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-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-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: 0.4890
|
| 21 |
+
- Rewards/chosen: -1.5856
|
| 22 |
+
- Rewards/rejected: -2.7832
|
| 23 |
+
- Rewards/accuracies: 0.7371
|
| 24 |
+
- Rewards/margins: 1.1976
|
| 25 |
+
- Logps/rejected: -526.5942
|
| 26 |
+
- Logps/chosen: -418.5211
|
| 27 |
+
- Logits/rejected: 3.6380
|
| 28 |
+
- Logits/chosen: 2.5080
|
| 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 |
+
| 0.6615 | 0.1143 | 50 | 0.6475 | -0.0448 | -0.1732 | 0.6595 | 0.1284 | -265.5944 | -264.4379 | -2.5754 | -2.6221 |
|
| 66 |
+
| 0.5507 | 0.2286 | 100 | 0.5504 | -0.7596 | -1.5622 | 0.6897 | 0.8026 | -404.4943 | -335.9164 | 0.2567 | -0.1843 |
|
| 67 |
+
| 0.521 | 0.3429 | 150 | 0.5234 | -0.9406 | -1.8921 | 0.7198 | 0.9516 | -437.4890 | -354.0176 | 1.8188 | 1.1097 |
|
| 68 |
+
| 0.5162 | 0.4571 | 200 | 0.5108 | -1.2309 | -2.2537 | 0.7328 | 1.0228 | -473.6453 | -383.0527 | 2.2567 | 1.2299 |
|
| 69 |
+
| 0.5186 | 0.5714 | 250 | 0.5027 | -1.3697 | -2.4832 | 0.7026 | 1.1135 | -496.5983 | -396.9323 | 2.8322 | 1.5762 |
|
| 70 |
+
| 0.4883 | 0.6857 | 300 | 0.4936 | -1.5404 | -2.7021 | 0.7371 | 1.1617 | -518.4875 | -413.9998 | 3.3276 | 2.1394 |
|
| 71 |
+
| 0.4714 | 0.8 | 350 | 0.4907 | -1.5101 | -2.6852 | 0.7457 | 1.1752 | -516.7975 | -410.9666 | 3.5416 | 2.3945 |
|
| 72 |
+
| 0.4977 | 0.9143 | 400 | 0.4890 | -1.5856 | -2.7832 | 0.7371 | 1.1976 | -526.5942 | -418.5211 | 3.6380 | 2.5080 |
|
| 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": 0.5399287248912611,
|
| 5 |
+
"train_runtime": 11756.1871,
|
| 6 |
+
"train_samples": 55999,
|
| 7 |
+
"train_samples_per_second": 4.763,
|
| 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": 0.9988571428571429,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5399287248912611,
|
| 5 |
+
"train_runtime": 11756.1871,
|
| 6 |
+
"train_samples": 55999,
|
| 7 |
+
"train_samples_per_second": 4.763,
|
| 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": 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": 6.122519788916474,
|
| 14 |
+
"learning_rate": 1.1363636363636363e-07,
|
| 15 |
+
"logits/chosen": -2.653984546661377,
|
| 16 |
+
"logits/rejected": -2.6347525119781494,
|
| 17 |
+
"logps/chosen": -287.884765625,
|
| 18 |
+
"logps/rejected": -266.53350830078125,
|
| 19 |
+
"loss": 0.6931,
|
| 20 |
+
"rewards/accuracies": 0.4625000059604645,
|
| 21 |
+
"rewards/chosen": 0.0003025080659426749,
|
| 22 |
+
"rewards/margins": 0.0007510354043915868,
|
| 23 |
+
"rewards/rejected": -0.0004485272220335901,
|
| 24 |
+
"step": 10
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.045714285714285714,
|
| 28 |
+
"grad_norm": 5.658247958378644,
|
| 29 |
+
"learning_rate": 2.2727272727272726e-07,
|
| 30 |
+
"logits/chosen": -2.655301332473755,
|
| 31 |
+
"logits/rejected": -2.6238667964935303,
|
| 32 |
+
"logps/chosen": -271.40631103515625,
|
| 33 |
+
"logps/rejected": -260.2054748535156,
|
| 34 |
+
"loss": 0.6924,
|
| 35 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 36 |
+
"rewards/chosen": 0.0030373495537787676,
|
| 37 |
+
"rewards/margins": 0.001921061659231782,
|
| 38 |
+
"rewards/rejected": 0.0011162875453010201,
|
| 39 |
+
"step": 20
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.06857142857142857,
|
| 43 |
+
"grad_norm": 5.381026859065415,
|
| 44 |
+
"learning_rate": 3.4090909090909085e-07,
|
| 45 |
+
"logits/chosen": -2.6506433486938477,
|
| 46 |
+
"logits/rejected": -2.55997896194458,
|
| 47 |
+
"logps/chosen": -271.79742431640625,
|
| 48 |
+
"logps/rejected": -254.66366577148438,
|
| 49 |
+
"loss": 0.689,
|
| 50 |
+
"rewards/accuracies": 0.731249988079071,
|
| 51 |
+
"rewards/chosen": 0.020913880318403244,
|
| 52 |
+
"rewards/margins": 0.012432877905666828,
|
| 53 |
+
"rewards/rejected": 0.008481001481413841,
|
| 54 |
+
"step": 30
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.09142857142857143,
|
| 58 |
+
"grad_norm": 4.9623708021127495,
|
| 59 |
+
"learning_rate": 4.545454545454545e-07,
|
| 60 |
+
"logits/chosen": -2.615001916885376,
|
| 61 |
+
"logits/rejected": -2.600529909133911,
|
| 62 |
+
"logps/chosen": -261.85809326171875,
|
| 63 |
+
"logps/rejected": -255.6659698486328,
|
| 64 |
+
"loss": 0.6791,
|
| 65 |
+
"rewards/accuracies": 0.699999988079071,
|
| 66 |
+
"rewards/chosen": 0.0391639769077301,
