|
| 784 |
+
"logits/rejected": -1.0091025829315186,
|
| 785 |
+
"logps/chosen": -0.5141640901565552,
|
| 786 |
+
"logps/rejected": -0.6968377828598022,
|
| 787 |
+
"loss": 2.1843,
|
| 788 |
+
"rewards/accuracies": 0.75,
|
| 789 |
+
"rewards/chosen": -5.141640663146973,
|
| 790 |
+
"rewards/margins": 1.8267381191253662,
|
| 791 |
+
"rewards/rejected": -6.96837854385376,
|
| 792 |
+
"semantic_entropy": 0.9718330502510071,
|
| 793 |
+
"step": 245
|
| 794 |
+
},
|
| 795 |
+
{
|
| 796 |
+
"epoch": 0.546448087431694,
|
| 797 |
+
"grad_norm": 70.83927823916859,
|
| 798 |
+
"learning_rate": 5.057327077024744e-07,
|
| 799 |
+
"logits/chosen": -1.146533727645874,
|
| 800 |
+
"logits/rejected": -1.1047497987747192,
|
| 801 |
+
"logps/chosen": -0.5028788447380066,
|
| 802 |
+
"logps/rejected": -0.6331702470779419,
|
| 803 |
+
"loss": 2.2718,
|
| 804 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 805 |
+
"rewards/chosen": -5.028789043426514,
|
| 806 |
+
"rewards/margins": 1.3029136657714844,
|
| 807 |
+
"rewards/rejected": -6.331702709197998,
|
| 808 |
+
"semantic_entropy": 0.9849420785903931,
|
| 809 |
+
"step": 250
|
| 810 |
+
},
|
| 811 |
+
{
|
| 812 |
+
"epoch": 0.5573770491803278,
|
| 813 |
+
"grad_norm": 78.29453714448445,
|
| 814 |
+
"learning_rate": 4.866249845720132e-07,
|
| 815 |
+
"logits/chosen": -1.1301579475402832,
|
| 816 |
+
"logits/rejected": -1.091973900794983,
|
| 817 |
+
"logps/chosen": -0.555388331413269,
|
| 818 |
+
"logps/rejected": -0.7187477946281433,
|
| 819 |
+
"loss": 2.1692,
|
| 820 |
+
"rewards/accuracies": 0.762499988079071,
|
| 821 |
+
"rewards/chosen": -5.553883075714111,
|
| 822 |
+
"rewards/margins": 1.633594274520874,
|
| 823 |
+
"rewards/rejected": -7.187478065490723,
|
| 824 |
+
"semantic_entropy": 0.9999436140060425,
|
| 825 |
+
"step": 255
|
| 826 |
+
},
|
| 827 |
+
{
|
| 828 |
+
"epoch": 0.5683060109289617,
|
| 829 |
+
"grad_norm": 74.66530559540921,
|
| 830 |
+
"learning_rate": 4.675367957273505e-07,
|
| 831 |
+
"logits/chosen": -1.096861720085144,
|
| 832 |
+
"logits/rejected": -1.0846450328826904,
|
| 833 |
+
"logps/chosen": -0.5131552815437317,
|
| 834 |
+
"logps/rejected": -0.671288251876831,
|
| 835 |
+
"loss": 2.1911,
|
| 836 |
+
"rewards/accuracies": 0.800000011920929,
|
| 837 |
+
"rewards/chosen": -5.131552219390869,
|
| 838 |
+
"rewards/margins": 1.5813300609588623,
|
| 839 |
+
"rewards/rejected": -6.712882041931152,
|
| 840 |
+
"semantic_entropy": 0.9927156567573547,
|
| 841 |
+
"step": 260
|
| 842 |
+
},
|
| 843 |
+
{
|
| 844 |
+
"epoch": 0.5792349726775956,
|
| 845 |
+
"grad_norm": 122.60471760919329,
|
| 846 |
+
"learning_rate": 4.4849601957642285e-07,
|
| 847 |
+
"logits/chosen": -1.124089002609253,
|
| 848 |
+
"logits/rejected": -1.0828189849853516,
|
| 849 |
+
"logps/chosen": -0.5186060070991516,
|
| 850 |
+
"logps/rejected": -0.6825847625732422,
|
| 851 |
+
"loss": 2.1964,
|
| 852 |
+
"rewards/accuracies": 0.762499988079071,
|
| 853 |
+
"rewards/chosen": -5.186059474945068,
|
| 854 |
+
"rewards/margins": 1.6397874355316162,
|
| 855 |
+
"rewards/rejected": -6.8258466720581055,
|
| 856 |
+
"semantic_entropy": 0.9880490303039551,
|
| 857 |
+
"step": 265
|
| 858 |
+
},
|
| 859 |
+
{
|
| 860 |
+
"epoch": 0.5901639344262295,
|
| 861 |
+
"grad_norm": 86.34568213442027,
|
| 862 |
+
"learning_rate": 4.295304652806592e-07,
|
| 863 |
+
"logits/chosen": -1.1392979621887207,
|
| 864 |
+
"logits/rejected": -1.1078673601150513,
|
| 865 |
+
"logps/chosen": -0.5172940492630005,
|
| 866 |
+
"logps/rejected": -0.699386477470398,
|
| 867 |
+
"loss": 2.0791,
|
| 868 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 869 |
+
"rewards/chosen": -5.172940254211426,
|
| 870 |
+
"rewards/margins": 1.8209247589111328,
|
| 871 |
+
"rewards/rejected": -6.993865013122559,
|
| 872 |
+
"semantic_entropy": 0.9866276979446411,
