Model save
Browse files- README.md +80 -0
- all_results.json +9 -0
- generation_config.json +6 -0
- train_results.json +9 -0
- trainer_state.json +815 -0
README.md
ADDED
|
@@ -0,0 +1,80 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
base_model: alignment-handbook/zephyr-7b-sft-full
|
| 4 |
+
tags:
|
| 5 |
+
- trl
|
| 6 |
+
- dpo
|
| 7 |
+
- generated_from_trainer
|
| 8 |
+
model-index:
|
| 9 |
+
- name: zephyr-7b-dpo-full-prometheus_consistent-reward-scale-1
|
| 10 |
+
results: []
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 14 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 15 |
+
|
| 16 |
+
# zephyr-7b-dpo-full-prometheus_consistent-reward-scale-1
|
| 17 |
+
|
| 18 |
+
This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full) on the None dataset.
|
| 19 |
+
It achieves the following results on the evaluation set:
|
| 20 |
+
- Loss: 0.5207
|
| 21 |
+
- Rewards/chosen: -1.2231
|
| 22 |
+
- Rewards/rejected: -2.4659
|
| 23 |
+
- Rewards/accuracies: 0.7716
|
| 24 |
+
- Rewards/margins: 1.2428
|
| 25 |
+
- Logps/rejected: -465.6644
|
| 26 |
+
- Logps/chosen: -397.9129
|
| 27 |
+
- Logits/rejected: 0.9533
|
| 28 |
+
- Logits/chosen: -0.4953
|
| 29 |
+
|
| 30 |
+
## Model description
|
| 31 |
+
|
| 32 |
+
More information needed
|
| 33 |
+
|
| 34 |
+
## Intended uses & limitations
|
| 35 |
+
|
| 36 |
+
More information needed
|
| 37 |
+
|
| 38 |
+
## Training and evaluation data
|
| 39 |
+
|
| 40 |
+
More information needed
|
| 41 |
+
|
| 42 |
+
## Training procedure
|
| 43 |
+
|
| 44 |
+
### Training hyperparameters
|
| 45 |
+
|
| 46 |
+
The following hyperparameters were used during training:
|
| 47 |
+
- learning_rate: 5e-07
|
| 48 |
+
- train_batch_size: 8
|
| 49 |
+
- eval_batch_size: 8
|
| 50 |
+
- seed: 55
|
| 51 |
+
- distributed_type: multi-GPU
|
| 52 |
+
- num_devices: 8
|
| 53 |
+
- gradient_accumulation_steps: 2
|
| 54 |
+
- total_train_batch_size: 128
|
| 55 |
+
- total_eval_batch_size: 64
|
| 56 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 57 |
+
- lr_scheduler_type: cosine
|
| 58 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 59 |
+
- num_epochs: 1
|
| 60 |
+
|
| 61 |
+
### Training results
|
| 62 |
+
|
| 63 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 64 |
+
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 65 |
+
| 0.6662 | 0.1143 | 50 | 0.6533 | -0.0057 | -0.1647 | 0.6853 | 0.1590 | -235.5430 | -276.1736 | -2.4378 | -2.5411 |
|
| 66 |
+
| 0.6064 | 0.2286 | 100 | 0.6013 | -0.6726 | -1.2005 | 0.7414 | 0.5279 | -339.1220 | -342.8641 | -2.3762 | -2.4982 |
|
| 67 |
+
| 0.5361 | 0.3429 | 150 | 0.5524 | -0.9597 | -2.0693 | 0.7629 | 1.1096 | -426.0058 | -371.5730 | 0.1212 | -0.9465 |
|
| 68 |
+
| 0.5408 | 0.4571 | 200 | 0.5365 | -0.9091 | -2.0574 | 0.7543 | 1.1483 | -424.8160 | -366.5098 | 0.2603 | -1.1779 |
|
| 69 |
+
| 0.5444 | 0.5714 | 250 | 0.5319 | -1.0642 | -2.2338 | 0.7672 | 1.1696 | -442.4582 | -382.0276 | 0.2351 | -0.8157 |
|
| 70 |
+
| 0.5195 | 0.6857 | 300 | 0.5224 | -1.0055 | -2.2442 | 0.7802 | 1.2387 | -443.4946 | -376.1569 | 0.5828 | -0.7317 |
|
| 71 |
+
| 0.5294 | 0.8 | 350 | 0.5203 | -1.1639 | -2.4063 | 0.7543 | 1.2424 | -459.7086 | -391.9946 | 0.8069 | -0.5934 |
|
| 72 |
+
| 0.5494 | 0.9143 | 400 | 0.5207 | -1.2231 | -2.4659 | 0.7716 | 1.2428 | -465.6644 | -397.9129 | 0.9533 | -0.4953 |
|
| 73 |
+
|
| 74 |
+
|
| 75 |
+
### Framework versions
|
| 76 |
+
|
| 77 |
+
- Transformers 4.44.0.dev0
|
| 78 |
+
- Pytorch 2.1.2
|
| 79 |
+
- Datasets 2.20.0
|
| 80 |
+
- Tokenizers 0.19.1
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9988571428571429,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5634412694577494,
|
| 5 |
+
"train_runtime": 11351.7318,
|
| 6 |
+
"train_samples": 55999,
|
| 7 |
+
"train_samples_per_second": 4.933,
|
| 8 |
+
"train_steps_per_second": 0.038
|
| 9 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 1,
|
| 4 |
+
"eos_token_id": 2,
|
| 5 |
+
"transformers_version": "4.44.0.dev0"
|
| 6 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9988571428571429,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5634412694577494,
|
| 5 |
+
"train_runtime": 11351.7318,
|
| 6 |
+
"train_samples": 55999,
|
| 7 |
+
"train_samples_per_second": 4.933,
|
| 8 |
+
"train_steps_per_second": 0.038
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,815 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.9988571428571429,
|
| 5 |
+
"eval_steps": 50,
|
| 6 |
+
"global_step": 437,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.022857142857142857,
|
| 13 |
+
"grad_norm": 6.7934202337025855,
|
| 14 |
+
"learning_rate": 1.1363636363636363e-07,
|
| 15 |
+
"logits/chosen": -2.700854778289795,
|
| 16 |
+
"logits/rejected": -2.6249899864196777,
|
| 17 |
+
"logps/chosen": -301.2730712890625,
