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-high-curriculum
|
| 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-high-curriculum
|
| 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: 0.5204
|
| 21 |
+
- Rewards/chosen: -0.8599
|
| 22 |
+
- Rewards/rejected: -1.8698
|
| 23 |
+
- Rewards/accuracies: 0.7716
|
| 24 |
+
- Rewards/margins: 1.0099
|
| 25 |
+
- Logps/rejected: -406.0525
|
| 26 |
+
- Logps/chosen: -361.5919
|
| 27 |
+
- Logits/rejected: 1.8489
|
| 28 |
+
- Logits/chosen: 0.5727
|
| 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 |
+
| 0.6604 | 0.1143 | 50 | 0.6510 | -0.0701 | -0.2214 | 0.7026 | 0.1513 | -241.2204 | -282.6137 | -2.3977 | -2.4995 |
|
| 66 |
+
| 0.6061 | 0.2286 | 100 | 0.5910 | -0.1695 | -0.5958 | 0.7629 | 0.4263 | -278.6604 | -292.5544 | -2.1026 | -2.2566 |
|
| 67 |
+
| 0.5667 | 0.3429 | 150 | 0.5483 | -0.6705 | -1.4559 | 0.7543 | 0.7854 | -364.6649 | -342.6550 | 0.2660 | -0.5061 |
|
| 68 |
+
| 0.5462 | 0.4571 | 200 | 0.5362 | -0.8961 | -1.7783 | 0.7759 | 0.8822 | -396.9079 | -365.2170 | 1.8660 | 0.8687 |
|
| 69 |
+
| 0.5166 | 0.5714 | 250 | 0.5322 | -0.7659 | -1.7273 | 0.7543 | 0.9613 | -391.8038 | -352.1977 | 1.6137 | 0.4218 |
|
| 70 |
+
| 0.52 | 0.6857 | 300 | 0.5290 | -0.8312 | -1.8309 | 0.7672 | 0.9996 | -402.1612 | -358.7278 | 1.5709 | 0.3575 |
|
| 71 |
+
| 0.5242 | 0.8 | 350 | 0.5237 | -0.9809 | -1.9725 | 0.7759 | 0.9916 | -416.3300 | -373.6958 | 2.1920 | 0.9155 |
|
| 72 |
+
| 0.5207 | 0.9143 | 400 | 0.5204 | -0.8599 | -1.8698 | 0.7716 | 1.0099 | -406.0525 | -361.5919 | 1.8489 | 0.5727 |
|
| 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": 0.5601670529258606,
|
| 5 |
+
"train_runtime": 11363.2534,
|
| 6 |
+
"train_samples": 55999,
|
| 7 |
+
"train_samples_per_second": 4.928,
|
| 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": 0.5601670529258606,
|
| 5 |
+
"train_runtime": 11363.2534,
|
| 6 |
+
"train_samples": 55999,
|
| 7 |
+
"train_samples_per_second": 4.928,
|
| 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": 7.972490696774519,
|
| 14 |
+
"learning_rate": 1.1363636363636363e-07,
|
| 15 |
+
"logits/chosen": -2.6948084831237793,
|
| 16 |
+
"logits/rejected": -2.647209405899048,
|
| 17 |
+
"logps/chosen": -302.53485107421875,
|
| 18 |
+
"logps/rejected": -291.70001220703125,
|
| 19 |
+
"loss": 0.6931,
|
| 20 |
+
"rewards/accuracies": 0.42500001192092896,
|
| 21 |
+
"rewards/chosen": 0.00010526743426453322,
|
| 22 |
+
"rewards/margins": 0.00048500439152121544,
|
| 23 |
+
"rewards/rejected": -0.0003797370009124279,
|
| 24 |
+
"step": 10
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.045714285714285714,
|
| 28 |
+
"grad_norm": 7.853520968040958,
|
| 29 |
+
"learning_rate": 2.2727272727272726e-07,
|
| 30 |
+
"logits/chosen": -2.6835687160491943,
|
| 31 |
+
"logits/rejected": -2.6186060905456543,
|
| 32 |
+
"logps/chosen": -289.909912109375,
|
| 33 |
+
"logps/rejected": -251.9034881591797,
|
| 34 |
+
"loss": 0.6925,
|
| 35 |
+
"rewards/accuracies": 0.53125,
|
| 36 |
+
"rewards/chosen": 0.00011107837781310081,
|
| 37 |
+
"rewards/margins": 0.001589515246450901,
|
| 38 |
+
"rewards/rejected": -0.0014784367522224784,
|
| 39 |
+
"step": 20
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.06857142857142857,
|
| 43 |
+
"grad_norm": 13.072730042036099,
|
| 44 |
+
"learning_rate": 3.4090909090909085e-07,
|
| 45 |
+
"logits/chosen": -2.643810272216797,
|
| 46 |
+
"logits/rejected": -2.581580638885498,
|
| 47 |
+
"logps/chosen": -293.62652587890625,
|
| 48 |
+
"logps/rejected": -246.8993377685547,
|
| 49 |
+
"loss": 0.6879,
|
| 50 |
+
"rewards/accuracies": 0.731249988079071,
|
| 51 |
+
"rewards/chosen": 0.00874017458409071,
|
| 52 |
+
"rewards/margins": 0.013179494068026543,
|
| 53 |
+
"rewards/rejected": -0.004439320415258408,
|
| 54 |
+
"step": 30
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.09142857142857143,
|
| 58 |
+
"grad_norm": 8.816267790782506,
|
| 59 |
+
"learning_rate": 4.545454545454545e-07,
|
| 60 |
+
"logits/chosen": -2.569484233856201,
|
| 61 |
+
"logits/rejected": -2.48750638961792,
|
| 62 |
+
"logps/chosen": -258.9664611816406,
|
| 63 |
+
"logps/rejected": -232.96670532226562,
|
| 64 |
+
"loss": 0.6773,
|
| 65 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 66 |
+
"rewards/chosen": 0.032418955117464066,
|
| 67 |
+
"rewards/margins": 0.03510286286473274,
|
| 68 |
+