|
| 873 |
+
"step": 270
|
| 874 |
+
},
|
| 875 |
+
{
|
| 876 |
+
"epoch": 0.6010928961748634,
|
| 877 |
+
"grad_norm": 58.20062582439532,
|
| 878 |
+
"learning_rate": 4.106678321395433e-07,
|
| 879 |
+
"logits/chosen": -1.1032135486602783,
|
| 880 |
+
"logits/rejected": -1.0302824974060059,
|
| 881 |
+
"logps/chosen": -0.5297619104385376,
|
| 882 |
+
"logps/rejected": -0.627161979675293,
|
| 883 |
+
"loss": 2.1916,
|
| 884 |
+
"rewards/accuracies": 0.643750011920929,
|
| 885 |
+
"rewards/chosen": -5.2976179122924805,
|
| 886 |
+
"rewards/margins": 0.9740018844604492,
|
| 887 |
+
"rewards/rejected": -6.271620750427246,
|
| 888 |
+
"semantic_entropy": 0.9879854917526245,
|
| 889 |
+
"step": 275
|
| 890 |
+
},
|
| 891 |
+
{
|
| 892 |
+
"epoch": 0.6120218579234973,
|
| 893 |
+
"grad_norm": 84.89751844734043,
|
| 894 |
+
"learning_rate": 3.9193566913562915e-07,
|
| 895 |
+
"logits/chosen": -1.0617036819458008,
|
| 896 |
+
"logits/rejected": -1.0624239444732666,
|
| 897 |
+
"logps/chosen": -0.5223734378814697,
|
| 898 |
+
"logps/rejected": -0.7275804281234741,
|
| 899 |
+
"loss": 2.1763,
|
| 900 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 901 |
+
"rewards/chosen": -5.223733901977539,
|
| 902 |
+
"rewards/margins": 2.0520694255828857,
|
| 903 |
+
"rewards/rejected": -7.275804042816162,
|
| 904 |
+
"semantic_entropy": 0.986528754234314,
|
| 905 |
+
"step": 280
|
| 906 |
+
},
|
| 907 |
+
{
|
| 908 |
+
"epoch": 0.6229508196721312,
|
| 909 |
+
"grad_norm": 72.19485427611927,
|
| 910 |
+
"learning_rate": 3.7336133469909623e-07,
|
| 911 |
+
"logits/chosen": -1.1958709955215454,
|
| 912 |
+
"logits/rejected": -1.1594369411468506,
|
| 913 |
+
"logps/chosen": -0.5087668895721436,
|
| 914 |
+
"logps/rejected": -0.702479898929596,
|
| 915 |
+
"loss": 2.1358,
|
| 916 |
+
"rewards/accuracies": 0.8125,
|
| 917 |
+
"rewards/chosen": -5.0876688957214355,
|
| 918 |
+
"rewards/margins": 1.9371296167373657,
|
| 919 |
+
"rewards/rejected": -7.024799346923828,
|
| 920 |
+
"semantic_entropy": 0.9978361129760742,
|
| 921 |
+
"step": 285
|
| 922 |
+
},
|
| 923 |
+
{
|
| 924 |
+
"epoch": 0.6338797814207651,
|
| 925 |
+
"grad_norm": 81.98995443073827,
|
| 926 |
+
"learning_rate": 3.549719567506076e-07,
|
| 927 |
+
"logits/chosen": -1.1317315101623535,
|
| 928 |
+
"logits/rejected": -1.0870287418365479,
|
| 929 |
+
"logps/chosen": -0.5346897840499878,
|
| 930 |
+
"logps/rejected": -0.6969183087348938,
|
| 931 |
+
"loss": 2.1523,
|
| 932 |
+
"rewards/accuracies": 0.800000011920929,
|
| 933 |
+
"rewards/chosen": -5.346898078918457,
|
| 934 |
+
"rewards/margins": 1.62228524684906,
|
| 935 |
+
"rewards/rejected": -6.969183444976807,
|
| 936 |
+
"semantic_entropy": 1.0123668909072876,
|
| 937 |
+
"step": 290
|
| 938 |
+
},
|
| 939 |
+
{
|
| 940 |
+
"epoch": 0.644808743169399,
|
| 941 |
+
"grad_norm": 67.90034025619751,
|
| 942 |
+
"learning_rate": 3.3679439308082774e-07,
|
| 943 |
+
"logits/chosen": -1.115994930267334,
|
| 944 |
+
"logits/rejected": -1.1152690649032593,
|
| 945 |
+
"logps/chosen": -0.5274439454078674,
|
| 946 |
+
"logps/rejected": -0.7239035367965698,
|
| 947 |
+
"loss": 1.9278,
|
| 948 |
+
"rewards/accuracies": 0.8125,
|
| 949 |
+
"rewards/chosen": -5.274438381195068,
|
| 950 |
+
"rewards/margins": 1.9645967483520508,
|
| 951 |
+
"rewards/rejected": -7.239035606384277,
|
| 952 |
+
"semantic_entropy": 1.0061827898025513,
|
| 953 |
+
"step": 295
|
| 954 |
+
},
|
| 955 |
+
{
|
| 956 |
+
"epoch": 0.6557377049180327,
|
| 957 |
+
"grad_norm": 70.85439265034472,
|
| 958 |
+
"learning_rate": 3.1885519212446716e-07,
|
| 959 |
+
"logits/chosen": -1.144639253616333,
|
| 960 |
+
"logits/rejected": -1.1228580474853516,
|
| 961 |
+
"logps/chosen": -0.542576253414154,
|
| 962 |
+
"logps/rejected": -0.7291213274002075,
|
| 963 |
+
"loss": 2.0159,