|
| 67 |
+
"rewards/margins": 0.04127795249223709,
|
| 68 |
+
"rewards/rejected": -0.0021139776799827814,
|
| 69 |
+
"step": 40
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.11428571428571428,
|
| 73 |
+
"grad_norm": 6.716802270370761,
|
| 74 |
+
"learning_rate": 4.997124959943201e-07,
|
| 75 |
+
"logits/chosen": -2.596109390258789,
|
| 76 |
+
"logits/rejected": -2.527108907699585,
|
| 77 |
+
"logps/chosen": -288.6273193359375,
|
| 78 |
+
"logps/rejected": -264.6972351074219,
|
| 79 |
+
"loss": 0.6615,
|
| 80 |
+
"rewards/accuracies": 0.762499988079071,
|
| 81 |
+
"rewards/chosen": 0.008229808881878853,
|
| 82 |
+
"rewards/margins": 0.11277468502521515,
|
| 83 |
+
"rewards/rejected": -0.10454487800598145,
|
| 84 |
+
"step": 50
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.11428571428571428,
|
| 88 |
+
"eval_logits/chosen": -2.622114658355713,
|
| 89 |
+
"eval_logits/rejected": -2.575448989868164,
|
| 90 |
+
"eval_logps/chosen": -264.4378662109375,
|
| 91 |
+
"eval_logps/rejected": -265.5943603515625,
|
| 92 |
+
"eval_loss": 0.6475189328193665,
|
| 93 |
+
"eval_rewards/accuracies": 0.6594827771186829,
|
| 94 |
+
"eval_rewards/chosen": -0.04476550966501236,
|
| 95 |
+
"eval_rewards/margins": 0.1284063458442688,
|
| 96 |
+
"eval_rewards/rejected": -0.17317183315753937,
|
| 97 |
+
"eval_runtime": 93.389,
|
| 98 |
+
"eval_samples_per_second": 19.606,
|
| 99 |
+
"eval_steps_per_second": 0.311,
|
| 100 |
+
"step": 50
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.13714285714285715,
|
| 104 |
+
"grad_norm": 10.665432103556162,
|
| 105 |
+
"learning_rate": 4.979579212164186e-07,
|
| 106 |
+
"logits/chosen": -2.6227073669433594,
|
| 107 |
+
"logits/rejected": -2.568916082382202,
|
| 108 |
+
"logps/chosen": -306.7157287597656,
|
| 109 |
+
"logps/rejected": -274.0307312011719,
|
| 110 |
+
"loss": 0.6358,
|
| 111 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 112 |
+
"rewards/chosen": -0.07229714095592499,
|
| 113 |
+
"rewards/margins": 0.2034287005662918,
|
| 114 |
+
"rewards/rejected": -0.2757258713245392,
|
| 115 |
+
"step": 60
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"epoch": 0.16,
|
| 119 |
+
"grad_norm": 14.460690059842785,
|
| 120 |
+
"learning_rate": 4.946196886175515e-07,
|
| 121 |
+
"logits/chosen": -2.6028428077697754,
|
| 122 |
+
"logits/rejected": -2.5148682594299316,
|
| 123 |
+
"logps/chosen": -322.69781494140625,
|
| 124 |
+
"logps/rejected": -301.1668395996094,
|
| 125 |
+
"loss": 0.6208,
|
| 126 |
+
"rewards/accuracies": 0.699999988079071,
|
| 127 |
+
"rewards/chosen": -0.1934281885623932,
|
| 128 |
+
"rewards/margins": 0.33089178800582886,
|
| 129 |
+
"rewards/rejected": -0.5243200063705444,
|
| 130 |
+
"step": 70
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 0.18285714285714286,
|
| 134 |
+
"grad_norm": 13.62881948148791,
|
| 135 |
+
"learning_rate": 4.897191188239667e-07,
|
| 136 |
+
"logits/chosen": -2.143632411956787,
|
| 137 |
+
"logits/rejected": -2.028029203414917,
|
| 138 |
+
"logps/chosen": -308.3243713378906,
|
| 139 |
+
"logps/rejected": -315.1154479980469,
|
| 140 |
+
"loss": 0.6067,
|
| 141 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 142 |
+
"rewards/chosen": -0.4112843871116638,
|
| 143 |
+
"rewards/margins": 0.27252885699272156,
|
| 144 |
+
"rewards/rejected": -0.683813214302063,
|
| 145 |
+
"step": 80
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"epoch": 0.2057142857142857,
|
| 149 |
+
"grad_norm": 14.003540136427612,
|
| 150 |
+
"learning_rate": 4.832875107981763e-07,
|
| 151 |
+
"logits/chosen": -0.7628787755966187,
|
| 152 |
+
"logits/rejected": -0.48756521940231323,
|
| 153 |
+
"logps/chosen": -307.3790588378906,
|
| 154 |
+
"logps/rejected": -373.88568115234375,
|
| 155 |
+
"loss": 0.5805,
|
| 156 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 157 |
+
"rewards/chosen": -0.657852053642273,
|
| 158 |
+
"rewards/margins": 0.5630446672439575,
|
| 159 |
+
"rewards/rejected": -1.22089684009552,
|
| 160 |
+
"step": 90
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 0.22857142857142856,
|
| 164 |
+
"grad_norm": 27.196144010198164,
|
| 165 |
+
"learning_rate": 4.753659419387223e-07,
|
| 166 |
+
"logits/chosen": -0.3219766616821289,
|
| 167 |
+
"logits/rejected": 0.034670639783144,
|
| 168 |
+
"logps/chosen": -338.6977233886719,
|
| 169 |
+
"logps/rejected": -383.44903564453125,
|
| 170 |
+