"rewards/rejected": -0.0026839044876396656,
|
| 69 |
+
"step": 40
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.11428571428571428,
|
| 73 |
+
"grad_norm": 8.526359525770381,
|
| 74 |
+
"learning_rate": 4.997124959943201e-07,
|
| 75 |
+
"logits/chosen": -2.4977941513061523,
|
| 76 |
+
"logits/rejected": -2.4374022483825684,
|
| 77 |
+
"logps/chosen": -270.9308166503906,
|
| 78 |
+
"logps/rejected": -256.2071838378906,
|
| 79 |
+
"loss": 0.6604,
|
| 80 |
+
"rewards/accuracies": 0.6875,
|
| 81 |
+
"rewards/chosen": -0.015730079263448715,
|
| 82 |
+
"rewards/margins": 0.07472670823335648,
|
| 83 |
+
"rewards/rejected": -0.09045679867267609,
|
| 84 |
+
"step": 50
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.11428571428571428,
|
| 88 |
+
"eval_logits/chosen": -2.4995484352111816,
|
| 89 |
+
"eval_logits/rejected": -2.397691488265991,
|
| 90 |
+
"eval_logps/chosen": -282.6136779785156,
|
| 91 |
+
"eval_logps/rejected": -241.22042846679688,
|
| 92 |
+
"eval_loss": 0.6509631276130676,
|
| 93 |
+
"eval_rewards/accuracies": 0.7025862336158752,
|
| 94 |
+
"eval_rewards/chosen": -0.07009723782539368,
|
| 95 |
+
"eval_rewards/margins": 0.15134572982788086,
|
| 96 |
+
"eval_rewards/rejected": -0.22144296765327454,
|
| 97 |
+
"eval_runtime": 92.1389,
|
| 98 |
+
"eval_samples_per_second": 19.872,
|
| 99 |
+
"eval_steps_per_second": 0.315,
|
| 100 |
+
"step": 50
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.13714285714285715,
|
| 104 |
+
"grad_norm": 10.81653462365135,
|
| 105 |
+
"learning_rate": 4.979579212164186e-07,
|
| 106 |
+
"logits/chosen": -2.5448544025421143,
|
| 107 |
+
"logits/rejected": -2.496579885482788,
|
| 108 |
+
"logps/chosen": -304.4873962402344,
|
| 109 |
+
"logps/rejected": -281.6524353027344,
|
| 110 |
+
"loss": 0.6418,
|
| 111 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 112 |
+
"rewards/chosen": -0.07611249387264252,
|
| 113 |
+
"rewards/margins": 0.13186688721179962,
|
| 114 |
+
"rewards/rejected": -0.20797939598560333,
|
| 115 |
+
"step": 60
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"epoch": 0.16,
|
| 119 |
+
"grad_norm": 14.155169380492403,
|
| 120 |
+
"learning_rate": 4.946196886175515e-07,
|
| 121 |
+
"logits/chosen": -2.5274157524108887,
|
| 122 |
+
"logits/rejected": -2.4085445404052734,
|
| 123 |
+
"logps/chosen": -319.63531494140625,
|
| 124 |
+
"logps/rejected": -284.4542541503906,
|
| 125 |
+
"loss": 0.6272,
|
| 126 |
+
"rewards/accuracies": 0.706250011920929,
|
| 127 |
+
"rewards/chosen": -0.15418021380901337,
|
| 128 |
+
"rewards/margins": 0.24344077706336975,
|
| 129 |
+
"rewards/rejected": -0.3976209759712219,
|
| 130 |
+
"step": 70
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 0.18285714285714286,
|
| 134 |
+
"grad_norm": 15.57036737948181,
|
| 135 |
+
"learning_rate": 4.897191188239667e-07,
|
| 136 |
+
"logits/chosen": -2.545555591583252,
|
| 137 |
+
"logits/rejected": -2.4691755771636963,
|
| 138 |
+
"logps/chosen": -302.2283935546875,
|
| 139 |
+
"logps/rejected": -310.9991149902344,
|
| 140 |
+
"loss": 0.6112,
|
| 141 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 142 |
+
"rewards/chosen": -0.242721289396286,
|
| 143 |
+
"rewards/margins": 0.20558595657348633,
|
| 144 |
+
"rewards/rejected": -0.44830721616744995,
|
| 145 |
+
"step": 80
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"epoch": 0.2057142857142857,
|
| 149 |
+
"grad_norm": 13.759558667047212,
|
| 150 |
+
"learning_rate": 4.832875107981763e-07,
|
| 151 |
+
"logits/chosen": -2.473137378692627,
|
| 152 |
+
"logits/rejected": -2.3435842990875244,
|
| 153 |
+
"logps/chosen": -291.57080078125,
|
| 154 |
+
"logps/rejected": -301.2342224121094,
|
| 155 |
+
"loss": 0.5962,
|
| 156 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 157 |
+
"rewards/chosen": -0.22334305942058563,
|
| 158 |
+
"rewards/margins": 0.335345596075058,
|
| 159 |
+
"rewards/rejected": -0.5586886405944824,
|
| 160 |
+
"step": 90
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 0.22857142857142856,
|
| 164 |
+
"grad_norm": 14.187129482186872,
|
| 165 |
+
"learning_rate": 4.753659419387223e-07,
|
| 166 |
+
"logits/chosen": -2.387033224105835,
|
| 167 |
+
"logits/rejected": -2.342754602432251,
|
| 168 |
+
"logps/chosen": -305.41363525390625,
|
| 169 |
+
"logps/rejected": -324.42626953125,
|
| 170 |
+
"loss": 0.6061,
|
| 171 |
+
"rewards/accuracies": 0.6875,
|
| 172 |
+