|
| 964 |
+
"rewards/accuracies": 0.8125,
|
| 965 |
+
"rewards/chosen": -5.42576265335083,
|
| 966 |
+
"rewards/margins": 1.865450143814087,
|
| 967 |
+
"rewards/rejected": -7.291213035583496,
|
| 968 |
+
"semantic_entropy": 0.9855409860610962,
|
| 969 |
+
"step": 300
|
| 970 |
+
},
|
| 971 |
+
{
|
| 972 |
+
"epoch": 0.6666666666666666,
|
| 973 |
+
"grad_norm": 76.04305212768787,
|
| 974 |
+
"learning_rate": 3.0118055418614295e-07,
|
| 975 |
+
"logits/chosen": -1.1450592279434204,
|
| 976 |
+
"logits/rejected": -1.0869606733322144,
|
| 977 |
+
"logps/chosen": -0.5319762229919434,
|
| 978 |
+
"logps/rejected": -0.7148723006248474,
|
| 979 |
+
"loss": 2.1436,
|
| 980 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 981 |
+
"rewards/chosen": -5.319762706756592,
|
| 982 |
+
"rewards/margins": 1.8289600610733032,
|
| 983 |
+
"rewards/rejected": -7.1487226486206055,
|
| 984 |
+
"semantic_entropy": 1.007246732711792,
|
| 985 |
+
"step": 305
|
| 986 |
+
},
|
| 987 |
+
{
|
| 988 |
+
"epoch": 0.6775956284153005,
|
| 989 |
+
"grad_norm": 108.42660481521786,
|
| 990 |
+
"learning_rate": 2.83796293174686e-07,
|
| 991 |
+
"logits/chosen": -1.0885827541351318,
|
| 992 |
+
"logits/rejected": -1.092543363571167,
|
| 993 |
+
"logps/chosen": -0.5401273369789124,
|
| 994 |
+
"logps/rejected": -0.748576283454895,
|
| 995 |
+
"loss": 2.2592,
|
| 996 |
+
"rewards/accuracies": 0.75,
|
| 997 |
+
"rewards/chosen": -5.401273250579834,
|
| 998 |
+
"rewards/margins": 2.084489107131958,
|
| 999 |
+
"rewards/rejected": -7.485762119293213,
|
| 1000 |
+
"semantic_entropy": 0.9948571920394897,
|
| 1001 |
+
"step": 310
|
| 1002 |
+
},
|
| 1003 |
+
{
|
| 1004 |
+
"epoch": 0.6885245901639344,
|
| 1005 |
+
"grad_norm": 70.86094179855972,
|
| 1006 |
+
"learning_rate": 2.6672779890178046e-07,
|
| 1007 |
+
"logits/chosen": -1.1491663455963135,
|
| 1008 |
+
"logits/rejected": -1.1490873098373413,
|
| 1009 |
+
"logps/chosen": -0.5631974935531616,
|
| 1010 |
+
"logps/rejected": -0.6903260946273804,
|
| 1011 |
+
"loss": 2.1308,
|
| 1012 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 1013 |
+
"rewards/chosen": -5.631974697113037,
|
| 1014 |
+
"rewards/margins": 1.2712849378585815,
|
| 1015 |
+
"rewards/rejected": -6.903260231018066,
|
| 1016 |
+
"semantic_entropy": 0.9926439523696899,
|
| 1017 |
+
"step": 315
|
| 1018 |
+
},
|
| 1019 |
+
{
|
| 1020 |
+
"epoch": 0.6994535519125683,
|
| 1021 |
+
"grad_norm": 57.13767602235213,
|
| 1022 |
+
"learning_rate": 2.500000000000001e-07,
|
| 1023 |
+
"logits/chosen": -1.2022249698638916,
|
| 1024 |
+
"logits/rejected": -1.1517468690872192,
|
| 1025 |
+
"logps/chosen": -0.5420448184013367,
|
| 1026 |
+
"logps/rejected": -0.747357964515686,
|
| 1027 |
+
"loss": 2.0999,
|
| 1028 |
+
"rewards/accuracies": 0.824999988079071,
|
| 1029 |
+
"rewards/chosen": -5.42044734954834,
|
| 1030 |
+
"rewards/margins": 2.053131580352783,
|
| 1031 |
+
"rewards/rejected": -7.473579406738281,
|
| 1032 |
+
"semantic_entropy": 1.00899338722229,
|
| 1033 |
+
"step": 320
|
| 1034 |
+
},
|
| 1035 |
+
{
|
| 1036 |
+
"epoch": 0.7103825136612022,
|
| 1037 |
+
"grad_norm": 82.09122654285451,
|
| 1038 |
+
"learning_rate": 2.3363732751439923e-07,
|
| 1039 |
+
"logits/chosen": -1.1618945598602295,
|
| 1040 |
+
"logits/rejected": -1.143754243850708,
|
| 1041 |
+
"logps/chosen": -0.5291402339935303,
|
| 1042 |
+
"logps/rejected": -0.7228410243988037,
|
| 1043 |
+
"loss": 2.0999,
|
| 1044 |
+
"rewards/accuracies": 0.78125,
|
| 1045 |
+
"rewards/chosen": -5.291402339935303,
|
| 1046 |
+
"rewards/margins": 1.9370079040527344,
|
| 1047 |
+
"rewards/rejected": -7.228410243988037,
|
| 1048 |
+
"semantic_entropy": 1.0088245868682861,
|
| 1049 |
+
"step": 325
|
| 1050 |
+
},
|
| 1051 |
+
{
|
| 1052 |
+
"epoch": 0.7213114754098361,
|
| 1053 |
+
"grad_norm": 62.18032792736575,