|
| 18 |
+
"logps/rejected": -281.7767028808594,
|
| 19 |
+
"loss": 0.6931,
|
| 20 |
+
"rewards/accuracies": 0.44999998807907104,
|
| 21 |
+
"rewards/chosen": 0.00013105585821904242,
|
| 22 |
+
"rewards/margins": 0.0004016496823169291,
|
| 23 |
+
"rewards/rejected": -0.0002705938240978867,
|
| 24 |
+
"step": 10
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.045714285714285714,
|
| 28 |
+
"grad_norm": 5.3206971599426955,
|
| 29 |
+
"learning_rate": 2.2727272727272726e-07,
|
| 30 |
+
"logits/chosen": -2.6406970024108887,
|
| 31 |
+
"logits/rejected": -2.6055245399475098,
|
| 32 |
+
"logps/chosen": -278.92767333984375,
|
| 33 |
+
"logps/rejected": -254.66806030273438,
|
| 34 |
+
"loss": 0.6925,
|
| 35 |
+
"rewards/accuracies": 0.5625,
|
| 36 |
+
"rewards/chosen": 0.0024350073654204607,
|
| 37 |
+
"rewards/margins": 0.001425002934411168,
|
| 38 |
+
"rewards/rejected": 0.0010100044310092926,
|
| 39 |
+
"step": 20
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.06857142857142857,
|
| 43 |
+
"grad_norm": 5.928979160384808,
|
| 44 |
+
"learning_rate": 3.4090909090909085e-07,
|
| 45 |
+
"logits/chosen": -2.63812518119812,
|
| 46 |
+
"logits/rejected": -2.617269992828369,
|
| 47 |
+
"logps/chosen": -263.2200622558594,
|
| 48 |
+
"logps/rejected": -263.3509216308594,
|
| 49 |
+
"loss": 0.6884,
|
| 50 |
+
"rewards/accuracies": 0.606249988079071,
|
| 51 |
+
"rewards/chosen": 0.014629853889346123,
|
| 52 |
+
"rewards/margins": 0.00820869579911232,
|
| 53 |
+
"rewards/rejected": 0.0064211576245725155,
|
| 54 |
+
"step": 30
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.09142857142857143,
|
| 58 |
+
"grad_norm": 6.666624768443373,
|
| 59 |
+
"learning_rate": 4.545454545454545e-07,
|
| 60 |
+
"logits/chosen": -2.649786949157715,
|
| 61 |
+
"logits/rejected": -2.586179733276367,
|
| 62 |
+
"logps/chosen": -290.2237548828125,
|
| 63 |
+
"logps/rejected": -268.332763671875,
|
| 64 |
+
"loss": 0.6786,
|
| 65 |
+
"rewards/accuracies": 0.59375,
|
| 66 |
+
"rewards/chosen": 0.03773483633995056,
|
| 67 |
+
"rewards/margins": 0.0447491891682148,
|
| 68 |
+
"rewards/rejected": -0.0070143528282642365,
|
| 69 |
+
"step": 40
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.11428571428571428,
|
| 73 |
+
"grad_norm": 9.747697215916363,
|
| 74 |
+
"learning_rate": 4.997124959943201e-07,
|
| 75 |
+
"logits/chosen": -2.6796483993530273,
|
| 76 |
+
"logits/rejected": -2.5985970497131348,
|
| 77 |
+
"logps/chosen": -293.6056213378906,
|
| 78 |
+
"logps/rejected": -254.311767578125,
|
| 79 |
+
"loss": 0.6662,
|
| 80 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 81 |
+
"rewards/chosen": 0.027046790346503258,
|
| 82 |
+
"rewards/margins": 0.10152111202478409,
|
| 83 |
+
"rewards/rejected": -0.07447431981563568,
|
| 84 |
+
"step": 50
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.11428571428571428,
|
| 88 |
+
"eval_logits/chosen": -2.541074275970459,
|
| 89 |
+
"eval_logits/rejected": -2.4378066062927246,
|
| 90 |
+
"eval_logps/chosen": -276.17364501953125,
|
| 91 |
+
"eval_logps/rejected": -235.5430145263672,
|
| 92 |
+
"eval_loss": 0.6532636284828186,
|
| 93 |
+
"eval_rewards/accuracies": 0.6853448152542114,
|
| 94 |
+
"eval_rewards/chosen": -0.005697119049727917,
|
| 95 |
+
"eval_rewards/margins": 0.1589716672897339,
|
| 96 |
+
"eval_rewards/rejected": -0.16466876864433289,
|
| 97 |
+
"eval_runtime": 91.4644,
|
| 98 |
+
"eval_samples_per_second": 20.019,
|
| 99 |
+
"eval_steps_per_second": 0.317,
|
| 100 |
+
"step": 50
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.13714285714285715,
|
| 104 |
+
"grad_norm": 9.11809665360445,
|
| 105 |
+
"learning_rate": 4.979579212164186e-07,
|
| 106 |
+
"logits/chosen": -2.579230308532715,
|
| 107 |
+
"logits/rejected": -2.47153639793396,
|
| 108 |
+
"logps/chosen": -293.18316650390625,
|
| 109 |
+
"logps/rejected": -275.13836669921875,
|
| 110 |
+
"loss": 0.646,
|
| 111 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 112 |
+
"rewards/chosen": -0.12677232921123505,
|
| 113 |
+
"rewards/margins": 0.13909009099006653,
|
| 114 |
+
"rewards/rejected": -0.2658624053001404,
|
| 115 |
+
"step": 60
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"epoch": 0.16,
|
| 119 |
+
"grad_norm": 9.033724858243865,
|
| 120 |
+
"learning_rate": 4.946196886175515e-07,
|
| 121 |
+
"logits/chosen": -2.587209701538086,
|
| 122 |
+
"logits/rejected": -2.537858247756958,
|
| 123 |
+
"logps/chosen": -293.44488525390625,
|
| 124 |
+
"logps/rejected": -300.06134033203125,
|
| 125 |
+
"loss": 0.6244,
|
| 126 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 127 |
+
"rewards/chosen": -0.1766660511493683,
|
| 128 |
+
"rewards/margins": 0.2276792824268341,
|
| 129 |
+
"rewards/rejected": -0.4043453335762024,
|
| 130 |
+
"step": 70
