Model save
Browse files- README.md +74 -0
- all_results.json +9 -0
- generation_config.json +8 -0
- train_results.json +9 -0
- trainer_state.json +1453 -0
README.md
ADDED
|
@@ -0,0 +1,74 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: transformers
|
| 3 |
+
license: gemma
|
| 4 |
+
base_model: google/gemma-2-9b-it
|
| 5 |
+
tags:
|
| 6 |
+
- trl
|
| 7 |
+
- simpo
|
| 8 |
+
- generated_from_trainer
|
| 9 |
+
model-index:
|
| 10 |
+
- name: simpo
|
| 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 |
+
# simpo
|
| 18 |
+
|
| 19 |
+
This model is a fine-tuned version of [google/gemma-2-9b-it](https://huggingface.co/google/gemma-2-9b-it) on an unknown dataset.
|
| 20 |
+
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 2.7580
|
| 22 |
+
- Rewards/chosen: -18.9526
|
| 23 |
+
- Rewards/rejected: -23.9387
|
| 24 |
+
- Rewards/accuracies: 0.7705
|
| 25 |
+
- Rewards/margins: 4.9861
|
| 26 |
+
- Logps/rejected: -2.3939
|
| 27 |
+
- Logps/chosen: -1.8953
|
| 28 |
+
- Logits/rejected: -14.4321
|
| 29 |
+
- Logits/chosen: -14.5024
|
| 30 |
+
|
| 31 |
+
## Model description
|
| 32 |
+
|
| 33 |
+
More information needed
|
| 34 |
+
|
| 35 |
+
## Intended uses & limitations
|
| 36 |
+
|
| 37 |
+
More information needed
|
| 38 |
+
|
| 39 |
+
## Training and evaluation data
|
| 40 |
+
|
| 41 |
+
More information needed
|
| 42 |
+
|
| 43 |
+
## Training procedure
|
| 44 |
+
|
| 45 |
+
### Training hyperparameters
|
| 46 |
+
|
| 47 |
+
The following hyperparameters were used during training:
|
| 48 |
+
- learning_rate: 8e-07
|
| 49 |
+
- train_batch_size: 2
|
| 50 |
+
- eval_batch_size: 4
|
| 51 |
+
- seed: 42
|
| 52 |
+
- distributed_type: multi-GPU
|
| 53 |
+
- num_devices: 4
|
| 54 |
+
- gradient_accumulation_steps: 16
|
| 55 |
+
- total_train_batch_size: 128
|
| 56 |
+
- total_eval_batch_size: 16
|
| 57 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 58 |
+
- lr_scheduler_type: cosine
|
| 59 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 60 |
+
- num_epochs: 1
|
| 61 |
+
|
| 62 |
+
### Training results
|
| 63 |
+
|
| 64 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 65 |
+
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 66 |
+
| 2.7196 | 0.8594 | 400 | 2.7580 | -18.9526 | -23.9387 | 0.7705 | 4.9861 | -2.3939 | -1.8953 | -14.4321 | -14.5024 |
|
| 67 |
+
|
| 68 |
+
|
| 69 |
+
### Framework versions
|
| 70 |
+
|
| 71 |
+
- Transformers 4.44.2
|
| 72 |
+
- Pytorch 2.7.0+cu128
|
| 73 |
+
- Datasets 2.18.0
|
| 74 |
+
- Tokenizers 0.19.1
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9990600241708071,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 3.541660060677477,
|
| 5 |
+
"train_runtime": 9046.6245,
|
| 6 |
+
"train_samples": 59569,
|
| 7 |
+
"train_samples_per_second": 6.585,
|
| 8 |
+
"train_steps_per_second": 0.051
|
| 9 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 2,
|
| 4 |
+
"cache_implementation": "hybrid",
|
| 5 |
+
"eos_token_id": 1,
|
| 6 |
+
"pad_token_id": 0,
|
| 7 |
+
"transformers_version": "4.44.2"
|
| 8 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9990600241708071,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 3.541660060677477,
|
| 5 |
+
"train_runtime": 9046.6245,
|
| 6 |
+
"train_samples": 59569,
|
| 7 |
+
"train_samples_per_second": 6.585,
|
| 8 |
+
"train_steps_per_second": 0.051
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,1453 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.9990600241708071,
|
| 5 |
+
"eval_steps": 400,
|
| 6 |
+
"global_step": 465,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.010742580905062441,
|
| 13 |
+
"grad_norm": 97.31650637788948,
|
| 14 |
+
"learning_rate": 8.51063829787234e-08,
|
| 15 |
+
"logits/chosen": -10.386978149414062,
|
| 16 |
+
"logits/rejected": -10.301819801330566,
|
| 17 |
+
"logps/chosen": -0.9842015504837036,
|
| 18 |
+
"logps/rejected": -0.9797419309616089,
|
| 19 |
+
"loss": 6.1427,
|
| 20 |
+
"rewards/accuracies": 0.5062500238418579,
|
| 21 |
+
"rewards/chosen": -9.84201431274414,
|
| 22 |
+
"rewards/margins": -0.044595133513212204,
|
| 23 |
+
"rewards/rejected": -9.797419548034668,
|
| 24 |
+
"step": 5
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.021485161810124883,
|
| 28 |
+
"grad_norm": 120.32296756999746,
|
| 29 |
+
"learning_rate": 1.702127659574468e-07,
|
| 30 |
+
"logits/chosen": -10.528864860534668,
|
| 31 |
+
"logits/rejected": -10.550542831420898,
|
| 32 |
+
"logps/chosen": -1.1053581237792969,
|
| 33 |
+
"logps/rejected": -1.0148571729660034,
|
| 34 |
+
"loss": 5.8489,
|
| 35 |
+
"rewards/accuracies": 0.5062500238418579,
|
| 36 |
+
"rewards/chosen": -11.053580284118652,
|
| 37 |
+
"rewards/margins": -0.9050087928771973,
|
| 38 |
+
"rewards/rejected": -10.148571968078613,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.03222774271518732,
|
| 43 |
+
"grad_norm": 63.03552892163084,
|
| 44 |
+
"learning_rate": 2.553191489361702e-07,
|
| 45 |
+
"logits/chosen": -10.048526763916016,
|
| 46 |
+
"logits/rejected": -9.960186004638672,
|
| 47 |
+
"logps/chosen": -1.2764005661010742,
|
| 48 |
+
"logps/rejected": -1.1620838642120361,
|
| 49 |
+
"loss": 6.2427,
|
| 50 |
+
"rewards/accuracies": 0.4749999940395355,
|
| 51 |
+
"rewards/chosen": -12.764005661010742,
|
| 52 |
+
"rewards/margins": -1.1431667804718018,
|
| 53 |
+
"rewards/rejected": -11.62083911895752,
|
| 54 |
+
"step": 15
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.042970323620249766,
|
| 58 |
+
"grad_norm": 115.65630638049038,
|
| 59 |
+
"learning_rate": 3.404255319148936e-07,
|
| 60 |
+
"logits/chosen": -9.938735961914062,
|
| 61 |
+
"logits/rejected": -10.032970428466797,
|
| 62 |
+
"logps/chosen": -0.886467456817627,
|
| 63 |
+
"logps/rejected": -0.9640073776245117,
|
| 64 |
+
"loss": 5.5509,
|
| 65 |
+
"rewards/accuracies": 0.4749999940395355,
|
| 66 |
+
"rewards/chosen": -8.86467456817627,
|
| 67 |
+
"rewards/margins": 0.775399386882782,
|
| 68 |
+
"rewards/rejected": -9.640073776245117,
|
| 69 |
+
"step": 20
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.05371290452531221,
|
| 73 |
+
"grad_norm": 96.51643381412,
|
| 74 |
+
"learning_rate": 4.25531914893617e-07,
|
| 75 |
+
"logits/chosen": -9.94281005859375,
|
| 76 |
+
"logits/rejected": -9.81358814239502,
|
| 77 |
+
"logps/chosen": -0.8847745060920715,
|
| 78 |
+
"logps/rejected": -0.9943971633911133,
|
| 79 |
+
"loss": 5.2621,
|
| 80 |
+
"rewards/accuracies": 0.4937500059604645,
|
| 81 |
+
"rewards/chosen": -8.84774398803711,
|
| 82 |
+
"rewards/margins": 1.0962274074554443,
|
| 83 |
+
"rewards/rejected": -9.943971633911133,
|
| 84 |
+
"step": 25
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.06445548543037465,
|
| 88 |
+
"grad_norm": 79.98872357640575,
|
| 89 |
+
"learning_rate": 5.106382978723404e-07,
|
| 90 |
+
"logits/chosen": -9.344173431396484,
|
| 91 |
+
"logits/rejected": -9.224719047546387,
|
| 92 |
+
"logps/chosen": -0.6412209272384644,
|
| 93 |
+
"logps/rejected": -0.6745115518569946,
|
| 94 |
+
"loss": 5.0537,
|
| 95 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 96 |
+
"rewards/chosen": -6.412209510803223,
|
| 97 |
+
"rewards/margins": 0.3329065442085266,
|
| 98 |
+
"rewards/rejected": -6.745115756988525,
|
| 99 |
+
"step": 30
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.07519806633543709,
|
| 103 |
+
"grad_norm": 90.98736128158838,
|
| 104 |
+
"learning_rate": 5.957446808510638e-07,
|
| 105 |
+
"logits/chosen": -8.797516822814941,
|
| 106 |
+
"logits/rejected": -8.775238037109375,
|
| 107 |
+
"logps/chosen": -0.7022872567176819,
|
| 108 |
+
"logps/rejected": -0.6628175973892212,
|
| 109 |
+
"loss": 5.2242,
|
| 110 |
+
"rewards/accuracies": 0.5375000238418579,
|
| 111 |
+
"rewards/chosen": -7.0228729248046875,
|
| 112 |
+
"rewards/margins": -0.3946966230869293,
|
| 113 |
+
"rewards/rejected": -6.628176212310791,
|
| 114 |
+
"step": 35
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.08594064724049953,
|
| 118 |
+
"grad_norm": 85.27765487288175,
|
| 119 |
+