"loss": 0.5507,
|
| 171 |
+
"rewards/accuracies": 0.706250011920929,
|
| 172 |
+
"rewards/chosen": -0.7744790315628052,
|
| 173 |
+
"rewards/margins": 0.7537156343460083,
|
| 174 |
+
"rewards/rejected": -1.528194546699524,
|
| 175 |
+
"step": 100
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"epoch": 0.22857142857142856,
|
| 179 |
+
"eval_logits/chosen": -0.1843080073595047,
|
| 180 |
+
"eval_logits/rejected": 0.2567289173603058,
|
| 181 |
+
"eval_logps/chosen": -335.9164123535156,
|
| 182 |
+
"eval_logps/rejected": -404.4942626953125,
|
| 183 |
+
"eval_loss": 0.5504318475723267,
|
| 184 |
+
"eval_rewards/accuracies": 0.6896551847457886,
|
| 185 |
+
"eval_rewards/chosen": -0.7595511674880981,
|
| 186 |
+
"eval_rewards/margins": 0.8026195764541626,
|
| 187 |
+
"eval_rewards/rejected": -1.5621706247329712,
|
| 188 |
+
"eval_runtime": 94.019,
|
| 189 |
+
"eval_samples_per_second": 19.475,
|
| 190 |
+
"eval_steps_per_second": 0.308,
|
| 191 |
+
"step": 100
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.25142857142857145,
|
| 195 |
+
"grad_norm": 20.09944606036678,
|
| 196 |
+
"learning_rate": 4.660050057270191e-07,
|
| 197 |
+
"logits/chosen": 0.22648504376411438,
|
| 198 |
+
"logits/rejected": 0.9535753130912781,
|
| 199 |
+
"logps/chosen": -380.6333923339844,
|
| 200 |
+
"logps/rejected": -404.15185546875,
|
| 201 |
+
"loss": 0.5258,
|
| 202 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 203 |
+
"rewards/chosen": -0.7437217831611633,
|
| 204 |
+
"rewards/margins": 0.9534850120544434,
|
| 205 |
+
"rewards/rejected": -1.6972068548202515,
|
| 206 |
+
"step": 110
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.2742857142857143,
|
| 210 |
+
"grad_norm": 18.922641880361788,
|
| 211 |
+
"learning_rate": 4.5526448859687144e-07,
|
| 212 |
+
"logits/chosen": 0.726629376411438,
|
| 213 |
+
"logits/rejected": 1.2900234460830688,
|
| 214 |
+
"logps/chosen": -386.16644287109375,
|
| 215 |
+
"logps/rejected": -446.7608947753906,
|
| 216 |
+
"loss": 0.5563,
|
| 217 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 218 |
+
"rewards/chosen": -1.1333491802215576,
|
| 219 |
+
"rewards/margins": 0.8371905088424683,
|
| 220 |
+
"rewards/rejected": -1.9705396890640259,
|
| 221 |
+
"step": 120
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.29714285714285715,
|
| 225 |
+
"grad_norm": 18.05413983916241,
|
| 226 |
+
"learning_rate": 4.432129880904388e-07,
|
| 227 |
+
"logits/chosen": 1.176317572593689,
|
| 228 |
+
"logits/rejected": 1.9044733047485352,
|
| 229 |
+
"logps/chosen": -381.20587158203125,
|
| 230 |
+
"logps/rejected": -410.0304260253906,
|
| 231 |
+
"loss": 0.5387,
|
| 232 |
+
"rewards/accuracies": 0.643750011920929,
|
| 233 |
+
"rewards/chosen": -1.07724130153656,
|
| 234 |
+
"rewards/margins": 0.6341379284858704,
|
| 235 |
+
"rewards/rejected": -1.7113792896270752,
|
| 236 |
+
"step": 130
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 0.32,
|
| 240 |
+
"grad_norm": 16.462235206322415,
|
| 241 |
+
"learning_rate": 4.299274747394055e-07,
|
| 242 |
+
"logits/chosen": 1.4854234457015991,
|
| 243 |
+
"logits/rejected": 2.0871422290802,
|
| 244 |
+
"logps/chosen": -424.3130798339844,
|
| 245 |
+
"logps/rejected": -455.3707580566406,
|
| 246 |
+
"loss": 0.5369,
|
| 247 |
+
"rewards/accuracies": 0.706250011920929,
|
| 248 |
+
"rewards/chosen": -1.0648014545440674,
|
| 249 |
+
"rewards/margins": 0.886120617389679,
|
| 250 |
+
"rewards/rejected": -1.9509223699569702,
|
| 251 |
+
"step": 140
|
| 252 |
+
},
|
| 253 |
+
{
|
| 254 |
+
"epoch": 0.34285714285714286,
|
| 255 |
+
"grad_norm": 28.494521300429405,
|
| 256 |
+
"learning_rate": 4.1549280046953653e-07,
|
| 257 |
+
"logits/chosen": 1.2995119094848633,
|
| 258 |
+
"logits/rejected": 1.8770694732666016,
|
| 259 |
+
"logps/chosen": -386.13653564453125,
|
| 260 |
+
"logps/rejected": -437.58349609375,
|
| 261 |
+
"loss": 0.521,
|
| 262 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 263 |
+
"rewards/chosen": -0.9008251428604126,
|
| 264 |
+
"rewards/margins": 0.8192182779312134,
|
| 265 |
+
"rewards/rejected": -1.7200431823730469,
|
| 266 |
+
"step": 150
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 0.34285714285714286,
|
| 270 |
+
"eval_logits/chosen": 1.1096802949905396,
|
| 271 |
+
"eval_logits/rejected": 1.8187731504440308,
|
| 272 |
+
"eval_logps/chosen": -354.0176086425781,
|
| 273 |
+
"eval_logps/rejected": -437.4890441894531,