"rewards/chosen": -0.39434295892715454,
|
| 173 |
+
"rewards/margins": 0.3070516288280487,
|
| 174 |
+
"rewards/rejected": -0.7013946175575256,
|
| 175 |
+
"step": 100
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"epoch": 0.22857142857142856,
|
| 179 |
+
"eval_logits/chosen": -2.2565627098083496,
|
| 180 |
+
"eval_logits/rejected": -2.1026110649108887,
|
| 181 |
+
"eval_logps/chosen": -292.554443359375,
|
| 182 |
+
"eval_logps/rejected": -278.6604309082031,
|
| 183 |
+
"eval_loss": 0.5910175442695618,
|
| 184 |
+
"eval_rewards/accuracies": 0.7629310488700867,
|
| 185 |
+
"eval_rewards/chosen": -0.16950508952140808,
|
| 186 |
+
"eval_rewards/margins": 0.4263380169868469,
|
| 187 |
+
"eval_rewards/rejected": -0.5958431363105774,
|
| 188 |
+
"eval_runtime": 90.8579,
|
| 189 |
+
"eval_samples_per_second": 20.152,
|
| 190 |
+
"eval_steps_per_second": 0.319,
|
| 191 |
+
"step": 100
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.25142857142857145,
|
| 195 |
+
"grad_norm": 18.43194931012707,
|
| 196 |
+
"learning_rate": 4.660050057270191e-07,
|
| 197 |
+
"logits/chosen": -1.704791784286499,
|
| 198 |
+
"logits/rejected": -1.5881969928741455,
|
| 199 |
+
"logps/chosen": -346.3254699707031,
|
| 200 |
+
"logps/rejected": -369.74151611328125,
|
| 201 |
+
"loss": 0.5948,
|
| 202 |
+
"rewards/accuracies": 0.706250011920929,
|
| 203 |
+
"rewards/chosen": -0.3935784697532654,
|
| 204 |
+
"rewards/margins": 0.3640627861022949,
|
| 205 |
+
"rewards/rejected": -0.7576413154602051,
|
| 206 |
+
"step": 110
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.2742857142857143,
|
| 210 |
+
"grad_norm": 27.602915778458193,
|
| 211 |
+
"learning_rate": 4.5526448859687144e-07,
|
| 212 |
+
"logits/chosen": -0.836137592792511,
|
| 213 |
+
"logits/rejected": -0.3214734196662903,
|
| 214 |
+
"logps/chosen": -368.4119873046875,
|
| 215 |
+
"logps/rejected": -379.9471435546875,
|
| 216 |
+
"loss": 0.5674,
|
| 217 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 218 |
+
"rewards/chosen": -0.751334011554718,
|
| 219 |
+
"rewards/margins": 0.5190311670303345,
|
| 220 |
+
"rewards/rejected": -1.2703649997711182,
|
| 221 |
+
"step": 120
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.29714285714285715,
|
| 225 |
+
"grad_norm": 34.95862509810705,
|
| 226 |
+
"learning_rate": 4.432129880904388e-07,
|
| 227 |
+
"logits/chosen": -0.9399459958076477,
|
| 228 |
+
"logits/rejected": -0.630912184715271,
|
| 229 |
+
"logps/chosen": -333.36175537109375,
|
| 230 |
+
"logps/rejected": -364.6360778808594,
|
| 231 |
+
"loss": 0.5493,
|
| 232 |
+
"rewards/accuracies": 0.699999988079071,
|
| 233 |
+
"rewards/chosen": -0.7244097590446472,
|
| 234 |
+
"rewards/margins": 0.5231384038925171,
|
| 235 |
+
"rewards/rejected": -1.24754798412323,
|
| 236 |
+
"step": 130
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 0.32,
|
| 240 |
+
"grad_norm": 24.79721344090085,
|
| 241 |
+
"learning_rate": 4.299274747394055e-07,
|
| 242 |
+
"logits/chosen": -0.3657013773918152,
|
| 243 |
+
"logits/rejected": 0.007754585240036249,
|
| 244 |
+
"logps/chosen": -404.93963623046875,
|
| 245 |
+
"logps/rejected": -424.1830139160156,
|
| 246 |
+
"loss": 0.5551,
|
| 247 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 248 |
+
"rewards/chosen": -1.0948762893676758,
|
| 249 |
+
"rewards/margins": 0.4250301420688629,
|
| 250 |
+
"rewards/rejected": -1.5199064016342163,
|
| 251 |
+
"step": 140
|
| 252 |
+
},
|
| 253 |
+
{
|
| 254 |
+
"epoch": 0.34285714285714286,
|
| 255 |
+
"grad_norm": 23.423480181820228,
|
| 256 |
+
"learning_rate": 4.1549280046953653e-07,
|
| 257 |
+
"logits/chosen": -0.3730842173099518,
|
| 258 |
+
"logits/rejected": -0.07465118169784546,
|
| 259 |
+
"logps/chosen": -375.44451904296875,
|
| 260 |
+
"logps/rejected": -412.56964111328125,
|
| 261 |
+
"loss": 0.5667,
|
| 262 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 263 |
+
"rewards/chosen": -0.7702018022537231,
|
| 264 |
+
"rewards/margins": 0.5271822214126587,
|
| 265 |
+
"rewards/rejected": -1.2973840236663818,
|
| 266 |
+
"step": 150
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 0.34285714285714286,
|
| 270 |
+
"eval_logits/chosen": -0.5060973167419434,
|
| 271 |
+
"eval_logits/rejected": 0.26600033044815063,
|
| 272 |
+
"eval_logps/chosen": -342.6549987792969,
|
| 273 |
+
"eval_logps/rejected": -364.66485595703125,
|
| 274 |
+