"learning_rate": 6.808510638297872e-07,
|
| 120 |
+
"logits/chosen": -8.165349006652832,
|
| 121 |
+
"logits/rejected": -7.939520835876465,
|
| 122 |
+
"logps/chosen": -0.6976481676101685,
|
| 123 |
+
"logps/rejected": -0.7118233442306519,
|
| 124 |
+
"loss": 4.8954,
|
| 125 |
+
"rewards/accuracies": 0.5687500238418579,
|
| 126 |
+
"rewards/chosen": -6.976480960845947,
|
| 127 |
+
"rewards/margins": 0.14175233244895935,
|
| 128 |
+
"rewards/rejected": -7.118234157562256,
|
| 129 |
+
"step": 40
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.09668322814556198,
|
| 133 |
+
"grad_norm": 60.0810198108563,
|
| 134 |
+
"learning_rate": 7.659574468085107e-07,
|
| 135 |
+
"logits/chosen": -8.638801574707031,
|
| 136 |
+
"logits/rejected": -8.301128387451172,
|
| 137 |
+
"logps/chosen": -0.5912588834762573,
|
| 138 |
+
"logps/rejected": -0.643363356590271,
|
| 139 |
+
"loss": 4.5924,
|
| 140 |
+
"rewards/accuracies": 0.625,
|
| 141 |
+
"rewards/chosen": -5.912589073181152,
|
| 142 |
+
"rewards/margins": 0.5210448503494263,
|
| 143 |
+
"rewards/rejected": -6.433633327484131,
|
| 144 |
+
"step": 45
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.10742580905062442,
|
| 148 |
+
"grad_norm": 56.311443110110645,
|
| 149 |
+
"learning_rate": 7.998983280184396e-07,
|
| 150 |
+
"logits/chosen": -8.66537857055664,
|
| 151 |
+
"logits/rejected": -8.455618858337402,
|
| 152 |
+
"logps/chosen": -0.532910943031311,
|
| 153 |
+
"logps/rejected": -0.5861265063285828,
|
| 154 |
+
"loss": 4.6603,
|
| 155 |
+
"rewards/accuracies": 0.581250011920929,
|
| 156 |
+
"rewards/chosen": -5.3291096687316895,
|
| 157 |
+
"rewards/margins": 0.5321555733680725,
|
| 158 |
+
"rewards/rejected": -5.861265659332275,
|
| 159 |
+
"step": 50
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.11816838995568685,
|
| 163 |
+
"grad_norm": 48.22964437945852,
|
| 164 |
+
"learning_rate": 7.992771864078597e-07,
|
| 165 |
+
"logits/chosen": -8.251328468322754,
|
| 166 |
+
"logits/rejected": -8.19440746307373,
|
| 167 |
+
"logps/chosen": -0.6145357489585876,
|
| 168 |
+
"logps/rejected": -0.7057743072509766,
|
| 169 |
+
"loss": 4.6648,
|
| 170 |
+
"rewards/accuracies": 0.5625,
|
| 171 |
+
"rewards/chosen": -6.145357131958008,
|
| 172 |
+
"rewards/margins": 0.9123857617378235,
|
| 173 |
+
"rewards/rejected": -7.057743072509766,
|
| 174 |
+
"step": 55
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 0.1289109708607493,
|
| 178 |
+
"grad_norm": 58.650484172589515,
|
| 179 |
+
"learning_rate": 7.980922636120897e-07,
|
| 180 |
+
"logits/chosen": -8.88660717010498,
|
| 181 |
+
"logits/rejected": -8.566266059875488,
|
| 182 |
+
"logps/chosen": -0.5707553625106812,
|
| 183 |
+
"logps/rejected": -0.6651050448417664,
|
| 184 |
+
"loss": 4.4872,
|
| 185 |
+
"rewards/accuracies": 0.581250011920929,
|
| 186 |
+
"rewards/chosen": -5.707553863525391,
|
| 187 |
+
"rewards/margins": 0.9434973001480103,
|
| 188 |
+
"rewards/rejected": -6.6510515213012695,
|
| 189 |
+
"step": 60
|
| 190 |
+
},
|
| 191 |
+
{
|
| 192 |
+
"epoch": 0.13965355176581173,
|
| 193 |
+
"grad_norm": 50.701519232478496,
|
| 194 |
+
"learning_rate": 7.963452327474534e-07,
|
| 195 |
+
"logits/chosen": -8.946069717407227,
|
| 196 |
+
"logits/rejected": -8.80902099609375,
|
| 197 |
+
"logps/chosen": -0.6769601702690125,
|
| 198 |
+
"logps/rejected": -0.7762205600738525,
|
| 199 |
+
"loss": 4.6697,
|
| 200 |
+
"rewards/accuracies": 0.5687500238418579,
|
| 201 |
+
"rewards/chosen": -6.769601345062256,
|
| 202 |
+
"rewards/margins": 0.9926047325134277,
|
| 203 |
+
"rewards/rejected": -7.762206077575684,
|
| 204 |
+
"step": 65
|
| 205 |
+
},
|
| 206 |
+
{
|
| 207 |
+
"epoch": 0.15039613267087418,
|
| 208 |
+
"grad_norm": 55.73800060229348,
|
| 209 |
+
"learning_rate": 7.940385606293987e-07,
|
| 210 |
+
"logits/chosen": -8.857608795166016,
|
| 211 |
+
"logits/rejected": -8.605030059814453,
|
| 212 |
+
"logps/chosen": -0.6864951252937317,
|
| 213 |
+
"logps/rejected": -0.7849037647247314,
|
| 214 |
+
"loss": 4.579,
|
| 215 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 216 |
+
"rewards/chosen": -6.864950656890869,
|
| 217 |
+
"rewards/margins": 0.9840868711471558,
|
| 218 |
+
"rewards/rejected": -7.849038124084473,
|
| 219 |
+
"step": 70
|
| 220 |
+
},
|
| 221 |
+
{
|
| 222 |
+
"epoch": 0.16113871357593662,
|
| 223 |
+
"grad_norm": 61.45336329452104,
|
| 224 |
+
"learning_rate": 7.911755042893434e-07,
|
| 225 |
+
"logits/chosen": -9.510955810546875,
|
| 226 |
+
"logits/rejected": -9.195249557495117,
|
| 227 |
+
"logps/chosen": -0.6112648248672485,
|
| 228 |
+
"logps/rejected": -0.7157038450241089,
|
| 229 |
+
"loss": 4.4206,
|
| 230 |
+
"rewards/accuracies": 0.637499988079071,
|
| 231 |
+
"rewards/chosen": -6.112648010253906,
|
| 232 |
+
"rewards/margins": 1.044390320777893,
|
| 233 |
+
"rewards/rejected": -7.157038688659668,
|
| 234 |
+
"step": 75
|
| 235 |
+
},
|
| 236 |
+
{
|
| 237 |
+
"epoch": 0.17188129448099906,
|
| 238 |
+
"grad_norm": 63.05991232357819,
|
| 239 |
+
"learning_rate": 7.877601063757321e-07,
|
| 240 |
+
"logits/chosen": -8.919134140014648,
|
| 241 |
+
"logits/rejected": -8.694350242614746,
|
| 242 |
+
"logps/chosen": -0.6809953451156616,
|
| 243 |
+
"logps/rejected": -0.8292111158370972,
|
| 244 |
+
"loss": 4.3782,
|
| 245 |
+
"rewards/accuracies": 0.675000011920929,
|
| 246 |
+
"rewards/chosen": -6.809953212738037,
|
| 247 |
+
"rewards/margins": 1.4821574687957764,
|
| 248 |
+
"rewards/rejected": -8.29211139678955,
|
| 249 |
+
"step": 80
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 0.1826238753860615,
|
| 253 |
+
"grad_norm": 80.14192606614782,
|
| 254 |
+
"learning_rate": 7.837971894457989e-07,
|
| 255 |
+
"logits/chosen": -9.203702926635742,
|
| 256 |
+
"logits/rejected": -8.941718101501465,
|
| 257 |
+
"logps/chosen": -0.6751728653907776,
|
| 258 |
+
"logps/rejected": -0.8182951807975769,
|
| 259 |
+
"loss": 4.6537,
|
| 260 |
+
"rewards/accuracies": 0.606249988079071,
|
| 261 |
+
"rewards/chosen": -6.751728057861328,
|
| 262 |
+
"rewards/margins": 1.4312243461608887,
|
| 263 |
+
"rewards/rejected": -8.182951927185059,
|
| 264 |
+
"step": 85
|
| 265 |
+
},
|
| 266 |
+
{
|
| 267 |
+
"epoch": 0.19336645629112395,
|
| 268 |
+
"grad_norm": 57.0890392656848,
|
| 269 |
+
"learning_rate": 7.792923491560942e-07,
|
| 270 |
+
"logits/chosen": -8.762784004211426,
|
| 271 |
+
"logits/rejected": -8.619606018066406,
|
| 272 |
+
"logps/chosen": -0.6792052388191223,
|
| 273 |
+
"logps/rejected": -0.782240092754364,
|
| 274 |
+
"loss": 4.4488,
|
| 275 |
+
"rewards/accuracies": 0.612500011920929,
|
| 276 |
+
"rewards/chosen": -6.792051792144775,
|
| 277 |
+
"rewards/margins": 1.0303497314453125,
|
| 278 |
+
"rewards/rejected": -7.822402000427246,
|
| 279 |
+
"step": 90
|
| 280 |
+
},
|
| 281 |
+
{
|
| 282 |
+
"epoch": 0.2041090371961864,
|
| 283 |
+
"grad_norm": 62.86804177211711,
|
| 284 |
+
"learning_rate": 7.742519463613926e-07,
|
| 285 |
+
"logits/chosen": -9.201288223266602,
|
| 286 |
+
"logits/rejected": -9.108887672424316,
|
| 287 |
+
"logps/chosen": -0.6846643090248108,
|
| 288 |
+
"logps/rejected": -0.8157285451889038,
|
| 289 |
+
"loss": 4.2734,
|
| 290 |
+
"rewards/accuracies": 0.574999988079071,
|
| 291 |
+
"rewards/chosen": -6.846642971038818,
|
| 292 |
+
"rewards/margins": 1.3106427192687988,
|
| 293 |
+
"rewards/rejected": -8.157285690307617,
|
| 294 |
+
"step": 95
|
| 295 |
+
},
|
| 296 |
+
{
|
| 297 |
+
"epoch": 0.21485161810124884,
|
| 298 |
+
"grad_norm": 156.13974754079638,
|
| 299 |
+
"learning_rate": 7.68683098133138e-07,
|
| 300 |
+
"logits/chosen": -8.922677040100098,
|
| 301 |
+
"logits/rejected": -8.721606254577637,
|
| 302 |
+
"logps/chosen": -0.7626782655715942,
|
| 303 |
+
"logps/rejected": -0.8957880139350891,
|
| 304 |
+
"loss": 4.4059,
|
| 305 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 306 |
+
"rewards/chosen": -7.626782417297363,
|
| 307 |
+
"rewards/margins": 1.3310966491699219,
|
| 308 |
+
"rewards/rejected": -8.957880020141602,
|
| 309 |
+
"step": 100