|
| 1054 |
+
"learning_rate": 2.1766367922083283e-07,
|
| 1055 |
+
"logits/chosen": -1.112157940864563,
|
| 1056 |
+
"logits/rejected": -1.0798307657241821,
|
| 1057 |
+
"logps/chosen": -0.4986172318458557,
|
| 1058 |
+
"logps/rejected": -0.7466678023338318,
|
| 1059 |
+
"loss": 2.0623,
|
| 1060 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 1061 |
+
"rewards/chosen": -4.986172676086426,
|
| 1062 |
+
"rewards/margins": 2.4805047512054443,
|
| 1063 |
+
"rewards/rejected": -7.466677188873291,
|
| 1064 |
+
"semantic_entropy": 0.98463374376297,
|
| 1065 |
+
"step": 330
|
| 1066 |
+
},
|
| 1067 |
+
{
|
| 1068 |
+
"epoch": 0.73224043715847,
|
| 1069 |
+
"grad_norm": 77.93323322539872,
|
| 1070 |
+
"learning_rate": 2.021023847231202e-07,
|
| 1071 |
+
"logits/chosen": -1.1002051830291748,
|
| 1072 |
+
"logits/rejected": -1.0612647533416748,
|
| 1073 |
+
"logps/chosen": -0.5647180080413818,
|
| 1074 |
+
"logps/rejected": -0.758574366569519,
|
| 1075 |
+
"loss": 2.0578,
|
| 1076 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 1077 |
+
"rewards/chosen": -5.647180557250977,
|
| 1078 |
+
"rewards/margins": 1.938563346862793,
|
| 1079 |
+
"rewards/rejected": -7.5857439041137695,
|
| 1080 |
+
"semantic_entropy": 0.9851423501968384,
|
| 1081 |
+
"step": 335
|
| 1082 |
+
},
|
| 1083 |
+
{
|
| 1084 |
+
"epoch": 0.7431693989071039,
|
| 1085 |
+
"grad_norm": 86.17026425537334,
|
| 1086 |
+
"learning_rate": 1.869761713800254e-07,
|
| 1087 |
+
"logits/chosen": -1.1088799238204956,
|
| 1088 |
+
"logits/rejected": -1.064263939857483,
|
| 1089 |
+
"logps/chosen": -0.5516521334648132,
|
| 1090 |
+
"logps/rejected": -0.7183089256286621,
|
| 1091 |
+
"loss": 2.1191,
|
| 1092 |
+
"rewards/accuracies": 0.71875,
|
| 1093 |
+
"rewards/chosen": -5.516521453857422,
|
| 1094 |
+
"rewards/margins": 1.6665668487548828,
|
| 1095 |
+
"rewards/rejected": -7.183088779449463,
|
| 1096 |
+
"semantic_entropy": 0.9953676462173462,
|
| 1097 |
+
"step": 340
|
| 1098 |
+
},
|
| 1099 |
+
{
|
| 1100 |
+
"epoch": 0.7540983606557377,
|
| 1101 |
+
"grad_norm": 80.58633018790611,
|
| 1102 |
+
"learning_rate": 1.7230713111182164e-07,
|
| 1103 |
+
"logits/chosen": -1.156589150428772,
|
| 1104 |
+
"logits/rejected": -1.1543285846710205,
|
| 1105 |
+
"logps/chosen": -0.5463498830795288,
|
| 1106 |
+
"logps/rejected": -0.7534428238868713,
|
| 1107 |
+
"loss": 2.1553,
|
| 1108 |
+
"rewards/accuracies": 0.8125,
|
| 1109 |
+
"rewards/chosen": -5.463499546051025,
|
| 1110 |
+
"rewards/margins": 2.0709292888641357,
|
| 1111 |
+
"rewards/rejected": -7.534428596496582,
|
| 1112 |
+
"semantic_entropy": 0.9916456341743469,
|
| 1113 |
+
"step": 345
|
| 1114 |
+
},
|
| 1115 |
+
{
|
| 1116 |
+
"epoch": 0.7650273224043715,
|
| 1117 |
+
"grad_norm": 89.6142113445473,
|
| 1118 |
+
"learning_rate": 1.5811668813491696e-07,
|
| 1119 |
+
"logits/chosen": -1.1436890363693237,
|
| 1120 |
+
"logits/rejected": -1.124874234199524,
|
| 1121 |
+
"logps/chosen": -0.5143482685089111,
|
| 1122 |
+
"logps/rejected": -0.6786841154098511,
|
| 1123 |
+
"loss": 2.0898,
|
| 1124 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 1125 |
+
"rewards/chosen": -5.143482685089111,
|
| 1126 |
+
"rewards/margins": 1.643358588218689,
|
| 1127 |
+
"rewards/rejected": -6.786840915679932,
|
| 1128 |
+
"semantic_entropy": 0.9892457127571106,
|
| 1129 |
+
"step": 350
|
| 1130 |
+
},
|
| 1131 |
+
{
|
| 1132 |
+
"epoch": 0.7759562841530054,
|
| 1133 |
+
"grad_norm": 66.31246023221092,
|
| 1134 |
+
"learning_rate": 1.4442556767166369e-07,
|
| 1135 |
+
"logits/chosen": -1.1231715679168701,
|
| 1136 |
+
"logits/rejected": -1.0935585498809814,
|
| 1137 |
+
"logps/chosen": -0.5266932845115662,
|
| 1138 |
+
"logps/rejected": -0.6977185010910034,
|
| 1139 |
+
"loss": 2.0776,
|
| 1140 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1141 |