"eval_loss": 0.5483084321022034,
|
| 275 |
+
"eval_rewards/accuracies": 0.7543103694915771,
|
| 276 |
+
"eval_rewards/chosen": -0.6705105304718018,
|
| 277 |
+
"eval_rewards/margins": 0.7853766083717346,
|
| 278 |
+
"eval_rewards/rejected": -1.4558870792388916,
|
| 279 |
+
"eval_runtime": 91.3845,
|
| 280 |
+
"eval_samples_per_second": 20.036,
|
| 281 |
+
"eval_steps_per_second": 0.317,
|
| 282 |
+
"step": 150
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 0.3657142857142857,
|
| 286 |
+
"grad_norm": 27.18613053303789,
|
| 287 |
+
"learning_rate": 4.000011566683401e-07,
|
| 288 |
+
"logits/chosen": -0.1914953887462616,
|
| 289 |
+
"logits/rejected": 0.2764665484428406,
|
| 290 |
+
"logps/chosen": -377.5716247558594,
|
| 291 |
+
"logps/rejected": -388.8448486328125,
|
| 292 |
+
"loss": 0.526,
|
| 293 |
+
"rewards/accuracies": 0.706250011920929,
|
| 294 |
+
"rewards/chosen": -0.9134766459465027,
|
| 295 |
+
"rewards/margins": 0.5826240181922913,
|
| 296 |
+
"rewards/rejected": -1.496100664138794,
|
| 297 |
+
"step": 160
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"epoch": 0.38857142857142857,
|
| 301 |
+
"grad_norm": 21.65805367669265,
|
| 302 |
+
"learning_rate": 3.8355148537705047e-07,
|
| 303 |
+
"logits/chosen": -0.4353760778903961,
|
| 304 |
+
"logits/rejected": 0.20893268287181854,
|
| 305 |
+
"logps/chosen": -373.71319580078125,
|
| 306 |
+
"logps/rejected": -383.9519958496094,
|
| 307 |
+
"loss": 0.5395,
|
| 308 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 309 |
+
"rewards/chosen": -0.7510371208190918,
|
| 310 |
+
"rewards/margins": 0.6987054347991943,
|
| 311 |
+
"rewards/rejected": -1.4497425556182861,
|
| 312 |
+
"step": 170
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 0.4114285714285714,
|
| 316 |
+
"grad_norm": 22.20737720981617,
|
| 317 |
+
"learning_rate": 3.662488473675315e-07,
|
| 318 |
+
"logits/chosen": -0.09447278827428818,
|
| 319 |
+
"logits/rejected": 0.7217932939529419,
|
| 320 |
+
"logps/chosen": -409.1649169921875,
|
| 321 |
+
"logps/rejected": -408.90557861328125,
|
| 322 |
+
"loss": 0.5495,
|
| 323 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 324 |
+
"rewards/chosen": -0.959947943687439,
|
| 325 |
+
"rewards/margins": 0.5610560774803162,
|
| 326 |
+
"rewards/rejected": -1.5210039615631104,
|
| 327 |
+
"step": 180
|
| 328 |
+
},
|
| 329 |
+
{
|
| 330 |
+
"epoch": 0.4342857142857143,
|
| 331 |
+
"grad_norm": 30.321606839159987,
|
| 332 |
+
"learning_rate": 3.48203751140067e-07,
|
| 333 |
+
"logits/chosen": 0.33530089259147644,
|
| 334 |
+
"logits/rejected": 0.8107538223266602,
|
| 335 |
+
"logps/chosen": -362.1139221191406,
|
| 336 |
+
"logps/rejected": -387.8373107910156,
|
| 337 |
+
"loss": 0.5503,
|
| 338 |
+
"rewards/accuracies": 0.668749988079071,
|
| 339 |
+
"rewards/chosen": -0.8776267766952515,
|
| 340 |
+
"rewards/margins": 0.4715437889099121,
|
| 341 |
+
"rewards/rejected": -1.3491705656051636,
|
| 342 |
+
"step": 190
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 0.45714285714285713,
|
| 346 |
+
"grad_norm": 25.0679630404246,
|
| 347 |
+
"learning_rate": 3.2953144712759537e-07,
|
| 348 |
+
"logits/chosen": 0.8955745697021484,
|
| 349 |
+
"logits/rejected": 1.5546410083770752,
|
| 350 |
+
"logps/chosen": -390.03497314453125,
|
| 351 |
+
"logps/rejected": -439.87127685546875,
|
| 352 |
+
"loss": 0.5462,
|
| 353 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 354 |
+
"rewards/chosen": -0.9517244100570679,
|
| 355 |
+
"rewards/margins": 0.6701112985610962,
|
| 356 |
+
"rewards/rejected": -1.621835708618164,
|
| 357 |
+
"step": 200
|
| 358 |
+
},
|
| 359 |
+
{
|
| 360 |
+
"epoch": 0.45714285714285713,
|
| 361 |
+
"eval_logits/chosen": 0.8686795234680176,
|
| 362 |
+
"eval_logits/rejected": 1.8659789562225342,
|
| 363 |
+
"eval_logps/chosen": -365.217041015625,
|
| 364 |
+
"eval_logps/rejected": -396.9079284667969,
|
| 365 |
+
"eval_loss": 0.5362085103988647,
|
| 366 |
+
"eval_rewards/accuracies": 0.7758620977401733,
|
| 367 |
+
"eval_rewards/chosen": -0.8961309194564819,
|
| 368 |
+
"eval_rewards/margins": 0.8821869492530823,
|
| 369 |
+
"eval_rewards/rejected": -1.7783178091049194,
|
| 370 |
+
"eval_runtime": 92.8978,
|
| 371 |
+
"eval_samples_per_second": 19.71,
|
| 372 |
+
"eval_steps_per_second": 0.312,
|
| 373 |
+
"step": 200
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"epoch": 0.48,
|
| 377 |