|
| 310 |
+
},
|
| 311 |
+
{
|
| 312 |
+
"epoch": 0.22559419900631128,
|
| 313 |
+
"grad_norm": 105.18853752160656,
|
| 314 |
+
"learning_rate": 7.625936677101051e-07,
|
| 315 |
+
"logits/chosen": -9.266253471374512,
|
| 316 |
+
"logits/rejected": -9.21663761138916,
|
| 317 |
+
"logps/chosen": -0.7982575297355652,
|
| 318 |
+
"logps/rejected": -0.824684739112854,
|
| 319 |
+
"loss": 4.4496,
|
| 320 |
+
"rewards/accuracies": 0.5625,
|
| 321 |
+
"rewards/chosen": -7.9825758934021,
|
| 322 |
+
"rewards/margins": 0.2642715573310852,
|
| 323 |
+
"rewards/rejected": -8.246847152709961,
|
| 324 |
+
"step": 105
|
| 325 |
+
},
|
| 326 |
+
{
|
| 327 |
+
"epoch": 0.2363367799113737,
|
| 328 |
+
"grad_norm": 55.591763777771575,
|
| 329 |
+
"learning_rate": 7.559922533954731e-07,
|
| 330 |
+
"logits/chosen": -9.533427238464355,
|
| 331 |
+
"logits/rejected": -9.415986061096191,
|
| 332 |
+
"logps/chosen": -0.760522723197937,
|
| 333 |
+
"logps/rejected": -0.880321204662323,
|
| 334 |
+
"loss": 4.2742,
|
| 335 |
+
"rewards/accuracies": 0.581250011920929,
|
| 336 |
+
"rewards/chosen": -7.605226039886475,
|
| 337 |
+
"rewards/margins": 1.1979866027832031,
|
| 338 |
+
"rewards/rejected": -8.803213119506836,
|
| 339 |
+
"step": 110
|
| 340 |
+
},
|
| 341 |
+
{
|
| 342 |
+
"epoch": 0.24707936081643614,
|
| 343 |
+
"grad_norm": 68.68376533729582,
|
| 344 |
+
"learning_rate": 7.488881764159808e-07,
|
| 345 |
+
"logits/chosen": -10.321605682373047,
|
| 346 |
+
"logits/rejected": -10.032452583312988,
|
| 347 |
+
"logps/chosen": -0.7515496015548706,
|
| 348 |
+
"logps/rejected": -0.8424334526062012,
|
| 349 |
+
"loss": 4.146,
|
| 350 |
+
"rewards/accuracies": 0.637499988079071,
|
| 351 |
+
"rewards/chosen": -7.515496730804443,
|
| 352 |
+
"rewards/margins": 0.9088379740715027,
|
| 353 |
+
"rewards/rejected": -8.424333572387695,
|
| 354 |
+
"step": 115
|
| 355 |
+
},
|
| 356 |
+
{
|
| 357 |
+
"epoch": 0.2578219417214986,
|
| 358 |
+
"grad_norm": 84.12336331826019,
|
| 359 |
+
"learning_rate": 7.412914677603135e-07,
|
| 360 |
+
"logits/chosen": -9.94493293762207,
|
| 361 |
+
"logits/rejected": -9.746753692626953,
|
| 362 |
+
"logps/chosen": -0.8637507557868958,
|
| 363 |
+
"logps/rejected": -0.9913690686225891,
|
| 364 |
+
"loss": 4.2196,
|
| 365 |
+
"rewards/accuracies": 0.59375,
|
| 366 |
+
"rewards/chosen": -8.637508392333984,
|
| 367 |
+
"rewards/margins": 1.2761822938919067,
|
| 368 |
+
"rewards/rejected": -9.913690567016602,
|
| 369 |
+
"step": 120
|
| 370 |
+
},
|
| 371 |
+
{
|
| 372 |
+
"epoch": 0.26856452262656105,
|
| 373 |
+
"grad_norm": 65.56256589725982,
|
| 374 |
+
"learning_rate": 7.332128540153017e-07,
|
| 375 |
+
"logits/chosen": -11.014134407043457,
|
| 376 |
+
"logits/rejected": -10.658555030822754,
|
| 377 |
+
"logps/chosen": -0.7951505184173584,
|
| 378 |
+
"logps/rejected": -0.9381753206253052,
|
| 379 |
+
"loss": 4.0983,
|
| 380 |
+
"rewards/accuracies": 0.65625,
|
| 381 |
+
"rewards/chosen": -7.951504707336426,
|
| 382 |
+
"rewards/margins": 1.430248498916626,
|
| 383 |
+
"rewards/rejected": -9.381753921508789,
|
| 384 |
+
"step": 125
|
| 385 |
+
},
|
| 386 |
+
{
|
| 387 |
+
"epoch": 0.27930710353162347,
|
| 388 |
+
"grad_norm": 65.20658081437091,
|
| 389 |
+
"learning_rate": 7.246637422199322e-07,
|
| 390 |
+
"logits/chosen": -10.825661659240723,
|
| 391 |
+
"logits/rejected": -10.685798645019531,
|
| 392 |
+
"logps/chosen": -0.867350697517395,
|
| 393 |
+
"logps/rejected": -1.0283457040786743,
|
| 394 |
+
"loss": 4.0614,
|
| 395 |
+
"rewards/accuracies": 0.675000011920929,
|
| 396 |
+
"rewards/chosen": -8.673507690429688,
|
| 397 |
+
"rewards/margins": 1.6099493503570557,
|
| 398 |
+
"rewards/rejected": -10.283456802368164,
|
| 399 |
+
"step": 130
|
| 400 |
+
},
|
| 401 |
+
{
|
| 402 |
+
"epoch": 0.29004968443668594,
|
| 403 |
+
"grad_norm": 71.60093594488256,
|
| 404 |
+
"learning_rate": 7.156562037585574e-07,
|
| 405 |
+
"logits/chosen": -11.534868240356445,
|
| 406 |
+
"logits/rejected": -11.279884338378906,
|
| 407 |
+
"logps/chosen": -0.8583852648735046,
|
| 408 |
+
"logps/rejected": -1.0596177577972412,
|
| 409 |
+
"loss": 3.9342,
|
| 410 |
+
"rewards/accuracies": 0.71875,
|
| 411 |
+
"rewards/chosen": -8.583852767944336,
|
| 412 |
+
"rewards/margins": 2.012324810028076,
|
| 413 |
+
"rewards/rejected": -10.59617805480957,
|
| 414 |
+
"step": 135
|
| 415 |
+
},
|
| 416 |
+
{
|
| 417 |
+
"epoch": 0.30079226534174835,
|
| 418 |
+
"grad_norm": 90.84373325771121,
|
| 419 |
+
"learning_rate": 7.062029573160467e-07,
|
| 420 |
+
"logits/chosen": -12.217193603515625,
|
| 421 |
+
"logits/rejected": -11.932929039001465,
|
| 422 |
+
"logps/chosen": -0.9700864553451538,
|
| 423 |
+
"logps/rejected": -1.183528184890747,
|
| 424 |
+
"loss": 3.69,
|
| 425 |
+
"rewards/accuracies": 0.737500011920929,
|
| 426 |
+
"rewards/chosen": -9.700864791870117,
|
| 427 |
+
"rewards/margins": 2.1344172954559326,
|
| 428 |
+
"rewards/rejected": -11.835283279418945,
|
| 429 |
+
"step": 140
|
| 430 |
+
},
|
| 431 |
+
{
|
| 432 |
+
"epoch": 0.31153484624681077,
|
| 433 |
+
"grad_norm": 108.66029102385438,
|
| 434 |
+
"learning_rate": 6.963173509189455e-07,
|
| 435 |
+
"logits/chosen": -13.757959365844727,
|
| 436 |
+
"logits/rejected": -13.603772163391113,
|
| 437 |
+
"logps/chosen": -1.109348177909851,
|
| 438 |
+
"logps/rejected": -1.3448617458343506,
|
| 439 |
+
"loss": 3.808,
|
| 440 |
+
"rewards/accuracies": 0.59375,
|
| 441 |
+
"rewards/chosen": -11.093481063842773,
|
| 442 |
+
"rewards/margins": 2.355137348175049,
|
| 443 |
+
"rewards/rejected": -13.448617935180664,
|
| 444 |
+
"step": 145
|
| 445 |
+
},
|
| 446 |
+
{
|
| 447 |
+
"epoch": 0.32227742715187324,
|
| 448 |
+
"grad_norm": 120.43718591165803,
|
| 449 |
+
"learning_rate": 6.860133430880024e-07,
|
| 450 |
+
"logits/chosen": -14.138008117675781,
|
| 451 |
+
"logits/rejected": -13.760258674621582,
|
| 452 |
+
"logps/chosen": -1.2876828908920288,
|
| 453 |
+
"logps/rejected": -1.5249125957489014,
|
| 454 |
+
"loss": 3.7404,
|
| 455 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 456 |
+
"rewards/chosen": -12.876829147338867,
|
| 457 |
+
"rewards/margins": 2.3722963333129883,
|
| 458 |
+
"rewards/rejected": -15.249125480651855,
|
| 459 |
+
"step": 150
|
| 460 |
+
},
|
| 461 |
+
{
|
| 462 |
+
"epoch": 0.33302000805693566,
|
| 463 |
+
"grad_norm": 110.47332053227805,
|
| 464 |
+
"learning_rate": 6.753054831286747e-07,
|
| 465 |
+
"logits/chosen": -14.456746101379395,
|
| 466 |
+
"logits/rejected": -14.32475471496582,
|
| 467 |
+
"logps/chosen": -1.3068922758102417,
|
| 468 |
+
"logps/rejected": -1.516629934310913,
|
| 469 |
+
"loss": 3.8826,
|
| 470 |
+
"rewards/accuracies": 0.6875,
|
| 471 |
+
"rewards/chosen": -13.06892204284668,
|
| 472 |
+
"rewards/margins": 2.097377300262451,
|
| 473 |
+
"rewards/rejected": -15.166299819946289,
|
| 474 |
+
"step": 155
|
| 475 |
+
},
|
| 476 |
+
{
|
| 477 |
+
"epoch": 0.34376258896199813,
|
| 478 |
+
"grad_norm": 280.21180497264857,
|
| 479 |
+
"learning_rate": 6.642088905874433e-07,
|
| 480 |
+
"logits/chosen": -14.917425155639648,
|
| 481 |
+
"logits/rejected": -14.61968994140625,
|
| 482 |
+
"logps/chosen": -1.3311243057250977,
|
| 483 |
+
"logps/rejected": -1.5419546365737915,
|
| 484 |
+
"loss": 3.7542,
|
| 485 |
+
"rewards/accuracies": 0.706250011920929,
|
| 486 |
+
"rewards/chosen": -13.311243057250977,
|
| 487 |
+
"rewards/margins": 2.108302593231201,
|
| 488 |
+
"rewards/rejected": -15.41954517364502,
|
| 489 |
+
"step": 160
|
| 490 |
+
},
|
| 491 |
+
{
|
| 492 |
+
"epoch": 0.35450516986706054,
|
| 493 |
+
"grad_norm": 97.70985018206724,
|
| 494 |
+
"learning_rate": 6.527392339029455e-07,
|
| 495 |
+
"logits/chosen": -14.492767333984375,
|
| 496 |
+
"logits/rejected": -14.438491821289062,
|
| 497 |
+
"logps/chosen": -1.2500733137130737,
|
| 498 |
+
"logps/rejected": -1.525674819946289,
|
| 499 |
+
"loss": 3.4458,
|
| 500 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 501 |
+
"rewards/chosen": -12.500734329223633,
|
