Model save
Browse files- README.md +79 -0
- all_results.json +9 -0
- generation_config.json +6 -0
- train_results.json +9 -0
- trainer_state.json +739 -0
README.md
ADDED
|
@@ -0,0 +1,79 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 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-high-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-prometheus-high-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 the None dataset.
|
| 19 |
+
It achieves the following results on the evaluation set:
|
| 20 |
+
- Loss: 0.5829
|
| 21 |
+
- Rewards/chosen: -2.8437
|
| 22 |
+
- Rewards/rejected: -4.4971
|
| 23 |
+
- Rewards/accuracies: 0.7629
|
| 24 |
+
- Rewards/margins: 1.6534
|
| 25 |
+
- Logps/rejected: -697.9909
|
| 26 |
+
- Logps/chosen: -544.3318
|
| 27 |
+
- Logits/rejected: 4.2569
|
| 28 |
+
- Logits/chosen: 2.5320
|
| 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 |
+
| 0.5554 | 0.3802 | 50 | 0.6140 | -0.1671 | -0.4180 | 0.6897 | 0.2509 | -290.0737 | -276.6687 | -2.5786 | -2.6244 |
|
| 66 |
+
| 0.3147 | 0.7605 | 100 | 0.5617 | -1.1542 | -2.1116 | 0.7328 | 0.9574 | -459.4361 | -375.3786 | 1.5949 | 0.7404 |
|
| 67 |
+
| 0.214 | 1.1407 | 150 | 0.5560 | -1.2961 | -2.4277 | 0.7457 | 1.1316 | -491.0475 | -389.5718 | 2.0487 | 0.8077 |
|
| 68 |
+
| 0.1866 | 1.5209 | 200 | 0.5364 | -1.4494 | -2.6242 | 0.7414 | 1.1748 | -510.6940 | -404.8973 | 2.0297 | 0.7929 |
|
| 69 |
+
| 0.1899 | 1.9011 | 250 | 0.5391 | -1.7883 | -3.0786 | 0.7457 | 1.2902 | -556.1323 | -438.7930 | 2.5714 | 1.2524 |
|
| 70 |
+
| 0.1083 | 2.2814 | 300 | 0.5584 | -2.1009 | -3.5313 | 0.7629 | 1.4304 | -601.4120 | -470.0527 | 3.6920 | 2.2903 |
|
| 71 |
+
| 0.0955 | 2.6616 | 350 | 0.5829 | -2.8437 | -4.4971 | 0.7629 | 1.6534 | -697.9909 | -544.3318 | 4.2569 | 2.5320 |
|
| 72 |
+
|
| 73 |
+
|
| 74 |
+
### Framework versions
|
| 75 |
+
|
| 76 |
+
- Transformers 4.44.0.dev0
|
| 77 |
+
- Pytorch 2.1.2
|
| 78 |
+
- Datasets 2.20.0
|
| 79 |
+
- Tokenizers 0.19.1
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 2.988593155893536,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.2546291965564699,
|
| 5 |
+
"train_runtime": 10353.4149,
|
| 6 |
+
"train_samples": 16800,
|
| 7 |
+
"train_samples_per_second": 4.868,
|
| 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": 2.988593155893536,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.2546291965564699,
|
| 5 |
+
"train_runtime": 10353.4149,
|
| 6 |
+
"train_samples": 16800,
|
| 7 |
+
"train_samples_per_second": 4.868,
|
| 8 |
+
"train_steps_per_second": 0.038
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,739 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 2.988593155893536,
|
| 5 |
+
"eval_steps": 50,
|
| 6 |
+
"global_step": 393,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.07604562737642585,
|
| 13 |
+
"grad_norm": 10.680026826860546,
|
| 14 |
+
"learning_rate": 1.25e-07,
|
| 15 |
+
"logits/chosen": -2.563544511795044,
|
| 16 |
+
"logits/rejected": -2.4952266216278076,
|
| 17 |
+
"logps/chosen": -273.53424072265625,
|
| 18 |
+
"logps/rejected": -239.1501922607422,
|
| 19 |
+
"loss": 0.6932,
|
| 20 |
+
"rewards/accuracies": 0.38749998807907104,
|
| 21 |
+
"rewards/chosen": 0.00019869348034262657,