|
| 274 |
+
"eval_loss": 0.5233712196350098,
|
| 275 |
+
"eval_rewards/accuracies": 0.7198275923728943,
|
| 276 |
+
"eval_rewards/chosen": -0.9405630826950073,
|
| 277 |
+
"eval_rewards/margins": 0.9515554308891296,
|
| 278 |
+
"eval_rewards/rejected": -1.8921183347702026,
|
| 279 |
+
"eval_runtime": 95.0191,
|
| 280 |
+
"eval_samples_per_second": 19.27,
|
| 281 |
+
"eval_steps_per_second": 0.305,
|
| 282 |
+
"step": 150
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 0.3657142857142857,
|
| 286 |
+
"grad_norm": 19.256664361058306,
|
| 287 |
+
"learning_rate": 4.000011566683401e-07,
|
| 288 |
+
"logits/chosen": 1.1306709051132202,
|
| 289 |
+
"logits/rejected": 1.9139797687530518,
|
| 290 |
+
"logps/chosen": -414.0078125,
|
| 291 |
+
"logps/rejected": -468.516357421875,
|
| 292 |
+
"loss": 0.511,
|
| 293 |
+
"rewards/accuracies": 0.737500011920929,
|
| 294 |
+
"rewards/chosen": -1.0485427379608154,
|
| 295 |
+
"rewards/margins": 0.9393121600151062,
|
| 296 |
+
"rewards/rejected": -1.9878549575805664,
|
| 297 |
+
"step": 160
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"epoch": 0.38857142857142857,
|
| 301 |
+
"grad_norm": 24.708541517576087,
|
| 302 |
+
"learning_rate": 3.8355148537705047e-07,
|
| 303 |
+
"logits/chosen": 1.1292288303375244,
|
| 304 |
+
"logits/rejected": 1.7640559673309326,
|
| 305 |
+
"logps/chosen": -385.39031982421875,
|
| 306 |
+
"logps/rejected": -465.3546447753906,
|
| 307 |
+
"loss": 0.5204,
|
| 308 |
+
"rewards/accuracies": 0.731249988079071,
|
| 309 |
+
"rewards/chosen": -0.91693514585495,
|
| 310 |
+
"rewards/margins": 0.99778813123703,
|
| 311 |
+
"rewards/rejected": -1.9147233963012695,
|
| 312 |
+
"step": 170
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 0.4114285714285714,
|
| 316 |
+
"grad_norm": 19.42104847899406,
|
| 317 |
+
"learning_rate": 3.662488473675315e-07,
|
| 318 |
+
"logits/chosen": 1.3040357828140259,
|
| 319 |
+
"logits/rejected": 2.3787357807159424,
|
| 320 |
+
"logps/chosen": -400.56915283203125,
|
| 321 |
+
"logps/rejected": -451.06890869140625,
|
| 322 |
+
"loss": 0.519,
|
| 323 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 324 |
+
"rewards/chosen": -0.9812885522842407,
|
| 325 |
+
"rewards/margins": 1.0648243427276611,
|
| 326 |
+
"rewards/rejected": -2.0461127758026123,
|
| 327 |
+
"step": 180
|
| 328 |
+
},
|
| 329 |
+
{
|
| 330 |
+
"epoch": 0.4342857142857143,
|
| 331 |
+
"grad_norm": 23.726334837975923,
|
| 332 |
+
"learning_rate": 3.48203751140067e-07,
|
| 333 |
+
"logits/chosen": 1.348807692527771,
|
| 334 |
+
"logits/rejected": 2.427119731903076,
|
| 335 |
+
"logps/chosen": -415.06341552734375,
|
| 336 |
+
"logps/rejected": -491.79290771484375,
|
| 337 |
+
"loss": 0.517,
|
| 338 |
+
"rewards/accuracies": 0.793749988079071,
|
| 339 |
+
"rewards/chosen": -1.542277455329895,
|
| 340 |
+
"rewards/margins": 1.0263307094573975,
|
| 341 |
+
"rewards/rejected": -2.568608283996582,
|
| 342 |
+
"step": 190
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 0.45714285714285713,
|
| 346 |
+
"grad_norm": 19.830821329672027,
|
| 347 |
+
"learning_rate": 3.2953144712759537e-07,
|
| 348 |
+
"logits/chosen": 1.331316351890564,
|
| 349 |
+
"logits/rejected": 2.264349937438965,
|
| 350 |
+
"logps/chosen": -363.25860595703125,
|
| 351 |
+
"logps/rejected": -451.0553283691406,
|
| 352 |
+
"loss": 0.5162,
|
| 353 |
+
"rewards/accuracies": 0.737500011920929,
|
| 354 |
+
"rewards/chosen": -1.1292212009429932,
|
| 355 |
+
"rewards/margins": 0.9935659170150757,
|
| 356 |
+
"rewards/rejected": -2.1227869987487793,
|
| 357 |
+
"step": 200
|
| 358 |
+
},
|
| 359 |
+
{
|
| 360 |
+
"epoch": 0.45714285714285713,
|
| 361 |
+
"eval_logits/chosen": 1.2298647165298462,
|
| 362 |
+
"eval_logits/rejected": 2.2567214965820312,
|
| 363 |
+
"eval_logps/chosen": -383.05267333984375,
|
| 364 |
+
"eval_logps/rejected": -473.64532470703125,
|
| 365 |
+
"eval_loss": 0.5107715129852295,
|
| 366 |
+
"eval_rewards/accuracies": 0.732758641242981,
|
| 367 |
+
"eval_rewards/chosen": -1.230913519859314,
|
| 368 |
+
"eval_rewards/margins": 1.0227679014205933,
|
| 369 |
+
"eval_rewards/rejected": -2.2536816596984863,
|
| 370 |
+
"eval_runtime": 94.5713,
|
| 371 |
+
"eval_samples_per_second": 19.361,
|
| 372 |
+
"eval_steps_per_second": 0.307,
|
| 373 |
+
"step": 200
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"epoch": 0.48,
|
