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 +800 -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-ultrabin-low-margin-3-epochs
|
| 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-ultrabin-low-margin-3-epochs
|
| 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 an unknown dataset.
|
| 19 |
+
It achieves the following results on the evaluation set:
|
| 20 |
+
- Loss: 0.6813
|
| 21 |
+
- Rewards/chosen: -1.9317
|
| 22 |
+
- Rewards/rejected: -2.2848
|
| 23 |
+
- Rewards/accuracies: 0.6797
|
| 24 |
+
- Rewards/margins: 0.3531
|
| 25 |
+
- Logps/rejected: -491.1398
|
| 26 |
+
- Logps/chosen: -455.8003
|
| 27 |
+
- Logits/rejected: -0.1808
|
| 28 |
+
- Logits/chosen: -0.3104
|
| 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: 3
|
| 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 |
+
| 0.6871 | 0.3484 | 50 | 0.6798 | -0.0015 | -0.0300 | 0.5977 | 0.0285 | -265.6574 | -262.7750 | -2.5993 | -2.6328 |
|
| 66 |
+
| 0.6724 | 0.6969 | 100 | 0.6721 | -0.0365 | -0.0949 | 0.5938 | 0.0584 | -272.1548 | -266.2806 | -2.4994 | -2.5340 |
|
| 67 |
+
| 0.6047 | 1.0453 | 150 | 0.6797 | -0.1660 | -0.2332 | 0.5898 | 0.0673 | -285.9855 | -279.2270 | -2.5025 | -2.5443 |
|
| 68 |
+
| 0.5265 | 1.3937 | 200 | 0.6762 | -0.5743 | -0.7331 | 0.6719 | 0.1588 | -335.9708 | -320.0576 | -2.2718 | -2.3328 |
|
| 69 |
+
| 0.4984 | 1.7422 | 250 | 0.6732 | -1.2121 | -1.4445 | 0.6562 | 0.2325 | -407.1154 | -383.8381 | -1.4451 | -1.5433 |
|
| 70 |
+
| 0.3569 | 2.0906 | 300 | 0.6527 | -1.3455 | -1.6681 | 0.6758 | 0.3226 | -429.4680 | -397.1805 | -0.8708 | -0.9999 |
|
| 71 |
+
| 0.3329 | 2.4390 | 350 | 0.6840 | -1.9045 | -2.2570 | 0.6602 | 0.3525 | -488.3670 | -453.0816 | -0.1084 | -0.2447 |
|
| 72 |
+
| 0.3368 | 2.7875 | 400 | 0.6813 | -1.9317 | -2.2848 | 0.6797 | 0.3531 | -491.1398 | -455.8003 | -0.1808 | -0.3104 |
|
| 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": 2.989547038327526,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5141507484418251,
|
| 5 |
+
"train_runtime": 11380.2127,
|
| 6 |
+
"train_samples": 18340,
|
| 7 |
+
"train_samples_per_second": 4.835,
|
| 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": 2.989547038327526,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5141507484418251,
|
| 5 |
+
"train_runtime": 11380.2127,
|
| 6 |
+
"train_samples": 18340,
|
| 7 |
+
"train_samples_per_second": 4.835,
|
| 8 |
+
"train_steps_per_second": 0.038
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,800 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 2.989547038327526,
|
| 5 |
+
"eval_steps": 50,
|
| 6 |
+
"global_step": 429,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.06968641114982578,
|
| 13 |
+
"grad_norm": 8.229574175695925,
|
| 14 |
+
"learning_rate": 1.1627906976744186e-07,
|
| 15 |
+
"logits/chosen": -2.7034733295440674,
|
| 16 |
+
"logits/rejected": -2.7302405834198,
|
| 17 |
+
"logps/chosen": -301.81427001953125,
|
| 18 |
+
"logps/rejected": -331.369140625,
|
| 19 |
+
"loss": 0.6932,
|
| 20 |
+
"rewards/accuracies": 0.4375,
|
| 21 |
+
"rewards/chosen": 0.0005243912455625832,
|
| 22 |
+
"rewards/margins": 0.0006931144162081182,
|
| 23 |
+
"rewards/rejected": -0.00016872311243787408,
|
| 24 |
+
"step": 10
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.13937282229965156,
|
| 28 |
+
"grad_norm": 7.845336267328567,
|
| 29 |
+
"learning_rate": 2.3255813953488372e-07,
|
| 30 |
+
"logits/chosen": -2.762924909591675,
|
| 31 |
+
"logits/rejected": -2.7517170906066895,
|
| 32 |
+
"logps/chosen": -320.4908752441406,
|
| 33 |
+
"logps/rejected": -314.0067138671875,
|
| 34 |
+
"loss": 0.6931,
|
| 35 |
+
"rewards/accuracies": 0.4625000059604645,
|
| 36 |
+
"rewards/chosen": 0.00030491696088574827,
|
| 37 |
+
"rewards/margins": -0.00010313927487004548,
|
| 38 |
+
"rewards/rejected": 0.00040805633761920035,
|
| 39 |
+
"step": 20
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.20905923344947736,
|
| 43 |
+
"grad_norm": 9.709940066560216,
|
| 44 |
+
"learning_rate": 3.4883720930232557e-07,
|
| 45 |
+
"logits/chosen": -2.702960252761841,
|
| 46 |
+
"logits/rejected": -2.6881918907165527,
|
| 47 |
+
"logps/chosen": -295.65521240234375,
|
| 48 |
+
"logps/rejected": -309.1160888671875,
|
| 49 |
+
"loss": 0.6926,
|
| 50 |
+
"rewards/accuracies": 0.5562499761581421,
|
| 51 |
+
"rewards/chosen": 0.005805189721286297,
|
| 52 |
+
"rewards/margins": -1.923509444168303e-05,
|
| 53 |
+
"rewards/rejected": 0.005824424792081118,
|
| 54 |
+
"step": 30
