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-gpt_consistent-reward-scale-1-rpo-gamma-05
|
| 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-gpt_consistent-reward-scale-1-rpo-gamma-05
|
| 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: 0.0087
|
| 21 |
+
- Rewards/chosen: -0.0012
|
| 22 |
+
- Rewards/rejected: -0.1544
|
| 23 |
+
- Rewards/accuracies: 0.7672
|
| 24 |
+
- Rewards/margins: 0.1532
|
| 25 |
+
- Logps/rejected: -261.9635
|
| 26 |
+
- Logps/chosen: -285.2108
|
| 27 |
+
- Logits/rejected: -2.3836
|
| 28 |
+
- Logits/chosen: -2.4800
|
| 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 |
+
| 0.012 | 0.1147 | 50 | 0.0113 | 0.0778 | -0.0076 | 0.7026 | 0.0854 | -247.2809 | -277.3098 | -2.5021 | -2.5787 |
|
| 66 |
+
| 0.011 | 0.2294 | 100 | 0.0100 | 0.0419 | -0.0781 | 0.7069 | 0.1200 | -254.3327 | -280.9024 | -2.3370 | -2.4392 |
|
| 67 |
+
| 0.0104 | 0.3440 | 150 | 0.0098 | -0.0076 | -0.1403 | 0.7198 | 0.1326 | -260.5493 | -285.8531 | -2.3956 | -2.4944 |
|
| 68 |
+
| 0.0096 | 0.4587 | 200 | 0.0093 | 0.0289 | -0.1110 | 0.7931 | 0.1399 | -257.6194 | -282.1954 | -2.4209 | -2.5140 |
|
| 69 |
+
| 0.0094 | 0.5734 | 250 | 0.0089 | -0.0113 | -0.1675 | 0.7802 | 0.1562 | -263.2694 | -286.2172 | -2.3630 | -2.4591 |
|
| 70 |
+
| 0.0096 | 0.6881 | 300 | 0.0088 | -0.0133 | -0.1638 | 0.7845 | 0.1505 | -262.9025 | -286.4156 | -2.3821 | -2.4772 |
|
| 71 |
+
| 0.0096 | 0.8028 | 350 | 0.0087 | -0.0056 | -0.1617 | 0.7802 | 0.1561 | -262.6906 | -285.6459 | -2.3851 | -2.4814 |
|
| 72 |
+
| 0.0093 | 0.9174 | 400 | 0.0087 | -0.0012 | -0.1544 | 0.7672 | 0.1532 | -261.9635 | -285.2108 | -2.3836 | -2.4800 |
|
| 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": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.010265434613673512,
|
| 5 |
+
"train_runtime": 11699.9785,
|
| 6 |
+
"train_samples": 55758,
|
| 7 |
+
"train_samples_per_second": 4.766,
|
| 8 |
+
"train_steps_per_second": 0.037
|
| 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": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.010265434613673512,
|
| 5 |
+
"train_runtime": 11699.9785,
|
| 6 |
+
"train_samples": 55758,
|
| 7 |
+
"train_samples_per_second": 4.766,
|
| 8 |
+
"train_steps_per_second": 0.037
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,815 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 1.0,
|
| 5 |
+
"eval_steps": 50,
|
| 6 |
+
"global_step": 436,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.022935779816513763,
|
| 13 |
+
"grad_norm": 1.642045632140662,
|
| 14 |
+
"learning_rate": 1.1363636363636363e-07,
|
| 15 |
+
"logits/chosen": -2.6192917823791504,
|
| 16 |
+
"logits/rejected": -2.5524227619171143,
|
| 17 |
+
"logps/chosen": -265.41119384765625,
|
| 18 |
+
"logps/rejected": -236.11862182617188,
|
| 19 |
+
"loss": 0.0154,
|
| 20 |
+
"rewards/accuracies": 0.4000000059604645,
|
| 21 |
+
"rewards/chosen": 0.00034645519917830825,
|
| 22 |
+
"rewards/margins": 0.00023277592845261097,
|
| 23 |
+
"rewards/rejected": 0.00011367930710548535,
|
| 24 |
+
"step": 10
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.045871559633027525,
|
| 28 |
+
"grad_norm": 1.5167637902207143,
|
| 29 |
+
"learning_rate": 2.2727272727272726e-07,
|
| 30 |
+
"logits/chosen": -2.657719135284424,
|
| 31 |
+
"logits/rejected": -2.5759785175323486,
|
| 32 |
+
"logps/chosen": -298.7945556640625,
|
| 33 |
+
"logps/rejected": -274.304443359375,
|
| 34 |
+
"loss": 0.0155,
|
| 35 |
+
"rewards/accuracies": 0.625,
|
| 36 |
+
"rewards/chosen": -9.255408076569438e-05,
|
| 37 |
+
"rewards/margins": 0.0018523468170315027,
|
| 38 |
+
"rewards/rejected": -0.0019449004903435707,
|
| 39 |
+
"step": 20
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.06880733944954129,
|
| 43 |
+
"grad_norm": 1.3745597654790793,
|
| 44 |
+
"learning_rate": 3.4090909090909085e-07,
|
| 45 |
+
"logits/chosen": -2.6762423515319824,
|
| 46 |
+
"logits/rejected": -2.6026246547698975,
|
| 47 |
+
"logps/chosen": -290.37896728515625,
|
| 48 |
+
"logps/rejected": -234.3507080078125,
|
| 49 |
+
"loss": 0.0146,
|
| 50 |
+
"rewards/accuracies": 0.71875,
|
| 51 |
+
"rewards/chosen": 0.009538007900118828,
|
| 52 |
+
"rewards/margins": 0.013226142153143883,
|
| 53 |
+
"rewards/rejected": -0.0036881337873637676,
|
| 54 |
+
"step": 30
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.09174311926605505,
|
| 58 |
+
"grad_norm": 1.272192316985563,
|
| 59 |
+
"learning_rate": 4.545454545454545e-07,
|
| 60 |
+
"logits/chosen": -2.660547971725464,