| 377 |
+
"grad_norm": 22.298035621509914,
|
| 378 |
+
"learning_rate": 3.103511916141658e-07,
|
| 379 |
+
"logits/chosen": 1.5307371616363525,
|
| 380 |
+
"logits/rejected": 2.729835033416748,
|
| 381 |
+
"logps/chosen": -424.962890625,
|
| 382 |
+
"logps/rejected": -514.0426025390625,
|
| 383 |
+
"loss": 0.5071,
|
| 384 |
+
"rewards/accuracies": 0.75,
|
| 385 |
+
"rewards/chosen": -1.3911924362182617,
|
| 386 |
+
"rewards/margins": 1.1286324262619019,
|
| 387 |
+
"rewards/rejected": -2.5198252201080322,
|
| 388 |
+
"step": 210
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 0.5028571428571429,
|
| 392 |
+
"grad_norm": 14.694863409372582,
|
| 393 |
+
"learning_rate": 2.9078548506882117e-07,
|
| 394 |
+
"logits/chosen": 1.7180284261703491,
|
| 395 |
+
"logits/rejected": 2.68153715133667,
|
| 396 |
+
"logps/chosen": -447.5650939941406,
|
| 397 |
+
"logps/rejected": -545.7744140625,
|
| 398 |
+
"loss": 0.5181,
|
| 399 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 400 |
+
"rewards/chosen": -1.4318263530731201,
|
| 401 |
+
"rewards/margins": 1.1245338916778564,
|
| 402 |
+
"rewards/rejected": -2.5563602447509766,
|
| 403 |
+
"step": 220
|
| 404 |
+
},
|
| 405 |
+
{
|
| 406 |
+
"epoch": 0.5257142857142857,
|
| 407 |
+
"grad_norm": 21.866604136555903,
|
| 408 |
+
"learning_rate": 2.709592897595191e-07,
|
| 409 |
+
"logits/chosen": 1.7002506256103516,
|
| 410 |
+
"logits/rejected": 2.5147674083709717,
|
| 411 |
+
"logps/chosen": -424.66607666015625,
|
| 412 |
+
"logps/rejected": -503.14837646484375,
|
| 413 |
+
"loss": 0.539,
|
| 414 |
+
"rewards/accuracies": 0.706250011920929,
|
| 415 |
+
"rewards/chosen": -1.3987152576446533,
|
| 416 |
+
"rewards/margins": 0.8508893251419067,
|
| 417 |
+
"rewards/rejected": -2.2496042251586914,
|
| 418 |
+
"step": 230
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"epoch": 0.5485714285714286,
|
| 422 |
+
"grad_norm": 23.461922171727267,
|
| 423 |
+
"learning_rate": 2.509992316440332e-07,
|
| 424 |
+
"logits/chosen": 2.7161831855773926,
|
| 425 |
+
"logits/rejected": 3.887934446334839,
|
| 426 |
+
"logps/chosen": -481.4347229003906,
|
| 427 |
+
"logps/rejected": -544.0018310546875,
|
| 428 |
+
"loss": 0.5234,
|
| 429 |
+
"rewards/accuracies": 0.75,
|
| 430 |
+
"rewards/chosen": -1.737464189529419,
|
| 431 |
+
"rewards/margins": 1.0862486362457275,
|
| 432 |
+
"rewards/rejected": -2.8237128257751465,
|
| 433 |
+
"step": 240
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 0.5714285714285714,
|
| 437 |
+
"grad_norm": 23.055743791457775,
|
| 438 |
+
"learning_rate": 2.3103279163519918e-07,
|
| 439 |
+
"logits/chosen": 2.5470688343048096,
|
| 440 |
+
"logits/rejected": 3.076658248901367,
|
| 441 |
+
"logps/chosen": -456.06646728515625,
|
| 442 |
+
"logps/rejected": -522.9896850585938,
|
| 443 |
+
"loss": 0.5186,
|
| 444 |
+
"rewards/accuracies": 0.643750011920929,
|
| 445 |
+
"rewards/chosen": -1.7880340814590454,
|
| 446 |
+
"rewards/margins": 0.6319095492362976,
|
| 447 |
+
"rewards/rejected": -2.419943332672119,
|
| 448 |
+
"step": 250
|
| 449 |
+
},
|
| 450 |
+
{
|
| 451 |
+
"epoch": 0.5714285714285714,
|
| 452 |
+
"eval_logits/chosen": 1.5761785507202148,
|
| 453 |
+
"eval_logits/rejected": 2.8322033882141113,
|
| 454 |
+
"eval_logps/chosen": -396.9323425292969,
|
| 455 |
+
"eval_logps/rejected": -496.5982971191406,
|
| 456 |
+
"eval_loss": 0.5027161836624146,
|
| 457 |
+
"eval_rewards/accuracies": 0.7025862336158752,
|
| 458 |
+
"eval_rewards/chosen": -1.3697103261947632,
|
| 459 |
+
"eval_rewards/margins": 1.1135003566741943,
|
| 460 |
+
"eval_rewards/rejected": -2.483210563659668,
|
| 461 |
+
"eval_runtime": 93.6382,
|
| 462 |
+
"eval_samples_per_second": 19.554,
|
| 463 |
+
"eval_steps_per_second": 0.31,
|
| 464 |
+
"step": 250
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 0.5942857142857143,
|
| 468 |
+
"grad_norm": 23.966486101527014,
|
| 469 |
+
"learning_rate": 2.1118749140573358e-07,
|
| 470 |
+
"logits/chosen": 2.044908046722412,
|
| 471 |
+
"logits/rejected": 3.2726006507873535,
|
| 472 |
+
"logps/chosen": -411.2015075683594,
|
| 473 |
+
"logps/rejected": -446.53857421875,
|
| 474 |
+
"loss": 0.514,
|
| 475 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 476 |
+
"rewards/chosen": -1.3537694215774536,
|
| 477 |
+
"rewards/margins": 0.8869277834892273,
|
| 478 |
+
"rewards/rejected": -2.240697145462036,
|
| 479 |
+