+
"rewards/chosen": -5.266932487487793,
|
| 1142 |
+
"rewards/margins": 1.7102525234222412,
|
| 1143 |
+
"rewards/rejected": -6.9771857261657715,
|
| 1144 |
+
"semantic_entropy": 1.0020959377288818,
|
| 1145 |
+
"step": 355
|
| 1146 |
+
},
|
| 1147 |
+
{
|
| 1148 |
+
"epoch": 0.7868852459016393,
|
| 1149 |
+
"grad_norm": 83.110534751007,
|
| 1150 |
+
"learning_rate": 1.312537656810549e-07,
|
| 1151 |
+
"logits/chosen": -1.0739078521728516,
|
| 1152 |
+
"logits/rejected": -1.0743262767791748,
|
| 1153 |
+
"logps/chosen": -0.5362976789474487,
|
| 1154 |
+
"logps/rejected": -0.714411735534668,
|
| 1155 |
+
"loss": 2.1365,
|
| 1156 |
+
"rewards/accuracies": 0.737500011920929,
|
| 1157 |
+
"rewards/chosen": -5.362977504730225,
|
| 1158 |
+
"rewards/margins": 1.7811400890350342,
|
| 1159 |
+
"rewards/rejected": -7.144117832183838,
|
| 1160 |
+
"semantic_entropy": 0.9827717542648315,
|
| 1161 |
+
"step": 360
|
| 1162 |
+
},
|
| 1163 |
+
{
|
| 1164 |
+
"epoch": 0.7978142076502732,
|
| 1165 |
+
"grad_norm": 141.91877356203636,
|
| 1166 |
+
"learning_rate": 1.1862051965451214e-07,
|
| 1167 |
+
"logits/chosen": -1.1579176187515259,
|
| 1168 |
+
"logits/rejected": -1.1566094160079956,
|
| 1169 |
+
"logps/chosen": -0.5423863530158997,
|
| 1170 |
+
"logps/rejected": -0.7343495488166809,
|
| 1171 |
+
"loss": 2.054,
|
| 1172 |
+
"rewards/accuracies": 0.824999988079071,
|
| 1173 |
+
"rewards/chosen": -5.423863410949707,
|
| 1174 |
+
"rewards/margins": 1.9196319580078125,
|
| 1175 |
+
"rewards/rejected": -7.3434953689575195,
|
| 1176 |
+
"semantic_entropy": 1.011482834815979,
|
| 1177 |
+
"step": 365
|
| 1178 |
+
},
|
| 1179 |
+
{
|
| 1180 |
+
"epoch": 0.8087431693989071,
|
| 1181 |
+
"grad_norm": 78.59703524378126,
|
| 1182 |
+
"learning_rate": 1.0654428051942138e-07,
|
| 1183 |
+
"logits/chosen": -1.165038824081421,
|
| 1184 |
+
"logits/rejected": -1.1290335655212402,
|
| 1185 |
+
"logps/chosen": -0.5577388405799866,
|
| 1186 |
+
"logps/rejected": -0.7825115919113159,
|
| 1187 |
+
"loss": 2.1829,
|
| 1188 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 1189 |
+
"rewards/chosen": -5.577388286590576,
|
| 1190 |
+
"rewards/margins": 2.247727632522583,
|
| 1191 |
+
"rewards/rejected": -7.825116157531738,
|
| 1192 |
+
"semantic_entropy": 1.0018393993377686,
|
| 1193 |
+
"step": 370
|
| 1194 |
+
},
|
| 1195 |
+
{
|
| 1196 |
+
"epoch": 0.819672131147541,
|
| 1197 |
+
"grad_norm": 63.6353735816688,
|
| 1198 |
+
"learning_rate": 9.504268569144763e-08,
|
| 1199 |
+
"logits/chosen": -1.2013657093048096,
|
| 1200 |
+
"logits/rejected": -1.1338837146759033,
|
| 1201 |
+
"logps/chosen": -0.5316141843795776,
|
| 1202 |
+
"logps/rejected": -0.7204681038856506,
|
| 1203 |
+
"loss": 2.0911,
|
| 1204 |
+
"rewards/accuracies": 0.762499988079071,
|
| 1205 |
+
"rewards/chosen": -5.316141605377197,
|
| 1206 |
+
"rewards/margins": 1.8885393142700195,
|
| 1207 |
+
"rewards/rejected": -7.204681396484375,
|
| 1208 |
+
"semantic_entropy": 1.0054622888565063,
|
| 1209 |
+
"step": 375
|
| 1210 |
+
},
|
| 1211 |
+
{
|
| 1212 |
+
"epoch": 0.8306010928961749,
|
| 1213 |
+
"grad_norm": 64.71607446988344,
|
| 1214 |
+
"learning_rate": 8.413253331499049e-08,
|
| 1215 |
+
"logits/chosen": -1.0807088613510132,
|
| 1216 |
+
"logits/rejected": -1.102399230003357,
|
| 1217 |
+
"logps/chosen": -0.549530029296875,
|
| 1218 |
+
"logps/rejected": -0.7361005544662476,
|
| 1219 |
+
"loss": 2.0374,
|
| 1220 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1221 |
+
"rewards/chosen": -5.49530029296875,
|
| 1222 |
+
"rewards/margins": 1.8657052516937256,
|
| 1223 |
+
"rewards/rejected": -7.361004829406738,
|
| 1224 |
+
"semantic_entropy": 1.0044524669647217,
|
| 1225 |
+
"step": 380
|
| 1226 |
+
},
|
| 1227 |
+
{
|
| 1228 |
+
"epoch": 0.8415300546448088,
|
| 1229 |