|
| 61 |
+
"logits/rejected": -2.6108529567718506,
|
| 62 |
+
"logps/chosen": -280.96484375,
|
| 63 |
+
"logps/rejected": -267.6105041503906,
|
| 64 |
+
"loss": 0.013,
|
| 65 |
+
"rewards/accuracies": 0.6875,
|
| 66 |
+
"rewards/chosen": 0.04204345494508743,
|
| 67 |
+
"rewards/margins": 0.04021826013922691,
|
| 68 |
+
"rewards/rejected": 0.0018251972505822778,
|
| 69 |
+
"step": 40
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.11467889908256881,
|
| 73 |
+
"grad_norm": 1.4079236984792325,
|
| 74 |
+
"learning_rate": 4.997110275491701e-07,
|
| 75 |
+
"logits/chosen": -2.6261672973632812,
|
| 76 |
+
"logits/rejected": -2.6208655834198,
|
| 77 |
+
"logps/chosen": -289.76519775390625,
|
| 78 |
+
"logps/rejected": -299.06353759765625,
|
| 79 |
+
"loss": 0.012,
|
| 80 |
+
"rewards/accuracies": 0.699999988079071,
|
| 81 |
+
"rewards/chosen": 0.06486029922962189,
|
| 82 |
+
"rewards/margins": 0.06365373730659485,
|
| 83 |
+
"rewards/rejected": 0.0012065758928656578,
|
| 84 |
+
"step": 50
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.11467889908256881,
|
| 88 |
+
"eval_logits/chosen": -2.5786819458007812,
|
| 89 |
+
"eval_logits/rejected": -2.502084732055664,
|
| 90 |
+
"eval_logps/chosen": -277.3097839355469,
|
| 91 |
+
"eval_logps/rejected": -247.28094482421875,
|
| 92 |
+
"eval_loss": 0.011314952746033669,
|
| 93 |
+
"eval_rewards/accuracies": 0.7025862336158752,
|
| 94 |
+
"eval_rewards/chosen": 0.07780314981937408,
|
| 95 |
+
"eval_rewards/margins": 0.08538833260536194,
|
| 96 |
+
"eval_rewards/rejected": -0.007585177198052406,
|
| 97 |
+
"eval_runtime": 94.9097,
|
| 98 |
+
"eval_samples_per_second": 19.155,
|
| 99 |
+
"eval_steps_per_second": 0.306,
|
| 100 |
+
"step": 50
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.13761467889908258,
|
| 104 |
+
"grad_norm": 1.418614516322105,
|
| 105 |
+
"learning_rate": 4.979475034558115e-07,
|
| 106 |
+
"logits/chosen": -2.5812747478485107,
|
| 107 |
+
"logits/rejected": -2.5265355110168457,
|
| 108 |
+
"logps/chosen": -284.41473388671875,
|
| 109 |
+
"logps/rejected": -259.66094970703125,
|
| 110 |
+
"loss": 0.0112,
|
| 111 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 112 |
+
"rewards/chosen": 0.05069952458143234,
|
| 113 |
+
"rewards/margins": 0.0717436671257019,
|
| 114 |
+
"rewards/rejected": -0.021044140681624413,
|
| 115 |
+
"step": 60
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"epoch": 0.16055045871559634,
|
| 119 |
+
"grad_norm": 1.4648752438020702,
|
| 120 |
+
"learning_rate": 4.945923025551788e-07,
|
| 121 |
+
"logits/chosen": -2.5097055435180664,
|
| 122 |
+
"logits/rejected": -2.4729437828063965,
|
| 123 |
+
"logps/chosen": -319.59161376953125,
|
| 124 |
+
"logps/rejected": -265.26019287109375,
|
| 125 |
+
"loss": 0.0115,
|
| 126 |
+
"rewards/accuracies": 0.737500011920929,
|
| 127 |
+
"rewards/chosen": 0.07979384809732437,
|
| 128 |
+
"rewards/margins": 0.1183195561170578,
|
| 129 |
+
"rewards/rejected": -0.03852573037147522,
|
| 130 |
+
"step": 70
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 0.1834862385321101,
|
| 134 |
+
"grad_norm": 1.379325651626498,
|
| 135 |
+
"learning_rate": 4.896669632591651e-07,
|
| 136 |
+
"logits/chosen": -2.4763333797454834,
|
| 137 |
+
"logits/rejected": -2.385080099105835,
|
| 138 |
+
"logps/chosen": -278.93988037109375,
|
| 139 |
+
"logps/rejected": -255.78781127929688,
|
| 140 |
+
"loss": 0.011,
|
| 141 |
+
"rewards/accuracies": 0.706250011920929,
|
| 142 |
+
"rewards/chosen": 0.06909944862127304,
|
| 143 |
+
"rewards/margins": 0.10041693598031998,
|
| 144 |
+
"rewards/rejected": -0.031317487359046936,
|
| 145 |
+
"step": 80
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"epoch": 0.20642201834862386,
|
| 149 |
+
"grad_norm": 1.3303176283339873,
|
| 150 |
+
"learning_rate": 4.832031033425662e-07,
|
| 151 |
+
"logits/chosen": -2.423882246017456,
|
| 152 |
+
"logits/rejected": -2.376399278640747,
|
| 153 |
+
"logps/chosen": -280.73907470703125,
|
| 154 |
+
"logps/rejected": -252.6663360595703,
|
| 155 |
+
"loss": 0.0103,
|
| 156 |
+
"rewards/accuracies": 0.71875,
|
| 157 |
+
"rewards/chosen": 0.06854326277971268,
|
| 158 |
+
"rewards/margins": 0.12719407677650452,
|
| 159 |
+
"rewards/rejected": -0.058650821447372437,
|
| 160 |
+
"step": 90
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 0.22935779816513763,
|
| 164 |
+
"grad_norm": 1.4785994754129348,
|
| 165 |
+
"learning_rate": 4.752422169756047e-07,
|
| 166 |
+