+
"grad_norm": 23.925735012710426,
|
| 378 |
+
"learning_rate": 3.103511916141658e-07,
|
| 379 |
+
"logits/chosen": 1.040865182876587,
|
| 380 |
+
"logits/rejected": 1.5942493677139282,
|
| 381 |
+
"logps/chosen": -361.822509765625,
|
| 382 |
+
"logps/rejected": -391.01483154296875,
|
| 383 |
+
"loss": 0.5463,
|
| 384 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 385 |
+
"rewards/chosen": -0.948311984539032,
|
| 386 |
+
"rewards/margins": 0.5272454619407654,
|
| 387 |
+
"rewards/rejected": -1.475557565689087,
|
| 388 |
+
"step": 210
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 0.5028571428571429,
|
| 392 |
+
"grad_norm": 27.94400965826576,
|
| 393 |
+
"learning_rate": 2.9078548506882117e-07,
|
| 394 |
+
"logits/chosen": 0.28927069902420044,
|
| 395 |
+
"logits/rejected": 1.0230720043182373,
|
| 396 |
+
"logps/chosen": -370.9271545410156,
|
| 397 |
+
"logps/rejected": -401.31414794921875,
|
| 398 |
+
"loss": 0.5362,
|
| 399 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 400 |
+
"rewards/chosen": -0.8033777475357056,
|
| 401 |
+
"rewards/margins": 0.7804895639419556,
|
| 402 |
+
"rewards/rejected": -1.5838673114776611,
|
| 403 |
+
"step": 220
|
| 404 |
+
},
|
| 405 |
+
{
|
| 406 |
+
"epoch": 0.5257142857142857,
|
| 407 |
+
"grad_norm": 21.142419258207983,
|
| 408 |
+
"learning_rate": 2.709592897595191e-07,
|
| 409 |
+
"logits/chosen": 0.07398202270269394,
|
| 410 |
+
"logits/rejected": 0.4865712523460388,
|
| 411 |
+
"logps/chosen": -304.87457275390625,
|
| 412 |
+
"logps/rejected": -344.7480163574219,
|
| 413 |
+
"loss": 0.5399,
|
| 414 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 415 |
+
"rewards/chosen": -0.6992014050483704,
|
| 416 |
+
"rewards/margins": 0.4801247715950012,
|
| 417 |
+
"rewards/rejected": -1.1793261766433716,
|
| 418 |
+
"step": 230
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"epoch": 0.5485714285714286,
|
| 422 |
+
"grad_norm": 19.343230158857015,
|
| 423 |
+
"learning_rate": 2.509992316440332e-07,
|
| 424 |
+
"logits/chosen": 0.2862880229949951,
|
| 425 |
+
"logits/rejected": 0.9607356190681458,
|
| 426 |
+
"logps/chosen": -330.4361877441406,
|
| 427 |
+
"logps/rejected": -371.2057189941406,
|
| 428 |
+
"loss": 0.5228,
|
| 429 |
+
"rewards/accuracies": 0.6875,
|
| 430 |
+
"rewards/chosen": -0.8042975664138794,
|
| 431 |
+
"rewards/margins": 0.6706650853157043,
|
| 432 |
+
"rewards/rejected": -1.4749624729156494,
|
| 433 |
+
"step": 240
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 0.5714285714285714,
|
| 437 |
+
"grad_norm": 35.596035717324796,
|
| 438 |
+
"learning_rate": 2.3103279163519918e-07,
|
| 439 |
+
"logits/chosen": 0.43575239181518555,
|
| 440 |
+
"logits/rejected": 1.2122005224227905,
|
| 441 |
+
"logps/chosen": -406.7032775878906,
|
| 442 |
+
"logps/rejected": -422.3251037597656,
|
| 443 |
+
"loss": 0.5166,
|
| 444 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 445 |
+
"rewards/chosen": -0.8864845037460327,
|
| 446 |
+
"rewards/margins": 0.6702730059623718,
|
| 447 |
+
"rewards/rejected": -1.5567573308944702,
|
| 448 |
+
"step": 250
|
| 449 |
+
},
|
| 450 |
+
{
|
| 451 |
+
"epoch": 0.5714285714285714,
|
| 452 |
+
"eval_logits/chosen": 0.42181965708732605,
|
| 453 |
+
"eval_logits/rejected": 1.6137125492095947,
|
| 454 |
+
"eval_logps/chosen": -352.1977233886719,
|
| 455 |
+
"eval_logps/rejected": -391.8038330078125,
|
| 456 |
+
"eval_loss": 0.5322425961494446,
|
| 457 |
+
"eval_rewards/accuracies": 0.7543103694915771,
|
| 458 |
+
"eval_rewards/chosen": -0.7659374475479126,
|
| 459 |
+
"eval_rewards/margins": 0.9613397121429443,
|
| 460 |
+
"eval_rewards/rejected": -1.7272770404815674,
|
| 461 |
+
"eval_runtime": 92.7968,
|
| 462 |
+
"eval_samples_per_second": 19.731,
|
| 463 |
+
"eval_steps_per_second": 0.313,
|
| 464 |
+
"step": 250
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 0.5942857142857143,
|
| 468 |
+
"grad_norm": 23.35941549954683,
|
| 469 |
+
"learning_rate": 2.1118749140573358e-07,
|
| 470 |
+
"logits/chosen": 0.34059277176856995,
|
| 471 |
+
"logits/rejected": 1.041529893875122,
|
| 472 |
+
"logps/chosen": -385.17755126953125,
|
| 473 |
+
"logps/rejected": -448.1366271972656,
|
| 474 |
+
"loss": 0.5247,
|
| 475 |
+
"rewards/accuracies": 0.800000011920929,
|
| 476 |
+
"rewards/chosen": -0.8077054023742676,
|
| 477 |
+
"rewards/margins": 0.8099581599235535,
|
| 478 |
+