| 502 |
+
"rewards/margins": 2.756016254425049,
|
| 503 |
+
"rewards/rejected": -15.256749153137207,
|
| 504 |
+
"step": 165
|
| 505 |
+
},
|
| 506 |
+
{
|
| 507 |
+
"epoch": 0.365247750772123,
|
| 508 |
+
"grad_norm": 115.13192096245145,
|
| 509 |
+
"learning_rate": 6.409127082820689e-07,
|
| 510 |
+
"logits/chosen": -14.65284252166748,
|
| 511 |
+
"logits/rejected": -14.798192977905273,
|
| 512 |
+
"logps/chosen": -1.573756217956543,
|
| 513 |
+
"logps/rejected": -1.85759699344635,
|
| 514 |
+
"loss": 3.6649,
|
| 515 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 516 |
+
"rewards/chosen": -15.73755931854248,
|
| 517 |
+
"rewards/margins": 2.838409900665283,
|
| 518 |
+
"rewards/rejected": -18.575969696044922,
|
| 519 |
+
"step": 170
|
| 520 |
+
},
|
| 521 |
+
{
|
| 522 |
+
"epoch": 0.37599033167718543,
|
| 523 |
+
"grad_norm": 120.51325866275113,
|
| 524 |
+
"learning_rate": 6.287460128322457e-07,
|
| 525 |
+
"logits/chosen": -15.47950553894043,
|
| 526 |
+
"logits/rejected": -15.095005989074707,
|
| 527 |
+
"logps/chosen": -1.6204099655151367,
|
| 528 |
+
"logps/rejected": -1.9557273387908936,
|
| 529 |
+
"loss": 3.4241,
|
| 530 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 531 |
+
"rewards/chosen": -16.2041015625,
|
| 532 |
+
"rewards/margins": 3.3531746864318848,
|
| 533 |
+
"rewards/rejected": -19.557273864746094,
|
| 534 |
+
"step": 175
|
| 535 |
+
},
|
| 536 |
+
{
|
| 537 |
+
"epoch": 0.3867329125822479,
|
| 538 |
+
"grad_norm": 177.63227168183283,
|
| 539 |
+
"learning_rate": 6.16256326982239e-07,
|
| 540 |
+
"logits/chosen": -15.96112060546875,
|
| 541 |
+
"logits/rejected": -15.974037170410156,
|
| 542 |
+
"logps/chosen": -1.5117028951644897,
|
| 543 |
+
"logps/rejected": -1.9151155948638916,
|
| 544 |
+
"loss": 3.4091,
|
| 545 |
+
"rewards/accuracies": 0.731249988079071,
|
| 546 |
+
"rewards/chosen": -15.117030143737793,
|
| 547 |
+
"rewards/margins": 4.034125328063965,
|
| 548 |
+
"rewards/rejected": -19.151155471801758,
|
| 549 |
+
"step": 180
|
| 550 |
+
},
|
| 551 |
+
{
|
| 552 |
+
"epoch": 0.3974754934873103,
|
| 553 |
+
"grad_norm": 119.1370485384488,
|
| 554 |
+
"learning_rate": 6.034612862247114e-07,
|
| 555 |
+
"logits/chosen": -15.26576042175293,
|
| 556 |
+
"logits/rejected": -15.074376106262207,
|
| 557 |
+
"logps/chosen": -1.525244951248169,
|
| 558 |
+
"logps/rejected": -1.833696722984314,
|
| 559 |
+
"loss": 3.105,
|
| 560 |
+
"rewards/accuracies": 0.75,
|
| 561 |
+
"rewards/chosen": -15.252447128295898,
|
| 562 |
+
"rewards/margins": 3.0845184326171875,
|
| 563 |
+
"rewards/rejected": -18.336965560913086,
|
| 564 |
+
"step": 185
|
| 565 |
+
},
|
| 566 |
+
{
|
| 567 |
+
"epoch": 0.4082180743923728,
|
| 568 |
+
"grad_norm": 115.37836093687133,
|
| 569 |
+
"learning_rate": 5.903789572148295e-07,
|
| 570 |
+
"logits/chosen": -14.956746101379395,
|
| 571 |
+
"logits/rejected": -14.613385200500488,
|
| 572 |
+
"logps/chosen": -1.6238969564437866,
|
| 573 |
+
"logps/rejected": -2.0761537551879883,
|
| 574 |
+
"loss": 3.3782,
|
| 575 |
+
"rewards/accuracies": 0.71875,
|
| 576 |
+
"rewards/chosen": -16.238969802856445,
|
| 577 |
+
"rewards/margins": 4.522566795349121,
|
| 578 |
+
"rewards/rejected": -20.76153564453125,
|
| 579 |
+
"step": 190
|
| 580 |
+
},
|
| 581 |
+
{
|
| 582 |
+
"epoch": 0.4189606552974352,
|
| 583 |
+
"grad_norm": 129.58691495445163,
|
| 584 |
+
"learning_rate": 5.770278122600662e-07,
|
| 585 |
+
"logits/chosen": -15.031437873840332,
|
| 586 |
+
"logits/rejected": -14.86229133605957,
|
| 587 |
+
"logps/chosen": -1.686156988143921,
|
| 588 |
+
"logps/rejected": -2.002657890319824,
|
| 589 |
+
"loss": 3.5752,
|
| 590 |
+
"rewards/accuracies": 0.675000011920929,
|
| 591 |
+
"rewards/chosen": -16.861570358276367,
|
| 592 |
+
"rewards/margins": 3.165009021759033,
|
| 593 |
+
"rewards/rejected": -20.026578903198242,
|
| 594 |
+
"step": 195
|
| 595 |
+
},
|
| 596 |
+
{
|
| 597 |
+
"epoch": 0.4297032362024977,
|
| 598 |
+
"grad_norm": 103.32755261822129,
|
| 599 |
+
"learning_rate": 5.634267032372192e-07,
|
| 600 |
+
"logits/chosen": -15.254351615905762,
|
| 601 |
+
"logits/rejected": -15.037300109863281,
|
| 602 |
+
"logps/chosen": -1.6978883743286133,
|
| 603 |
+
"logps/rejected": -2.0401151180267334,
|
| 604 |
+
"loss": 3.3862,
|
| 605 |
+
"rewards/accuracies": 0.737500011920929,
|
| 606 |
+
"rewards/chosen": -16.978879928588867,
|
| 607 |
+
"rewards/margins": 3.4222705364227295,
|
| 608 |
+
"rewards/rejected": -20.401153564453125,
|
| 609 |
+
"step": 200
|
| 610 |
+
},
|
| 611 |
+
{
|
| 612 |
+
"epoch": 0.4404458171075601,
|
| 613 |
+
"grad_norm": 128.218280047368,
|
| 614 |
+
"learning_rate": 5.495948349734758e-07,
|
| 615 |
+
"logits/chosen": -14.857444763183594,
|
| 616 |
+
"logits/rejected": -14.940084457397461,
|
| 617 |
+
"logps/chosen": -1.7693378925323486,
|
| 618 |
+
"logps/rejected": -2.061185836791992,
|
| 619 |
+
"loss": 3.1236,
|
| 620 |
+
"rewards/accuracies": 0.668749988079071,
|
| 621 |
+
"rewards/chosen": -17.693378448486328,
|
| 622 |
+
"rewards/margins": 2.91848087310791,
|
| 623 |
+
"rewards/rejected": -20.611858367919922,
|
| 624 |
+
"step": 205
|
| 625 |
+
},
|
| 626 |
+
{
|
| 627 |
+
"epoch": 0.45118839801262256,
|
| 628 |
+
"grad_norm": 107.84849265892646,
|
| 629 |
+
"learning_rate": 5.355517381291105e-07,
|
| 630 |
+
"logits/chosen": -16.1163330078125,
|
| 631 |
+
"logits/rejected": -15.79603099822998,
|
| 632 |
+
"logps/chosen": -1.7783616781234741,
|
| 633 |
+
"logps/rejected": -2.141852378845215,
|
| 634 |
+
"loss": 3.1842,
|
| 635 |
+
"rewards/accuracies": 0.71875,
|
| 636 |
+
"rewards/chosen": -17.783615112304688,
|
| 637 |
+
"rewards/margins": 3.6349079608917236,
|
| 638 |
+
"rewards/rejected": -21.41852378845215,
|
| 639 |
+
"step": 210
|
| 640 |
+
},
|
| 641 |
+
{
|
| 642 |
+
"epoch": 0.461930978917685,
|
| 643 |
+
"grad_norm": 108.45122155813243,
|
| 644 |
+
"learning_rate": 5.21317241620105e-07,
|
| 645 |
+
"logits/chosen": -16.612058639526367,
|
| 646 |
+
"logits/rejected": -16.40911293029785,
|
| 647 |
+
"logps/chosen": -1.7623794078826904,
|
| 648 |
+
"logps/rejected": -2.1722538471221924,
|
| 649 |
+
"loss": 3.4158,
|
| 650 |
+
"rewards/accuracies": 0.737500011920929,
|
| 651 |
+
"rewards/chosen": -17.623790740966797,
|
| 652 |
+
"rewards/margins": 4.0987443923950195,
|
| 653 |
+
"rewards/rejected": -21.722537994384766,
|
| 654 |
+
"step": 215
|
| 655 |
+
},
|
| 656 |
+
{
|
| 657 |
+
"epoch": 0.4726735598227474,
|
| 658 |
+
"grad_norm": 115.68303627040952,
|
| 659 |
+
"learning_rate": 5.069114446196291e-07,
|
| 660 |
+
"logits/chosen": -14.818249702453613,
|
| 661 |
+
"logits/rejected": -14.463000297546387,
|
| 662 |
+
"logps/chosen": -1.832088828086853,
|
| 663 |
+
"logps/rejected": -2.3377938270568848,
|
| 664 |
+
"loss": 3.1208,
|
| 665 |
+
"rewards/accuracies": 0.8125,
|
| 666 |
+
"rewards/chosen": -18.32088851928711,
|
| 667 |
+
"rewards/margins": 5.057046890258789,
|
| 668 |
+
"rewards/rejected": -23.37793731689453,
|
| 669 |
+
"step": 220
|
| 670 |
+
},
|
| 671 |
+
{
|
| 672 |
+
"epoch": 0.48341614072780986,
|
| 673 |
+
"grad_norm": 115.29167221000085,
|
| 674 |
+
"learning_rate": 4.923546881779183e-07,
|
| 675 |
+
"logits/chosen": -15.626416206359863,
|
| 676 |
+
"logits/rejected": -15.386268615722656,
|
| 677 |
+
"logps/chosen": -1.7103182077407837,
|
| 678 |
+
"logps/rejected": -2.0725884437561035,
|
| 679 |
+
"loss": 3.0943,
|
| 680 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 681 |
+
"rewards/chosen": -17.103179931640625,
|
| 682 |
+
"rewards/margins": 3.622704267501831,
|
| 683 |
+
"rewards/rejected": -20.72588539123535,
|
| 684 |
+
"step": 225
|
| 685 |
+
},
|
| 686 |
+
{
|
| 687 |
+
"epoch": 0.4941587216328723,
|
| 688 |
+
"grad_norm": 110.55749043958713,
|
| 689 |
+
"learning_rate": 4.776675265006186e-07,
|
| 690 |
+
"logits/chosen": -15.542366027832031,
|
| 691 |
+
"logits/rejected": -15.470932006835938,
|
| 692 |
+
"logps/chosen": -1.7697646617889404,
|
| 693 |
+
"logps/rejected": -2.1862690448760986,