|
| 22 |
+
"rewards/margins": -1.28610986394051e-06,
|
| 23 |
+
"rewards/rejected": 0.00019997954950667918,
|
| 24 |
+
"step": 10
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.1520912547528517,
|
| 28 |
+
"grad_norm": 8.993807887778436,
|
| 29 |
+
"learning_rate": 2.5e-07,
|
| 30 |
+
"logits/chosen": -2.580817222595215,
|
| 31 |
+
"logits/rejected": -2.462310314178467,
|
| 32 |
+
"logps/chosen": -269.5621337890625,
|
| 33 |
+
"logps/rejected": -258.239990234375,
|
| 34 |
+
"loss": 0.6892,
|
| 35 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 36 |
+
"rewards/chosen": 0.00650571146979928,
|
| 37 |
+
"rewards/margins": 0.008498361334204674,
|
| 38 |
+
"rewards/rejected": -0.0019926498644053936,
|
| 39 |
+
"step": 20
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.22813688212927757,
|
| 43 |
+
"grad_norm": 8.92132470459759,
|
| 44 |
+
"learning_rate": 3.75e-07,
|
| 45 |
+
"logits/chosen": -2.6156678199768066,
|
| 46 |
+
"logits/rejected": -2.4716460704803467,
|
| 47 |
+
"logps/chosen": -295.3561096191406,
|
| 48 |
+
"logps/rejected": -241.26644897460938,
|
| 49 |
+
"loss": 0.6709,
|
| 50 |
+
"rewards/accuracies": 0.84375,
|
| 51 |
+
"rewards/chosen": 0.040296901017427444,
|
| 52 |
+
"rewards/margins": 0.05595223233103752,
|
| 53 |
+
"rewards/rejected": -0.015655336901545525,
|
| 54 |
+
"step": 30
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.3041825095057034,
|
| 58 |
+
"grad_norm": 9.266525353467978,
|
| 59 |
+
"learning_rate": 5e-07,
|
| 60 |
+
"logits/chosen": -2.6184380054473877,
|
| 61 |
+
"logits/rejected": -2.513917922973633,
|
| 62 |
+
"logps/chosen": -254.7106475830078,
|
| 63 |
+
"logps/rejected": -234.47219848632812,
|
| 64 |
+
"loss": 0.6239,
|
| 65 |
+
"rewards/accuracies": 0.8125,
|
| 66 |
+
"rewards/chosen": 0.028813377022743225,
|
| 67 |
+
"rewards/margins": 0.14417138695716858,
|
| 68 |
+
"rewards/rejected": -0.11535801738500595,
|
| 69 |
+
"step": 40
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.38022813688212925,
|
| 73 |
+
"grad_norm": 12.173416216409002,
|
| 74 |
+
"learning_rate": 4.990105959637203e-07,
|
| 75 |
+
"logits/chosen": -2.546933650970459,
|
| 76 |
+
"logits/rejected": -2.4483542442321777,
|
| 77 |
+
"logps/chosen": -243.9954376220703,
|
| 78 |
+
"logps/rejected": -256.17242431640625,
|
| 79 |
+
"loss": 0.5554,
|
| 80 |
+
"rewards/accuracies": 0.800000011920929,
|
| 81 |
+
"rewards/chosen": -0.08808934688568115,
|
| 82 |
+
"rewards/margins": 0.2857440114021301,
|
| 83 |
+
"rewards/rejected": -0.3738333582878113,
|
| 84 |
+
"step": 50
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.38022813688212925,
|
| 88 |
+
"eval_logits/chosen": -2.62440824508667,
|
| 89 |
+
"eval_logits/rejected": -2.5786266326904297,
|
| 90 |
+
"eval_logps/chosen": -276.6687316894531,
|
| 91 |
+
"eval_logps/rejected": -290.07366943359375,
|
| 92 |
+
"eval_loss": 0.6140323281288147,
|
| 93 |
+
"eval_rewards/accuracies": 0.6896551847457886,
|
| 94 |
+
"eval_rewards/chosen": -0.16707415878772736,
|
| 95 |
+
"eval_rewards/margins": 0.2508907914161682,
|
| 96 |
+
"eval_rewards/rejected": -0.4179649353027344,
|
| 97 |
+
"eval_runtime": 94.0274,
|
| 98 |
+
"eval_samples_per_second": 19.473,
|