"logits/chosen": -2.3845717906951904,
|
| 167 |
+
"logits/rejected": -2.3276596069335938,
|
| 168 |
+
"logps/chosen": -268.87396240234375,
|
| 169 |
+
"logps/rejected": -280.0634460449219,
|
| 170 |
+
"loss": 0.011,
|
| 171 |
+
"rewards/accuracies": 0.71875,
|
| 172 |
+
"rewards/chosen": 0.035507336258888245,
|
| 173 |
+
"rewards/margins": 0.11685723066329956,
|
| 174 |
+
"rewards/rejected": -0.08134988695383072,
|
| 175 |
+
"step": 100
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"epoch": 0.22935779816513763,
|
| 179 |
+
"eval_logits/chosen": -2.4392149448394775,
|
| 180 |
+
"eval_logits/rejected": -2.3370442390441895,
|
| 181 |
+
"eval_logps/chosen": -280.90240478515625,
|
| 182 |
+
"eval_logps/rejected": -254.33267211914062,
|
| 183 |
+
"eval_loss": 0.009970244951546192,
|
| 184 |
+
"eval_rewards/accuracies": 0.7068965435028076,
|
| 185 |
+
"eval_rewards/chosen": 0.04187687486410141,
|
| 186 |
+
"eval_rewards/margins": 0.1199791207909584,
|
| 187 |
+
"eval_rewards/rejected": -0.078102245926857,
|
| 188 |
+
"eval_runtime": 95.3397,
|
| 189 |
+
"eval_samples_per_second": 19.069,
|
| 190 |
+
"eval_steps_per_second": 0.304,
|
| 191 |
+
"step": 100
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.25229357798165136,
|
| 195 |
+
"grad_norm": 2.0224189698687702,
|
| 196 |
+
"learning_rate": 4.658354083558188e-07,
|
| 197 |
+
"logits/chosen": -2.4632654190063477,
|
| 198 |
+
"logits/rejected": -2.354820966720581,
|
| 199 |
+
"logps/chosen": -263.564453125,
|
| 200 |
+
"logps/rejected": -241.26791381835938,
|
| 201 |
+
"loss": 0.0113,
|
| 202 |
+
"rewards/accuracies": 0.675000011920929,
|
| 203 |
+
"rewards/chosen": 0.01623677834868431,
|
| 204 |
+
"rewards/margins": 0.09158362448215485,
|
| 205 |
+
"rewards/rejected": -0.07534684240818024,
|
| 206 |
+
"step": 110
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.27522935779816515,
|
| 210 |
+
"grad_norm": 1.4135396422800603,
|
| 211 |
+
"learning_rate": 4.550430636492389e-07,
|
| 212 |
+
"logits/chosen": -2.382753610610962,
|
| 213 |
+
"logits/rejected": -2.350785970687866,
|
| 214 |
+
"logps/chosen": -275.2984619140625,
|
| 215 |
+
"logps/rejected": -260.0333557128906,
|
| 216 |
+
"loss": 0.0102,
|
| 217 |
+
"rewards/accuracies": 0.6875,
|
| 218 |
+
"rewards/chosen": 0.005208671558648348,
|
| 219 |
+
"rewards/margins": 0.0909029170870781,
|
| 220 |
+
"rewards/rejected": -0.08569424599409103,
|
| 221 |
+
"step": 120
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.2981651376146789,
|
| 225 |
+
"grad_norm": 2.470336886918338,
|
| 226 |
+
"learning_rate": 4.429344633468004e-07,
|
| 227 |
+
"logits/chosen": -2.404059648513794,
|
| 228 |
+
"logits/rejected": -2.3569350242614746,
|
| 229 |
+
"logps/chosen": -254.55068969726562,
|
| 230 |
+
"logps/rejected": -252.3956756591797,
|
| 231 |
+
"loss": 0.0105,
|
| 232 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 233 |
+
"rewards/chosen": 0.016294140368700027,
|
| 234 |
+
"rewards/margins": 0.11412493139505386,
|
| 235 |
+
"rewards/rejected": -0.09783079475164413,
|
| 236 |
+
"step": 130
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 0.3211009174311927,
|
| 240 |
+
"grad_norm": 3.815944054063513,
|
| 241 |
+
"learning_rate": 4.2958733752443187e-07,
|
| 242 |
+
"logits/chosen": -2.4113519191741943,
|
| 243 |
+
"logits/rejected": -2.340986728668213,
|
| 244 |
+
"logps/chosen": -272.99420166015625,
|
| 245 |
+
"logps/rejected": -231.37026977539062,
|
| 246 |
+
"loss": 0.0123,
|
| 247 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 248 |
+
"rewards/chosen": 0.0319095216691494,
|
| 249 |
+
"rewards/margins": 0.10264714062213898,
|
| 250 |
+
"rewards/rejected": -0.07073761522769928,
|
| 251 |
+
"step": 140
|
| 252 |
+
},
|
| 253 |
+
{
|
| 254 |
+
"epoch": 0.3440366972477064,
|
| 255 |
+
"grad_norm": 1.3170543399714232,
|
| 256 |
+
"learning_rate": 4.150873668617898e-07,
|
| 257 |
+
"logits/chosen": -2.454432249069214,
|
| 258 |
+
"logits/rejected": -2.370666265487671,
|
| 259 |
+
"logps/chosen": -268.79541015625,
|
| 260 |
+
"logps/rejected": -247.0717315673828,
|
| 261 |
+
"loss": 0.0104,
|
| 262 |
+
"rewards/accuracies": 0.675000011920929,
|
| 263 |
+
"rewards/chosen": -0.022518452256917953,
|
| 264 |
+
"rewards/margins": 0.11632315069437027,
|
| 265 |
+
"rewards/rejected": -0.13884159922599792,
|
| 266 |
+
"step": 150
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 0.3440366972477064,
|
| 270 |
+
"eval_logits/chosen": -2.494401454925537,
|
| 271 |
+
"eval_logits/rejected": -2.39561128616333,