|
| 694 |
+
"loss": 3.2388,
|
| 695 |
+
"rewards/accuracies": 0.706250011920929,
|
| 696 |
+
"rewards/chosen": -17.69764518737793,
|
| 697 |
+
"rewards/margins": 4.165043830871582,
|
| 698 |
+
"rewards/rejected": -21.862689971923828,
|
| 699 |
+
"step": 230
|
| 700 |
+
},
|
| 701 |
+
{
|
| 702 |
+
"epoch": 0.5049013025379347,
|
| 703 |
+
"grad_norm": 194.28560949567938,
|
| 704 |
+
"learning_rate": 4.62870697926156e-07,
|
| 705 |
+
"logits/chosen": -14.893994331359863,
|
| 706 |
+
"logits/rejected": -15.048649787902832,
|
| 707 |
+
"logps/chosen": -1.764272928237915,
|
| 708 |
+
"logps/rejected": -2.2773966789245605,
|
| 709 |
+
"loss": 3.2608,
|
| 710 |
+
"rewards/accuracies": 0.71875,
|
| 711 |
+
"rewards/chosen": -17.642728805541992,
|
| 712 |
+
"rewards/margins": 5.13123893737793,
|
| 713 |
+
"rewards/rejected": -22.773967742919922,
|
| 714 |
+
"step": 235
|
| 715 |
+
},
|
| 716 |
+
{
|
| 717 |
+
"epoch": 0.5156438834429972,
|
| 718 |
+
"grad_norm": 118.02806259112292,
|
| 719 |
+
"learning_rate": 4.479850956431092e-07,
|
| 720 |
+
"logits/chosen": -15.161096572875977,
|
| 721 |
+
"logits/rejected": -15.109652519226074,
|
| 722 |
+
"logps/chosen": -1.7125499248504639,
|
| 723 |
+
"logps/rejected": -2.1439616680145264,
|
| 724 |
+
"loss": 3.0925,
|
| 725 |
+
"rewards/accuracies": 0.731249988079071,
|
| 726 |
+
"rewards/chosen": -17.125499725341797,
|
| 727 |
+
"rewards/margins": 4.31411600112915,
|
| 728 |
+
"rewards/rejected": -21.43961524963379,
|
| 729 |
+
"step": 240
|
| 730 |
+
},
|
| 731 |
+
{
|
| 732 |
+
"epoch": 0.5263864643480596,
|
| 733 |
+
"grad_norm": 94.41744321870145,
|
| 734 |
+
"learning_rate": 4.33031738188933e-07,
|
| 735 |
+
"logits/chosen": -15.63347053527832,
|
| 736 |
+
"logits/rejected": -15.267684936523438,
|
| 737 |
+
"logps/chosen": -1.6985079050064087,
|
| 738 |
+
"logps/rejected": -2.109764575958252,
|
| 739 |
+
"loss": 3.2248,
|
| 740 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 741 |
+
"rewards/chosen": -16.98508071899414,
|
| 742 |
+
"rewards/margins": 4.1125664710998535,
|
| 743 |
+
"rewards/rejected": -21.097644805908203,
|
| 744 |
+
"step": 245
|
| 745 |
+
},
|
| 746 |
+
{
|
| 747 |
+
"epoch": 0.5371290452531221,
|
| 748 |
+
"grad_norm": 132.41327050594964,
|
| 749 |
+
"learning_rate": 4.180317397716889e-07,
|
| 750 |
+
"logits/chosen": -15.816108703613281,
|
| 751 |
+
"logits/rejected": -15.996536254882812,
|
| 752 |
+
"logps/chosen": -1.7325900793075562,
|
| 753 |
+
"logps/rejected": -2.2646021842956543,
|
| 754 |
+
"loss": 3.1082,
|
| 755 |
+
"rewards/accuracies": 0.737500011920929,
|
| 756 |
+
"rewards/chosen": -17.32590103149414,
|
| 757 |
+
"rewards/margins": 5.320119857788086,
|
| 758 |
+
"rewards/rejected": -22.646020889282227,
|
| 759 |
+
"step": 250
|
| 760 |
+
},
|
| 761 |
+
{
|
| 762 |
+
"epoch": 0.5478716261581845,
|
| 763 |
+
"grad_norm": 123.8588013852755,
|
| 764 |
+
"learning_rate": 4.030062804566888e-07,
|
| 765 |
+
"logits/chosen": -15.66382122039795,
|
| 766 |
+
"logits/rejected": -15.521852493286133,
|
| 767 |
+
"logps/chosen": -1.703418493270874,
|
| 768 |
+
"logps/rejected": -2.1070733070373535,
|
| 769 |
+
"loss": 3.0582,
|
| 770 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 771 |
+
"rewards/chosen": -17.03418731689453,
|
| 772 |
+
"rewards/margins": 4.036545753479004,
|
| 773 |
+
"rewards/rejected": -21.07073211669922,
|
| 774 |
+
"step": 255
|
| 775 |
+
},
|
| 776 |
+
{
|
| 777 |
+
"epoch": 0.5586142070632469,
|
| 778 |
+
"grad_norm": 100.71973130430227,
|
| 779 |
+
"learning_rate": 3.8797657626014614e-07,
|
| 780 |
+
"logits/chosen": -15.945584297180176,
|
| 781 |
+
"logits/rejected": -15.955207824707031,
|
| 782 |
+
"logps/chosen": -1.7251708507537842,
|
| 783 |
+
"logps/rejected": -2.1078107357025146,
|
| 784 |
+
"loss": 3.1785,
|
| 785 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 786 |
+
"rewards/chosen": -17.251708984375,
|
| 787 |
+
"rewards/margins": 3.826399326324463,
|
| 788 |
+
"rewards/rejected": -21.078109741210938,
|
| 789 |
+
"step": 260
|
| 790 |
+
},
|
| 791 |
+
{
|
| 792 |
+
"epoch": 0.5693567879683094,
|
| 793 |
+
"grad_norm": 118.85717581252126,
|
| 794 |
+
"learning_rate": 3.729638491920669e-07,
|
| 795 |
+
"logits/chosen": -15.352853775024414,
|
| 796 |
+
"logits/rejected": -15.46752643585205,
|
| 797 |
+
"logps/chosen": -1.7334463596343994,
|
| 798 |
+
"logps/rejected": -2.135951519012451,
|
| 799 |
+
"loss": 3.1444,
|
| 800 |
+
"rewards/accuracies": 0.75,
|
| 801 |
+
"rewards/chosen": -17.334461212158203,
|
| 802 |
+
"rewards/margins": 4.025053977966309,
|
| 803 |
+
"rewards/rejected": -21.359514236450195,
|
| 804 |
+
"step": 265
|
| 805 |
+
},
|
| 806 |
+
{
|
| 807 |
+
"epoch": 0.5800993688733719,
|
| 808 |
+
"grad_norm": 85.82493905943672,
|
| 809 |
+
"learning_rate": 3.5798929729067464e-07,
|
| 810 |
+
"logits/chosen": -16.411659240722656,
|
| 811 |
+
"logits/rejected": -16.199129104614258,
|
| 812 |
+
"logps/chosen": -1.7729793787002563,
|
| 813 |
+
"logps/rejected": -2.220877170562744,
|
| 814 |
+
"loss": 2.7332,
|
| 815 |
+
"rewards/accuracies": 0.8125,
|
| 816 |
+
"rewards/chosen": -17.729793548583984,
|
| 817 |
+
"rewards/margins": 4.478978157043457,
|
| 818 |
+
"rewards/rejected": -22.208770751953125,
|
| 819 |
+
"step": 270
|
| 820 |
+
},
|
| 821 |
+
{
|
| 822 |
+
"epoch": 0.5908419497784343,
|
| 823 |
+
"grad_norm": 106.03649240710887,
|
| 824 |
+
"learning_rate": 3.4307406469068595e-07,
|
| 825 |
+
"logits/chosen": -14.978113174438477,
|
| 826 |
+
"logits/rejected": -15.032865524291992,
|
| 827 |
+
"logps/chosen": -1.809618592262268,
|
| 828 |
+
"logps/rejected": -2.3406805992126465,
|
| 829 |
+
"loss": 2.9025,
|
| 830 |
+
"rewards/accuracies": 0.78125,
|
| 831 |
+
"rewards/chosen": -18.096187591552734,
|
| 832 |
+
"rewards/margins": 5.3106207847595215,
|
| 833 |
+
"rewards/rejected": -23.40680694580078,
|
| 834 |
+
"step": 275
|
| 835 |
+
},
|
| 836 |
+
{
|
| 837 |
+
"epoch": 0.6015845306834967,
|
| 838 |
+
"grad_norm": 111.84746493906809,
|
| 839 |
+
"learning_rate": 3.282392117676968e-07,
|
| 840 |
+
"logits/chosen": -15.695220947265625,
|
| 841 |
+
"logits/rejected": -15.893925666809082,
|
| 842 |
+
"logps/chosen": -1.9052400588989258,
|
| 843 |
+
"logps/rejected": -2.4178624153137207,
|
| 844 |
+
"loss": 2.8972,
|
| 845 |
+
"rewards/accuracies": 0.793749988079071,
|
| 846 |
+
"rewards/chosen": -19.052398681640625,
|
| 847 |
+
"rewards/margins": 5.126224040985107,
|
| 848 |
+
"rewards/rejected": -24.178625106811523,
|
| 849 |
+
"step": 280
|
| 850 |
+
},
|
| 851 |
+
{
|
| 852 |
+
"epoch": 0.6123271115885591,
|
| 853 |
+
"grad_norm": 103.57050178439866,
|
| 854 |
+
"learning_rate": 3.135056854008371e-07,
|
| 855 |
+
"logits/chosen": -15.808944702148438,
|
| 856 |
+
"logits/rejected": -15.778024673461914,
|
| 857 |
+
"logps/chosen": -1.7815015316009521,
|
| 858 |
+
"logps/rejected": -2.271728515625,
|
| 859 |
+
"loss": 2.8431,
|
| 860 |
+
"rewards/accuracies": 0.762499988079071,
|
| 861 |
+
"rewards/chosen": -17.81501579284668,
|
| 862 |
+
"rewards/margins": 4.9022698402404785,
|
| 863 |
+
"rewards/rejected": -22.717287063598633,
|
| 864 |
+
"step": 285
|
| 865 |
+
},
|
| 866 |
+
{
|
| 867 |
+
"epoch": 0.6230696924936215,
|
| 868 |
+
"grad_norm": 108.56282250963646,
|
| 869 |
+
"learning_rate": 2.988942893956833e-07,
|
| 870 |
+
"logits/chosen": -15.920690536499023,
|
| 871 |
+
"logits/rejected": -15.81785774230957,
|
| 872 |
+
"logps/chosen": -1.8119595050811768,
|
| 873 |
+
"logps/rejected": -2.2470898628234863,
|
| 874 |
+
"loss": 3.0512,
|
| 875 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 876 |
+
"rewards/chosen": -18.11959457397461,
|
| 877 |
+
"rewards/margins": 4.351304054260254,
|
| 878 |
+
"rewards/rejected": -22.470901489257812,
|
| 879 |
+
"step": 290
|
| 880 |
+
},
|
| 881 |
+
{
|
| 882 |
+
"epoch": 0.6338122733986841,
|
| 883 |
+
"grad_norm": 116.17030678566253,
|
| 884 |
+
"learning_rate": 2.844256551091911e-07,
|
| 885 |
+
"logits/chosen": -16.46514320373535,