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 0.18285714285714286,
|
| 134 |
+
"grad_norm": 9.3700719469498,
|
| 135 |
+
"learning_rate": 4.897191188239667e-07,
|
| 136 |
+
"logits/chosen": -2.622152805328369,
|
| 137 |
+
"logits/rejected": -2.571810483932495,
|
| 138 |
+
"logps/chosen": -285.61334228515625,
|
| 139 |
+
"logps/rejected": -306.8240051269531,
|
| 140 |
+
"loss": 0.6129,
|
| 141 |
+
"rewards/accuracies": 0.731249988079071,
|
| 142 |
+
"rewards/chosen": -0.1858430951833725,
|
| 143 |
+
"rewards/margins": 0.3293076157569885,
|
| 144 |
+
"rewards/rejected": -0.5151506662368774,
|
| 145 |
+
"step": 80
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"epoch": 0.2057142857142857,
|
| 149 |
+
"grad_norm": 12.801415984987877,
|
| 150 |
+
"learning_rate": 4.832875107981763e-07,
|
| 151 |
+
"logits/chosen": -2.6656365394592285,
|
| 152 |
+
"logits/rejected": -2.609879732131958,
|
| 153 |
+
"logps/chosen": -298.3469543457031,
|
| 154 |
+
"logps/rejected": -315.2607421875,
|
| 155 |
+
"loss": 0.6171,
|
| 156 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 157 |
+
"rewards/chosen": -0.2206166684627533,
|
| 158 |
+
"rewards/margins": 0.3765110969543457,
|
| 159 |
+
"rewards/rejected": -0.5971277952194214,
|
| 160 |
+
"step": 90
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 0.22857142857142856,
|
| 164 |
+
"grad_norm": 13.512709674743487,
|
| 165 |
+
"learning_rate": 4.753659419387223e-07,
|
| 166 |
+
"logits/chosen": -2.659526824951172,
|
| 167 |
+
"logits/rejected": -2.5662050247192383,
|
| 168 |
+
"logps/chosen": -337.0499267578125,
|
| 169 |
+
"logps/rejected": -323.8541564941406,
|
| 170 |
+
"loss": 0.6064,
|
| 171 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 172 |
+
"rewards/chosen": -0.5066351294517517,
|
| 173 |
+
"rewards/margins": 0.39739224314689636,
|
| 174 |
+
"rewards/rejected": -0.9040273427963257,
|
| 175 |
+
"step": 100
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"epoch": 0.22857142857142856,
|
| 179 |
+
"eval_logits/chosen": -2.4981839656829834,
|
| 180 |
+
"eval_logits/rejected": -2.3761825561523438,
|
| 181 |
+
"eval_logps/chosen": -342.8641357421875,
|
| 182 |
+
"eval_logps/rejected": -339.1220397949219,
|
| 183 |
+
"eval_loss": 0.601341187953949,
|
| 184 |
+
"eval_rewards/accuracies": 0.7413793206214905,
|
| 185 |
+
"eval_rewards/chosen": -0.6726016402244568,
|
| 186 |
+
"eval_rewards/margins": 0.5278574824333191,
|
| 187 |
+
"eval_rewards/rejected": -1.2004590034484863,
|
| 188 |
+
"eval_runtime": 90.7394,
|
| 189 |
+
"eval_samples_per_second": 20.179,
|
| 190 |
+
"eval_steps_per_second": 0.32,
|
| 191 |
+
"step": 100
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.25142857142857145,
|
| 195 |
+
"grad_norm": 19.006376300807332,
|
| 196 |
+
"learning_rate": 4.660050057270191e-07,
|
| 197 |
+
"logits/chosen": -2.255047559738159,
|
| 198 |
+
"logits/rejected": -2.1392579078674316,
|
| 199 |
+
"logps/chosen": -385.914794921875,
|
| 200 |
+
"logps/rejected": -409.34942626953125,
|
| 201 |
+
"loss": 0.5827,
|
| 202 |
+
"rewards/accuracies": 0.643750011920929,
|
| 203 |
+
"rewards/chosen": -0.6961116194725037,
|
| 204 |
+
"rewards/margins": 0.4182862639427185,
|
| 205 |
+
"rewards/rejected": -1.1143978834152222,
|
| 206 |
+
"step": 110
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.2742857142857143,
|
| 210 |
+
"grad_norm": 18.203031953893095,
|
| 211 |
+
"learning_rate": 4.5526448859687144e-07,
|
| 212 |
+
"logits/chosen": -1.6530084609985352,
|
| 213 |
+
"logits/rejected": -1.3287270069122314,
|
| 214 |
+
"logps/chosen": -383.32232666015625,
|
| 215 |
+
"logps/rejected": -358.1570129394531,
|
| 216 |
+
"loss": 0.5761,
|
| 217 |
+
"rewards/accuracies": 0.706250011920929,
|
| 218 |
+
"rewards/chosen": -0.7241784334182739,
|
| 219 |
+
"rewards/margins": 0.5402495861053467,
|
| 220 |
+
"rewards/rejected": -1.264427900314331,
|
| 221 |
+
"step": 120
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.29714285714285715,
|
| 225 |
+
"grad_norm": 25.870445845824126,
|
| 226 |
+
"learning_rate": 4.432129880904388e-07,
|
| 227 |
+
"logits/chosen": -0.8325661420822144,
|
| 228 |
+
"logits/rejected": -0.3734116554260254,
|
| 229 |
+
"logps/chosen": -404.2379455566406,
|
| 230 |
+
"logps/rejected": -406.59088134765625,
|
| 231 |
+
"loss": 0.5471,
|
| 232 |
+
"rewards/accuracies": 0.606249988079071,
|
| 233 |
+
"rewards/chosen": -0.9908201098442078,
|
| 234 |
+
"rewards/margins": 0.5703855752944946,
|
| 235 |
+
"rewards/rejected": -1.5612057447433472,
|
| 236 |
+
"step": 130
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 0.32,
|
| 240 |
+
"grad_norm": 24.51686978994005,
|
| 241 |
+
"learning_rate": 4.299274747394055e-07,
|
| 242 |
+
"logits/chosen": -1.1053862571716309,
|
| 243 |
+
"logits/rejected": -0.7209962606430054,