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.2787456445993031,
|
| 58 |
+
"grad_norm": 7.6570471523786505,
|
| 59 |
+
"learning_rate": 4.6511627906976743e-07,
|
| 60 |
+
"logits/chosen": -2.6843132972717285,
|
| 61 |
+
"logits/rejected": -2.6987385749816895,
|
| 62 |
+
"logps/chosen": -289.13720703125,
|
| 63 |
+
"logps/rejected": -296.2955322265625,
|
| 64 |
+
"loss": 0.6902,
|
| 65 |
+
"rewards/accuracies": 0.5562499761581421,
|
| 66 |
+
"rewards/chosen": 0.010863080620765686,
|
| 67 |
+
"rewards/margins": 0.004013759084045887,
|
| 68 |
+
"rewards/rejected": 0.006849322468042374,
|
| 69 |
+
"step": 40
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.34843205574912894,
|
| 73 |
+
"grad_norm": 7.937140028593177,
|
| 74 |
+
"learning_rate": 4.995943852340362e-07,
|
| 75 |
+
"logits/chosen": -2.6573901176452637,
|
| 76 |
+
"logits/rejected": -2.66682767868042,
|
| 77 |
+
"logps/chosen": -327.9737854003906,
|
| 78 |
+
"logps/rejected": -309.1617431640625,
|
| 79 |
+
"loss": 0.6871,
|
| 80 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 81 |
+
"rewards/chosen": 0.006907278206199408,
|
| 82 |
+
"rewards/margins": 0.010998749174177647,
|
| 83 |
+
"rewards/rejected": -0.004091470967978239,
|
| 84 |
+
"step": 50
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.34843205574912894,
|
| 88 |
+
"eval_logits/chosen": -2.632770538330078,
|
| 89 |
+
"eval_logits/rejected": -2.5992560386657715,
|
| 90 |
+
"eval_logps/chosen": -262.7750244140625,
|
| 91 |
+
"eval_logps/rejected": -265.65740966796875,
|
| 92 |
+
"eval_loss": 0.6798189878463745,
|
| 93 |
+
"eval_rewards/accuracies": 0.59765625,
|
| 94 |
+
"eval_rewards/chosen": -0.001450682058930397,
|
| 95 |
+
"eval_rewards/margins": 0.02850232645869255,
|
| 96 |
+
"eval_rewards/rejected": -0.02995300479233265,
|
| 97 |
+
"eval_runtime": 103.8301,
|
| 98 |
+
"eval_samples_per_second": 19.262,
|
| 99 |
+
"eval_steps_per_second": 0.308,
|
| 100 |
+
"step": 50
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.4181184668989547,
|
| 104 |
+
"grad_norm": 9.169375441320238,
|
| 105 |
+
"learning_rate": 4.976108685115826e-07,
|
| 106 |
+
"logits/chosen": -2.6978957653045654,
|
| 107 |
+
"logits/rejected": -2.672510862350464,
|
| 108 |
+
"logps/chosen": -300.9576110839844,
|
| 109 |
+
"logps/rejected": -301.5975341796875,
|
| 110 |
+
"loss": 0.6817,
|
| 111 |
+
"rewards/accuracies": 0.5625,
|
| 112 |
+
"rewards/chosen": -0.012056882493197918,
|
| 113 |
+
"rewards/margins": 0.029647041112184525,
|
| 114 |
+
"rewards/rejected": -0.04170392453670502,
|
| 115 |
+
"step": 60
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"epoch": 0.4878048780487805,
|
| 119 |
+
"grad_norm": 9.090563868123345,
|
| 120 |
+
"learning_rate": 4.939880644182383e-07,
|
| 121 |
+
"logits/chosen": -2.6447739601135254,
|
| 122 |
+
"logits/rejected": -2.6350605487823486,
|
| 123 |
+
"logps/chosen": -327.4669494628906,
|
| 124 |
+
"logps/rejected": -307.3121643066406,
|
| 125 |
+
"loss": 0.6832,
|
| 126 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 127 |
+
"rewards/chosen": -0.017700044438242912,
|
| 128 |
+
"rewards/margins": 0.042197179049253464,
|
| 129 |
+
"rewards/rejected": -0.05989723280072212,
|
| 130 |
+
"step": 70
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 0.5574912891986062,
|
| 134 |
+
"grad_norm": 10.380422622809006,
|
| 135 |
+
"learning_rate": 4.887499574302625e-07,
|
| 136 |
+
"logits/chosen": -2.6520066261291504,
|
| 137 |
+
"logits/rejected": -2.61989164352417,
|
| 138 |
+
"logps/chosen": -279.8045349121094,
|
| 139 |
+
"logps/rejected": -279.3233337402344,
|
| 140 |
+
"loss": 0.6793,
|
| 141 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 142 |
+
"rewards/chosen": 0.08412985503673553,
|
| 143 |
+
"rewards/margins": 0.039986394345760345,
|
| 144 |
+
"rewards/rejected": 0.044143468141555786,
|
| 145 |
+
"step": 80
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"epoch": 0.627177700348432,
|
| 149 |
+
"grad_norm": 13.408348648381818,
|
| 150 |
+
"learning_rate": 4.819312260037522e-07,
|
| 151 |
+
"logits/chosen": -2.5424866676330566,
|
| 152 |
+
"logits/rejected": -2.5196774005889893,
|
| 153 |
+
"logps/chosen": -313.853515625,
|
| 154 |
+
"logps/rejected": -311.6841125488281,
|
| 155 |
+
"loss": 0.6786,
|
| 156 |
+
"rewards/accuracies": 0.612500011920929,
|
| 157 |
+
"rewards/chosen": -0.08200756460428238,
|
| 158 |
+
"rewards/margins": 0.05388826131820679,
|
| 159 |
+
"rewards/rejected": -0.13589580357074738,