| 99 |
+
"eval_steps_per_second": 0.308,
|
| 100 |
+
"step": 50
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.45627376425855515,
|
| 104 |
+
"grad_norm": 24.13668429607915,
|
| 105 |
+
"learning_rate": 4.960502152176573e-07,
|
| 106 |
+
"logits/chosen": -2.1383702754974365,
|
| 107 |
+
"logits/rejected": -2.0003161430358887,
|
| 108 |
+
"logps/chosen": -272.81854248046875,
|
| 109 |
+
"logps/rejected": -323.1360168457031,
|
| 110 |
+
"loss": 0.4773,
|
| 111 |
+
"rewards/accuracies": 0.8812500238418579,
|
| 112 |
+
"rewards/chosen": -0.23167654871940613,
|
| 113 |
+
"rewards/margins": 0.8607891798019409,
|
| 114 |
+
"rewards/rejected": -1.0924659967422485,
|
| 115 |
+
"step": 60
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"epoch": 0.532319391634981,
|
| 119 |
+
"grad_norm": 40.013780890248455,
|
| 120 |
+
"learning_rate": 4.911422898630837e-07,
|
| 121 |
+
"logits/chosen": -1.2432479858398438,
|
| 122 |
+
"logits/rejected": -0.6645946502685547,
|
| 123 |
+
"logps/chosen": -334.7701721191406,
|
| 124 |
+
"logps/rejected": -392.15496826171875,
|
| 125 |
+
"loss": 0.3889,
|
| 126 |
+
"rewards/accuracies": 0.856249988079071,
|
| 127 |
+
"rewards/chosen": -0.5091000199317932,
|
| 128 |
+
"rewards/margins": 1.100501298904419,
|
| 129 |
+
"rewards/rejected": -1.6096012592315674,
|
| 130 |
+
"step": 70
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 0.6083650190114068,
|
| 134 |
+
"grad_norm": 31.64851560561826,
|
| 135 |
+
"learning_rate": 4.84325667269244e-07,
|
| 136 |
+
"logits/chosen": 0.15650656819343567,
|
| 137 |
+
"logits/rejected": 0.751736044883728,
|
| 138 |
+
"logps/chosen": -311.51922607421875,
|
| 139 |
+
"logps/rejected": -406.87860107421875,
|
| 140 |
+
"loss": 0.3708,
|
| 141 |
+
"rewards/accuracies": 0.84375,
|
| 142 |
+
"rewards/chosen": -0.7787342667579651,
|
| 143 |
+
"rewards/margins": 1.0774469375610352,
|
| 144 |
+
"rewards/rejected": -1.8561809062957764,
|
| 145 |
+
"step": 80
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"epoch": 0.6844106463878327,
|
| 149 |
+
"grad_norm": 39.73186343074897,
|
| 150 |
+
"learning_rate": 4.7565430258740336e-07,
|
| 151 |
+
"logits/chosen": 0.31803542375564575,
|
| 152 |
+
"logits/rejected": 1.7618499994277954,
|
| 153 |
+
"logps/chosen": -330.3303527832031,
|
| 154 |
+
"logps/rejected": -454.4091796875,
|
| 155 |
+
"loss": 0.3342,
|
| 156 |
+
"rewards/accuracies": 0.893750011920929,
|
| 157 |
+
"rewards/chosen": -0.8064614534378052,
|
| 158 |
+
"rewards/margins": 1.6557552814483643,
|
| 159 |
+
"rewards/rejected": -2.46221661567688,
|
| 160 |
+
"step": 90
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 0.7604562737642585,
|
| 164 |
+
"grad_norm": 26.592175518902643,
|
| 165 |
+
"learning_rate": 4.6519683168329195e-07,
|
| 166 |
+
"logits/chosen": 0.48759278655052185,
|
| 167 |
+
"logits/rejected": 2.2304329872131348,
|
| 168 |
+
"logps/chosen": -379.5733947753906,
|
| 169 |
+
"logps/rejected": -496.21636962890625,
|
| 170 |
+
"loss": 0.3147,
|
| 171 |
+
"rewards/accuracies": 0.893750011920929,
|
| 172 |
+
"rewards/chosen": -0.8934251070022583,
|
| 173 |
+
"rewards/margins": 1.828790307044983,
|
| 174 |
+
"rewards/rejected": -2.722215414047241,
|
| 175 |