|
| 244 |
+
"logps/chosen": -378.8704833984375,
|
| 245 |
+
"logps/rejected": -408.87982177734375,
|
| 246 |
+
"loss": 0.5662,
|
| 247 |
+
"rewards/accuracies": 0.699999988079071,
|
| 248 |
+
"rewards/chosen": -0.634228527545929,
|
| 249 |
+
"rewards/margins": 0.735455334186554,
|
| 250 |
+
"rewards/rejected": -1.3696839809417725,
|
| 251 |
+
"step": 140
|
| 252 |
+
},
|
| 253 |
+
{
|
| 254 |
+
"epoch": 0.34285714285714286,
|
| 255 |
+
"grad_norm": 20.058403951755405,
|
| 256 |
+
"learning_rate": 4.1549280046953653e-07,
|
| 257 |
+
"logits/chosen": -1.0304594039916992,
|
| 258 |
+
"logits/rejected": -0.4146367907524109,
|
| 259 |
+
"logps/chosen": -371.8462829589844,
|
| 260 |
+
"logps/rejected": -442.0589904785156,
|
| 261 |
+
"loss": 0.5361,
|
| 262 |
+
"rewards/accuracies": 0.699999988079071,
|
| 263 |
+
"rewards/chosen": -0.848603367805481,
|
| 264 |
+
"rewards/margins": 0.8089972734451294,
|
| 265 |
+
"rewards/rejected": -1.6576006412506104,
|
| 266 |
+
"step": 150
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 0.34285714285714286,
|
| 270 |
+
"eval_logits/chosen": -0.9465453028678894,
|
| 271 |
+
"eval_logits/rejected": 0.12122771143913269,
|
| 272 |
+
"eval_logps/chosen": -371.5729675292969,
|
| 273 |
+
"eval_logps/rejected": -426.00579833984375,
|
| 274 |
+
"eval_loss": 0.5524057149887085,
|
| 275 |
+
"eval_rewards/accuracies": 0.7629310488700867,
|
| 276 |
+
"eval_rewards/chosen": -0.9596903324127197,
|
| 277 |
+
"eval_rewards/margins": 1.1096062660217285,
|
| 278 |
+
"eval_rewards/rejected": -2.069296360015869,
|
| 279 |
+
"eval_runtime": 90.5989,
|
| 280 |
+
"eval_samples_per_second": 20.21,
|
| 281 |
+
"eval_steps_per_second": 0.32,
|
| 282 |
+
"step": 150
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 0.3657142857142857,
|
| 286 |
+
"grad_norm": 20.887944941915286,
|
| 287 |
+
"learning_rate": 4.000011566683401e-07,
|
| 288 |
+
"logits/chosen": -1.1443428993225098,
|
| 289 |
+
"logits/rejected": -0.4119592607021332,
|
| 290 |
+
"logps/chosen": -398.99407958984375,
|
| 291 |
+
"logps/rejected": -444.39227294921875,
|
| 292 |
+
"loss": 0.5491,
|
| 293 |
+
"rewards/accuracies": 0.71875,
|
| 294 |
+
"rewards/chosen": -0.9965378642082214,
|
| 295 |
+
"rewards/margins": 0.8909576535224915,
|
| 296 |
+
"rewards/rejected": -1.8874953985214233,
|
| 297 |
+
"step": 160
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"epoch": 0.38857142857142857,
|
| 301 |
+
"grad_norm": 26.245115570494136,
|
| 302 |
+
"learning_rate": 3.8355148537705047e-07,
|
| 303 |
+
"logits/chosen": -1.1884818077087402,
|
| 304 |
+
"logits/rejected": -0.4392818808555603,
|
| 305 |
+
"logps/chosen": -405.96453857421875,
|
| 306 |
+
"logps/rejected": -431.96875,
|
| 307 |
+
"loss": 0.545,
|
| 308 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 309 |
+
"rewards/chosen": -1.0185964107513428,
|
| 310 |
+
"rewards/margins": 0.643780529499054,
|
| 311 |
+
"rewards/rejected": -1.6623769998550415,
|
| 312 |
+
"step": 170
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 0.4114285714285714,
|
| 316 |
+
"grad_norm": 16.97689595709094,
|
| 317 |
+
"learning_rate": 3.662488473675315e-07,
|
| 318 |
+
"logits/chosen": -1.2794054746627808,
|
| 319 |
+
"logits/rejected": -0.2185342013835907,
|
| 320 |
+
"logps/chosen": -429.20098876953125,
|
| 321 |
+
"logps/rejected": -465.3219299316406,
|
| 322 |
+
"loss": 0.554,
|
| 323 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 324 |
+
"rewards/chosen": -0.9902114868164062,
|
| 325 |
+
"rewards/margins": 1.0635021924972534,
|
| 326 |
+
"rewards/rejected": -2.053713798522949,
|
| 327 |
+
"step": 180
|
| 328 |
+
},
|
| 329 |
+
{
|
| 330 |
+
"epoch": 0.4342857142857143,
|
| 331 |
+
"grad_norm": 15.630086574400098,
|
| 332 |
+
"learning_rate": 3.48203751140067e-07,
|
| 333 |
+
"logits/chosen": -1.3240091800689697,
|
| 334 |
+
"logits/rejected": -0.43907278776168823,
|
| 335 |
+
"logps/chosen": -380.5593566894531,
|
| 336 |
+
"logps/rejected": -413.9124450683594,
|
| 337 |
+
"loss": 0.5389,
|
| 338 |
+
"rewards/accuracies": 0.706250011920929,
|
| 339 |
+
"rewards/chosen": -1.043017864227295,
|
| 340 |
+
"rewards/margins": 0.7511247992515564,
|
| 341 |
+
"rewards/rejected": -1.794142484664917,
|
| 342 |
+
"step": 190
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 0.45714285714285713,
|
| 346 |
+
"grad_norm": 17.980745825993548,
|
| 347 |
+
"learning_rate": 3.2953144712759537e-07,
|
| 348 |
+
"logits/chosen": -1.4564930200576782,
|
| 349 |
+
"logits/rejected": -0.43932801485061646,
|
| 350 |
+
"logps/chosen": -365.29913330078125,
|
| 351 |
+
"logps/rejected": -422.6018981933594,
|
| 352 |
+
"loss": 0.5408,
|
| 353 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 354 |
+
"rewards/chosen": -0.9695641398429871,
|
| 355 |
+