|
| 160 |
+
"step": 90
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 0.6968641114982579,
|
| 164 |
+
"grad_norm": 10.057954318267676,
|
| 165 |
+
"learning_rate": 4.7357701298877766e-07,
|
| 166 |
+
"logits/chosen": -2.5143790245056152,
|
| 167 |
+
"logits/rejected": -2.5063555240631104,
|
| 168 |
+
"logps/chosen": -316.9302978515625,
|
| 169 |
+
"logps/rejected": -336.63006591796875,
|
| 170 |
+
"loss": 0.6724,
|
| 171 |
+
"rewards/accuracies": 0.5062500238418579,
|
| 172 |
+
"rewards/chosen": -0.2273595631122589,
|
| 173 |
+
"rewards/margins": 0.014442856423556805,
|
| 174 |
+
"rewards/rejected": -0.2418024092912674,
|
| 175 |
+
"step": 100
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"epoch": 0.6968641114982579,
|
| 179 |
+
"eval_logits/chosen": -2.5339930057525635,
|
| 180 |
+
"eval_logits/rejected": -2.4994165897369385,
|
| 181 |
+
"eval_logps/chosen": -266.2806091308594,
|
| 182 |
+
"eval_logps/rejected": -272.15484619140625,
|
| 183 |
+
"eval_loss": 0.6720507144927979,
|
| 184 |
+
"eval_rewards/accuracies": 0.59375,
|
| 185 |
+
"eval_rewards/chosen": -0.03650704026222229,
|
| 186 |
+
"eval_rewards/margins": 0.05842053145170212,
|
| 187 |
+
"eval_rewards/rejected": -0.094927579164505,
|
| 188 |
+
"eval_runtime": 104.7452,
|
| 189 |
+
"eval_samples_per_second": 19.094,
|
| 190 |
+
"eval_steps_per_second": 0.306,
|
| 191 |
+
"step": 100
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.7665505226480837,
|
| 195 |
+
"grad_norm": 9.469148029484852,
|
| 196 |
+
"learning_rate": 4.637426267648599e-07,
|
| 197 |
+
"logits/chosen": -2.615734338760376,
|
| 198 |
+
"logits/rejected": -2.6145644187927246,
|
| 199 |
+
"logps/chosen": -302.81866455078125,
|
| 200 |
+
"logps/rejected": -306.31378173828125,
|
| 201 |
+
"loss": 0.6774,
|
| 202 |
+
"rewards/accuracies": 0.5375000238418579,
|
| 203 |
+
"rewards/chosen": 0.011746838688850403,
|
| 204 |
+
"rewards/margins": 0.01884249597787857,
|
| 205 |
+
"rewards/rejected": -0.007095657289028168,
|
| 206 |
+
"step": 110
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.8362369337979094,
|
| 210 |
+
"grad_norm": 8.913491607759642,
|
| 211 |
+
"learning_rate": 4.5249317507639726e-07,
|
| 212 |
+
"logits/chosen": -2.541506290435791,
|
| 213 |
+
"logits/rejected": -2.5293049812316895,
|
| 214 |
+
"logps/chosen": -251.04202270507812,
|
| 215 |
+
"logps/rejected": -270.9402770996094,
|
| 216 |
+
"loss": 0.6769,
|
| 217 |
+
"rewards/accuracies": 0.53125,
|
| 218 |
+
"rewards/chosen": -0.024001404643058777,
|
| 219 |
+
"rewards/margins": 0.0283985435962677,
|
| 220 |
+
"rewards/rejected": -0.052399951964616776,
|
| 221 |
+
"step": 120
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.9059233449477352,
|
| 225 |
+
"grad_norm": 10.141409304343389,
|
| 226 |
+
"learning_rate": 4.399031339922038e-07,
|
| 227 |
+
"logits/chosen": -2.622816801071167,
|
| 228 |
+
"logits/rejected": -2.620767116546631,
|
| 229 |
+
"logps/chosen": -305.6891174316406,
|
| 230 |
+
"logps/rejected": -307.5406188964844,
|
| 231 |
+
"loss": 0.673,
|
| 232 |
+
"rewards/accuracies": 0.53125,
|
| 233 |
+
"rewards/chosen": -0.04585634917020798,
|
| 234 |
+
"rewards/margins": 0.06582097709178925,
|
| 235 |
+
"rewards/rejected": -0.11167732626199722,
|
| 236 |
+
"step": 130
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 0.975609756097561,
|
| 240 |
+
"grad_norm": 8.960281893519205,
|
| 241 |
+
"learning_rate": 4.2605585484282636e-07,
|
| 242 |
+
"logits/chosen": -2.6274361610412598,
|
| 243 |
+
"logits/rejected": -2.6125521659851074,
|
| 244 |
+
"logps/chosen": -332.13348388671875,
|
| 245 |
+
"logps/rejected": -312.84503173828125,
|
| 246 |
+
"loss": 0.6721,
|
| 247 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 248 |
+
"rewards/chosen": -0.06096304580569267,
|
| 249 |
+
"rewards/margins": 0.05341456085443497,
|
| 250 |
+
"rewards/rejected": -0.11437759548425674,
|
| 251 |
+
"step": 140
|
| 252 |
+
},
|
| 253 |
+
{
|
| 254 |
+
"epoch": 1.0452961672473868,
|
| 255 |
+
"grad_norm": 9.355042641000702,
|
| 256 |
+
"learning_rate": 4.110430123999227e-07,
|
| 257 |
+
"logits/chosen": -2.6396541595458984,
|
| 258 |
+
"logits/rejected": -2.5929980278015137,
|
| 259 |
+
"logps/chosen": -314.0408630371094,
|
| 260 |
+
"logps/rejected": -329.1007080078125,
|
| 261 |
+
"loss": 0.6047,
|
| 262 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 263 |
+
"rewards/chosen": -0.0018128050724044442,
|
| 264 |
+
"rewards/margins": 0.18482232093811035,