+
"grad_norm": 67.03443291040514,
|
| 1230 |
+
"learning_rate": 7.382975772939865e-08,
|
| 1231 |
+
"logits/chosen": -1.1790930032730103,
|
| 1232 |
+
"logits/rejected": -1.1615774631500244,
|
| 1233 |
+
"logps/chosen": -0.590388834476471,
|
| 1234 |
+
"logps/rejected": -0.7754439115524292,
|
| 1235 |
+
"loss": 2.1706,
|
| 1236 |
+
"rewards/accuracies": 0.84375,
|
| 1237 |
+
"rewards/chosen": -5.903887748718262,
|
| 1238 |
+
"rewards/margins": 1.850551962852478,
|
| 1239 |
+
"rewards/rejected": -7.754439353942871,
|
| 1240 |
+
"semantic_entropy": 1.0115418434143066,
|
| 1241 |
+
"step": 385
|
| 1242 |
+
},
|
| 1243 |
+
{
|
| 1244 |
+
"epoch": 0.8524590163934426,
|
| 1245 |
+
"grad_norm": 103.88519596336293,
|
| 1246 |
+
"learning_rate": 6.414940619677734e-08,
|
| 1247 |
+
"logits/chosen": -1.166526436805725,
|
| 1248 |
+
"logits/rejected": -1.1425046920776367,
|
| 1249 |
+
"logps/chosen": -0.5350316762924194,
|
| 1250 |
+
"logps/rejected": -0.7514439821243286,
|
| 1251 |
+
"loss": 2.1188,
|
| 1252 |
+
"rewards/accuracies": 0.831250011920929,
|
| 1253 |
+
"rewards/chosen": -5.350315570831299,
|
| 1254 |
+
"rewards/margins": 2.164124011993408,
|
| 1255 |
+
"rewards/rejected": -7.514439582824707,
|
| 1256 |
+
"semantic_entropy": 1.0120559930801392,
|
| 1257 |
+
"step": 390
|
| 1258 |
+
},
|
| 1259 |
+
{
|
| 1260 |
+
"epoch": 0.8633879781420765,
|
| 1261 |
+
"grad_norm": 71.31872002268815,
|
| 1262 |
+
"learning_rate": 5.5105616925376296e-08,
|
| 1263 |
+
"logits/chosen": -1.1460245847702026,
|
| 1264 |
+
"logits/rejected": -1.1267726421356201,
|
| 1265 |
+
"logps/chosen": -0.5442631244659424,
|
| 1266 |
+
"logps/rejected": -0.6925864815711975,
|
| 1267 |
+
"loss": 2.0494,
|
| 1268 |
+
"rewards/accuracies": 0.768750011920929,
|
| 1269 |
+
"rewards/chosen": -5.442631721496582,
|
| 1270 |
+
"rewards/margins": 1.4832336902618408,
|
| 1271 |
+
"rewards/rejected": -6.925864219665527,
|
| 1272 |
+
"semantic_entropy": 1.0200846195220947,
|
| 1273 |
+
"step": 395
|
| 1274 |
+
},
|
| 1275 |
+
{
|
| 1276 |
+
"epoch": 0.8743169398907104,
|
| 1277 |
+
"grad_norm": 75.02341041681119,
|
| 1278 |
+
"learning_rate": 4.6711598420656976e-08,
|
| 1279 |
+
"logits/chosen": -1.0774163007736206,
|
| 1280 |
+
"logits/rejected": -1.0491969585418701,
|
| 1281 |
+
"logps/chosen": -0.5711244344711304,
|
| 1282 |
+
"logps/rejected": -0.7721945643424988,
|
| 1283 |
+
"loss": 2.0424,
|
| 1284 |
+
"rewards/accuracies": 0.824999988079071,
|
| 1285 |
+
"rewards/chosen": -5.711243629455566,
|
| 1286 |
+
"rewards/margins": 2.010702133178711,
|
| 1287 |
+
"rewards/rejected": -7.721946716308594,
|
| 1288 |
+
"semantic_entropy": 0.9843025207519531,
|
| 1289 |
+
"step": 400
|
| 1290 |
+
},
|
| 1291 |
+
{
|
| 1292 |
+
"epoch": 0.8743169398907104,
|
| 1293 |
+
"eval_logits/chosen": -1.373081088066101,
|
| 1294 |
+
"eval_logits/rejected": -1.332649827003479,
|
| 1295 |
+
"eval_logps/chosen": -0.5314387083053589,
|
| 1296 |
+
"eval_logps/rejected": -0.7141891121864319,
|
| 1297 |
+
"eval_loss": 2.0935795307159424,
|
| 1298 |
+
"eval_rewards/accuracies": 0.7771084308624268,
|
| 1299 |
+
"eval_rewards/chosen": -5.314386367797852,
|
| 1300 |
+
"eval_rewards/margins": 1.8275047540664673,
|
| 1301 |
+
"eval_rewards/rejected": -7.1418914794921875,
|
| 1302 |
+
"eval_runtime": 37.9946,
|
| 1303 |
+
"eval_samples_per_second": 34.689,
|
| 1304 |
+
"eval_semantic_entropy": 0.9976296424865723,
|
| 1305 |
+
"eval_steps_per_second": 2.185,
|
| 1306 |
+
"step": 400
|
| 1307 |
+
},
|
| 1308 |
+
{
|
| 1309 |
+
"epoch": 0.8852459016393442,
|
| 1310 |
+
"grad_norm": 68.55432141696805,
|
| 1311 |
+
"learning_rate": 3.897961019419516e-08,
|
| 1312 |
+
"logits/chosen": -1.1141546964645386,
|
| 1313 |
+
"logits/rejected": -1.046690583229065,
|
| 1314 |
+
"logps/chosen": -0.5127943754196167,
|
| 1315 |