"rewards/rejected": -1.6176636219024658,
|
| 479 |
+
"step": 260
|
| 480 |
+
},
|
| 481 |
+
{
|
| 482 |
+
"epoch": 0.6171428571428571,
|
| 483 |
+
"grad_norm": 20.969034525593848,
|
| 484 |
+
"learning_rate": 1.9159007893272703e-07,
|
| 485 |
+
"logits/chosen": 0.7277138829231262,
|
| 486 |
+
"logits/rejected": 1.7486486434936523,
|
| 487 |
+
"logps/chosen": -373.39337158203125,
|
| 488 |
+
"logps/rejected": -425.66802978515625,
|
| 489 |
+
"loss": 0.5179,
|
| 490 |
+
"rewards/accuracies": 0.800000011920929,
|
| 491 |
+
"rewards/chosen": -0.9260489344596863,
|
| 492 |
+
"rewards/margins": 0.8373041152954102,
|
| 493 |
+
"rewards/rejected": -1.7633531093597412,
|
| 494 |
+
"step": 270
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 0.64,
|
| 498 |
+
"grad_norm": 23.247855138801246,
|
| 499 |
+
"learning_rate": 1.7236571898357766e-07,
|
| 500 |
+
"logits/chosen": 1.1543443202972412,
|
| 501 |
+
"logits/rejected": 1.953770637512207,
|
| 502 |
+
"logps/chosen": -383.1849365234375,
|
| 503 |
+
"logps/rejected": -455.0911560058594,
|
| 504 |
+
"loss": 0.5114,
|
| 505 |
+
"rewards/accuracies": 0.737500011920929,
|
| 506 |
+
"rewards/chosen": -1.180774450302124,
|
| 507 |
+
"rewards/margins": 0.7801226377487183,
|
| 508 |
+
"rewards/rejected": -1.9608970880508423,
|
| 509 |
+
"step": 280
|
| 510 |
+
},
|
| 511 |
+
{
|
| 512 |
+
"epoch": 0.6628571428571428,
|
| 513 |
+
"grad_norm": 24.724486484729947,
|
| 514 |
+
"learning_rate": 1.5363719371356882e-07,
|
| 515 |
+
"logits/chosen": 0.6572316884994507,
|
| 516 |
+
"logits/rejected": 1.695812463760376,
|
| 517 |
+
"logps/chosen": -416.1396484375,
|
| 518 |
+
"logps/rejected": -455.5172424316406,
|
| 519 |
+
"loss": 0.5263,
|
| 520 |
+
"rewards/accuracies": 0.75,
|
| 521 |
+
"rewards/chosen": -1.0608103275299072,
|
| 522 |
+
"rewards/margins": 0.8319962620735168,
|
| 523 |
+
"rewards/rejected": -1.8928064107894897,
|
| 524 |
+
"step": 290
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"epoch": 0.6857142857142857,
|
| 528 |
+
"grad_norm": 29.971468773313187,
|
| 529 |
+
"learning_rate": 1.3552411848071565e-07,
|
| 530 |
+
"logits/chosen": 0.6122870445251465,
|
| 531 |
+
"logits/rejected": 1.7082029581069946,
|
| 532 |
+
"logps/chosen": -430.2732849121094,
|
| 533 |
+
"logps/rejected": -462.9169006347656,
|
| 534 |
+
"loss": 0.52,
|
| 535 |
+
"rewards/accuracies": 0.71875,
|
| 536 |
+
"rewards/chosen": -0.9619089961051941,
|
| 537 |
+
"rewards/margins": 0.8669039607048035,
|
| 538 |
+
"rewards/rejected": -1.8288129568099976,
|
| 539 |
+
"step": 300
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 0.6857142857142857,
|
| 543 |
+
"eval_logits/chosen": 0.35753339529037476,
|
| 544 |
+
"eval_logits/rejected": 1.570934534072876,
|
| 545 |
+
"eval_logps/chosen": -358.7277526855469,
|
| 546 |
+
"eval_logps/rejected": -402.1611633300781,
|
| 547 |
+
"eval_loss": 0.5289955139160156,
|
| 548 |
+
"eval_rewards/accuracies": 0.767241358757019,
|
| 549 |
+
"eval_rewards/chosen": -0.8312381505966187,
|
| 550 |
+
"eval_rewards/margins": 0.9996121525764465,
|
| 551 |
+
"eval_rewards/rejected": -1.8308502435684204,
|
| 552 |
+
"eval_runtime": 91.4265,
|
| 553 |
+
"eval_samples_per_second": 20.027,
|
| 554 |
+
"eval_steps_per_second": 0.317,
|
| 555 |
+
"step": 300
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 0.7085714285714285,
|
| 559 |
+
"grad_norm": 24.12999361369321,
|
| 560 |
+
"learning_rate": 1.1814217788631473e-07,
|
| 561 |
+
"logits/chosen": -0.16139774024486542,
|
| 562 |
+
"logits/rejected": 0.688940167427063,
|
| 563 |
+
"logps/chosen": -382.5923767089844,
|
| 564 |
+
"logps/rejected": -401.3270568847656,
|
| 565 |
+
"loss": 0.5354,
|
| 566 |
+
"rewards/accuracies": 0.75,
|
| 567 |
+
"rewards/chosen": -0.7693190574645996,
|
| 568 |
+
"rewards/margins": 0.7588919997215271,
|
| 569 |
+
"rewards/rejected": -1.528210997581482,
|
| 570 |
+
"step": 310
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"epoch": 0.7314285714285714,
|
| 574 |
+
"grad_norm": 26.54925451004345,
|
| 575 |
+
"learning_rate": 1.0160238692045331e-07,
|
| 576 |
+
"logits/chosen": -0.06717093288898468,
|
| 577 |
+
"logits/rejected": 0.6305860280990601,
|
| 578 |
+
"logps/chosen": -369.2928161621094,
|
| 579 |
+
"logps/rejected": -412.58697509765625,
|
| 580 |
+
"loss": 0.5477,
|
| 581 |
+
"rewards/accuracies": 0.737500011920929,
|
| 582 |
+
"rewards/chosen": -0.7303933501243591,