|
| 265 |
+
"rewards/rejected": -0.18663513660430908,
|
| 266 |
+
"step": 150
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 1.0452961672473868,
|
| 270 |
+
"eval_logits/chosen": -2.5443286895751953,
|
| 271 |
+
"eval_logits/rejected": -2.5025253295898438,
|
| 272 |
+
"eval_logps/chosen": -279.22698974609375,
|
| 273 |
+
"eval_logps/rejected": -285.9855041503906,
|
| 274 |
+
"eval_loss": 0.6797215938568115,
|
| 275 |
+
"eval_rewards/accuracies": 0.58984375,
|
| 276 |
+
"eval_rewards/chosen": -0.16597062349319458,
|
| 277 |
+
"eval_rewards/margins": 0.06726360321044922,
|
| 278 |
+
"eval_rewards/rejected": -0.233234241604805,
|
| 279 |
+
"eval_runtime": 105.0322,
|
| 280 |
+
"eval_samples_per_second": 19.042,
|
| 281 |
+
"eval_steps_per_second": 0.305,
|
| 282 |
+
"step": 150
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 1.1149825783972125,
|
| 286 |
+
"grad_norm": 9.102186975626122,
|
| 287 |
+
"learning_rate": 3.9496399795098266e-07,
|
| 288 |
+
"logits/chosen": -2.611131191253662,
|
| 289 |
+
"logits/rejected": -2.577604293823242,
|
| 290 |
+
"logps/chosen": -355.5671081542969,
|
| 291 |
+
"logps/rejected": -354.0496520996094,
|
| 292 |
+
"loss": 0.5626,
|
| 293 |
+
"rewards/accuracies": 0.8187500238418579,
|
| 294 |
+
"rewards/chosen": -0.13642819225788116,
|
| 295 |
+
"rewards/margins": 0.3638991117477417,
|
| 296 |
+
"rewards/rejected": -0.5003272294998169,
|
| 297 |
+
"step": 160
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"epoch": 1.1846689895470384,
|
| 301 |
+
"grad_norm": 10.030074101152461,
|
| 302 |
+
"learning_rate": 3.779252612874913e-07,
|
| 303 |
+
"logits/chosen": -2.52521014213562,
|
| 304 |
+
"logits/rejected": -2.461188793182373,
|
| 305 |
+
"logps/chosen": -291.77777099609375,
|
| 306 |
+
"logps/rejected": -298.83343505859375,
|
| 307 |
+
"loss": 0.5548,
|
| 308 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 309 |
+
"rewards/chosen": -0.029370594769716263,
|
| 310 |
+
"rewards/margins": 0.332529217004776,
|
| 311 |
+
"rewards/rejected": -0.36189982295036316,
|
| 312 |
+
"step": 170
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 1.254355400696864,
|
| 316 |
+
"grad_norm": 10.899159922293036,
|
| 317 |
+
"learning_rate": 3.60039605962848e-07,
|
| 318 |
+
"logits/chosen": -2.48093843460083,
|
| 319 |
+
"logits/rejected": -2.453683376312256,
|
| 320 |
+
"logps/chosen": -336.80316162109375,
|
| 321 |
+
"logps/rejected": -371.10650634765625,
|
| 322 |
+
"loss": 0.5374,
|
| 323 |
+
"rewards/accuracies": 0.831250011920929,
|
| 324 |
+
"rewards/chosen": -0.12406754493713379,
|
| 325 |
+
"rewards/margins": 0.418355792760849,
|
| 326 |
+
"rewards/rejected": -0.5424233675003052,
|
| 327 |
+
"step": 180
|
| 328 |
+
},
|
| 329 |
+
{
|
| 330 |
+
"epoch": 1.32404181184669,
|
| 331 |
+
"grad_norm": 10.743619932904148,
|
| 332 |
+
"learning_rate": 3.414254424857272e-07,
|
| 333 |
+
"logits/chosen": -2.402945041656494,
|
| 334 |
+
"logits/rejected": -2.4313735961914062,
|
| 335 |
+
"logps/chosen": -327.4953308105469,
|
| 336 |
+
"logps/rejected": -387.18035888671875,
|
| 337 |
+
"loss": 0.5441,
|
| 338 |
+
"rewards/accuracies": 0.875,
|
| 339 |
+
"rewards/chosen": -0.2883428931236267,
|
| 340 |
+
"rewards/margins": 0.4703540802001953,
|
| 341 |
+
"rewards/rejected": -0.758696973323822,
|
| 342 |
+
"step": 190
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 1.3937282229965158,
|
| 346 |
+
"grad_norm": 13.29972806953709,
|
| 347 |
+
"learning_rate": 3.2220600439305403e-07,
|
| 348 |
+
"logits/chosen": -2.356320858001709,
|
| 349 |
+
"logits/rejected": -2.3801255226135254,
|
| 350 |
+
"logps/chosen": -323.59130859375,
|
| 351 |
+
"logps/rejected": -368.29425048828125,
|
| 352 |
+
"loss": 0.5265,
|
| 353 |
+
"rewards/accuracies": 0.8187500238418579,
|
| 354 |
+
"rewards/chosen": -0.3764956295490265,
|
| 355 |
+
"rewards/margins": 0.45888185501098633,
|
| 356 |
+
"rewards/rejected": -0.8353773951530457,
|
| 357 |
+
"step": 200
|
| 358 |
+
},
|
| 359 |
+
{
|
| 360 |
+
"epoch": 1.3937282229965158,
|
| 361 |
+
"eval_logits/chosen": -2.332766532897949,
|
| 362 |
+
"eval_logits/rejected": -2.2717788219451904,
|
| 363 |
+
"eval_logps/chosen": -320.0576477050781,
|
| 364 |
+
"eval_logps/rejected": -335.9707946777344,
|
| 365 |
+
"eval_loss": 0.6762288808822632,
|
| 366 |
+
"eval_rewards/accuracies": 0.671875,
|
| 367 |
+
"eval_rewards/chosen": -0.5742772817611694,
|
| 368 |
+
"eval_rewards/margins": 0.15881015360355377,
|
| 369 |
+