"rewards/margins": 0.9676488041877747,
|
| 356 |
+
"rewards/rejected": -1.9372127056121826,
|
| 357 |
+
"step": 200
|
| 358 |
+
},
|
| 359 |
+
{
|
| 360 |
+
"epoch": 0.45714285714285713,
|
| 361 |
+
"eval_logits/chosen": -1.1779377460479736,
|
| 362 |
+
"eval_logits/rejected": 0.2602512240409851,
|
| 363 |
+
"eval_logps/chosen": -366.50982666015625,
|
| 364 |
+
"eval_logps/rejected": -424.8160095214844,
|
| 365 |
+
"eval_loss": 0.5365040898323059,
|
| 366 |
+
"eval_rewards/accuracies": 0.7543103694915771,
|
| 367 |
+
"eval_rewards/chosen": -0.9090585112571716,
|
| 368 |
+
"eval_rewards/margins": 1.148340106010437,
|
| 369 |
+
"eval_rewards/rejected": -2.057398557662964,
|
| 370 |
+
"eval_runtime": 92.0351,
|
| 371 |
+
"eval_samples_per_second": 19.895,
|
| 372 |
+
"eval_steps_per_second": 0.315,
|
| 373 |
+
"step": 200
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"epoch": 0.48,
|
| 377 |
+
"grad_norm": 20.920453502633716,
|
| 378 |
+
"learning_rate": 3.103511916141658e-07,
|
| 379 |
+
"logits/chosen": -0.5911210775375366,
|
| 380 |
+
"logits/rejected": 0.41120481491088867,
|
| 381 |
+
"logps/chosen": -377.15191650390625,
|
| 382 |
+
"logps/rejected": -446.093017578125,
|
| 383 |
+
"loss": 0.5341,
|
| 384 |
+
"rewards/accuracies": 0.706250011920929,
|
| 385 |
+
"rewards/chosen": -1.13698410987854,
|
| 386 |
+
"rewards/margins": 0.8711468577384949,
|
| 387 |
+
"rewards/rejected": -2.0081307888031006,
|
| 388 |
+
"step": 210
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 0.5028571428571429,
|
| 392 |
+
"grad_norm": 22.32799037201365,
|
| 393 |
+
"learning_rate": 2.9078548506882117e-07,
|
| 394 |
+
"logits/chosen": -0.42250823974609375,
|
| 395 |
+
"logits/rejected": 0.5733365416526794,
|
| 396 |
+
"logps/chosen": -427.79559326171875,
|
| 397 |
+
"logps/rejected": -463.9886779785156,
|
| 398 |
+
"loss": 0.5589,
|
| 399 |
+
"rewards/accuracies": 0.731249988079071,
|
| 400 |
+
"rewards/chosen": -1.4272958040237427,
|
| 401 |
+
"rewards/margins": 0.7507564425468445,
|
| 402 |
+
"rewards/rejected": -2.1780521869659424,
|
| 403 |
+
"step": 220
|
| 404 |
+
},
|
| 405 |
+
{
|
| 406 |
+
"epoch": 0.5257142857142857,
|
| 407 |
+
"grad_norm": 22.22957333517813,
|
| 408 |
+
"learning_rate": 2.709592897595191e-07,
|
| 409 |
+
"logits/chosen": -0.6746516823768616,
|
| 410 |
+
"logits/rejected": 0.0753551721572876,
|
| 411 |
+
"logps/chosen": -399.5505676269531,
|
| 412 |
+
"logps/rejected": -437.2046813964844,
|
| 413 |
+
"loss": 0.5344,
|
| 414 |
+
"rewards/accuracies": 0.706250011920929,
|
| 415 |
+
"rewards/chosen": -1.184914231300354,
|
| 416 |
+
"rewards/margins": 0.794265627861023,
|
| 417 |
+
"rewards/rejected": -1.9791797399520874,
|
| 418 |
+
"step": 230
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"epoch": 0.5485714285714286,
|
| 422 |
+
"grad_norm": 22.46549138618286,
|
| 423 |
+
"learning_rate": 2.509992316440332e-07,
|
| 424 |
+
"logits/chosen": -0.9359372854232788,
|
| 425 |
+
"logits/rejected": 0.04701079800724983,
|
| 426 |
+
"logps/chosen": -403.68194580078125,
|
| 427 |
+
"logps/rejected": -491.6949157714844,
|
| 428 |
+
"loss": 0.5315,
|
| 429 |
+
"rewards/accuracies": 0.768750011920929,
|
| 430 |
+
"rewards/chosen": -1.049721360206604,
|
| 431 |
+
"rewards/margins": 1.1530095338821411,
|
| 432 |
+
"rewards/rejected": -2.202730894088745,
|
| 433 |
+
"step": 240
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 0.5714285714285714,
|
| 437 |
+
"grad_norm": 18.11548443644794,
|
| 438 |
+
"learning_rate": 2.3103279163519918e-07,
|
| 439 |
+
"logits/chosen": -0.7540313601493835,
|
| 440 |
+
"logits/rejected": -0.2726219594478607,
|
| 441 |
+
"logps/chosen": -388.8069763183594,
|
| 442 |
+
"logps/rejected": -476.65692138671875,
|
| 443 |
+
"loss": 0.5444,
|
| 444 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 445 |
+
"rewards/chosen": -1.0818074941635132,
|
| 446 |
+
"rewards/margins": 0.9827569127082825,
|
| 447 |
+
"rewards/rejected": -2.0645642280578613,
|
| 448 |
+
"step": 250
|
| 449 |
+
},
|
| 450 |
+
{
|
| 451 |
+
"epoch": 0.5714285714285714,
|
| 452 |
+
"eval_logits/chosen": -0.8157382011413574,
|
| 453 |
+
"eval_logits/rejected": 0.23507556319236755,
|
| 454 |
+
"eval_logps/chosen": -382.0275573730469,
|
| 455 |
+
"eval_logps/rejected": -442.4582214355469,
|
| 456 |
+
"eval_loss": 0.5319008827209473,
|
| 457 |
+
"eval_rewards/accuracies": 0.767241358757019,
|
| 458 |
+
"eval_rewards/chosen": -1.0642356872558594,
|
| 459 |
+
"eval_rewards/margins": 1.16958487033844,
|
| 460 |
+
"eval_rewards/rejected": -2.233820676803589,
|
| 461 |
+
"eval_runtime": 91.9506,
|
| 462 |
+
"eval_samples_per_second": 19.913,
|
| 463 |
+
"eval_steps_per_second": 0.315,
|
| 464 |
+
"step": 250
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 0.5942857142857143,