|
| 583 |
+
"rewards/margins": 0.7106696367263794,
|
| 584 |
+
"rewards/rejected": -1.4410629272460938,
|
| 585 |
+
"step": 320
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 0.7542857142857143,
|
| 589 |
+
"grad_norm": 24.915758947808765,
|
| 590 |
+
"learning_rate": 8.601038193139438e-08,
|
| 591 |
+
"logits/chosen": 0.2589293122291565,
|
| 592 |
+
"logits/rejected": 1.0479013919830322,
|
| 593 |
+
"logps/chosen": -345.1177673339844,
|
| 594 |
+
"logps/rejected": -404.4964294433594,
|
| 595 |
+
"loss": 0.5268,
|
| 596 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 597 |
+
"rewards/chosen": -0.8461502194404602,
|
| 598 |
+
"rewards/margins": 0.694037914276123,
|
| 599 |
+
"rewards/rejected": -1.540188193321228,
|
| 600 |
+
"step": 330
|
| 601 |
+
},
|
| 602 |
+
{
|
| 603 |
+
"epoch": 0.7771428571428571,
|
| 604 |
+
"grad_norm": 25.634054231200437,
|
| 605 |
+
"learning_rate": 7.146574594727572e-08,
|
| 606 |
+
"logits/chosen": 0.4899168908596039,
|
| 607 |
+
"logits/rejected": 1.3734686374664307,
|
| 608 |
+
"logps/chosen": -405.3310546875,
|
| 609 |
+
"logps/rejected": -441.66229248046875,
|
| 610 |
+
"loss": 0.5244,
|
| 611 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 612 |
+
"rewards/chosen": -0.8923100233078003,
|
| 613 |
+
"rewards/margins": 0.7486563920974731,
|
| 614 |
+
"rewards/rejected": -1.6409664154052734,
|
| 615 |
+
"step": 340
|
| 616 |
+
},
|
| 617 |
+
{
|
| 618 |
+
"epoch": 0.8,
|
| 619 |
+
"grad_norm": 22.913432536942167,
|
| 620 |
+
"learning_rate": 5.8061372659157306e-08,
|
| 621 |
+
"logits/chosen": 0.6021924614906311,
|
| 622 |
+
"logits/rejected": 1.459370732307434,
|
| 623 |
+
"logps/chosen": -393.2467041015625,
|
| 624 |
+
"logps/rejected": -454.7303161621094,
|
| 625 |
+
"loss": 0.5242,
|
| 626 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 627 |
+
"rewards/chosen": -1.0371372699737549,
|
| 628 |
+
"rewards/margins": 0.7981128096580505,
|
| 629 |
+
"rewards/rejected": -1.8352502584457397,
|
| 630 |
+
"step": 350
|
| 631 |
+
},
|
| 632 |
+
{
|
| 633 |
+
"epoch": 0.8,
|
| 634 |
+
"eval_logits/chosen": 0.9154686331748962,
|
| 635 |
+
"eval_logits/rejected": 2.1920182704925537,
|
| 636 |
+
"eval_logps/chosen": -373.6957702636719,
|
| 637 |
+
"eval_logps/rejected": -416.3299560546875,
|
| 638 |
+
"eval_loss": 0.5237171053886414,
|
| 639 |
+
"eval_rewards/accuracies": 0.7758620977401733,
|
| 640 |
+
"eval_rewards/chosen": -0.9809183478355408,
|
| 641 |
+
"eval_rewards/margins": 0.9916198253631592,
|
| 642 |
+
"eval_rewards/rejected": -1.9725382328033447,
|
| 643 |
+
"eval_runtime": 90.3705,
|
| 644 |
+
"eval_samples_per_second": 20.261,
|
| 645 |
+
"eval_steps_per_second": 0.321,
|
| 646 |
+
"step": 350
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"epoch": 0.8228571428571428,
|
| 650 |
+
"grad_norm": 21.470834511474877,
|
| 651 |
+
"learning_rate": 4.5882873127531614e-08,
|
| 652 |
+
"logits/chosen": 1.230400562286377,
|
| 653 |
+
"logits/rejected": 1.8857431411743164,
|
| 654 |
+
"logps/chosen": -349.6596984863281,
|
| 655 |
+
"logps/rejected": -409.2167663574219,
|
| 656 |
+
"loss": 0.5193,
|
| 657 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 658 |
+
"rewards/chosen": -1.1211602687835693,
|
| 659 |
+
"rewards/margins": 0.669862687587738,
|
| 660 |
+
"rewards/rejected": -1.7910230159759521,
|
| 661 |
+
"step": 360
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 0.8457142857142858,
|
| 665 |
+
"grad_norm": 29.520887318743274,
|
| 666 |
+
"learning_rate": 3.500802900154412e-08,
|
| 667 |
+
"logits/chosen": 0.82422935962677,
|
| 668 |
+
"logits/rejected": 1.839655876159668,
|
| 669 |
+
"logps/chosen": -406.80926513671875,
|
| 670 |
+
"logps/rejected": -428.79949951171875,
|
| 671 |
+
"loss": 0.5243,
|
| 672 |
+
"rewards/accuracies": 0.71875,
|
| 673 |
+
"rewards/chosen": -1.1053653955459595,
|
| 674 |
+
"rewards/margins": 0.7291324734687805,
|
| 675 |
+
"rewards/rejected": -1.8344980478286743,
|
| 676 |
+
"step": 370
|
| 677 |
+
},
|
| 678 |
+
{
|
| 679 |
+
"epoch": 0.8685714285714285,
|
| 680 |
+
"grad_norm": 24.74940690379961,
|
| 681 |
+
"learning_rate": 2.550629574310309e-08,
|
| 682 |
+
"logits/chosen": 0.5695315599441528,
|
| 683 |
+
"logits/rejected": 1.3380911350250244,
|
| 684 |
+
"logps/chosen": -394.51263427734375,
|
| 685 |
+
"logps/rejected": -444.4217224121094,
|
| 686 |
+