"eval_rewards/rejected": -0.733087420463562,
|
| 370 |
+
"eval_runtime": 103.8887,
|
| 371 |
+
"eval_samples_per_second": 19.251,
|
| 372 |
+
"eval_steps_per_second": 0.308,
|
| 373 |
+
"step": 200
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"epoch": 1.4634146341463414,
|
| 377 |
+
"grad_norm": 15.19540789655145,
|
| 378 |
+
"learning_rate": 3.025085323925175e-07,
|
| 379 |
+
"logits/chosen": -2.238861083984375,
|
| 380 |
+
"logits/rejected": -2.186549663543701,
|
| 381 |
+
"logps/chosen": -322.6126403808594,
|
| 382 |
+
"logps/rejected": -375.65850830078125,
|
| 383 |
+
"loss": 0.5251,
|
| 384 |
+
"rewards/accuracies": 0.831250011920929,
|
| 385 |
+
"rewards/chosen": -0.4552794396877289,
|
| 386 |
+
"rewards/margins": 0.5174066424369812,
|
| 387 |
+
"rewards/rejected": -0.9726861119270325,
|
| 388 |
+
"step": 210
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 1.533101045296167,
|
| 392 |
+
"grad_norm": 14.440109215320728,
|
| 393 |
+
"learning_rate": 2.8246343197594046e-07,
|
| 394 |
+
"logits/chosen": -2.1742804050445557,
|
| 395 |
+
"logits/rejected": -2.0851190090179443,
|
| 396 |
+
"logps/chosen": -388.40509033203125,
|
| 397 |
+
"logps/rejected": -404.67388916015625,
|
| 398 |
+
"loss": 0.5096,
|
| 399 |
+
"rewards/accuracies": 0.78125,
|
| 400 |
+
"rewards/chosen": -0.5906688570976257,
|
| 401 |
+
"rewards/margins": 0.48450303077697754,
|
| 402 |
+
"rewards/rejected": -1.075171947479248,
|
| 403 |
+
"step": 220
|
| 404 |
+
},
|
| 405 |
+
{
|
| 406 |
+
"epoch": 1.6027874564459932,
|
| 407 |
+
"grad_norm": 16.964461935174956,
|
| 408 |
+
"learning_rate": 2.622034100804566e-07,
|
| 409 |
+
"logits/chosen": -1.8995920419692993,
|
| 410 |
+
"logits/rejected": -2.022343397140503,
|
| 411 |
+
"logps/chosen": -324.2644958496094,
|
| 412 |
+
"logps/rejected": -404.7146911621094,
|
| 413 |
+
"loss": 0.5032,
|
| 414 |
+
"rewards/accuracies": 0.8187500238418579,
|
| 415 |
+
"rewards/chosen": -0.6232214570045471,
|
| 416 |
+
"rewards/margins": 0.5319386720657349,
|
| 417 |
+
"rewards/rejected": -1.1551600694656372,
|
| 418 |
+
"step": 230
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"epoch": 1.6724738675958188,
|
| 422 |
+
"grad_norm": 27.37981327191052,
|
| 423 |
+
"learning_rate": 2.418625965131574e-07,
|
| 424 |
+
"logits/chosen": -1.7780876159667969,
|
| 425 |
+
"logits/rejected": -1.6987870931625366,
|
| 426 |
+
"logps/chosen": -376.9257507324219,
|
| 427 |
+
"logps/rejected": -409.15887451171875,
|
| 428 |
+
"loss": 0.5097,
|
| 429 |
+
"rewards/accuracies": 0.768750011920929,
|
| 430 |
+
"rewards/chosen": -0.7337759137153625,
|
| 431 |
+
"rewards/margins": 0.56010901927948,
|
| 432 |
+
"rewards/rejected": -1.2938848733901978,
|
| 433 |
+
"step": 240
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 1.7421602787456445,
|
| 437 |
+
"grad_norm": 18.86770347379352,
|
| 438 |
+
"learning_rate": 2.2157565595574668e-07,
|
| 439 |
+
"logits/chosen": -1.6350816488265991,
|
| 440 |
+
"logits/rejected": -1.648374319076538,
|
| 441 |
+
"logps/chosen": -389.5054931640625,
|
| 442 |
+
"logps/rejected": -428.66802978515625,
|
| 443 |
+
"loss": 0.4984,
|
| 444 |
+
"rewards/accuracies": 0.831250011920929,
|
| 445 |
+
"rewards/chosen": -0.949749767780304,
|
| 446 |
+
"rewards/margins": 0.5309340953826904,
|
| 447 |
+
"rewards/rejected": -1.4806839227676392,
|
| 448 |
+
"step": 250
|
| 449 |
+
},
|
| 450 |
+
{
|
| 451 |
+
"epoch": 1.7421602787456445,
|
| 452 |
+
"eval_logits/chosen": -1.543340802192688,
|
| 453 |
+
"eval_logits/rejected": -1.445103645324707,
|
| 454 |
+
"eval_logps/chosen": -383.83807373046875,
|
| 455 |
+
"eval_logps/rejected": -407.1153869628906,
|
| 456 |
+
"eval_loss": 0.6731657981872559,
|
| 457 |
+
"eval_rewards/accuracies": 0.65625,
|
| 458 |
+
"eval_rewards/chosen": -1.2120810747146606,
|
| 459 |
+
"eval_rewards/margins": 0.23245161771774292,
|
| 460 |
+
"eval_rewards/rejected": -1.4445327520370483,
|
| 461 |
+
"eval_runtime": 104.3313,
|
| 462 |
+
"eval_samples_per_second": 19.17,
|
| 463 |
+
"eval_steps_per_second": 0.307,
|
| 464 |
+
"step": 250
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 1.8118466898954704,
|
| 468 |
+
"grad_norm": 17.331986790708136,
|
| 469 |
+
"learning_rate": 2.0147689642810138e-07,
|
| 470 |
+
"logits/chosen": -1.6317332983016968,
|
| 471 |
+
"logits/rejected": -1.572344183921814,
|
| 472 |
+
"logps/chosen": -410.221923828125,
|
| 473 |
+
"logps/rejected": -480.1829528808594,
|
| 474 |
+
"loss": 0.4902,
|