|
| 886 |
+
"logits/rejected": -16.6345272064209,
|
| 887 |
+
"logps/chosen": -1.7750489711761475,
|
| 888 |
+
"logps/rejected": -2.3242697715759277,
|
| 889 |
+
"loss": 2.8258,
|
| 890 |
+
"rewards/accuracies": 0.8187500238418579,
|
| 891 |
+
"rewards/chosen": -17.75048828125,
|
| 892 |
+
"rewards/margins": 5.492205619812012,
|
| 893 |
+
"rewards/rejected": -23.242694854736328,
|
| 894 |
+
"step": 295
|
| 895 |
+
},
|
| 896 |
+
{
|
| 897 |
+
"epoch": 0.6445548543037465,
|
| 898 |
+
"grad_norm": 147.68751040852177,
|
| 899 |
+
"learning_rate": 2.7012021231812664e-07,
|
| 900 |
+
"logits/chosen": -16.046567916870117,
|
| 901 |
+
"logits/rejected": -15.983491897583008,
|
| 902 |
+
"logps/chosen": -1.897653579711914,
|
| 903 |
+
"logps/rejected": -2.281114339828491,
|
| 904 |
+
"loss": 3.2387,
|
| 905 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 906 |
+
"rewards/chosen": -18.97653579711914,
|
| 907 |
+
"rewards/margins": 3.834609270095825,
|
| 908 |
+
"rewards/rejected": -22.811145782470703,
|
| 909 |
+
"step": 300
|
| 910 |
+
},
|
| 911 |
+
{
|
| 912 |
+
"epoch": 0.6552974352088089,
|
| 913 |
+
"grad_norm": 124.9827211839036,
|
| 914 |
+
"learning_rate": 2.5599816037212954e-07,
|
| 915 |
+
"logits/chosen": -15.513954162597656,
|
| 916 |
+
"logits/rejected": -15.608304023742676,
|
| 917 |
+
"logps/chosen": -1.8266518115997314,
|
| 918 |
+
"logps/rejected": -2.34920072555542,
|
| 919 |
+
"loss": 2.8842,
|
| 920 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 921 |
+
"rewards/chosen": -18.266515731811523,
|
| 922 |
+
"rewards/margins": 5.225489616394043,
|
| 923 |
+
"rewards/rejected": -23.492008209228516,
|
| 924 |
+
"step": 305
|
| 925 |
+
},
|
| 926 |
+
{
|
| 927 |
+
"epoch": 0.6660400161138713,
|
| 928 |
+
"grad_norm": 141.73868750375132,
|
| 929 |
+
"learning_rate": 2.4207943967214064e-07,
|
| 930 |
+
"logits/chosen": -16.29332160949707,
|
| 931 |
+
"logits/rejected": -16.148832321166992,
|
| 932 |
+
"logps/chosen": -1.9530408382415771,
|
| 933 |
+
"logps/rejected": -2.4292497634887695,
|
| 934 |
+
"loss": 3.0684,
|
| 935 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 936 |
+
"rewards/chosen": -19.530406951904297,
|
| 937 |
+
"rewards/margins": 4.762091636657715,
|
| 938 |
+
"rewards/rejected": -24.292499542236328,
|
| 939 |
+
"step": 310
|
| 940 |
+
},
|
| 941 |
+
{
|
| 942 |
+
"epoch": 0.6767825970189338,
|
| 943 |
+
"grad_norm": 172.98236669812187,
|
| 944 |
+
"learning_rate": 2.2838370351446547e-07,
|
| 945 |
+
"logits/chosen": -15.80296802520752,
|
| 946 |
+
"logits/rejected": -15.732707023620605,
|
| 947 |
+
"logps/chosen": -1.7977100610733032,
|
| 948 |
+
"logps/rejected": -2.3419437408447266,
|
| 949 |
+
"loss": 2.7361,
|
| 950 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 951 |
+
"rewards/chosen": -17.977100372314453,
|
| 952 |
+
"rewards/margins": 5.4423370361328125,
|
| 953 |
+
"rewards/rejected": -23.419437408447266,
|
| 954 |
+
"step": 315
|
| 955 |
+
},
|
| 956 |
+
{
|
| 957 |
+
"epoch": 0.6875251779239963,
|
| 958 |
+
"grad_norm": 112.41993171792654,
|
| 959 |
+
"learning_rate": 2.1493029034023188e-07,
|
| 960 |
+
"logits/chosen": -15.622350692749023,
|
| 961 |
+
"logits/rejected": -15.62098503112793,
|
| 962 |
+
"logps/chosen": -1.784433126449585,
|
| 963 |
+
"logps/rejected": -2.267364025115967,
|
| 964 |
+
"loss": 2.8725,
|
| 965 |
+
"rewards/accuracies": 0.737500011920929,
|
| 966 |
+
"rewards/chosen": -17.844331741333008,
|
| 967 |
+
"rewards/margins": 4.829309940338135,
|
| 968 |
+
"rewards/rejected": -22.673641204833984,
|
| 969 |
+
"step": 320
|
| 970 |
+
},
|
| 971 |
+
{
|
| 972 |
+
"epoch": 0.6982677588290587,
|
| 973 |
+
"grad_norm": 116.72973302505498,
|
| 974 |
+
"learning_rate": 2.0173819642942376e-07,
|
| 975 |
+
"logits/chosen": -15.983564376831055,
|
| 976 |
+
"logits/rejected": -15.785995483398438,
|
| 977 |
+
"logps/chosen": -1.9481559991836548,
|
| 978 |
+
"logps/rejected": -2.5938241481781006,
|
| 979 |
+
"loss": 2.9776,
|
| 980 |
+
"rewards/accuracies": 0.762499988079071,
|
| 981 |
+
"rewards/chosen": -19.48155975341797,
|
| 982 |
+
"rewards/margins": 6.456681251525879,
|
| 983 |
+
"rewards/rejected": -25.938241958618164,
|
| 984 |
+
"step": 325
|
| 985 |
+
},
|
| 986 |
+
{
|
| 987 |
+
"epoch": 0.7090103397341211,
|
| 988 |
+
"grad_norm": 137.081227030888,
|
| 989 |
+
"learning_rate": 1.888260490780485e-07,
|
| 990 |
+
"logits/chosen": -15.733128547668457,
|
| 991 |
+
"logits/rejected": -15.935976028442383,
|
| 992 |
+
"logps/chosen": -1.8248291015625,
|
| 993 |
+
"logps/rejected": -2.2822577953338623,
|
| 994 |
+
"loss": 3.0336,
|
| 995 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 996 |
+
"rewards/chosen": -18.248287200927734,
|
| 997 |
+
"rewards/margins": 4.574288368225098,
|
| 998 |
+
"rewards/rejected": -22.82257652282715,
|
| 999 |
+
"step": 330
|
| 1000 |
+
},
|
| 1001 |
+
{
|
| 1002 |
+
"epoch": 0.7197529206391836,
|
| 1003 |
+
"grad_norm": 117.59237011545949,
|
| 1004 |
+
"learning_rate": 1.7621208029631078e-07,
|
| 1005 |
+
"logits/chosen": -15.677896499633789,
|
| 1006 |
+
"logits/rejected": -15.52332878112793,
|
| 1007 |
+
"logps/chosen": -1.8683006763458252,
|
| 1008 |
+
"logps/rejected": -2.448673725128174,
|
| 1009 |
+
"loss": 2.8842,
|
| 1010 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1011 |
+
"rewards/chosen": -18.68300437927246,
|
| 1012 |
+
"rewards/margins": 5.803731918334961,
|
| 1013 |
+
"rewards/rejected": -24.486736297607422,
|
| 1014 |
+
"step": 335
|
| 1015 |
+
},
|
| 1016 |
+
{
|
| 1017 |
+
"epoch": 0.730495501544246,
|
| 1018 |
+
"grad_norm": 122.52464571458833,
|
| 1019 |
+
"learning_rate": 1.6391410106493227e-07,
|
| 1020 |
+
"logits/chosen": -15.322489738464355,
|
| 1021 |
+
"logits/rejected": -15.156359672546387,
|
| 1022 |
+
"logps/chosen": -1.9856983423233032,
|
| 1023 |
+
"logps/rejected": -2.5366251468658447,
|
| 1024 |
+
"loss": 2.8321,
|
| 1025 |
+
"rewards/accuracies": 0.768750011920929,
|
| 1026 |
+
"rewards/chosen": -19.856985092163086,
|
| 1027 |
+
"rewards/margins": 5.509265422821045,
|
| 1028 |
+
"rewards/rejected": -25.36625099182129,
|
| 1029 |
+
"step": 340
|
| 1030 |
+
},
|
| 1031 |
+
{
|
| 1032 |
+
"epoch": 0.7412380824493084,
|
| 1033 |
+
"grad_norm": 98.13439288099542,
|
| 1034 |
+
"learning_rate": 1.5194947618596673e-07,
|
| 1035 |
+
"logits/chosen": -16.58279037475586,
|
| 1036 |
+
"logits/rejected": -16.51984405517578,
|
| 1037 |
+
"logps/chosen": -1.9558055400848389,
|
| 1038 |
+
"logps/rejected": -2.502556324005127,
|
| 1039 |
+
"loss": 3.0276,
|
| 1040 |
+
"rewards/accuracies": 0.793749988079071,
|
| 1041 |
+
"rewards/chosen": -19.558055877685547,
|
| 1042 |
+
"rewards/margins": 5.467510223388672,
|
| 1043 |
+
"rewards/rejected": -25.025564193725586,
|
| 1044 |
+
"step": 345
|
| 1045 |
+
},
|
| 1046 |
+
{
|
| 1047 |
+
"epoch": 0.7519806633543709,
|
| 1048 |
+
"grad_norm": 104.49522716837973,
|
| 1049 |
+
"learning_rate": 1.4033509976362083e-07,
|
| 1050 |
+
"logits/chosen": -15.616808891296387,
|
| 1051 |
+
"logits/rejected": -15.737129211425781,
|
| 1052 |
+
"logps/chosen": -1.9377790689468384,
|
| 1053 |
+
"logps/rejected": -2.3541884422302246,
|
| 1054 |
+
"loss": 2.9578,
|
| 1055 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 1056 |
+
"rewards/chosen": -19.377790451049805,
|
| 1057 |
+
"rewards/margins": 4.164095401763916,
|
| 1058 |
+
"rewards/rejected": -23.541885375976562,
|
| 1059 |
+
"step": 350
|
| 1060 |
+
},
|
| 1061 |
+
{
|
| 1062 |
+
"epoch": 0.7627232442594333,
|
| 1063 |
+
"grad_norm": 107.6884489363759,
|
| 1064 |
+
"learning_rate": 1.2908737134970363e-07,
|
| 1065 |
+
"logits/chosen": -14.991783142089844,
|
| 1066 |
+
"logits/rejected": -15.030191421508789,
|
| 1067 |
+
"logps/chosen": -1.8485435247421265,
|
| 1068 |
+
"logps/rejected": -2.38105845451355,
|
| 1069 |
+
"loss": 3.0742,
|
| 1070 |
+
"rewards/accuracies": 0.737500011920929,
|
| 1071 |
+
"rewards/chosen": -18.485435485839844,
|
| 1072 |
+
"rewards/margins": 5.325149059295654,