|
| 468 |
+
"grad_norm": 23.815191161905986,
|
| 469 |
+
"learning_rate": 2.1118749140573358e-07,
|
| 470 |
+
"logits/chosen": -0.47870174050331116,
|
| 471 |
+
"logits/rejected": -0.004892909433692694,
|
| 472 |
+
"logps/chosen": -419.97027587890625,
|
| 473 |
+
"logps/rejected": -487.68896484375,
|
| 474 |
+
"loss": 0.5427,
|
| 475 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 476 |
+
"rewards/chosen": -1.4018356800079346,
|
| 477 |
+
"rewards/margins": 0.7416240572929382,
|
| 478 |
+
"rewards/rejected": -2.1434597969055176,
|
| 479 |
+
"step": 260
|
| 480 |
+
},
|
| 481 |
+
{
|
| 482 |
+
"epoch": 0.6171428571428571,
|
| 483 |
+
"grad_norm": 22.07331008404508,
|
| 484 |
+
"learning_rate": 1.9159007893272703e-07,
|
| 485 |
+
"logits/chosen": -0.11218120157718658,
|
| 486 |
+
"logits/rejected": 1.1597168445587158,
|
| 487 |
+
"logps/chosen": -407.2730712890625,
|
| 488 |
+
"logps/rejected": -467.1512756347656,
|
| 489 |
+
"loss": 0.5208,
|
| 490 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 491 |
+
"rewards/chosen": -1.3474724292755127,
|
| 492 |
+
"rewards/margins": 1.0189656019210815,
|
| 493 |
+
"rewards/rejected": -2.366438150405884,
|
| 494 |
+
"step": 270
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 0.64,
|
| 498 |
+
"grad_norm": 26.61085213163748,
|
| 499 |
+
"learning_rate": 1.7236571898357766e-07,
|
| 500 |
+
"logits/chosen": -0.20857758820056915,
|
| 501 |
+
"logits/rejected": 0.7698697447776794,
|
| 502 |
+
"logps/chosen": -405.8653869628906,
|
| 503 |
+
"logps/rejected": -496.41900634765625,
|
| 504 |
+
"loss": 0.5278,
|
| 505 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 506 |
+
"rewards/chosen": -1.3282725811004639,
|
| 507 |
+
"rewards/margins": 1.0048491954803467,
|
| 508 |
+
"rewards/rejected": -2.3331217765808105,
|
| 509 |
+
"step": 280
|
| 510 |
+
},
|
| 511 |
+
{
|
| 512 |
+
"epoch": 0.6628571428571428,
|
| 513 |
+
"grad_norm": 24.592234854455963,
|
| 514 |
+
"learning_rate": 1.5363719371356882e-07,
|
| 515 |
+
"logits/chosen": -0.3117966651916504,
|
| 516 |
+
"logits/rejected": 0.6535851359367371,
|
| 517 |
+
"logps/chosen": -433.9137268066406,
|
| 518 |
+
"logps/rejected": -489.1080017089844,
|
| 519 |
+
"loss": 0.5288,
|
| 520 |
+
"rewards/accuracies": 0.75,
|
| 521 |
+
"rewards/chosen": -1.3300788402557373,
|
| 522 |
+
"rewards/margins": 0.9055169820785522,
|
| 523 |
+
"rewards/rejected": -2.235595703125,
|
| 524 |
+
"step": 290
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"epoch": 0.6857142857142857,
|
| 528 |
+
"grad_norm": 17.15967829902915,
|
| 529 |
+
"learning_rate": 1.3552411848071565e-07,
|
| 530 |
+
"logits/chosen": -0.4298080503940582,
|
| 531 |
+
"logits/rejected": 0.8657402992248535,
|
| 532 |
+
"logps/chosen": -430.40216064453125,
|
| 533 |
+
"logps/rejected": -487.58544921875,
|
| 534 |
+
"loss": 0.5195,
|
| 535 |
+
"rewards/accuracies": 0.75,
|
| 536 |
+
"rewards/chosen": -1.2849299907684326,
|
| 537 |
+
"rewards/margins": 1.0645029544830322,
|
| 538 |
+
"rewards/rejected": -2.349433183670044,
|
| 539 |
+
"step": 300
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 0.6857142857142857,
|
| 543 |
+
"eval_logits/chosen": -0.7317452430725098,
|
| 544 |
+
"eval_logits/rejected": 0.5828115940093994,
|
| 545 |
+
"eval_logps/chosen": -376.1568603515625,
|
| 546 |
+
"eval_logps/rejected": -443.4945983886719,
|
| 547 |
+
"eval_loss": 0.522352933883667,
|
| 548 |
+
"eval_rewards/accuracies": 0.7801724076271057,
|
| 549 |
+
"eval_rewards/chosen": -1.0055286884307861,
|
| 550 |
+
"eval_rewards/margins": 1.2386558055877686,
|
| 551 |
+
"eval_rewards/rejected": -2.2441844940185547,
|
| 552 |
+
"eval_runtime": 90.6437,
|
| 553 |
+
"eval_samples_per_second": 20.2,
|
| 554 |
+
"eval_steps_per_second": 0.32,
|
| 555 |
+
"step": 300
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 0.7085714285714285,
|
| 559 |
+
"grad_norm": 19.670890656220536,
|
| 560 |
+
"learning_rate": 1.1814217788631473e-07,
|
| 561 |
+
"logits/chosen": -0.6274710893630981,
|
| 562 |
+
"logits/rejected": 0.25394701957702637,
|
| 563 |
+
"logps/chosen": -375.80413818359375,
|
| 564 |
+
"logps/rejected": -445.390625,
|
| 565 |
+
"loss": 0.5308,
|
| 566 |
+
"rewards/accuracies": 0.6875,
|
| 567 |
+
"rewards/chosen": -1.1950000524520874,
|
| 568 |
+
"rewards/margins": 0.8869529962539673,
|
| 569 |
+
"rewards/rejected": -2.0819530487060547,
|
| 570 |
+
"step": 310
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"epoch": 0.7314285714285714,
|
| 574 |
+
"grad_norm": 19.334843719063908,
|
| 575 |
+
"learning_rate": 1.0160238692045331e-07,
|
| 576 |
+
"logits/chosen": -0.4475599229335785,
|
| 577 |
+
"logits/rejected": 0.30151477456092834,
|
| 578 |
+
"logps/chosen": -378.7860107421875,
|
| 579 |
+
"logps/rejected": -447.75927734375,