+
"step": 100
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"epoch": 0.7604562737642585,
|
| 179 |
+
"eval_logits/chosen": 0.7404455542564392,
|
| 180 |
+
"eval_logits/rejected": 1.5949417352676392,
|
| 181 |
+
"eval_logps/chosen": -375.37860107421875,
|
| 182 |
+
"eval_logps/rejected": -459.43609619140625,
|
| 183 |
+
"eval_loss": 0.5617097020149231,
|
| 184 |
+
"eval_rewards/accuracies": 0.732758641242981,
|
| 185 |
+
"eval_rewards/chosen": -1.1541731357574463,
|
| 186 |
+
"eval_rewards/margins": 0.957415759563446,
|
| 187 |
+
"eval_rewards/rejected": -2.111588716506958,
|
| 188 |
+
"eval_runtime": 93.8291,
|
| 189 |
+
"eval_samples_per_second": 19.514,
|
| 190 |
+
"eval_steps_per_second": 0.309,
|
| 191 |
+
"step": 100
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.8365019011406845,
|
| 195 |
+
"grad_norm": 33.75286446177569,
|
| 196 |
+
"learning_rate": 4.530360278682841e-07,
|
| 197 |
+
"logits/chosen": 0.4949573576450348,
|
| 198 |
+
"logits/rejected": 1.976785659790039,
|
| 199 |
+
"logps/chosen": -353.00445556640625,
|
| 200 |
+
"logps/rejected": -476.89678955078125,
|
| 201 |
+
"loss": 0.3642,
|
| 202 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 203 |
+
"rewards/chosen": -0.8960109949111938,
|
| 204 |
+
"rewards/margins": 1.611445665359497,
|
| 205 |
+
"rewards/rejected": -2.5074567794799805,
|
| 206 |
+
"step": 110
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.9125475285171103,
|
| 210 |
+
"grad_norm": 25.522512035563206,
|
| 211 |
+
"learning_rate": 4.3926814672941166e-07,
|
| 212 |
+
"logits/chosen": 0.5252203941345215,
|
| 213 |
+
"logits/rejected": 2.202810525894165,
|
| 214 |
+
"logps/chosen": -359.41876220703125,
|
| 215 |
+
"logps/rejected": -494.12969970703125,
|
| 216 |
+
"loss": 0.318,
|
| 217 |
+
"rewards/accuracies": 0.8687499761581421,
|
| 218 |
+
"rewards/chosen": -1.0113165378570557,
|
| 219 |
+
"rewards/margins": 1.6037089824676514,
|
| 220 |
+
"rewards/rejected": -2.615025520324707,
|
| 221 |
+
"step": 120
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.9885931558935361,
|
| 225 |
+
"grad_norm": 29.14776533666164,
|
| 226 |
+
"learning_rate": 4.240021642440332e-07,
|
| 227 |
+
"logits/chosen": -0.04642736166715622,
|
| 228 |
+
"logits/rejected": 1.8715919256210327,
|
| 229 |
+
"logps/chosen": -352.2441101074219,
|
| 230 |
+
"logps/rejected": -486.38970947265625,
|
| 231 |
+
"loss": 0.3174,
|
| 232 |
+
"rewards/accuracies": 0.875,
|
| 233 |
+
"rewards/chosen": -0.9902883768081665,
|
| 234 |
+
"rewards/margins": 1.7977688312530518,
|
| 235 |
+
"rewards/rejected": -2.7880570888519287,
|
| 236 |
+
"step": 130
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 1.064638783269962,
|
| 240 |
+
"grad_norm": 32.26612817404267,
|
| 241 |
+
"learning_rate": 4.073589142096592e-07,
|
| 242 |
+
"logits/chosen": 0.5508675575256348,
|
| 243 |
+
"logits/rejected": 2.800567150115967,
|
| 244 |
+
"logps/chosen": -352.32440185546875,
|
| 245 |
+
"logps/rejected": -517.09228515625,
|
| 246 |
+
"loss": 0.2352,
|
| 247 |
+
"rewards/accuracies": 0.90625,
|
| 248 |
+
"rewards/chosen": -0.9236076474189758,
|
| 249 |
+
"rewards/margins": 2.010570764541626,
|
| 250 |
+
"rewards/rejected": -2.934178590774536,