"step": 260
|
| 480 |
+
},
|
| 481 |
+
{
|
| 482 |
+
"epoch": 0.6171428571428571,
|
| 483 |
+
"grad_norm": 19.203425356174254,
|
| 484 |
+
"learning_rate": 1.9159007893272703e-07,
|
| 485 |
+
"logits/chosen": 2.4072651863098145,
|
| 486 |
+
"logits/rejected": 3.3252644538879395,
|
| 487 |
+
"logps/chosen": -457.94830322265625,
|
| 488 |
+
"logps/rejected": -485.1842346191406,
|
| 489 |
+
"loss": 0.5102,
|
| 490 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 491 |
+
"rewards/chosen": -1.6982141733169556,
|
| 492 |
+
"rewards/margins": 0.6931914687156677,
|
| 493 |
+
"rewards/rejected": -2.3914055824279785,
|
| 494 |
+
"step": 270
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 0.64,
|
| 498 |
+
"grad_norm": 17.06179349785014,
|
| 499 |
+
"learning_rate": 1.7236571898357766e-07,
|
| 500 |
+
"logits/chosen": 1.9787667989730835,
|
| 501 |
+
"logits/rejected": 3.012749195098877,
|
| 502 |
+
"logps/chosen": -375.3614807128906,
|
| 503 |
+
"logps/rejected": -471.38397216796875,
|
| 504 |
+
"loss": 0.514,
|
| 505 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 506 |
+
"rewards/chosen": -1.2563382387161255,
|
| 507 |
+
"rewards/margins": 1.122116208076477,
|
| 508 |
+
"rewards/rejected": -2.3784542083740234,
|
| 509 |
+
"step": 280
|
| 510 |
+
},
|
| 511 |
+
{
|
| 512 |
+
"epoch": 0.6628571428571428,
|
| 513 |
+
"grad_norm": 22.052326742230484,
|
| 514 |
+
"learning_rate": 1.5363719371356882e-07,
|
| 515 |
+
"logits/chosen": 2.3124561309814453,
|
| 516 |
+
"logits/rejected": 3.362464189529419,
|
| 517 |
+
"logps/chosen": -446.782958984375,
|
| 518 |
+
"logps/rejected": -505.8094177246094,
|
| 519 |
+
"loss": 0.496,
|
| 520 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 521 |
+
"rewards/chosen": -1.6085598468780518,
|
| 522 |
+
"rewards/margins": 0.9179994463920593,
|
| 523 |
+
"rewards/rejected": -2.5265591144561768,
|
| 524 |
+
"step": 290
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"epoch": 0.6857142857142857,
|
| 528 |
+
"grad_norm": 21.6246754175424,
|
| 529 |
+
"learning_rate": 1.3552411848071565e-07,
|
| 530 |
+
"logits/chosen": 2.0221407413482666,
|
| 531 |
+
"logits/rejected": 3.554128646850586,
|
| 532 |
+
"logps/chosen": -473.49298095703125,
|
| 533 |
+
"logps/rejected": -550.6085815429688,
|
| 534 |
+
"loss": 0.4883,
|
| 535 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 536 |
+
"rewards/chosen": -1.6014286279678345,
|
| 537 |
+
"rewards/margins": 1.2605220079421997,
|
| 538 |
+
"rewards/rejected": -2.861950397491455,
|
| 539 |
+
"step": 300
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 0.6857142857142857,
|
| 543 |
+
"eval_logits/chosen": 2.1394236087799072,
|
| 544 |
+
"eval_logits/rejected": 3.3275508880615234,
|
| 545 |
+
"eval_logps/chosen": -413.999755859375,
|
| 546 |
+
"eval_logps/rejected": -518.487548828125,
|
| 547 |
+
"eval_loss": 0.4935988783836365,
|
| 548 |
+
"eval_rewards/accuracies": 0.7370689511299133,
|
| 549 |
+
"eval_rewards/chosen": -1.5403847694396973,
|
| 550 |
+
"eval_rewards/margins": 1.1617193222045898,
|
| 551 |
+
"eval_rewards/rejected": -2.702104091644287,
|
| 552 |
+
"eval_runtime": 93.5025,
|
| 553 |
+
"eval_samples_per_second": 19.582,
|
| 554 |
+
"eval_steps_per_second": 0.31,
|
| 555 |
+
"step": 300
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 0.7085714285714285,
|
| 559 |
+
"grad_norm": 20.645058605492473,
|
| 560 |
+
"learning_rate": 1.1814217788631473e-07,
|
| 561 |
+
"logits/chosen": 1.9714610576629639,
|
| 562 |
+
"logits/rejected": 3.4288382530212402,
|
| 563 |
+
"logps/chosen": -459.2438049316406,
|
| 564 |
+
"logps/rejected": -542.9857177734375,
|
| 565 |
+
"loss": 0.5034,
|
| 566 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 567 |
+
"rewards/chosen": -1.459855079650879,
|
| 568 |
+
"rewards/margins": 1.2939279079437256,
|
| 569 |
+
"rewards/rejected": -2.7537827491760254,
|
| 570 |
+
"step": 310
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"epoch": 0.7314285714285714,
|
| 574 |
+
"grad_norm": 28.46682682513433,
|
| 575 |
+
"learning_rate": 1.0160238692045331e-07,
|
| 576 |
+
"logits/chosen": 2.388679265975952,
|
| 577 |
+
"logits/rejected": 3.5533275604248047,
|
| 578 |
+
"logps/chosen": -448.905029296875,
|
| 579 |
+
"logps/rejected": -526.8614501953125,
|
| 580 |
+
"loss": 0.5066,
|
| 581 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 582 |
+
"rewards/chosen": -1.5178719758987427,