|
| 580 |
+
"loss": 0.5382,
|
| 581 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 582 |
+
"rewards/chosen": -1.345036506652832,
|
| 583 |
+
"rewards/margins": 0.7226740717887878,
|
| 584 |
+
"rewards/rejected": -2.0677103996276855,
|
| 585 |
+
"step": 320
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 0.7542857142857143,
|
| 589 |
+
"grad_norm": 21.07812644783168,
|
| 590 |
+
"learning_rate": 8.601038193139438e-08,
|
| 591 |
+
"logits/chosen": -0.6940028071403503,
|
| 592 |
+
"logits/rejected": 0.2311718910932541,
|
| 593 |
+
"logps/chosen": -427.7415466308594,
|
| 594 |
+
"logps/rejected": -477.0531311035156,
|
| 595 |
+
"loss": 0.5302,
|
| 596 |
+
"rewards/accuracies": 0.78125,
|
| 597 |
+
"rewards/chosen": -1.2852336168289185,
|
| 598 |
+
"rewards/margins": 1.0074186325073242,
|
| 599 |
+
"rewards/rejected": -2.2926523685455322,
|
| 600 |
+
"step": 330
|
| 601 |
+
},
|
| 602 |
+
{
|
| 603 |
+
"epoch": 0.7771428571428571,
|
| 604 |
+
"grad_norm": 18.64039907134669,
|
| 605 |
+
"learning_rate": 7.146574594727572e-08,
|
| 606 |
+
"logits/chosen": -0.3057669997215271,
|
| 607 |
+
"logits/rejected": 0.44727668166160583,
|
| 608 |
+
"logps/chosen": -404.18865966796875,
|
| 609 |
+
"logps/rejected": -483.5855407714844,
|
| 610 |
+
"loss": 0.5216,
|
| 611 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 612 |
+
"rewards/chosen": -1.3823763132095337,
|
| 613 |
+
"rewards/margins": 1.0332415103912354,
|
| 614 |
+
"rewards/rejected": -2.4156179428100586,
|
| 615 |
+
"step": 340
|
| 616 |
+
},
|
| 617 |
+
{
|
| 618 |
+
"epoch": 0.8,
|
| 619 |
+
"grad_norm": 22.5426523083621,
|
| 620 |
+
"learning_rate": 5.8061372659157306e-08,
|
| 621 |
+
"logits/chosen": -0.7059189081192017,
|
| 622 |
+
"logits/rejected": 0.17293286323547363,
|
| 623 |
+
"logps/chosen": -422.30352783203125,
|
| 624 |
+
"logps/rejected": -467.66473388671875,
|
| 625 |
+
"loss": 0.5294,
|
| 626 |
+
"rewards/accuracies": 0.699999988079071,
|
| 627 |
+
"rewards/chosen": -1.325963020324707,
|
| 628 |
+
"rewards/margins": 0.8077083826065063,
|
| 629 |
+
"rewards/rejected": -2.133671522140503,
|
| 630 |
+
"step": 350
|
| 631 |
+
},
|
| 632 |
+
{
|
| 633 |
+
"epoch": 0.8,
|
| 634 |
+
"eval_logits/chosen": -0.5933817625045776,
|
| 635 |
+
"eval_logits/rejected": 0.806897759437561,
|
| 636 |
+
"eval_logps/chosen": -391.99456787109375,
|
| 637 |
+
"eval_logps/rejected": -459.7085876464844,
|
| 638 |
+
"eval_loss": 0.5202531814575195,
|
| 639 |
+
"eval_rewards/accuracies": 0.7543103694915771,
|
| 640 |
+
"eval_rewards/chosen": -1.163906455039978,
|
| 641 |
+
"eval_rewards/margins": 1.2424179315567017,
|
| 642 |
+
"eval_rewards/rejected": -2.406324625015259,
|
| 643 |
+
"eval_runtime": 91.3368,
|
| 644 |
+
"eval_samples_per_second": 20.047,
|
| 645 |
+
"eval_steps_per_second": 0.318,
|
| 646 |
+
"step": 350
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"epoch": 0.8228571428571428,
|
| 650 |
+
"grad_norm": 17.956675630723836,
|
| 651 |
+
"learning_rate": 4.5882873127531614e-08,
|
| 652 |
+
"logits/chosen": -0.7362015247344971,
|
| 653 |
+
"logits/rejected": 0.41978272795677185,
|
| 654 |
+
"logps/chosen": -408.91522216796875,
|
| 655 |
+
"logps/rejected": -478.81451416015625,
|
| 656 |
+
"loss": 0.5169,
|
| 657 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 658 |
+
"rewards/chosen": -1.2351018190383911,
|
| 659 |
+
"rewards/margins": 1.0421779155731201,
|
| 660 |
+
"rewards/rejected": -2.2772796154022217,
|
| 661 |
+
"step": 360
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 0.8457142857142858,
|
| 665 |
+
"grad_norm": 19.593030414418095,
|
| 666 |
+
"learning_rate": 3.500802900154412e-08,
|
| 667 |
+
"logits/chosen": -0.6445934176445007,
|
| 668 |
+
"logits/rejected": 0.46942490339279175,
|
| 669 |
+
"logps/chosen": -387.9786071777344,
|
| 670 |
+
"logps/rejected": -464.98126220703125,
|
| 671 |
+
"loss": 0.5188,
|
| 672 |
+
"rewards/accuracies": 0.75,
|
| 673 |
+
"rewards/chosen": -1.1691720485687256,
|
| 674 |
+
"rewards/margins": 1.096484899520874,
|
| 675 |
+
"rewards/rejected": -2.2656567096710205,
|
| 676 |
+
"step": 370
|
| 677 |
+
},
|
| 678 |
+
{
|
| 679 |
+
"epoch": 0.8685714285714285,
|
| 680 |
+
"grad_norm": 32.43514788557893,
|
| 681 |
+
"learning_rate": 2.550629574310309e-08,
|
| 682 |
+
"logits/chosen": -0.766352117061615,
|
| 683 |
+
"logits/rejected": 0.4028262197971344,
|
| 684 |
+
"logps/chosen": -461.1717834472656,
|
| 685 |
+
"logps/rejected": -484.5889587402344,
|
| 686 |
+
"loss": 0.5266,
|
| 687 |
+
"rewards/accuracies": 0.6875,
|
| 688 |
+
"rewards/chosen": -1.4311548471450806,
|
| 689 |
+
"rewards/margins": 0.8407408595085144,
|
| 690 |
+
"rewards/rejected": -2.2718958854675293,
|
| 691 |
+
"step": 380
|
| 692 |
+
},
|
| 693 |
+
{
|
| 694 |
+