|
| 272 |
+
"eval_logps/chosen": -285.85308837890625,
|
| 273 |
+
"eval_logps/rejected": -260.5493469238281,
|
| 274 |
+
"eval_loss": 0.009796149097383022,
|
| 275 |
+
"eval_rewards/accuracies": 0.7198275923728943,
|
| 276 |
+
"eval_rewards/chosen": -0.007629875559359789,
|
| 277 |
+
"eval_rewards/margins": 0.13263897597789764,
|
| 278 |
+
"eval_rewards/rejected": -0.14026884734630585,
|
| 279 |
+
"eval_runtime": 95.8939,
|
| 280 |
+
"eval_samples_per_second": 18.958,
|
| 281 |
+
"eval_steps_per_second": 0.302,
|
| 282 |
+
"step": 150
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 0.3669724770642202,
|
| 286 |
+
"grad_norm": 2.6585100320508737,
|
| 287 |
+
"learning_rate": 3.9952763262280397e-07,
|
| 288 |
+
"logits/chosen": -2.4578630924224854,
|
| 289 |
+
"logits/rejected": -2.3927180767059326,
|
| 290 |
+
"logps/chosen": -293.46124267578125,
|
| 291 |
+
"logps/rejected": -295.8601379394531,
|
| 292 |
+
"loss": 0.01,
|
| 293 |
+
"rewards/accuracies": 0.737500011920929,
|
| 294 |
+
"rewards/chosen": 0.030898962169885635,
|
| 295 |
+
"rewards/margins": 0.12981417775154114,
|
| 296 |
+
"rewards/rejected": -0.0989152044057846,
|
| 297 |
+
"step": 160
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"epoch": 0.38990825688073394,
|
| 301 |
+
"grad_norm": 1.3118055868117948,
|
| 302 |
+
"learning_rate": 3.8300801912883414e-07,
|
| 303 |
+
"logits/chosen": -2.504678249359131,
|
| 304 |
+
"logits/rejected": -2.4544835090637207,
|
| 305 |
+
"logps/chosen": -267.3302307128906,
|
| 306 |
+
"logps/rejected": -279.33062744140625,
|
| 307 |
+
"loss": 0.0091,
|
| 308 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 309 |
+
"rewards/chosen": 0.04270657151937485,
|
| 310 |
+
"rewards/margins": 0.13532397150993347,
|
| 311 |
+
"rewards/rejected": -0.09261739999055862,
|
| 312 |
+
"step": 170
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 0.41284403669724773,
|
| 316 |
+
"grad_norm": 1.4248945431420277,
|
| 317 |
+
"learning_rate": 3.6563457256020884e-07,
|
| 318 |
+
"logits/chosen": -2.509859561920166,
|
| 319 |
+
"logits/rejected": -2.4090194702148438,
|
| 320 |
+
"logps/chosen": -306.2326965332031,
|
| 321 |
+
"logps/rejected": -253.7351531982422,
|
| 322 |
+
"loss": 0.0096,
|
| 323 |
+
"rewards/accuracies": 0.75,
|
| 324 |
+
"rewards/chosen": -0.028747806325554848,
|
| 325 |
+
"rewards/margins": 0.123654805123806,
|
| 326 |
+
"rewards/rejected": -0.1524026244878769,
|
| 327 |
+
"step": 180
|
| 328 |
+
},
|
| 329 |
+
{
|
| 330 |
+
"epoch": 0.43577981651376146,
|
| 331 |
+
"grad_norm": 4.61575651504461,
|
| 332 |
+
"learning_rate": 3.475188202022617e-07,
|
| 333 |
+
"logits/chosen": -2.4155473709106445,
|
| 334 |
+
"logits/rejected": -2.3843960762023926,
|
| 335 |
+
"logps/chosen": -256.64117431640625,
|
| 336 |
+
"logps/rejected": -276.67425537109375,
|
| 337 |
+
"loss": 0.0105,
|
| 338 |
+
"rewards/accuracies": 0.78125,
|
| 339 |
+
"rewards/chosen": 0.014733311720192432,
|
| 340 |
+
"rewards/margins": 0.13803140819072723,
|
| 341 |
+
"rewards/rejected": -0.12329809367656708,
|
| 342 |
+
"step": 190
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 0.45871559633027525,
|
| 346 |
+
"grad_norm": 1.2837860780603194,
|
| 347 |
+
"learning_rate": 3.287770545059052e-07,
|
| 348 |
+
"logits/chosen": -2.529806613922119,
|
| 349 |
+
"logits/rejected": -2.4307055473327637,
|
| 350 |
+
"logps/chosen": -277.78741455078125,
|
| 351 |
+
"logps/rejected": -256.05511474609375,
|
| 352 |
+
"loss": 0.0096,
|
| 353 |
+
"rewards/accuracies": 0.737500011920929,
|
| 354 |
+
"rewards/chosen": 0.004942134954035282,
|
| 355 |
+
"rewards/margins": 0.12693332135677338,
|
| 356 |
+
"rewards/rejected": -0.12199117988348007,
|
| 357 |
+
"step": 200
|
| 358 |
+
},
|
| 359 |
+
{
|
| 360 |
+
"epoch": 0.45871559633027525,
|
| 361 |
+
"eval_logits/chosen": -2.514023542404175,
|
| 362 |
+
"eval_logits/rejected": -2.4208788871765137,
|
| 363 |
+
"eval_logps/chosen": -282.1954345703125,
|
| 364 |
+
"eval_logps/rejected": -257.61944580078125,
|
| 365 |
+
"eval_loss": 0.009310290217399597,
|
| 366 |
+
"eval_rewards/accuracies": 0.7931034564971924,
|
| 367 |
+
"eval_rewards/chosen": 0.028946416452527046,
|
| 368 |
+
"eval_rewards/margins": 0.13991650938987732,
|
| 369 |
+
"eval_rewards/rejected": -0.11097008734941483,
|
| 370 |
+
"eval_runtime": 96.1764,
|
| 371 |
+
"eval_samples_per_second": 18.903,
|
| 372 |
+
"eval_steps_per_second": 0.302,
|
| 373 |
+
"step": 200