|
| 1073 |
+
"rewards/rejected": -23.810585021972656,
|
| 1074 |
+
"step": 355
|
| 1075 |
+
},
|
| 1076 |
+
{
|
| 1077 |
+
"epoch": 0.7734658251644958,
|
| 1078 |
+
"grad_norm": 105.82348414404971,
|
| 1079 |
+
"learning_rate": 1.1822217278738515e-07,
|
| 1080 |
+
"logits/chosen": -15.954935073852539,
|
| 1081 |
+
"logits/rejected": -16.074337005615234,
|
| 1082 |
+
"logps/chosen": -1.9491569995880127,
|
| 1083 |
+
"logps/rejected": -2.3941268920898438,
|
| 1084 |
+
"loss": 3.0422,
|
| 1085 |
+
"rewards/accuracies": 0.75,
|
| 1086 |
+
"rewards/chosen": -19.4915714263916,
|
| 1087 |
+
"rewards/margins": 4.449699878692627,
|
| 1088 |
+
"rewards/rejected": -23.941268920898438,
|
| 1089 |
+
"step": 360
|
| 1090 |
+
},
|
| 1091 |
+
{
|
| 1092 |
+
"epoch": 0.7842084060695582,
|
| 1093 |
+
"grad_norm": 208.94228225878328,
|
| 1094 |
+
"learning_rate": 1.0775484578596241e-07,
|
| 1095 |
+
"logits/chosen": -16.27437973022461,
|
| 1096 |
+
"logits/rejected": -16.11089324951172,
|
| 1097 |
+
"logps/chosen": -1.8912540674209595,
|
| 1098 |
+
"logps/rejected": -2.4439680576324463,
|
| 1099 |
+
"loss": 2.7384,
|
| 1100 |
+
"rewards/accuracies": 0.78125,
|
| 1101 |
+
"rewards/chosen": -18.912538528442383,
|
| 1102 |
+
"rewards/margins": 5.5271430015563965,
|
| 1103 |
+
"rewards/rejected": -24.439682006835938,
|
| 1104 |
+
"step": 365
|
| 1105 |
+
},
|
| 1106 |
+
{
|
| 1107 |
+
"epoch": 0.7949509869746206,
|
| 1108 |
+
"grad_norm": 126.14591373204779,
|
| 1109 |
+
"learning_rate": 9.770017025829673e-08,
|
| 1110 |
+
"logits/chosen": -16.250118255615234,
|
| 1111 |
+
"logits/rejected": -16.38758087158203,
|
| 1112 |
+
"logps/chosen": -2.065514087677002,
|
| 1113 |
+
"logps/rejected": -2.72831654548645,
|
| 1114 |
+
"loss": 2.6227,
|
| 1115 |
+
"rewards/accuracies": 0.831250011920929,
|
| 1116 |
+
"rewards/chosen": -20.655139923095703,
|
| 1117 |
+
"rewards/margins": 6.628026008605957,
|
| 1118 |
+
"rewards/rejected": -27.283166885375977,
|
| 1119 |
+
"step": 370
|
| 1120 |
+
},
|
| 1121 |
+
{
|
| 1122 |
+
"epoch": 0.805693567879683,
|
| 1123 |
+
"grad_norm": 121.32471246816698,
|
| 1124 |
+
"learning_rate": 8.807234345151027e-08,
|
| 1125 |
+
"logits/chosen": -16.10807991027832,
|
| 1126 |
+
"logits/rejected": -16.029443740844727,
|
| 1127 |
+
"logps/chosen": -1.9544318914413452,
|
| 1128 |
+
"logps/rejected": -2.565009355545044,
|
| 1129 |
+
"loss": 2.899,
|
| 1130 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1131 |
+
"rewards/chosen": -19.544321060180664,
|
| 1132 |
+
"rewards/margins": 6.105769634246826,
|
| 1133 |
+
"rewards/rejected": -25.65009117126465,
|
| 1134 |
+
"step": 375
|
| 1135 |
+
},
|
| 1136 |
+
{
|
| 1137 |
+
"epoch": 0.8164361487847456,
|
| 1138 |
+
"grad_norm": 201.08985911319357,
|
| 1139 |
+
"learning_rate": 7.888495990040924e-08,
|
| 1140 |
+
"logits/chosen": -14.932947158813477,
|
| 1141 |
+
"logits/rejected": -14.924430847167969,
|
| 1142 |
+
"logps/chosen": -1.965070128440857,
|
| 1143 |
+
"logps/rejected": -2.5968446731567383,
|
| 1144 |
+
"loss": 2.9375,
|
| 1145 |
+
"rewards/accuracies": 0.78125,
|
| 1146 |
+
"rewards/chosen": -19.650699615478516,
|
| 1147 |
+
"rewards/margins": 6.317748069763184,
|
| 1148 |
+
"rewards/rejected": -25.968448638916016,
|
| 1149 |
+
"step": 380
|
| 1150 |
+
},
|
| 1151 |
+
{
|
| 1152 |
+
"epoch": 0.827178729689808,
|
| 1153 |
+
"grad_norm": 132.40169960863122,
|
| 1154 |
+
"learning_rate": 7.015099223193943e-08,
|
| 1155 |
+
"logits/chosen": -16.522253036499023,
|
| 1156 |
+
"logits/rejected": -16.4668025970459,
|
| 1157 |
+
"logps/chosen": -2.0363919734954834,
|
| 1158 |
+
"logps/rejected": -2.5673530101776123,
|
| 1159 |
+
"loss": 2.8305,
|
| 1160 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 1161 |
+
"rewards/chosen": -20.36391830444336,
|
| 1162 |
+
"rewards/margins": 5.309610366821289,
|
| 1163 |
+
"rewards/rejected": -25.67352867126465,
|
| 1164 |
+
"step": 385
|
| 1165 |
+
},
|
| 1166 |
+
{
|
| 1167 |
+
"epoch": 0.8379213105948704,
|
| 1168 |
+
"grad_norm": 113.91880111182708,
|
| 1169 |
+
"learning_rate": 6.188277284777857e-08,
|
| 1170 |
+
"logits/chosen": -15.97686767578125,
|
| 1171 |
+
"logits/rejected": -15.647547721862793,
|
| 1172 |
+
"logps/chosen": -2.0422708988189697,
|
| 1173 |
+
"logps/rejected": -2.549582004547119,
|
| 1174 |
+
"loss": 2.7972,
|
| 1175 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 1176 |
+
"rewards/chosen": -20.42270851135254,
|
| 1177 |
+
"rewards/margins": 5.073108673095703,
|
| 1178 |
+
"rewards/rejected": -25.49581527709961,
|
| 1179 |
+
"step": 390
|
| 1180 |
+
},
|
| 1181 |
+
{
|
| 1182 |
+
"epoch": 0.8486638914999328,
|
| 1183 |
+
"grad_norm": 126.36653904434415,
|
| 1184 |
+
"learning_rate": 5.409197651092965e-08,
|
| 1185 |
+
"logits/chosen": -15.692411422729492,
|
| 1186 |
+
"logits/rejected": -15.747018814086914,
|
| 1187 |
+
"logps/chosen": -2.10833477973938,
|
| 1188 |
+
"logps/rejected": -2.676734685897827,
|
| 1189 |
+
"loss": 2.6959,
|
| 1190 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 1191 |
+
"rewards/chosen": -21.083349227905273,
|
| 1192 |
+
"rewards/margins": 5.683998107910156,
|
| 1193 |
+
"rewards/rejected": -26.767345428466797,
|
| 1194 |
+
"step": 395
|
| 1195 |
+
},
|
| 1196 |
+
{
|
| 1197 |
+
"epoch": 0.8594064724049953,
|
| 1198 |
+
"grad_norm": 115.47783241869803,
|
| 1199 |
+
"learning_rate": 4.678960386090298e-08,
|
| 1200 |
+
"logits/chosen": -15.30811595916748,
|
| 1201 |
+
"logits/rejected": -15.540433883666992,
|
| 1202 |
+
"logps/chosen": -1.9201295375823975,
|
| 1203 |
+
"logps/rejected": -2.523301601409912,
|
| 1204 |
+
"loss": 2.7196,
|
| 1205 |
+
"rewards/accuracies": 0.793749988079071,
|
| 1206 |
+
"rewards/chosen": -19.2012939453125,
|
| 1207 |
+
"rewards/margins": 6.03171968460083,
|
| 1208 |
+
"rewards/rejected": -25.233013153076172,
|
| 1209 |
+
"step": 400
|
| 1210 |
+
},
|
| 1211 |
+
{
|
| 1212 |
+
"epoch": 0.8594064724049953,
|
| 1213 |
+
"eval_logits/chosen": -14.502357482910156,
|
| 1214 |
+
"eval_logits/rejected": -14.432116508483887,
|
| 1215 |
+
"eval_logps/chosen": -1.8952556848526,
|
| 1216 |
+
"eval_logps/rejected": -2.3938686847686768,
|
| 1217 |
+
"eval_loss": 2.7580387592315674,
|
| 1218 |
+
"eval_rewards/accuracies": 0.7704917788505554,
|
| 1219 |
+
"eval_rewards/chosen": -18.952556610107422,
|
| 1220 |
+
"eval_rewards/margins": 4.986130237579346,
|
| 1221 |
+
"eval_rewards/rejected": -23.938688278198242,
|
| 1222 |
+
"eval_runtime": 86.7509,
|
| 1223 |
+
"eval_samples_per_second": 22.374,
|
| 1224 |
+
"eval_steps_per_second": 1.406,
|
| 1225 |
+
"step": 400
|
| 1226 |
+
},
|
| 1227 |
+
{
|
| 1228 |
+
"epoch": 0.8701490533100578,
|
| 1229 |
+
"grad_norm": 112.75479415537629,
|
| 1230 |
+
"learning_rate": 3.998596588076366e-08,
|
| 1231 |
+
"logits/chosen": -15.262601852416992,
|
| 1232 |
+
"logits/rejected": -14.926340103149414,
|
| 1233 |
+
"logps/chosen": -1.9843738079071045,
|
| 1234 |
+
"logps/rejected": -2.531123638153076,
|
| 1235 |
+
"loss": 3.0297,
|
| 1236 |
+
"rewards/accuracies": 0.737500011920929,
|
| 1237 |
+
"rewards/chosen": -19.843738555908203,
|
| 1238 |
+
"rewards/margins": 5.467496871948242,
|
| 1239 |
+
"rewards/rejected": -25.311237335205078,
|
| 1240 |
+
"step": 405
|
| 1241 |
+
},
|
| 1242 |
+
{
|
| 1243 |
+
"epoch": 0.8808916342151202,
|
| 1244 |
+
"grad_norm": 124.72604723824772,
|
| 1245 |
+
"learning_rate": 3.3690669337976996e-08,
|
| 1246 |
+
"logits/chosen": -15.930806159973145,
|
| 1247 |
+
"logits/rejected": -15.931562423706055,
|
| 1248 |
+
"logps/chosen": -1.9564170837402344,
|
| 1249 |
+
"logps/rejected": -2.517925500869751,
|
| 1250 |
+
"loss": 2.6997,
|
| 1251 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 1252 |
+
"rewards/chosen": -19.564170837402344,
|
| 1253 |
+
"rewards/margins": 5.615079402923584,
|
| 1254 |
+
"rewards/rejected": -25.17925262451172,
|
| 1255 |
+
"step": 410
|
| 1256 |
+
},
|
| 1257 |
+
{
|
| 1258 |
+
"epoch": 0.8916342151201826,
|
| 1259 |
+
"grad_norm": 124.6339007044252,