"epoch": 0.8914285714285715,
|
| 695 |
+
"grad_norm": 24.475035112334364,
|
| 696 |
+
"learning_rate": 1.7438359028687983e-08,
|
| 697 |
+
"logits/chosen": -0.6120296716690063,
|
| 698 |
+
"logits/rejected": 0.2141704559326172,
|
| 699 |
+
"logps/chosen": -433.58111572265625,
|
| 700 |
+
"logps/rejected": -508.9109802246094,
|
| 701 |
+
"loss": 0.5291,
|
| 702 |
+
"rewards/accuracies": 0.737500011920929,
|
| 703 |
+
"rewards/chosen": -1.222081184387207,
|
| 704 |
+
"rewards/margins": 0.9193018674850464,
|
| 705 |
+
"rewards/rejected": -2.141382932662964,
|
| 706 |
+
"step": 390
|
| 707 |
+
},
|
| 708 |
+
{
|
| 709 |
+
"epoch": 0.9142857142857143,
|
| 710 |
+
"grad_norm": 32.810528214564755,
|
| 711 |
+
"learning_rate": 1.0855747162029361e-08,
|
| 712 |
+
"logits/chosen": -0.3861927390098572,
|
| 713 |
+
"logits/rejected": -0.02623150870203972,
|
| 714 |
+
"logps/chosen": -416.01171875,
|
| 715 |
+
"logps/rejected": -484.2184143066406,
|
| 716 |
+
"loss": 0.5494,
|
| 717 |
+
"rewards/accuracies": 0.731249988079071,
|
| 718 |
+
"rewards/chosen": -1.3727846145629883,
|
| 719 |
+
"rewards/margins": 0.8036632537841797,
|
| 720 |
+
"rewards/rejected": -2.176448106765747,
|
| 721 |
+
"step": 400
|
| 722 |
+
},
|
| 723 |
+
{
|
| 724 |
+
"epoch": 0.9142857142857143,
|
| 725 |
+
"eval_logits/chosen": -0.495297372341156,
|
| 726 |
+
"eval_logits/rejected": 0.9533113241195679,
|
| 727 |
+
"eval_logps/chosen": -397.91290283203125,
|
| 728 |
+
"eval_logps/rejected": -465.6643981933594,
|
| 729 |
+
"eval_loss": 0.5207196474075317,
|
| 730 |
+
"eval_rewards/accuracies": 0.7715517282485962,
|
| 731 |
+
"eval_rewards/chosen": -1.2230898141860962,
|
| 732 |
+
"eval_rewards/margins": 1.2427926063537598,
|
| 733 |
+
"eval_rewards/rejected": -2.4658823013305664,
|
| 734 |
+
"eval_runtime": 90.0935,
|
| 735 |
+
"eval_samples_per_second": 20.323,
|
| 736 |
+
"eval_steps_per_second": 0.322,
|
| 737 |
+
"step": 400
|
| 738 |
+
},
|
| 739 |
+
{
|
| 740 |
+
"epoch": 0.9371428571428572,
|
| 741 |
+
"grad_norm": 21.588729517678896,
|
| 742 |
+
"learning_rate": 5.8005019731033615e-09,
|
| 743 |
+
"logits/chosen": -0.6237171292304993,
|
| 744 |
+
"logits/rejected": 0.3366670608520508,
|
| 745 |
+
"logps/chosen": -432.98870849609375,
|
| 746 |
+
"logps/rejected": -489.3666076660156,
|
| 747 |
+
"loss": 0.519,
|
| 748 |
+
"rewards/accuracies": 0.731249988079071,
|
| 749 |
+
"rewards/chosen": -1.448087453842163,
|
| 750 |
+
"rewards/margins": 0.8482778668403625,
|
| 751 |
+
"rewards/rejected": -2.296365261077881,
|
| 752 |
+
"step": 410
|
| 753 |
+
},
|
| 754 |
+
{
|
| 755 |
+
"epoch": 0.96,
|
| 756 |
+
"grad_norm": 19.892103213679338,
|
| 757 |
+
"learning_rate": 2.3049103053431886e-09,
|
| 758 |
+
"logits/chosen": -0.632776141166687,
|
| 759 |
+
"logits/rejected": 0.6551538705825806,
|
| 760 |
+
"logps/chosen": -390.06964111328125,
|
| 761 |
+
"logps/rejected": -465.9344177246094,
|
| 762 |
+
"loss": 0.5239,
|
| 763 |
+
"rewards/accuracies": 0.768750011920929,
|
| 764 |
+
"rewards/chosen": -1.073215126991272,
|
| 765 |
+
"rewards/margins": 1.246260404586792,
|
| 766 |
+
"rewards/rejected": -2.3194754123687744,
|
| 767 |
+
"step": 420
|
| 768 |
+
},
|
| 769 |
+
{
|
| 770 |
+
"epoch": 0.9828571428571429,
|
| 771 |
+
"grad_norm": 19.60054815654678,
|
| 772 |
+
"learning_rate": 3.9129780600541397e-10,
|
| 773 |
+
"logits/chosen": -0.3492957055568695,
|
| 774 |
+
"logits/rejected": 0.4392589032649994,
|
| 775 |
+
"logps/chosen": -409.07745361328125,
|
| 776 |
+
"logps/rejected": -488.45751953125,
|
| 777 |
+
"loss": 0.5228,
|
| 778 |
+
"rewards/accuracies": 0.75,
|
| 779 |
+
"rewards/chosen": -1.2684558629989624,
|
| 780 |
+
"rewards/margins": 0.9794237017631531,
|
| 781 |
+
"rewards/rejected": -2.2478795051574707,
|
| 782 |
+
"step": 430
|
| 783 |
+
},
|
| 784 |
+
{
|
| 785 |
+
"epoch": 0.9988571428571429,
|
| 786 |
+
"step": 437,
|
| 787 |
+
"total_flos": 0.0,
|
| 788 |
+
"train_loss": 0.5634412694577494,
|
| 789 |
+
"train_runtime": 11351.7318,
|
| 790 |
+
"train_samples_per_second": 4.933,
|
| 791 |
+
"train_steps_per_second": 0.038
|
| 792 |
+
}
|
| 793 |
+
],
|
| 794 |
+
"logging_steps": 10,
|
| 795 |
+
"max_steps": 437,
|
| 796 |
+
"num_input_tokens_seen": 0,
|
| 797 |
+
"num_train_epochs": 1,
|
| 798 |
+
"save_steps": 100,
|
| 799 |
+
"stateful_callbacks": {
|
| 800 |
+
"TrainerControl": {
|
| 801 |
+
"args": {
|
| 802 |
+
"should_epoch_stop": false,
|
| 803 |
+
"should_evaluate": false,
|
| 804 |
+
"should_log": false,
|
| 805 |
+
"should_save": true,
|
| 806 |
+
"should_training_stop": true
|
| 807 |
+
},
|
| 808 |
+
"attributes": {}
|
| 809 |
+
}
|
| 810 |
+
},
|
| 811 |
+
"total_flos": 0.0,
|
| 812 |
+
"train_batch_size": 8,
|
| 813 |
+
"trial_name": null,
|
| 814 |
+
"trial_params": null
|
| 815 |
+
}
|