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"epoch": 0.481651376146789,
|
| 377 |
+
"grad_norm": 1.090616369349917,
|
| 378 |
+
"learning_rate": 3.0952958655864954e-07,
|
| 379 |
+
"logits/chosen": -2.4676451683044434,
|
| 380 |
+
"logits/rejected": -2.4505865573883057,
|
| 381 |
+
"logps/chosen": -270.9748840332031,
|
| 382 |
+
"logps/rejected": -265.1389465332031,
|
| 383 |
+
"loss": 0.0094,
|
| 384 |
+
"rewards/accuracies": 0.762499988079071,
|
| 385 |
+
"rewards/chosen": 0.023295128718018532,
|
| 386 |
+
"rewards/margins": 0.13126251101493835,
|
| 387 |
+
"rewards/rejected": -0.10796739161014557,
|
| 388 |
+
"step": 210
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 0.5045871559633027,
|
| 392 |
+
"grad_norm": 1.493461344772632,
|
| 393 |
+
"learning_rate": 2.898999737583448e-07,
|
| 394 |
+
"logits/chosen": -2.504462957382202,
|
| 395 |
+
"logits/rejected": -2.404648542404175,
|
| 396 |
+
"logps/chosen": -322.62689208984375,
|
| 397 |
+
"logps/rejected": -300.34686279296875,
|
| 398 |
+
"loss": 0.0091,
|
| 399 |
+
"rewards/accuracies": 0.78125,
|
| 400 |
+
"rewards/chosen": 0.00053420226322487,
|
| 401 |
+
"rewards/margins": 0.15430037677288055,
|
| 402 |
+
"rewards/rejected": -0.1537661850452423,
|
| 403 |
+
"step": 220
|
| 404 |
+
},
|
| 405 |
+
{
|
| 406 |
+
"epoch": 0.5275229357798165,
|
| 407 |
+
"grad_norm": 1.2922195153842637,
|
| 408 |
+
"learning_rate": 2.7001422664752333e-07,
|
| 409 |
+
"logits/chosen": -2.3982276916503906,
|
| 410 |
+
"logits/rejected": -2.3454272747039795,
|
| 411 |
+
"logps/chosen": -269.4321594238281,
|
| 412 |
+
"logps/rejected": -279.17926025390625,
|
| 413 |
+
"loss": 0.0093,
|
| 414 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 415 |
+
"rewards/chosen": -0.014886477962136269,
|
| 416 |
+
"rewards/margins": 0.12070702016353607,
|
| 417 |
+
"rewards/rejected": -0.1355935037136078,
|
| 418 |
+
"step": 230
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"epoch": 0.5504587155963303,
|
| 422 |
+
"grad_norm": 1.3338425925379636,
|
| 423 |
+
"learning_rate": 2.5e-07,
|
| 424 |
+
"logits/chosen": -2.4747557640075684,
|
| 425 |
+
"logits/rejected": -2.3822951316833496,
|
| 426 |
+
"logps/chosen": -288.63037109375,
|
| 427 |
+
"logps/rejected": -274.5633850097656,
|
| 428 |
+
"loss": 0.0095,
|
| 429 |
+
"rewards/accuracies": 0.762499988079071,
|
| 430 |
+
"rewards/chosen": -0.010984973981976509,
|
| 431 |
+
"rewards/margins": 0.12114904075860977,
|
| 432 |
+
"rewards/rejected": -0.13213400542736053,
|
| 433 |
+
"step": 240
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 0.573394495412844,
|
| 437 |
+
"grad_norm": 1.161806057366623,
|
| 438 |
+
"learning_rate": 2.2998577335247667e-07,
|
| 439 |
+
"logits/chosen": -2.513209819793701,
|
| 440 |
+
"logits/rejected": -2.424811363220215,
|
| 441 |
+
"logps/chosen": -303.5930480957031,
|
| 442 |
+
"logps/rejected": -273.9664611816406,
|
| 443 |
+
"loss": 0.0094,
|
| 444 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 445 |
+
"rewards/chosen": 0.002007619244977832,
|
| 446 |
+
"rewards/margins": 0.13723386824131012,
|
| 447 |
+
"rewards/rejected": -0.1352262645959854,
|
| 448 |
+
"step": 250
|
| 449 |
+
},
|
| 450 |
+
{
|
| 451 |
+
"epoch": 0.573394495412844,
|
| 452 |
+
"eval_logits/chosen": -2.4590651988983154,
|
| 453 |
+
"eval_logits/rejected": -2.362999200820923,
|
| 454 |
+
"eval_logps/chosen": -286.21722412109375,
|
| 455 |
+
"eval_logps/rejected": -263.2693786621094,
|
| 456 |
+
"eval_loss": 0.008912510238587856,
|
| 457 |
+
"eval_rewards/accuracies": 0.7801724076271057,
|
| 458 |
+
"eval_rewards/chosen": -0.011271164752542973,
|
| 459 |
+
"eval_rewards/margins": 0.15619821846485138,
|
| 460 |
+
"eval_rewards/rejected": -0.16746938228607178,
|
| 461 |
+
"eval_runtime": 94.78,
|
| 462 |
+
"eval_samples_per_second": 19.181,
|
| 463 |
+
"eval_steps_per_second": 0.306,
|
| 464 |
+
"step": 250
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 0.5963302752293578,
|
| 468 |
+
"grad_norm": 1.1995883831745127,
|
| 469 |
+
"learning_rate": 2.1010002624165524e-07,
|
| 470 |
+
"logits/chosen": -2.4454376697540283,
|
| 471 |
+
"logits/rejected": -2.407090902328491,
|
| 472 |
+
"logps/chosen": -273.8803405761719,
|
| 473 |
+
"logps/rejected": -292.88421630859375,
|
| 474 |
+
"loss": 0.0097,
|
| 475 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 476 |
+
"rewards/chosen": -0.03165815398097038,
|
| 477 |
+
"rewards/margins": 0.15324734151363373,
|
| 478 |
+
"rewards/rejected": -0.1849054992198944,