| 475 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 476 |
+
"rewards/chosen": -1.0687463283538818,
|
| 477 |
+
"rewards/margins": 0.6032934188842773,
|
| 478 |
+
"rewards/rejected": -1.6720397472381592,
|
| 479 |
+
"step": 260
|
| 480 |
+
},
|
| 481 |
+
{
|
| 482 |
+
"epoch": 1.8815331010452963,
|
| 483 |
+
"grad_norm": 17.843882545206057,
|
| 484 |
+
"learning_rate": 1.8169938011308233e-07,
|
| 485 |
+
"logits/chosen": -1.4750444889068604,
|
| 486 |
+
"logits/rejected": -1.417011022567749,
|
| 487 |
+
"logps/chosen": -395.61480712890625,
|
| 488 |
+
"logps/rejected": -437.10577392578125,
|
| 489 |
+
"loss": 0.492,
|
| 490 |
+
"rewards/accuracies": 0.800000011920929,
|
| 491 |
+
"rewards/chosen": -0.9328581094741821,
|
| 492 |
+
"rewards/margins": 0.6057752370834351,
|
| 493 |
+
"rewards/rejected": -1.538633108139038,
|
| 494 |
+
"step": 270
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 1.951219512195122,
|
| 498 |
+
"grad_norm": 19.896351603470077,
|
| 499 |
+
"learning_rate": 1.6237404242930697e-07,
|
| 500 |
+
"logits/chosen": -1.417770266532898,
|
| 501 |
+
"logits/rejected": -1.3385121822357178,
|
| 502 |
+
"logps/chosen": -375.705078125,
|
| 503 |
+
"logps/rejected": -404.3011169433594,
|
| 504 |
+
"loss": 0.4867,
|
| 505 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 506 |
+
"rewards/chosen": -0.9175931215286255,
|
| 507 |
+
"rewards/margins": 0.4988563060760498,
|
| 508 |
+
"rewards/rejected": -1.4164493083953857,
|
| 509 |
+
"step": 280
|
| 510 |
+
},
|
| 511 |
+
{
|
| 512 |
+
"epoch": 2.0209059233449476,
|
| 513 |
+
"grad_norm": 16.576113766725356,
|
| 514 |
+
"learning_rate": 1.4362882518398945e-07,
|
| 515 |
+
"logits/chosen": -1.3333556652069092,
|
| 516 |
+
"logits/rejected": -1.3687762022018433,
|
| 517 |
+
"logps/chosen": -405.6753234863281,
|
| 518 |
+
"logps/rejected": -468.34490966796875,
|
| 519 |
+
"loss": 0.4573,
|
| 520 |
+
"rewards/accuracies": 0.78125,
|
| 521 |
+
"rewards/chosen": -0.9316972494125366,
|
| 522 |
+
"rewards/margins": 0.6436307430267334,
|
| 523 |
+
"rewards/rejected": -1.5753281116485596,
|
| 524 |
+
"step": 290
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"epoch": 2.0905923344947737,
|
| 528 |
+
"grad_norm": 18.642417597691527,
|
| 529 |
+
"learning_rate": 1.2558782954473823e-07,
|
| 530 |
+
"logits/chosen": -1.1027063131332397,
|
| 531 |
+
"logits/rejected": -1.0710804462432861,
|
| 532 |
+
"logps/chosen": -400.2093505859375,
|
| 533 |
+
"logps/rejected": -480.50518798828125,
|
| 534 |
+
"loss": 0.3569,
|
| 535 |
+
"rewards/accuracies": 0.925000011920929,
|
| 536 |
+
"rewards/chosen": -0.9277191162109375,
|
| 537 |
+
"rewards/margins": 1.0768729448318481,
|
| 538 |
+
"rewards/rejected": -2.004591941833496,
|
| 539 |
+
"step": 300
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 2.0905923344947737,
|
| 543 |
+
"eval_logits/chosen": -0.9999401569366455,
|
| 544 |
+
"eval_logits/rejected": -0.870820164680481,
|
| 545 |
+
"eval_logps/chosen": -397.18048095703125,
|
| 546 |
+
"eval_logps/rejected": -429.4680480957031,
|
| 547 |
+
"eval_loss": 0.6527448892593384,
|
| 548 |
+
"eval_rewards/accuracies": 0.67578125,
|
| 549 |
+
"eval_rewards/chosen": -1.3455055952072144,
|
| 550 |
+
"eval_rewards/margins": 0.32255375385284424,
|
| 551 |
+
"eval_rewards/rejected": -1.6680593490600586,
|
| 552 |
+
"eval_runtime": 102.8366,
|
| 553 |
+
"eval_samples_per_second": 19.448,
|
| 554 |
+
"eval_steps_per_second": 0.311,
|
| 555 |
+
"step": 300
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 2.1602787456445993,
|
| 559 |
+
"grad_norm": 19.642342649385267,
|
| 560 |
+
"learning_rate": 1.0837049443799279e-07,
|
| 561 |
+
"logits/chosen": -0.9505928158760071,
|
| 562 |
+
"logits/rejected": -0.9347362518310547,
|
| 563 |
+
"logps/chosen": -373.06011962890625,
|
| 564 |
+
"logps/rejected": -473.77880859375,
|
| 565 |
+
"loss": 0.3689,
|
| 566 |
+
"rewards/accuracies": 0.875,
|
| 567 |
+
"rewards/chosen": -0.9760599136352539,
|
| 568 |
+
"rewards/margins": 1.0135242938995361,
|
| 569 |
+
"rewards/rejected": -1.98958420753479,
|
| 570 |
+
"step": 310
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"epoch": 2.229965156794425,
|
| 574 |
+
"grad_norm": 20.908278202953714,
|
| 575 |
+
"learning_rate": 9.209080581344306e-08,
|
| 576 |
+
"logits/chosen": -0.6433783173561096,
|
| 577 |
+
"logits/rejected": -0.6169986128807068,
|
| 578 |
+
"logps/chosen": -395.59600830078125,
|
| 579 |
+
"logps/rejected": -535.1781005859375,
|
| 580 |
+