|
| 251 |
+
"step": 140
|
| 252 |
+
},
|
| 253 |
+
{
|
| 254 |
+
"epoch": 1.1406844106463878,
|
| 255 |
+
"grad_norm": 24.555758805773927,
|
| 256 |
+
"learning_rate": 3.8947013181637624e-07,
|
| 257 |
+
"logits/chosen": 0.8321820497512817,
|
| 258 |
+
"logits/rejected": 3.650494337081909,
|
| 259 |
+
"logps/chosen": -331.7710266113281,
|
| 260 |
+
"logps/rejected": -498.3985900878906,
|
| 261 |
+
"loss": 0.214,
|
| 262 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 263 |
+
"rewards/chosen": -0.9579178094863892,
|
| 264 |
+
"rewards/margins": 2.1326422691345215,
|
| 265 |
+
"rewards/rejected": -3.0905601978302,
|
| 266 |
+
"step": 150
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 1.1406844106463878,
|
| 270 |
+
"eval_logits/chosen": 0.8076741695404053,
|
| 271 |
+
"eval_logits/rejected": 2.0487349033355713,
|
| 272 |
+
"eval_logps/chosen": -389.57183837890625,
|
| 273 |
+
"eval_logps/rejected": -491.0475158691406,
|
| 274 |
+
"eval_loss": 0.5559520125389099,
|
| 275 |
+
"eval_rewards/accuracies": 0.7456896305084229,
|
| 276 |
+
"eval_rewards/chosen": -1.2961053848266602,
|
| 277 |
+
"eval_rewards/margins": 1.1315982341766357,
|
| 278 |
+
"eval_rewards/rejected": -2.427703619003296,
|
| 279 |
+
"eval_runtime": 93.901,
|
| 280 |
+
"eval_samples_per_second": 19.499,
|
| 281 |
+
"eval_steps_per_second": 0.309,
|
| 282 |
+
"step": 150
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 1.2167300380228137,
|
| 286 |
+
"grad_norm": 23.812525631547885,
|
| 287 |
+
"learning_rate": 3.7047741093221656e-07,
|
| 288 |
+
"logits/chosen": 0.6415907740592957,
|
| 289 |
+
"logits/rejected": 3.418233871459961,
|
| 290 |
+
"logps/chosen": -337.47711181640625,
|
| 291 |
+
"logps/rejected": -565.9837036132812,
|
| 292 |
+
"loss": 0.1915,
|
| 293 |
+
"rewards/accuracies": 0.9312499761581421,
|
| 294 |
+
"rewards/chosen": -0.7956029176712036,
|
| 295 |
+
"rewards/margins": 2.5107922554016113,
|
| 296 |
+
"rewards/rejected": -3.306394577026367,
|
| 297 |
+
"step": 160
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"epoch": 1.2927756653992395,
|
| 301 |
+
"grad_norm": 22.027445926013243,
|
| 302 |
+
"learning_rate": 3.5053108335480205e-07,
|
| 303 |
+
"logits/chosen": 0.43230652809143066,
|
| 304 |
+
"logits/rejected": 3.4372222423553467,
|
| 305 |
+
"logps/chosen": -404.5794982910156,
|
| 306 |
+
"logps/rejected": -585.620849609375,
|
| 307 |
+
"loss": 0.2138,
|
| 308 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 309 |
+
"rewards/chosen": -1.1555755138397217,
|
| 310 |
+
"rewards/margins": 2.351931571960449,
|
| 311 |
+
"rewards/rejected": -3.50750732421875,
|
| 312 |
+
"step": 170
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 1.3688212927756653,
|
| 316 |
+
"grad_norm": 21.8184530010132,
|
| 317 |
+
"learning_rate": 3.29789028900245e-07,
|
| 318 |
+
"logits/chosen": 0.6227339506149292,
|
| 319 |
+
"logits/rejected": 3.597677707672119,
|
| 320 |
+
"logps/chosen": -414.3572692871094,
|
| 321 |
+
"logps/rejected": -602.4873657226562,
|
| 322 |
+
"loss": 0.2101,
|
| 323 |
+
"rewards/accuracies": 0.949999988079071,
|
| 324 |
+
"rewards/chosen": -1.1512380838394165,
|
| 325 |
+
"rewards/margins": 2.4040679931640625,
|