|
| 1260 |
+
"learning_rate": 2.7912603219609798e-08,
|
| 1261 |
+
"logits/chosen": -16.048858642578125,
|
| 1262 |
+
"logits/rejected": -15.944051742553711,
|
| 1263 |
+
"logps/chosen": -2.1005568504333496,
|
| 1264 |
+
"logps/rejected": -2.568063735961914,
|
| 1265 |
+
"loss": 2.6664,
|
| 1266 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1267 |
+
"rewards/chosen": -21.005565643310547,
|
| 1268 |
+
"rewards/margins": 4.675069332122803,
|
| 1269 |
+
"rewards/rejected": -25.680639266967773,
|
| 1270 |
+
"step": 415
|
| 1271 |
+
},
|
| 1272 |
+
{
|
| 1273 |
+
"epoch": 0.9023767960252451,
|
| 1274 |
+
"grad_norm": 121.69678463559275,
|
| 1275 |
+
"learning_rate": 2.265992618104029e-08,
|
| 1276 |
+
"logits/chosen": -16.162189483642578,
|
| 1277 |
+
"logits/rejected": -16.180706024169922,
|
| 1278 |
+
"logps/chosen": -2.018214702606201,
|
| 1279 |
+
"logps/rejected": -2.648829460144043,
|
| 1280 |
+
"loss": 2.7859,
|
| 1281 |
+
"rewards/accuracies": 0.8187500238418579,
|
| 1282 |
+
"rewards/chosen": -20.182147979736328,
|
| 1283 |
+
"rewards/margins": 6.306146621704102,
|
| 1284 |
+
"rewards/rejected": -26.488290786743164,
|
| 1285 |
+
"step": 420
|
| 1286 |
+
},
|
| 1287 |
+
{
|
| 1288 |
+
"epoch": 0.9131193769303075,
|
| 1289 |
+
"grad_norm": 267.04077532902875,
|
| 1290 |
+
"learning_rate": 1.7940055025900304e-08,
|
| 1291 |
+
"logits/chosen": -14.952715873718262,
|
| 1292 |
+
"logits/rejected": -14.8345308303833,
|
| 1293 |
+
"logps/chosen": -1.9776952266693115,
|
| 1294 |
+
"logps/rejected": -2.4361090660095215,
|
| 1295 |
+
"loss": 3.0349,
|
| 1296 |
+
"rewards/accuracies": 0.737500011920929,
|
| 1297 |
+
"rewards/chosen": -19.77695083618164,
|
| 1298 |
+
"rewards/margins": 4.584140777587891,
|
| 1299 |
+
"rewards/rejected": -24.361093521118164,
|
| 1300 |
+
"step": 425
|
| 1301 |
+
},
|
| 1302 |
+
{
|
| 1303 |
+
"epoch": 0.92386195783537,
|
| 1304 |
+
"grad_norm": 115.027170497356,
|
| 1305 |
+
"learning_rate": 1.3759654233514817e-08,
|
| 1306 |
+
"logits/chosen": -15.242410659790039,
|
| 1307 |
+
"logits/rejected": -15.459531784057617,
|
| 1308 |
+
"logps/chosen": -1.8886226415634155,
|
| 1309 |
+
"logps/rejected": -2.344308853149414,
|
| 1310 |
+
"loss": 2.7946,
|
| 1311 |
+
"rewards/accuracies": 0.737500011920929,
|
| 1312 |
+
"rewards/chosen": -18.8862247467041,
|
| 1313 |
+
"rewards/margins": 4.556865692138672,
|
| 1314 |
+
"rewards/rejected": -23.443092346191406,
|
| 1315 |
+
"step": 430
|
| 1316 |
+
},
|
| 1317 |
+
{
|
| 1318 |
+
"epoch": 0.9346045387404324,
|
| 1319 |
+
"grad_norm": 116.35861873175186,
|
| 1320 |
+
"learning_rate": 1.0124626548627402e-08,
|
| 1321 |
+
"logits/chosen": -15.868557929992676,
|
| 1322 |
+
"logits/rejected": -15.74053955078125,
|
| 1323 |
+
"logps/chosen": -1.9277899265289307,
|
| 1324 |
+
"logps/rejected": -2.5743651390075684,
|
| 1325 |
+
"loss": 2.7446,
|
| 1326 |
+
"rewards/accuracies": 0.8187500238418579,
|
| 1327 |
+
"rewards/chosen": -19.27789878845215,
|
| 1328 |
+
"rewards/margins": 6.46575403213501,
|
| 1329 |
+
"rewards/rejected": -25.743656158447266,
|
| 1330 |
+
"step": 435
|
| 1331 |
+
},
|
| 1332 |
+
{
|
| 1333 |
+
"epoch": 0.9453471196454948,
|
| 1334 |
+
"grad_norm": 101.71134576562835,
|
| 1335 |
+
"learning_rate": 7.040104646698042e-09,
|
| 1336 |
+
"logits/chosen": -15.439016342163086,
|
| 1337 |
+
"logits/rejected": -15.44897747039795,
|
| 1338 |
+
"logps/chosen": -1.959753394126892,
|
| 1339 |
+
"logps/rejected": -2.5274107456207275,
|
| 1340 |
+
"loss": 2.6695,
|
| 1341 |
+
"rewards/accuracies": 0.78125,
|
| 1342 |
+
"rewards/chosen": -19.597536087036133,
|
| 1343 |
+
"rewards/margins": 5.676573276519775,
|
| 1344 |
+
"rewards/rejected": -25.274110794067383,
|
| 1345 |
+
"step": 440
|
| 1346 |
+
},
|
| 1347 |
+
{
|
| 1348 |
+
"epoch": 0.9560897005505573,
|
| 1349 |
+
"grad_norm": 110.29407715801415,
|
| 1350 |
+
"learning_rate": 4.510443886542114e-09,
|
| 1351 |
+
"logits/chosen": -15.287280082702637,
|
| 1352 |
+
"logits/rejected": -15.34148120880127,
|
| 1353 |
+
"logps/chosen": -2.031892776489258,
|
| 1354 |
+
"logps/rejected": -2.656055450439453,
|
| 1355 |
+
"loss": 2.7797,
|
| 1356 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 1357 |
+
"rewards/chosen": -20.318927764892578,
|
| 1358 |
+
"rewards/margins": 6.2416229248046875,
|
| 1359 |
+
"rewards/rejected": -26.560550689697266,
|
| 1360 |
+
"step": 445
|
| 1361 |
+
},
|
| 1362 |
+
{
|
| 1363 |
+
"epoch": 0.9668322814556197,
|
| 1364 |
+
"grad_norm": 122.67920649488762,
|
| 1365 |
+
"learning_rate": 2.539216160544333e-09,
|
| 1366 |
+
"logits/chosen": -15.329913139343262,
|
| 1367 |
+
"logits/rejected": -15.057653427124023,
|
| 1368 |
+
"logps/chosen": -2.1148476600646973,
|
| 1369 |
+
"logps/rejected": -2.682511806488037,
|
| 1370 |
+
"loss": 2.8577,
|
| 1371 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1372 |
+
"rewards/chosen": -21.14847755432129,
|
| 1373 |
+
"rewards/margins": 5.676642894744873,
|
| 1374 |
+
"rewards/rejected": -26.825122833251953,
|
| 1375 |
+
"step": 450
|
| 1376 |
+
},
|
| 1377 |
+
{
|
| 1378 |
+
"epoch": 0.9775748623606821,
|
| 1379 |
+
"grad_norm": 127.18334450658871,
|
| 1380 |
+
"learning_rate": 1.1292048511303054e-09,
|
| 1381 |
+
"logits/chosen": -15.279156684875488,
|
| 1382 |
+
"logits/rejected": -15.283819198608398,
|
| 1383 |
+
"logps/chosen": -2.0526576042175293,
|
| 1384 |
+
"logps/rejected": -2.5599799156188965,
|
| 1385 |
+
"loss": 2.9411,
|
| 1386 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 1387 |
+
"rewards/chosen": -20.52657699584961,
|
| 1388 |
+
"rewards/margins": 5.0732221603393555,
|
| 1389 |
+
"rewards/rejected": -25.599796295166016,
|
| 1390 |
+
"step": 455
|
| 1391 |
+
},
|
| 1392 |
+
{
|
| 1393 |
+
"epoch": 0.9883174432657446,
|
| 1394 |
+
"grad_norm": 115.1645535617376,
|
| 1395 |
+
"learning_rate": 2.82400900618418e-10,
|
| 1396 |
+
"logits/chosen": -15.98609447479248,
|
| 1397 |
+
"logits/rejected": -15.959650993347168,
|
| 1398 |
+
"logps/chosen": -1.9472544193267822,
|
| 1399 |
+
"logps/rejected": -2.6097803115844727,
|
| 1400 |
+
"loss": 2.5164,
|
| 1401 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1402 |
+
"rewards/chosen": -19.47254180908203,
|
| 1403 |
+
"rewards/margins": 6.625257968902588,
|
| 1404 |
+
"rewards/rejected": -26.097803115844727,
|
| 1405 |
+
"step": 460
|
| 1406 |
+
},
|
| 1407 |
+
{
|
| 1408 |
+
"epoch": 0.9990600241708071,
|
| 1409 |
+
"grad_norm": 233.73237047558877,
|
| 1410 |
+
"learning_rate": 0.0,
|
| 1411 |
+
"logits/chosen": -15.412060737609863,
|
| 1412 |
+
"logits/rejected": -15.519399642944336,
|
| 1413 |
+
"logps/chosen": -1.9192397594451904,
|
| 1414 |
+
"logps/rejected": -2.4578652381896973,
|
| 1415 |
+
"loss": 2.5846,
|
| 1416 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1417 |
+
"rewards/chosen": -19.19239616394043,
|
| 1418 |
+
"rewards/margins": 5.386256217956543,
|
| 1419 |
+
"rewards/rejected": -24.57865333557129,
|
| 1420 |
+
"step": 465
|
| 1421 |
+
},
|
| 1422 |
+
{
|
| 1423 |
+
"epoch": 0.9990600241708071,
|
| 1424 |
+
"step": 465,
|
| 1425 |
+
"total_flos": 0.0,
|
| 1426 |
+
"train_loss": 3.541660060677477,
|
| 1427 |
+
"train_runtime": 9046.6245,
|
| 1428 |
+
"train_samples_per_second": 6.585,
|
| 1429 |
+
"train_steps_per_second": 0.051
|
| 1430 |
+
}
|
| 1431 |
+
],
|
| 1432 |
+
"logging_steps": 5,
|
| 1433 |
+
"max_steps": 465,
|
| 1434 |
+
"num_input_tokens_seen": 0,
|
| 1435 |
+
"num_train_epochs": 1,
|
| 1436 |
+
"save_steps": 1000000,
|
| 1437 |
+
"stateful_callbacks": {
|
| 1438 |
+
"TrainerControl": {
|
| 1439 |
+
"args": {
|
| 1440 |
+
"should_epoch_stop": false,
|
| 1441 |
+
"should_evaluate": false,
|
| 1442 |
+
"should_log": false,
|
| 1443 |
+
"should_save": true,
|
| 1444 |
+
"should_training_stop": true
|
| 1445 |
+
},
|
| 1446 |
+
"attributes": {}
|
| 1447 |
+
}
|
| 1448 |
+
},
|
| 1449 |
+
"total_flos": 0.0,
|
| 1450 |
+
"train_batch_size": 2,
|
| 1451 |
+
"trial_name": null,
|
| 1452 |
+
"trial_params": null
|
| 1453 |
+
}
|