"loss": 0.3438,
|
| 581 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 582 |
+
"rewards/chosen": -1.1918704509735107,
|
| 583 |
+
"rewards/margins": 1.1342499256134033,
|
| 584 |
+
"rewards/rejected": -2.326120615005493,
|
| 585 |
+
"step": 320
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 2.2996515679442506,
|
| 589 |
+
"grad_norm": 22.95505810018377,
|
| 590 |
+
"learning_rate": 7.685654200943378e-08,
|
| 591 |
+
"logits/chosen": -0.7587612867355347,
|
| 592 |
+
"logits/rejected": -0.6282288432121277,
|
| 593 |
+
"logps/chosen": -465.53985595703125,
|
| 594 |
+
"logps/rejected": -561.067626953125,
|
| 595 |
+
"loss": 0.3429,
|
| 596 |
+
"rewards/accuracies": 0.9375,
|
| 597 |
+
"rewards/chosen": -1.326777696609497,
|
| 598 |
+
"rewards/margins": 1.2314026355743408,
|
| 599 |
+
"rewards/rejected": -2.558180332183838,
|
| 600 |
+
"step": 330
|
| 601 |
+
},
|
| 602 |
+
{
|
| 603 |
+
"epoch": 2.3693379790940767,
|
| 604 |
+
"grad_norm": 23.693323873855398,
|
| 605 |
+
"learning_rate": 6.27685602153478e-08,
|
| 606 |
+
"logits/chosen": -0.5435560941696167,
|
| 607 |
+
"logits/rejected": -0.4107975959777832,
|
| 608 |
+
"logps/chosen": -466.288818359375,
|
| 609 |
+
"logps/rejected": -571.3019409179688,
|
| 610 |
+
"loss": 0.3396,
|
| 611 |
+
"rewards/accuracies": 0.8687499761581421,
|
| 612 |
+
"rewards/chosen": -1.5259153842926025,
|
| 613 |
+
"rewards/margins": 1.2022850513458252,
|
| 614 |
+
"rewards/rejected": -2.7282001972198486,
|
| 615 |
+
"step": 340
|
| 616 |
+
},
|
| 617 |
+
{
|
| 618 |
+
"epoch": 2.4390243902439024,
|
| 619 |
+
"grad_norm": 24.606275846522493,
|
| 620 |
+
"learning_rate": 4.992012875488669e-08,
|
| 621 |
+
"logits/chosen": -0.3131232261657715,
|
| 622 |
+
"logits/rejected": -0.2396572083234787,
|
| 623 |
+
"logps/chosen": -431.25140380859375,
|
| 624 |
+
"logps/rejected": -555.533203125,
|
| 625 |
+
"loss": 0.3329,
|
| 626 |
+
"rewards/accuracies": 0.90625,
|
| 627 |
+
"rewards/chosen": -1.5053646564483643,
|
| 628 |
+
"rewards/margins": 1.0768215656280518,
|
| 629 |
+
"rewards/rejected": -2.582186222076416,
|
| 630 |
+
"step": 350
|
| 631 |
+
},
|
| 632 |
+
{
|
| 633 |
+
"epoch": 2.4390243902439024,
|
| 634 |
+
"eval_logits/chosen": -0.24473249912261963,
|
| 635 |
+
"eval_logits/rejected": -0.10840671509504318,
|
| 636 |
+
"eval_logps/chosen": -453.08160400390625,
|
| 637 |
+
"eval_logps/rejected": -488.3669738769531,
|
| 638 |
+
"eval_loss": 0.6840001940727234,
|
| 639 |
+
"eval_rewards/accuracies": 0.66015625,
|
| 640 |
+
"eval_rewards/chosen": -1.904516339302063,
|
| 641 |
+
"eval_rewards/margins": 0.35253193974494934,
|
| 642 |
+
"eval_rewards/rejected": -2.2570483684539795,
|
| 643 |
+
"eval_runtime": 106.925,
|
| 644 |
+
"eval_samples_per_second": 18.705,
|
| 645 |
+
"eval_steps_per_second": 0.299,
|
| 646 |
+
"step": 350
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"epoch": 2.508710801393728,
|
| 650 |
+
"grad_norm": 27.353529501179093,
|
| 651 |
+
"learning_rate": 3.8396309610812086e-08,
|
| 652 |
+
"logits/chosen": -0.23779411613941193,
|
| 653 |
+
"logits/rejected": -0.15696097910404205,
|
| 654 |
+
"logps/chosen": -438.49298095703125,
|
| 655 |
+
"logps/rejected": -567.0818481445312,
|
| 656 |
+
"loss": 0.3351,
|
| 657 |
+
"rewards/accuracies": 0.918749988079071,
|
| 658 |
+
"rewards/chosen": -1.482677698135376,
|
| 659 |
+
"rewards/margins": 1.268634557723999,
|
| 660 |
+
"rewards/rejected": -2.751312255859375,
|
| 661 |
+
"step": 360
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 2.578397212543554,
|
| 665 |
+
"grad_norm": 26.645676298973676,
|
| 666 |
+
"learning_rate": 2.8273395279091005e-08,
|
| 667 |
+
"logits/chosen": -0.31052619218826294,
|
| 668 |
+
"logits/rejected": -0.2516113221645355,
|
| 669 |
+
"logps/chosen": -440.5013122558594,
|
| 670 |
+
"logps/rejected": -570.08056640625,
|
| 671 |
+
"loss": 0.3373,
|
| 672 |
+
"rewards/accuracies": 0.9312499761581421,
|
| 673 |
+
"rewards/chosen": -1.4818499088287354,
|
| 674 |
+
"rewards/margins": 1.3230129480361938,
|
| 675 |
+
"rewards/rejected": -2.8048629760742188,
|
| 676 |
+
"step": 370
|
| 677 |
+
},
|
| 678 |
+
{
|
| 679 |
+
"epoch": 2.64808362369338,
|
| 680 |
+
"grad_norm": 23.8312792481954,
|
| 681 |
+
"learning_rate": 1.9618403680707053e-08,
|
| 682 |
+
"logits/chosen": -0.3248611092567444,
|
| 683 |
+
"logits/rejected": -0.3839193284511566,
|
| 684 |
+
"logps/chosen": -462.94873046875,
|
| 685 |
+
"logps/rejected": -592.1238403320312,
|
| 686 |