+
"logps/rejected": -0.656581699848175,
|
| 1316 |
+
"loss": 1.9895,
|
| 1317 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 1318 |
+
"rewards/chosen": -5.127943992614746,
|
| 1319 |
+
"rewards/margins": 1.4378730058670044,
|
| 1320 |
+
"rewards/rejected": -6.565816402435303,
|
| 1321 |
+
"semantic_entropy": 1.010331630706787,
|
| 1322 |
+
"step": 405
|
| 1323 |
+
},
|
| 1324 |
+
{
|
| 1325 |
+
"epoch": 0.8961748633879781,
|
| 1326 |
+
"grad_norm": 78.80129315841073,
|
| 1327 |
+
"learning_rate": 3.192094485859526e-08,
|
| 1328 |
+
"logits/chosen": -1.1211316585540771,
|
| 1329 |
+
"logits/rejected": -1.1407296657562256,
|
| 1330 |
+
"logps/chosen": -0.5510164499282837,
|
| 1331 |
+
"logps/rejected": -0.7787143588066101,
|
| 1332 |
+
"loss": 2.0771,
|
| 1333 |
+
"rewards/accuracies": 0.762499988079071,
|
| 1334 |
+
"rewards/chosen": -5.5101637840271,
|
| 1335 |
+
"rewards/margins": 2.2769789695739746,
|
| 1336 |
+
"rewards/rejected": -7.787143707275391,
|
| 1337 |
+
"semantic_entropy": 0.9897591471672058,
|
| 1338 |
+
"step": 410
|
| 1339 |
+
},
|
| 1340 |
+
{
|
| 1341 |
+
"epoch": 0.907103825136612,
|
| 1342 |
+
"grad_norm": 63.07603955603986,
|
| 1343 |
+
"learning_rate": 2.5545911634565265e-08,
|
| 1344 |
+
"logits/chosen": -1.1598929166793823,
|
| 1345 |
+
"logits/rejected": -1.1571664810180664,
|
| 1346 |
+
"logps/chosen": -0.5616727471351624,
|
| 1347 |
+
"logps/rejected": -0.7835390567779541,
|
| 1348 |
+
"loss": 2.1028,
|
| 1349 |
+
"rewards/accuracies": 0.824999988079071,
|
| 1350 |
+
"rewards/chosen": -5.616726875305176,
|
| 1351 |
+
"rewards/margins": 2.2186641693115234,
|
| 1352 |
+
"rewards/rejected": -7.835390567779541,
|
| 1353 |
+
"semantic_entropy": 0.9976503252983093,
|
| 1354 |
+
"step": 415
|
| 1355 |
+
},
|
| 1356 |
+
{
|
| 1357 |
+
"epoch": 0.9180327868852459,
|
| 1358 |
+
"grad_norm": 89.74277499235377,
|
| 1359 |
+
"learning_rate": 1.9863821294241522e-08,
|
| 1360 |
+
"logits/chosen": -1.1581684350967407,
|
| 1361 |
+
"logits/rejected": -1.1270772218704224,
|
| 1362 |
+
"logps/chosen": -0.5167144536972046,
|
| 1363 |
+
"logps/rejected": -0.7354345321655273,
|
| 1364 |
+
"loss": 1.9853,
|
| 1365 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1366 |
+
"rewards/chosen": -5.167144298553467,
|
| 1367 |
+
"rewards/margins": 2.1872007846832275,
|
| 1368 |
+
"rewards/rejected": -7.354344844818115,
|
| 1369 |
+
"semantic_entropy": 1.002636194229126,
|
| 1370 |
+
"step": 420
|
| 1371 |
+
},
|
| 1372 |
+
{
|
| 1373 |
+
"epoch": 0.9289617486338798,
|
| 1374 |
+
"grad_norm": 71.82508706026904,
|
| 1375 |
+
"learning_rate": 1.4882972562753615e-08,
|
| 1376 |
+
"logits/chosen": -1.1459519863128662,
|
| 1377 |
+
"logits/rejected": -1.1357475519180298,
|
| 1378 |
+
"logps/chosen": -0.5768141746520996,
|
| 1379 |
+
"logps/rejected": -0.7887662053108215,
|
| 1380 |
+
"loss": 2.1228,
|
| 1381 |
+
"rewards/accuracies": 0.78125,
|
| 1382 |
+
"rewards/chosen": -5.768141746520996,
|
| 1383 |
+
"rewards/margins": 2.1195199489593506,
|
| 1384 |
+
"rewards/rejected": -7.887660980224609,
|
| 1385 |
+
"semantic_entropy": 0.9999387860298157,
|
| 1386 |
+
"step": 425
|
| 1387 |
+
},
|
| 1388 |
+
{
|
| 1389 |
+
"epoch": 0.9398907103825137,
|
| 1390 |
+
"grad_norm": 91.24129023810345,
|
| 1391 |
+
"learning_rate": 1.0610639997888915e-08,
|
| 1392 |
+
"logits/chosen": -1.0857610702514648,
|
| 1393 |
+
"logits/rejected": -1.0863049030303955,
|
| 1394 |
+
"logps/chosen": -0.520858883857727,
|
| 1395 |
+
"logps/rejected": -0.7221606969833374,
|
| 1396 |
+
"loss": 1.9229,
|
| 1397 |
+
"rewards/accuracies": 0.8687499761581421,
|
| 1398 |
+
"rewards/chosen": -5.208588600158691,
|
| 1399 |
+
"rewards/margins": 2.0130181312561035,
|
| 1400 |
+
"rewards/rejected": -7.221606254577637,
|
| 1401 |
+
"semantic_entropy": 1.0139762163162231,