|
| 479 |
+
"step": 260
|
| 480 |
+
},
|
| 481 |
+
{
|
| 482 |
+
"epoch": 0.6192660550458715,
|
| 483 |
+
"grad_norm": 1.236047360327654,
|
| 484 |
+
"learning_rate": 1.9047041344135043e-07,
|
| 485 |
+
"logits/chosen": -2.4361116886138916,
|
| 486 |
+
"logits/rejected": -2.4080004692077637,
|
| 487 |
+
"logps/chosen": -270.04998779296875,
|
| 488 |
+
"logps/rejected": -270.3052978515625,
|
| 489 |
+
"loss": 0.0091,
|
| 490 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 491 |
+
"rewards/chosen": -0.03987707942724228,
|
| 492 |
+
"rewards/margins": 0.14470525085926056,
|
| 493 |
+
"rewards/rejected": -0.18458232283592224,
|
| 494 |
+
"step": 270
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 0.6422018348623854,
|
| 498 |
+
"grad_norm": 1.3000373431490932,
|
| 499 |
+
"learning_rate": 1.7122294549409482e-07,
|
| 500 |
+
"logits/chosen": -2.512075424194336,
|
| 501 |
+
"logits/rejected": -2.448713779449463,
|
| 502 |
+
"logps/chosen": -273.0007019042969,
|
| 503 |
+
"logps/rejected": -280.35565185546875,
|
| 504 |
+
"loss": 0.0098,
|
| 505 |
+
"rewards/accuracies": 0.800000011920929,
|
| 506 |
+
"rewards/chosen": -1.8634367734193802e-05,
|
| 507 |
+
"rewards/margins": 0.16557954251766205,
|
| 508 |
+
"rewards/rejected": -0.16559818387031555,
|
| 509 |
+
"step": 280
|
| 510 |
+
},
|
| 511 |
+
{
|
| 512 |
+
"epoch": 0.6651376146788991,
|
| 513 |
+
"grad_norm": 1.354856941159808,
|
| 514 |
+
"learning_rate": 1.524811797977383e-07,
|
| 515 |
+
"logits/chosen": -2.493873119354248,
|
| 516 |
+
"logits/rejected": -2.416597843170166,
|
| 517 |
+
"logps/chosen": -287.20404052734375,
|
| 518 |
+
"logps/rejected": -267.6150817871094,
|
| 519 |
+
"loss": 0.009,
|
| 520 |
+
"rewards/accuracies": 0.768750011920929,
|
| 521 |
+
"rewards/chosen": 0.005267986096441746,
|
| 522 |
+
"rewards/margins": 0.1422511339187622,
|
| 523 |
+
"rewards/rejected": -0.13698314130306244,
|
| 524 |
+
"step": 290
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"epoch": 0.6880733944954128,
|
| 528 |
+
"grad_norm": 1.3081620144699164,
|
| 529 |
+
"learning_rate": 1.3436542743979125e-07,
|
| 530 |
+
"logits/chosen": -2.4964632987976074,
|
| 531 |
+
"logits/rejected": -2.472830057144165,
|
| 532 |
+
"logps/chosen": -307.45733642578125,
|
| 533 |
+
"logps/rejected": -271.1383361816406,
|
| 534 |
+
"loss": 0.0096,
|
| 535 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 536 |
+
"rewards/chosen": -0.015080097131431103,
|
| 537 |
+
"rewards/margins": 0.10770467668771744,
|
| 538 |
+
"rewards/rejected": -0.12278477847576141,
|
| 539 |
+
"step": 300
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 0.6880733944954128,
|
| 543 |
+
"eval_logits/chosen": -2.477182388305664,
|
| 544 |
+
"eval_logits/rejected": -2.382119655609131,
|
| 545 |
+
"eval_logps/chosen": -286.4156188964844,
|
| 546 |
+
"eval_logps/rejected": -262.90252685546875,
|
| 547 |
+
"eval_loss": 0.008760624565184116,
|
| 548 |
+
"eval_rewards/accuracies": 0.7844827771186829,
|
| 549 |
+
"eval_rewards/chosen": -0.013255205936729908,
|
| 550 |
+
"eval_rewards/margins": 0.15054550766944885,
|
| 551 |
+
"eval_rewards/rejected": -0.16380071640014648,
|
| 552 |
+
"eval_runtime": 95.1377,
|
| 553 |
+
"eval_samples_per_second": 19.109,
|
| 554 |
+
"eval_steps_per_second": 0.305,
|
| 555 |
+
"step": 300
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 0.7110091743119266,
|
| 559 |
+
"grad_norm": 1.3153487126066306,
|
| 560 |
+
"learning_rate": 1.1699198087116588e-07,
|
| 561 |
+
"logits/chosen": -2.5111746788024902,
|
| 562 |
+
"logits/rejected": -2.422899007797241,
|
| 563 |
+
"logps/chosen": -280.66143798828125,
|
| 564 |
+
"logps/rejected": -280.439697265625,
|
| 565 |
+
"loss": 0.0094,
|
| 566 |
+
"rewards/accuracies": 0.699999988079071,
|
| 567 |
+
"rewards/chosen": -0.026962900534272194,
|
| 568 |
+
"rewards/margins": 0.11293704807758331,
|
| 569 |
+
"rewards/rejected": -0.13989993929862976,
|
| 570 |
+
"step": 310
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"epoch": 0.7339449541284404,
|
| 574 |
+
"grad_norm": 1.337124035482138,
|
| 575 |
+
"learning_rate": 1.00472367377196e-07,
|
| 576 |
+
"logits/chosen": -2.4476630687713623,
|
| 577 |
+
"logits/rejected": -2.397982120513916,
|
| 578 |
+
"logps/chosen": -275.42584228515625,
|
| 579 |
+
"logps/rejected": -251.29122924804688,
|
| 580 |
+
"loss": 0.0093,
|
| 581 |
+
"rewards/accuracies": 0.75,
|
| 582 |
+
"rewards/chosen": -0.011938780546188354,
|
| 583 |
+