"loss": 0.5117,
|
| 687 |
+
"rewards/accuracies": 0.75,
|
| 688 |
+
"rewards/chosen": -1.0350737571716309,
|
| 689 |
+
"rewards/margins": 0.7248605489730835,
|
| 690 |
+
"rewards/rejected": -1.759934425354004,
|
| 691 |
+
"step": 380
|
| 692 |
+
},
|
| 693 |
+
{
|
| 694 |
+
"epoch": 0.8914285714285715,
|
| 695 |
+
"grad_norm": 22.68029569887221,
|
| 696 |
+
"learning_rate": 1.7438359028687983e-08,
|
| 697 |
+
"logits/chosen": 0.7584401965141296,
|
| 698 |
+
"logits/rejected": 1.7040205001831055,
|
| 699 |
+
"logps/chosen": -372.4088439941406,
|
| 700 |
+
"logps/rejected": -406.352783203125,
|
| 701 |
+
"loss": 0.513,
|
| 702 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 703 |
+
"rewards/chosen": -1.0060627460479736,
|
| 704 |
+
"rewards/margins": 0.7981821298599243,
|
| 705 |
+
"rewards/rejected": -1.8042447566986084,
|
| 706 |
+
"step": 390
|
| 707 |
+
},
|
| 708 |
+
{
|
| 709 |
+
"epoch": 0.9142857142857143,
|
| 710 |
+
"grad_norm": 28.690684821171704,
|
| 711 |
+
"learning_rate": 1.0855747162029361e-08,
|
| 712 |
+
"logits/chosen": 0.6766539812088013,
|
| 713 |
+
"logits/rejected": 1.2123651504516602,
|
| 714 |
+
"logps/chosen": -355.7047119140625,
|
| 715 |
+
"logps/rejected": -403.9508361816406,
|
| 716 |
+
"loss": 0.5207,
|
| 717 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 718 |
+
"rewards/chosen": -0.8698047399520874,
|
| 719 |
+
"rewards/margins": 0.698576807975769,
|
| 720 |
+
"rewards/rejected": -1.568381428718567,
|
| 721 |
+
"step": 400
|
| 722 |
+
},
|
| 723 |
+
{
|
| 724 |
+
"epoch": 0.9142857142857143,
|
| 725 |
+
"eval_logits/chosen": 0.5726661682128906,
|
| 726 |
+
"eval_logits/rejected": 1.8489381074905396,
|
| 727 |
+
"eval_logps/chosen": -361.5919189453125,
|
| 728 |
+
"eval_logps/rejected": -406.0525207519531,
|
| 729 |
+
"eval_loss": 0.5204084515571594,
|
| 730 |
+
"eval_rewards/accuracies": 0.7715517282485962,
|
| 731 |
+
"eval_rewards/chosen": -0.8598796129226685,
|
| 732 |
+
"eval_rewards/margins": 1.0098847150802612,
|
| 733 |
+
"eval_rewards/rejected": -1.8697644472122192,
|
| 734 |
+
"eval_runtime": 90.7052,
|
| 735 |
+
"eval_samples_per_second": 20.186,
|
| 736 |
+
"eval_steps_per_second": 0.32,
|
| 737 |
+
"step": 400
|
| 738 |
+
},
|
| 739 |
+
{
|
| 740 |
+
"epoch": 0.9371428571428572,
|
| 741 |
+
"grad_norm": 22.4897885356645,
|
| 742 |
+
"learning_rate": 5.8005019731033615e-09,
|
| 743 |
+
"logits/chosen": 0.5689116716384888,
|
| 744 |
+
"logits/rejected": 1.7866685390472412,
|
| 745 |
+
"logps/chosen": -360.1455383300781,
|
| 746 |
+
"logps/rejected": -413.405029296875,
|
| 747 |
+
"loss": 0.4922,
|
| 748 |
+
"rewards/accuracies": 0.78125,
|
| 749 |
+
"rewards/chosen": -0.8828630447387695,
|
| 750 |
+
"rewards/margins": 0.9393178820610046,
|
| 751 |
+
"rewards/rejected": -1.8221811056137085,
|
| 752 |
+
"step": 410
|
| 753 |
+
},
|
| 754 |
+
{
|
| 755 |
+
"epoch": 0.96,
|
| 756 |
+
"grad_norm": 24.590466446146614,
|
| 757 |
+
"learning_rate": 2.3049103053431886e-09,
|
| 758 |
+
"logits/chosen": 0.4758734107017517,
|
| 759 |
+
"logits/rejected": 1.4166442155838013,
|
| 760 |
+
"logps/chosen": -368.29669189453125,
|
| 761 |
+
"logps/rejected": -418.43438720703125,
|
| 762 |
+
"loss": 0.5034,
|
| 763 |
+
"rewards/accuracies": 0.75,
|
| 764 |
+
"rewards/chosen": -0.8787800669670105,
|
| 765 |
+
"rewards/margins": 0.8519768714904785,
|
| 766 |
+
"rewards/rejected": -1.7307569980621338,
|
| 767 |
+
"step": 420
|
| 768 |
+
},
|
| 769 |
+
{
|
| 770 |
+
"epoch": 0.9828571428571429,
|
| 771 |
+
"grad_norm": 29.700043608251338,
|
| 772 |
+
"learning_rate": 3.9129780600541397e-10,
|
| 773 |
+
"logits/chosen": 0.4247834086418152,
|
| 774 |
+
"logits/rejected": 1.170922040939331,
|
| 775 |
+
"logps/chosen": -397.17230224609375,
|
| 776 |
+
"logps/rejected": -453.1036682128906,
|
| 777 |
+
"loss": 0.5236,
|
| 778 |
+
"rewards/accuracies": 0.706250011920929,
|
| 779 |
+
"rewards/chosen": -0.9241801500320435,
|
| 780 |
+
"rewards/margins": 0.7625172138214111,
|
| 781 |
+
"rewards/rejected": -1.6866973638534546,
|
| 782 |
+
"step": 430
|
| 783 |
+
},
|
| 784 |
+
{
|
| 785 |
+
"epoch": 0.9988571428571429,
|
| 786 |
+
"step": 437,
|
| 787 |
+
"total_flos": 0.0,
|
| 788 |
+
"train_loss": 0.5601670529258606,
|
| 789 |
+
"train_runtime": 11363.2534,
|
| 790 |
+
"train_samples_per_second": 4.928,
|
| 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 |
+
}
|