|
| 1402 |
+
"step": 430
|
| 1403 |
+
},
|
| 1404 |
+
{
|
| 1405 |
+
"epoch": 0.9508196721311475,
|
| 1406 |
+
"grad_norm": 88.53510645254667,
|
| 1407 |
+
"learning_rate": 7.053063365559997e-09,
|
| 1408 |
+
"logits/chosen": -1.147918939590454,
|
| 1409 |
+
"logits/rejected": -1.1737738847732544,
|
| 1410 |
+
"logps/chosen": -0.5408393740653992,
|
| 1411 |
+
"logps/rejected": -0.7648274898529053,
|
| 1412 |
+
"loss": 2.0597,
|
| 1413 |
+
"rewards/accuracies": 0.84375,
|
| 1414 |
+
"rewards/chosen": -5.408394813537598,
|
| 1415 |
+
"rewards/margins": 2.2398805618286133,
|
| 1416 |
+
"rewards/rejected": -7.6482744216918945,
|
| 1417 |
+
"semantic_entropy": 1.0074278116226196,
|
| 1418 |
+
"step": 435
|
| 1419 |
+
},
|
| 1420 |
+
{
|
| 1421 |
+
"epoch": 0.9617486338797814,
|
| 1422 |
+
"grad_norm": 77.47275941246323,
|
| 1423 |
+
"learning_rate": 4.215438526591064e-09,
|
| 1424 |
+
"logits/chosen": -1.08914053440094,
|
| 1425 |
+
"logits/rejected": -1.0488555431365967,
|
| 1426 |
+
"logps/chosen": -0.5592411160469055,
|
| 1427 |
+
"logps/rejected": -0.7156537175178528,
|
| 1428 |
+
"loss": 2.0346,
|
| 1429 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 1430 |
+
"rewards/chosen": -5.592411518096924,
|
| 1431 |
+
"rewards/margins": 1.5641257762908936,
|
| 1432 |
+
"rewards/rejected": -7.156537055969238,
|
| 1433 |
+
"semantic_entropy": 0.9892482757568359,
|
| 1434 |
+
"step": 440
|
| 1435 |
+
},
|
| 1436 |
+
{
|
| 1437 |
+
"epoch": 0.9726775956284153,
|
| 1438 |
+
"grad_norm": 85.37542604306078,
|
| 1439 |
+
"learning_rate": 2.1019098481337426e-09,
|
| 1440 |
+
"logits/chosen": -1.1617168188095093,
|
| 1441 |
+
"logits/rejected": -1.1342122554779053,
|
| 1442 |
+
"logps/chosen": -0.5354763865470886,
|
| 1443 |
+
"logps/rejected": -0.7130267024040222,
|
| 1444 |
+
"loss": 2.0073,
|
| 1445 |
+
"rewards/accuracies": 0.8125,
|
| 1446 |
+
"rewards/chosen": -5.354763984680176,
|
| 1447 |
+
"rewards/margins": 1.7755035161972046,
|
| 1448 |
+
"rewards/rejected": -7.130267143249512,
|
| 1449 |
+
"semantic_entropy": 1.0182139873504639,
|
| 1450 |
+
"step": 445
|
| 1451 |
+
},
|
| 1452 |
+
{
|
| 1453 |
+
"epoch": 0.9836065573770492,
|
| 1454 |
+
"grad_norm": 92.55545564226749,
|
| 1455 |
+
"learning_rate": 7.155641507955445e-10,
|
| 1456 |
+
"logits/chosen": -1.0736119747161865,
|
| 1457 |
+
"logits/rejected": -1.0667097568511963,
|
| 1458 |
+
"logps/chosen": -0.5805756449699402,
|
| 1459 |
+
"logps/rejected": -0.7572126388549805,
|
| 1460 |
+
"loss": 2.146,
|
| 1461 |
+
"rewards/accuracies": 0.78125,
|
| 1462 |
+
"rewards/chosen": -5.805756568908691,
|
| 1463 |
+
"rewards/margins": 1.7663694620132446,
|
| 1464 |
+
"rewards/rejected": -7.5721259117126465,
|
| 1465 |
+
"semantic_entropy": 0.9835384488105774,
|
| 1466 |
+
"step": 450
|
| 1467 |
+
},
|
| 1468 |
+
{
|
| 1469 |
+
"epoch": 0.994535519125683,
|
| 1470 |
+
"grad_norm": 79.92510322372067,
|
| 1471 |
+
"learning_rate": 5.842620032053824e-11,
|
| 1472 |
+
"logits/chosen": -1.0938892364501953,
|
| 1473 |
+
"logits/rejected": -1.0882636308670044,
|
| 1474 |
+
"logps/chosen": -0.5764094591140747,
|
| 1475 |
+
"logps/rejected": -0.7290435433387756,
|
| 1476 |
+
"loss": 2.187,
|
| 1477 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 1478 |
+
"rewards/chosen": -5.764094829559326,
|
| 1479 |
+
"rewards/margins": 1.5263407230377197,
|
| 1480 |
+
"rewards/rejected": -7.290434837341309,
|
| 1481 |
+
"semantic_entropy": 1.0177193880081177,
|
| 1482 |
+
"step": 455
|
| 1483 |
+
},
|
| 1484 |
+
{
|
| 1485 |
+
"epoch": 0.9989071038251366,
|
| 1486 |
+
"step": 457,
|
| 1487 |
+
"total_flos": 0.0,
|
| 1488 |
+
"train_loss": 2.4655840506438875,
|
| 1489 |
+
"train_runtime": 5955.1851,
|
| 1490 |
+
"train_samples_per_second": 9.833,
|
| 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 |
+
}
|