"rewards/margins": 0.15607169270515442,
|
| 584 |
+
"rewards/rejected": -0.16801045835018158,
|
| 585 |
+
"step": 320
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 0.7568807339449541,
|
| 589 |
+
"grad_norm": 1.4113093599011106,
|
| 590 |
+
"learning_rate": 8.49126331382102e-08,
|
| 591 |
+
"logits/chosen": -2.435225009918213,
|
| 592 |
+
"logits/rejected": -2.386702537536621,
|
| 593 |
+
"logps/chosen": -275.8332214355469,
|
| 594 |
+
"logps/rejected": -260.52587890625,
|
| 595 |
+
"loss": 0.0095,
|
| 596 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 597 |
+
"rewards/chosen": -0.013110649771988392,
|
| 598 |
+
"rewards/margins": 0.12114731222391129,
|
| 599 |
+
"rewards/rejected": -0.1342579573392868,
|
| 600 |
+
"step": 330
|
| 601 |
+
},
|
| 602 |
+
{
|
| 603 |
+
"epoch": 0.7798165137614679,
|
| 604 |
+
"grad_norm": 1.366187901878708,
|
| 605 |
+
"learning_rate": 7.041266247556812e-08,
|
| 606 |
+
"logits/chosen": -2.521238327026367,
|
| 607 |
+
"logits/rejected": -2.4764904975891113,
|
| 608 |
+
"logps/chosen": -289.8477478027344,
|
| 609 |
+
"logps/rejected": -272.86456298828125,
|
| 610 |
+
"loss": 0.009,
|
| 611 |
+
"rewards/accuracies": 0.731249988079071,
|
| 612 |
+
"rewards/chosen": -0.01855117455124855,
|
| 613 |
+
"rewards/margins": 0.11025450378656387,
|
| 614 |
+
"rewards/rejected": -0.12880566716194153,
|
| 615 |
+
"step": 340
|
| 616 |
+
},
|
| 617 |
+
{
|
| 618 |
+
"epoch": 0.8027522935779816,
|
| 619 |
+
"grad_norm": 1.3145992203531836,
|
| 620 |
+
"learning_rate": 5.706553665319955e-08,
|
| 621 |
+
"logits/chosen": -2.4982964992523193,
|
| 622 |
+
"logits/rejected": -2.3964176177978516,
|
| 623 |
+
"logps/chosen": -277.12042236328125,
|
| 624 |
+
"logps/rejected": -253.37203979492188,
|
| 625 |
+
"loss": 0.0096,
|
| 626 |
+
"rewards/accuracies": 0.737500011920929,
|
| 627 |
+
"rewards/chosen": -0.026293223723769188,
|
| 628 |
+
"rewards/margins": 0.13081298768520355,
|
| 629 |
+
"rewards/rejected": -0.1571062207221985,
|
| 630 |
+
"step": 350
|
| 631 |
+
},
|
| 632 |
+
{
|
| 633 |
+
"epoch": 0.8027522935779816,
|
| 634 |
+
"eval_logits/chosen": -2.481409788131714,
|
| 635 |
+
"eval_logits/rejected": -2.3850882053375244,
|
| 636 |
+
"eval_logps/chosen": -285.64593505859375,
|
| 637 |
+
"eval_logps/rejected": -262.6905517578125,
|
| 638 |
+
"eval_loss": 0.008698553778231144,
|
| 639 |
+
"eval_rewards/accuracies": 0.7801724076271057,
|
| 640 |
+
"eval_rewards/chosen": -0.005558254197239876,
|
| 641 |
+
"eval_rewards/margins": 0.15612287819385529,
|
| 642 |
+
"eval_rewards/rejected": -0.1616811454296112,
|
| 643 |
+
"eval_runtime": 94.1358,
|
| 644 |
+
"eval_samples_per_second": 19.313,
|
| 645 |
+
"eval_steps_per_second": 0.308,
|
| 646 |
+
"step": 350
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"epoch": 0.8256880733944955,
|
| 650 |
+
"grad_norm": 1.1160006320842197,
|
| 651 |
+
"learning_rate": 4.4956936350761005e-08,
|
| 652 |
+
"logits/chosen": -2.4785516262054443,
|
| 653 |
+
"logits/rejected": -2.427774429321289,
|
| 654 |
+
"logps/chosen": -252.49581909179688,
|
| 655 |
+
"logps/rejected": -271.5592041015625,
|
| 656 |
+
"loss": 0.0088,
|
| 657 |
+
"rewards/accuracies": 0.731249988079071,
|
| 658 |
+
"rewards/chosen": -0.0017578303813934326,
|
| 659 |
+
"rewards/margins": 0.12846335768699646,
|
| 660 |
+
"rewards/rejected": -0.1302211880683899,
|
| 661 |
+
"step": 360
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 0.8486238532110092,
|
| 665 |
+
"grad_norm": 1.2680559242352918,
|
| 666 |
+
"learning_rate": 3.416459164418123e-08,
|
| 667 |
+
"logits/chosen": -2.514042854309082,
|
| 668 |
+
"logits/rejected": -2.4509449005126953,
|
| 669 |
+
"logps/chosen": -300.03057861328125,
|
| 670 |
+
"logps/rejected": -276.8949279785156,
|
| 671 |
+
"loss": 0.0087,
|
| 672 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 673 |
+
"rewards/chosen": -0.004460807889699936,
|
| 674 |
+
"rewards/margins": 0.13874222338199615,
|
| 675 |
+
"rewards/rejected": -0.1432030349969864,
|
| 676 |
+
"step": 370
|
| 677 |
+
},
|
| 678 |
+
{
|
| 679 |
+
"epoch": 0.8715596330275229,
|
| 680 |
+
"grad_norm": 1.5408590452374413,
|
| 681 |
+
"learning_rate": 2.475778302439524e-08,
|
| 682 |
+
"logits/chosen": -2.4833552837371826,
|
| 683 |
+
"logits/rejected": -2.4306118488311768,
|
| 684 |
+
"logps/chosen": -296.1810607910156,
|
| 685 |
+
"logps/rejected": -272.9661560058594,
|
| 686 |
+
"loss": 0.009,
|
| 687 |
+
"rewards/accuracies": 0.78125,
|
| 688 |
+