+
"loss": 0.34,
|
| 687 |
+
"rewards/accuracies": 0.9312499761581421,
|
| 688 |
+
"rewards/chosen": -1.5153675079345703,
|
| 689 |
+
"rewards/margins": 1.3629937171936035,
|
| 690 |
+
"rewards/rejected": -2.878361225128174,
|
| 691 |
+
"step": 380
|
| 692 |
+
},
|
| 693 |
+
{
|
| 694 |
+
"epoch": 2.7177700348432055,
|
| 695 |
+
"grad_norm": 25.615650685832552,
|
| 696 |
+
"learning_rate": 1.2488634475031761e-08,
|
| 697 |
+
"logits/chosen": -0.10789848864078522,
|
| 698 |
+
"logits/rejected": -0.0006875753169879317,
|
| 699 |
+
"logps/chosen": -432.10968017578125,
|
| 700 |
+
"logps/rejected": -547.6038208007812,
|
| 701 |
+
"loss": 0.3334,
|
| 702 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 703 |
+
"rewards/chosen": -1.524472951889038,
|
| 704 |
+
"rewards/margins": 1.2026252746582031,
|
| 705 |
+
"rewards/rejected": -2.727097988128662,
|
| 706 |
+
"step": 390
|
| 707 |
+
},
|
| 708 |
+
{
|
| 709 |
+
"epoch": 2.7874564459930316,
|
| 710 |
+
"grad_norm": 29.507902477027233,
|
| 711 |
+
"learning_rate": 6.9312897121466815e-09,
|
| 712 |
+
"logits/chosen": -0.16302387416362762,
|
| 713 |
+
"logits/rejected": -0.1776006668806076,
|
| 714 |
+
"logps/chosen": -456.0896911621094,
|
| 715 |
+
"logps/rejected": -589.2247924804688,
|
| 716 |
+
"loss": 0.3368,
|
| 717 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 718 |
+
"rewards/chosen": -1.5656400918960571,
|
| 719 |
+
"rewards/margins": 1.24466872215271,
|
| 720 |
+
"rewards/rejected": -2.8103089332580566,
|
| 721 |
+
"step": 400
|
| 722 |
+
},
|
| 723 |
+
{
|
| 724 |
+
"epoch": 2.7874564459930316,
|
| 725 |
+
"eval_logits/chosen": -0.3103621006011963,
|
| 726 |
+
"eval_logits/rejected": -0.1808159053325653,
|
| 727 |
+
"eval_logps/chosen": -455.8002624511719,
|
| 728 |
+
"eval_logps/rejected": -491.1397705078125,
|
| 729 |
+
"eval_loss": 0.681273877620697,
|
| 730 |
+
"eval_rewards/accuracies": 0.6796875,
|
| 731 |
+
"eval_rewards/chosen": -1.9317032098770142,
|
| 732 |
+
"eval_rewards/margins": 0.35307374596595764,
|
| 733 |
+
"eval_rewards/rejected": -2.2847771644592285,
|
| 734 |
+
"eval_runtime": 103.4584,
|
| 735 |
+
"eval_samples_per_second": 19.331,
|
| 736 |
+
"eval_steps_per_second": 0.309,
|
| 737 |
+
"step": 400
|
| 738 |
+
},
|
| 739 |
+
{
|
| 740 |
+
"epoch": 2.857142857142857,
|
| 741 |
+
"grad_norm": 23.881676504304508,
|
| 742 |
+
"learning_rate": 2.983161335556761e-09,
|
| 743 |
+
"logits/chosen": -0.464876651763916,
|
| 744 |
+
"logits/rejected": -0.40461286902427673,
|
| 745 |
+
"logps/chosen": -434.8916015625,
|
| 746 |
+
"logps/rejected": -576.1773681640625,
|
| 747 |
+
"loss": 0.3237,
|
| 748 |
+
"rewards/accuracies": 0.9375,
|
| 749 |
+
"rewards/chosen": -1.5066254138946533,
|
| 750 |
+
"rewards/margins": 1.3078076839447021,
|
| 751 |
+
"rewards/rejected": -2.8144326210021973,
|
| 752 |
+
"step": 410
|
| 753 |
+
},
|
| 754 |
+
{
|
| 755 |
+
"epoch": 2.926829268292683,
|
| 756 |
+
"grad_norm": 24.99330232547937,
|
| 757 |
+
"learning_rate": 6.703876041571077e-10,
|
| 758 |
+
"logits/chosen": -0.48426467180252075,
|
| 759 |
+
"logits/rejected": -0.1355866938829422,
|
| 760 |
+
"logps/chosen": -447.3482971191406,
|
| 761 |
+
"logps/rejected": -561.0531005859375,
|
| 762 |
+
"loss": 0.3235,
|
| 763 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 764 |
+
"rewards/chosen": -1.4588124752044678,
|
| 765 |
+
"rewards/margins": 1.2296662330627441,
|
| 766 |
+
"rewards/rejected": -2.688478946685791,
|
| 767 |
+
"step": 420
|
| 768 |
+
},
|
| 769 |
+
{
|
| 770 |
+
"epoch": 2.989547038327526,
|
| 771 |
+
"step": 429,
|
| 772 |
+
"total_flos": 0.0,
|
| 773 |
+
"train_loss": 0.5141507484418251,
|
| 774 |
+
"train_runtime": 11380.2127,
|
| 775 |
+
"train_samples_per_second": 4.835,
|
| 776 |
+
"train_steps_per_second": 0.038
|
| 777 |
+
}
|
| 778 |
+
],
|
| 779 |
+
"logging_steps": 10,
|
| 780 |
+
"max_steps": 429,
|
| 781 |
+
"num_input_tokens_seen": 0,
|
| 782 |
+
"num_train_epochs": 3,
|
| 783 |
+
"save_steps": 100,
|
| 784 |
+
"stateful_callbacks": {
|
| 785 |
+
"TrainerControl": {
|
| 786 |
+
"args": {
|
| 787 |
+
"should_epoch_stop": false,
|
| 788 |
+
"should_evaluate": false,
|
| 789 |
+
"should_log": false,
|
| 790 |
+
"should_save": true,
|
| 791 |
+
"should_training_stop": true
|
| 792 |
+
},
|
| 793 |
+
"attributes": {}
|
| 794 |
+
}
|
| 795 |
+
},
|
| 796 |
+
"total_flos": 0.0,
|
| 797 |
+
"train_batch_size": 8,
|
| 798 |
+
"trial_name": null,
|
| 799 |
+
"trial_params": null
|
| 800 |
+
}
|