|
| 583 |
+
"rewards/margins": 1.000969648361206,
|
| 584 |
+
"rewards/rejected": -2.5188417434692383,
|
| 585 |
+
"step": 320
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 0.7542857142857143,
|
| 589 |
+
"grad_norm": 21.745911663397,
|
| 590 |
+
"learning_rate": 8.601038193139438e-08,
|
| 591 |
+
"logits/chosen": 2.581099271774292,
|
| 592 |
+
"logits/rejected": 3.3041462898254395,
|
| 593 |
+
"logps/chosen": -381.9894714355469,
|
| 594 |
+
"logps/rejected": -467.76580810546875,
|
| 595 |
+
"loss": 0.5113,
|
| 596 |
+
"rewards/accuracies": 0.675000011920929,
|
| 597 |
+
"rewards/chosen": -1.3630098104476929,
|
| 598 |
+
"rewards/margins": 0.9106127023696899,
|
| 599 |
+
"rewards/rejected": -2.273622512817383,
|
| 600 |
+
"step": 330
|
| 601 |
+
},
|
| 602 |
+
{
|
| 603 |
+
"epoch": 0.7771428571428571,
|
| 604 |
+
"grad_norm": 22.95799998221957,
|
| 605 |
+
"learning_rate": 7.146574594727572e-08,
|
| 606 |
+
"logits/chosen": 2.3186068534851074,
|
| 607 |
+
"logits/rejected": 3.4312567710876465,
|
| 608 |
+
"logps/chosen": -391.4497985839844,
|
| 609 |
+
"logps/rejected": -469.642333984375,
|
| 610 |
+
"loss": 0.4892,
|
| 611 |
+
"rewards/accuracies": 0.78125,
|
| 612 |
+
"rewards/chosen": -1.4544594287872314,
|
| 613 |
+
"rewards/margins": 1.1061028242111206,
|
| 614 |
+
"rewards/rejected": -2.5605618953704834,
|
| 615 |
+
"step": 340
|
| 616 |
+
},
|
| 617 |
+
{
|
| 618 |
+
"epoch": 0.8,
|
| 619 |
+
"grad_norm": 29.21776715425853,
|
| 620 |
+
"learning_rate": 5.8061372659157306e-08,
|
| 621 |
+
"logits/chosen": 2.201559066772461,
|
| 622 |
+
"logits/rejected": 3.762953996658325,
|
| 623 |
+
"logps/chosen": -459.7745056152344,
|
| 624 |
+
"logps/rejected": -539.0149536132812,
|
| 625 |
+
"loss": 0.4714,
|
| 626 |
+
"rewards/accuracies": 0.78125,
|
| 627 |
+
"rewards/chosen": -1.4960135221481323,
|
| 628 |
+
"rewards/margins": 1.210325002670288,
|
| 629 |
+
"rewards/rejected": -2.70633864402771,
|
| 630 |
+
"step": 350
|
| 631 |
+
},
|
| 632 |
+
{
|
| 633 |
+
"epoch": 0.8,
|
| 634 |
+
"eval_logits/chosen": 2.394544839859009,
|
| 635 |
+
"eval_logits/rejected": 3.541562557220459,
|
| 636 |
+
"eval_logps/chosen": -410.9665832519531,
|
| 637 |
+
"eval_logps/rejected": -516.7974853515625,
|
| 638 |
+
"eval_loss": 0.4907006621360779,
|
| 639 |
+
"eval_rewards/accuracies": 0.7456896305084229,
|
| 640 |
+
"eval_rewards/chosen": -1.5100529193878174,
|
| 641 |
+
"eval_rewards/margins": 1.1751505136489868,
|
| 642 |
+
"eval_rewards/rejected": -2.6852033138275146,
|
| 643 |
+
"eval_runtime": 93.0638,
|
| 644 |
+
"eval_samples_per_second": 19.675,
|
| 645 |
+
"eval_steps_per_second": 0.312,
|
| 646 |
+
"step": 350
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"epoch": 0.8228571428571428,
|
| 650 |
+
"grad_norm": 22.43994807327437,
|
| 651 |
+
"learning_rate": 4.5882873127531614e-08,
|
| 652 |
+
"logits/chosen": 2.859044075012207,
|
| 653 |
+
"logits/rejected": 3.9125289916992188,
|
| 654 |
+
"logps/chosen": -440.3011169433594,
|
| 655 |
+
"logps/rejected": -565.3880615234375,
|
| 656 |
+
"loss": 0.4928,
|
| 657 |
+
"rewards/accuracies": 0.8187500238418579,
|
| 658 |
+
"rewards/chosen": -1.4779033660888672,
|
| 659 |
+
"rewards/margins": 1.5023510456085205,
|
| 660 |
+
"rewards/rejected": -2.9802544116973877,
|
| 661 |
+
"step": 360
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 0.8457142857142858,
|
| 665 |
+
"grad_norm": 22.645094236252362,
|
| 666 |
+
"learning_rate": 3.500802900154412e-08,
|
| 667 |
+
"logits/chosen": 2.81022310256958,
|
| 668 |
+
"logits/rejected": 3.5364959239959717,
|
| 669 |
+
"logps/chosen": -403.3380126953125,
|
| 670 |
+
"logps/rejected": -527.1024169921875,
|
| 671 |
+
"loss": 0.4891,
|
| 672 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 673 |
+
"rewards/chosen": -1.4733188152313232,
|
| 674 |
+
"rewards/margins": 1.1765823364257812,
|
| 675 |
+
"rewards/rejected": -2.6499013900756836,
|
| 676 |
+
"step": 370
|
| 677 |
+
},
|
| 678 |
+
{
|
| 679 |
+
"epoch": 0.8685714285714285,
|
| 680 |
+
"grad_norm": 24.474487723268208,
|
| 681 |
+
"learning_rate": 2.550629574310309e-08,
|
| 682 |
+
"logits/chosen": 3.3984062671661377,
|
| 683 |
+
"logits/rejected": 4.194012641906738,
|
| 684 |
+
"logps/chosen": -419.423583984375,
|
| 685 |
+
"logps/rejected": -518.23583984375,
|
| 686 |
+
"loss": 0.4926,
|
| 687 |
+
"rewards/accuracies": 0.6875,
|