"rewards/chosen": 0.001969636185094714,
|
| 689 |
+
"rewards/margins": 0.15506890416145325,
|
| 690 |
+
"rewards/rejected": -0.15309928357601166,
|
| 691 |
+
"step": 380
|
| 692 |
+
},
|
| 693 |
+
{
|
| 694 |
+
"epoch": 0.8944954128440367,
|
| 695 |
+
"grad_norm": 1.4629042335763474,
|
| 696 |
+
"learning_rate": 1.6796896657433805e-08,
|
| 697 |
+
"logits/chosen": -2.4795479774475098,
|
| 698 |
+
"logits/rejected": -2.3772130012512207,
|
| 699 |
+
"logps/chosen": -253.0847930908203,
|
| 700 |
+
"logps/rejected": -242.50405883789062,
|
| 701 |
+
"loss": 0.0095,
|
| 702 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 703 |
+
"rewards/chosen": -0.011273050680756569,
|
| 704 |
+
"rewards/margins": 0.12070544064044952,
|
| 705 |
+
"rewards/rejected": -0.13197848200798035,
|
| 706 |
+
"step": 390
|
| 707 |
+
},
|
| 708 |
+
{
|
| 709 |
+
"epoch": 0.9174311926605505,
|
| 710 |
+
"grad_norm": 1.3631415527821928,
|
| 711 |
+
"learning_rate": 1.0333036740834855e-08,
|
| 712 |
+
"logits/chosen": -2.4071569442749023,
|
| 713 |
+
"logits/rejected": -2.371851682662964,
|
| 714 |
+
"logps/chosen": -224.0278778076172,
|
| 715 |
+
"logps/rejected": -241.8699188232422,
|
| 716 |
+
"loss": 0.0093,
|
| 717 |
+
"rewards/accuracies": 0.731249988079071,
|
| 718 |
+
"rewards/chosen": 0.0053174905478954315,
|
| 719 |
+
"rewards/margins": 0.1318611204624176,
|
| 720 |
+
"rewards/rejected": -0.12654362618923187,
|
| 721 |
+
"step": 400
|
| 722 |
+
},
|
| 723 |
+
{
|
| 724 |
+
"epoch": 0.9174311926605505,
|
| 725 |
+
"eval_logits/chosen": -2.4800400733947754,
|
| 726 |
+
"eval_logits/rejected": -2.383610486984253,
|
| 727 |
+
"eval_logps/chosen": -285.21075439453125,
|
| 728 |
+
"eval_logps/rejected": -261.9634704589844,
|
| 729 |
+
"eval_loss": 0.008683313615620136,
|
| 730 |
+
"eval_rewards/accuracies": 0.767241358757019,
|
| 731 |
+
"eval_rewards/chosen": -0.0012068306095898151,
|
| 732 |
+
"eval_rewards/margins": 0.15320327877998352,
|
| 733 |
+
"eval_rewards/rejected": -0.15441007912158966,
|
| 734 |
+
"eval_runtime": 95.2651,
|
| 735 |
+
"eval_samples_per_second": 19.084,
|
| 736 |
+
"eval_steps_per_second": 0.304,
|
| 737 |
+
"step": 400
|
| 738 |
+
},
|
| 739 |
+
{
|
| 740 |
+
"epoch": 0.9403669724770642,
|
| 741 |
+
"grad_norm": 1.2959245163133988,
|
| 742 |
+
"learning_rate": 5.4076974448211685e-09,
|
| 743 |
+
"logits/chosen": -2.414137363433838,
|
| 744 |
+
"logits/rejected": -2.364811420440674,
|
| 745 |
+
"logps/chosen": -271.1864318847656,
|
| 746 |
+
"logps/rejected": -253.48568725585938,
|
| 747 |
+
"loss": 0.009,
|
| 748 |
+
"rewards/accuracies": 0.768750011920929,
|
| 749 |
+
"rewards/chosen": -0.0023452930618077517,
|
| 750 |
+
"rewards/margins": 0.1520446538925171,
|
| 751 |
+
"rewards/rejected": -0.15438991785049438,
|
| 752 |
+
"step": 410
|
| 753 |
+
},
|
| 754 |
+
{
|
| 755 |
+
"epoch": 0.963302752293578,
|
| 756 |
+
"grad_norm": 2.8287990040902504,
|
| 757 |
+
"learning_rate": 2.052496544188487e-09,
|
| 758 |
+
"logits/chosen": -2.4582016468048096,
|
| 759 |
+
"logits/rejected": -2.3712384700775146,
|
| 760 |
+
"logps/chosen": -260.9129638671875,
|
| 761 |
+
"logps/rejected": -261.7887878417969,
|
| 762 |
+
"loss": 0.0093,
|
| 763 |
+
"rewards/accuracies": 0.75,
|
| 764 |
+
"rewards/chosen": -0.014739753678441048,
|
| 765 |
+
"rewards/margins": 0.14437474310398102,
|
| 766 |
+
"rewards/rejected": -0.1591145098209381,
|
| 767 |
+
"step": 420
|
| 768 |
+
},
|
| 769 |
+
{
|
| 770 |
+
"epoch": 0.9862385321100917,
|
| 771 |
+
"grad_norm": 1.3866473112375508,
|
| 772 |
+
"learning_rate": 2.889724508297886e-10,
|
| 773 |
+
"logits/chosen": -2.474024772644043,
|
| 774 |
+
"logits/rejected": -2.357807159423828,
|
| 775 |
+
"logps/chosen": -303.5400695800781,
|
| 776 |
+
"logps/rejected": -256.8912658691406,
|
| 777 |
+
"loss": 0.0094,
|
| 778 |
+
"rewards/accuracies": 0.768750011920929,
|
| 779 |
+
"rewards/chosen": -0.003646157681941986,
|
| 780 |
+
"rewards/margins": 0.12490139901638031,
|
| 781 |
+
"rewards/rejected": -0.1285475790500641,
|
| 782 |
+
"step": 430
|
| 783 |
+
},
|
| 784 |
+
{
|
| 785 |
+
"epoch": 1.0,
|
| 786 |
+
"step": 436,
|
| 787 |
+
"total_flos": 0.0,
|
| 788 |
+
"train_loss": 0.010265434613673512,
|
| 789 |
+
"train_runtime": 11699.9785,
|
| 790 |
+
"train_samples_per_second": 4.766,
|
| 791 |
+
"train_steps_per_second": 0.037
|
| 792 |
+
}
|
| 793 |
+
],
|
| 794 |
+
"logging_steps": 10,
|
| 795 |
+
"max_steps": 436,
|
| 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 |
+
}
|