| 688 |
+
"rewards/chosen": -1.635619878768921,
|
| 689 |
+
"rewards/margins": 0.9691934585571289,
|
| 690 |
+
"rewards/rejected": -2.60481333732605,
|
| 691 |
+
"step": 380
|
| 692 |
+
},
|
| 693 |
+
{
|
| 694 |
+
"epoch": 0.8914285714285715,
|
| 695 |
+
"grad_norm": 25.84913647388076,
|
| 696 |
+
"learning_rate": 1.7438359028687983e-08,
|
| 697 |
+
"logits/chosen": 2.8808953762054443,
|
| 698 |
+
"logits/rejected": 3.839012861251831,
|
| 699 |
+
"logps/chosen": -421.871337890625,
|
| 700 |
+
"logps/rejected": -526.1373291015625,
|
| 701 |
+
"loss": 0.497,
|
| 702 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 703 |
+
"rewards/chosen": -1.5530052185058594,
|
| 704 |
+
"rewards/margins": 1.1946289539337158,
|
| 705 |
+
"rewards/rejected": -2.747633934020996,
|
| 706 |
+
"step": 390
|
| 707 |
+
},
|
| 708 |
+
{
|
| 709 |
+
"epoch": 0.9142857142857143,
|
| 710 |
+
"grad_norm": 28.299780272584496,
|
| 711 |
+
"learning_rate": 1.0855747162029361e-08,
|
| 712 |
+
"logits/chosen": 2.4124693870544434,
|
| 713 |
+
"logits/rejected": 3.1909306049346924,
|
| 714 |
+
"logps/chosen": -425.010498046875,
|
| 715 |
+
"logps/rejected": -537.2962646484375,
|
| 716 |
+
"loss": 0.4977,
|
| 717 |
+
"rewards/accuracies": 0.737500011920929,
|
| 718 |
+
"rewards/chosen": -1.5087743997573853,
|
| 719 |
+
"rewards/margins": 1.1911652088165283,
|
| 720 |
+
"rewards/rejected": -2.699939250946045,
|
| 721 |
+
"step": 400
|
| 722 |
+
},
|
| 723 |
+
{
|
| 724 |
+
"epoch": 0.9142857142857143,
|
| 725 |
+
"eval_logits/chosen": 2.507979393005371,
|
| 726 |
+
"eval_logits/rejected": 3.6379663944244385,
|
| 727 |
+
"eval_logps/chosen": -418.5211486816406,
|
| 728 |
+
"eval_logps/rejected": -526.59423828125,
|
| 729 |
+
"eval_loss": 0.4890148639678955,
|
| 730 |
+
"eval_rewards/accuracies": 0.7370689511299133,
|
| 731 |
+
"eval_rewards/chosen": -1.5855982303619385,
|
| 732 |
+
"eval_rewards/margins": 1.197571873664856,
|
| 733 |
+
"eval_rewards/rejected": -2.783169984817505,
|
| 734 |
+
"eval_runtime": 95.2901,
|
| 735 |
+
"eval_samples_per_second": 19.215,
|
| 736 |
+
"eval_steps_per_second": 0.304,
|
| 737 |
+
"step": 400
|
| 738 |
+
},
|
| 739 |
+
{
|
| 740 |
+
"epoch": 0.9371428571428572,
|
| 741 |
+
"grad_norm": 19.845237856376826,
|
| 742 |
+
"learning_rate": 5.8005019731033615e-09,
|
| 743 |
+
"logits/chosen": 2.7473673820495605,
|
| 744 |
+
"logits/rejected": 3.6581249237060547,
|
| 745 |
+
"logps/chosen": -407.8973083496094,
|
| 746 |
+
"logps/rejected": -511.965576171875,
|
| 747 |
+
"loss": 0.4863,
|
| 748 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 749 |
+
"rewards/chosen": -1.4759327173233032,
|
| 750 |
+
"rewards/margins": 1.0670742988586426,
|
| 751 |
+
"rewards/rejected": -2.5430071353912354,
|
| 752 |
+
"step": 410
|
| 753 |
+
},
|
| 754 |
+
{
|
| 755 |
+
"epoch": 0.96,
|
| 756 |
+
"grad_norm": 20.45469236378943,
|
| 757 |
+
"learning_rate": 2.3049103053431886e-09,
|
| 758 |
+
"logits/chosen": 2.3240396976470947,
|
| 759 |
+
"logits/rejected": 3.4506924152374268,
|
| 760 |
+
"logps/chosen": -462.15362548828125,
|
| 761 |
+
"logps/rejected": -587.3756103515625,
|
| 762 |
+
"loss": 0.4973,
|
| 763 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 764 |
+
"rewards/chosen": -1.514866590499878,
|
| 765 |
+
"rewards/margins": 1.4163448810577393,
|
| 766 |
+
"rewards/rejected": -2.931211471557617,
|
| 767 |
+
"step": 420
|
| 768 |
+
},
|
| 769 |
+
{
|
| 770 |
+
"epoch": 0.9828571428571429,
|
| 771 |
+
"grad_norm": 26.841510292915476,
|
| 772 |
+
"learning_rate": 3.9129780600541397e-10,
|
| 773 |
+
"logits/chosen": 2.6934516429901123,
|
| 774 |
+
"logits/rejected": 3.5782058238983154,
|
| 775 |
+
"logps/chosen": -432.9996643066406,
|
| 776 |
+
"logps/rejected": -534.2493896484375,
|
| 777 |
+
"loss": 0.4797,
|
| 778 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 779 |
+
"rewards/chosen": -1.4740568399429321,
|
| 780 |
+
"rewards/margins": 1.289287805557251,
|
| 781 |
+
"rewards/rejected": -2.7633445262908936,
|
| 782 |
+
"step": 430
|
| 783 |
+
},
|
| 784 |
+
{
|
| 785 |
+
"epoch": 0.9988571428571429,
|
| 786 |
+
"step": 437,
|
| 787 |
+
"total_flos": 0.0,
|
| 788 |
+
"train_loss": 0.5399287248912611,
|
| 789 |
+
"train_runtime": 11756.1871,
|
| 790 |
+
"train_samples_per_second": 4.763,
|
| 791 |
+
"train_steps_per_second": 0.037
|
| 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 |
+
}
|