| 326 |
+
"rewards/rejected": -3.5553059577941895,
|
| 327 |
+
"step": 180
|
| 328 |
+
},
|
| 329 |
+
{
|
| 330 |
+
"epoch": 1.4448669201520912,
|
| 331 |
+
"grad_norm": 30.492941117704806,
|
| 332 |
+
"learning_rate": 3.084154257477301e-07,
|
| 333 |
+
"logits/chosen": 1.2600494623184204,
|
| 334 |
+
"logits/rejected": 3.9732604026794434,
|
| 335 |
+
"logps/chosen": -392.4749450683594,
|
| 336 |
+
"logps/rejected": -593.7326049804688,
|
| 337 |
+
"loss": 0.1835,
|
| 338 |
+
"rewards/accuracies": 0.893750011920929,
|
| 339 |
+
"rewards/chosen": -1.3306286334991455,
|
| 340 |
+
"rewards/margins": 2.4590394496917725,
|
| 341 |
+
"rewards/rejected": -3.7896676063537598,
|
| 342 |
+
"step": 190
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 1.5209125475285172,
|
| 346 |
+
"grad_norm": 18.187832229568688,
|
| 347 |
+
"learning_rate": 2.865794509310888e-07,
|
| 348 |
+
"logits/chosen": 1.334993839263916,
|
| 349 |
+
"logits/rejected": 4.3924736976623535,
|
| 350 |
+
"logps/chosen": -390.69781494140625,
|
| 351 |
+
"logps/rejected": -616.49951171875,
|
| 352 |
+
"loss": 0.1866,
|
| 353 |
+
"rewards/accuracies": 0.9375,
|
| 354 |
+
"rewards/chosen": -1.2437045574188232,
|
| 355 |
+
"rewards/margins": 2.7270522117614746,
|
| 356 |
+
"rewards/rejected": -3.9707565307617188,
|
| 357 |
+
"step": 200
|
| 358 |
+
},
|
| 359 |
+
{
|
| 360 |
+
"epoch": 1.5209125475285172,
|
| 361 |
+
"eval_logits/chosen": 0.7928547859191895,
|
| 362 |
+
"eval_logits/rejected": 2.0297083854675293,
|
| 363 |
+
"eval_logps/chosen": -404.8973388671875,
|
| 364 |
+
"eval_logps/rejected": -510.6939697265625,
|
| 365 |
+
"eval_loss": 0.5364186763763428,
|
| 366 |
+
"eval_rewards/accuracies": 0.7413793206214905,
|
| 367 |
+
"eval_rewards/chosen": -1.4493602514266968,
|
| 368 |
+
"eval_rewards/margins": 1.1748069524765015,
|
| 369 |
+
"eval_rewards/rejected": -2.624166965484619,
|
| 370 |
+
"eval_runtime": 93.5108,
|
| 371 |
+
"eval_samples_per_second": 19.581,
|
| 372 |
+
"eval_steps_per_second": 0.31,
|
| 373 |
+
"step": 200
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"epoch": 1.5969581749049429,
|
| 377 |
+
"grad_norm": 30.203312935912965,
|
| 378 |
+
"learning_rate": 2.644539412632786e-07,
|
| 379 |
+
"logits/chosen": 0.7376717329025269,
|
| 380 |
+
"logits/rejected": 3.46760630607605,
|
| 381 |
+
"logps/chosen": -381.1366882324219,
|
| 382 |
+
"logps/rejected": -576.8548583984375,
|
| 383 |
+
"loss": 0.2112,
|
| 384 |
+
"rewards/accuracies": 0.918749988079071,
|
| 385 |
+
"rewards/chosen": -1.2630369663238525,
|
| 386 |
+
"rewards/margins": 2.4375810623168945,
|
| 387 |
+
"rewards/rejected": -3.700617551803589,
|
| 388 |
+
"step": 210
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 1.673003802281369,
|
| 392 |
+
"grad_norm": 34.60056674030066,
|
| 393 |
+
"learning_rate": 2.422140252928601e-07,
|
| 394 |
+
"logits/chosen": 1.3862824440002441,
|
| 395 |
+
"logits/rejected": 3.9012253284454346,
|
| 396 |
+
"logps/chosen": -363.8896484375,
|
| 397 |
+
"logps/rejected": -638.3570556640625,
|
| 398 |
+
"loss": 0.2026,
|
| 399 |
+
"rewards/accuracies": 0.949999988079071,
|
| 400 |
+
"rewards/chosen": -1.3880947828292847,
|
| 401 |
+
"rewards/margins": 2.823024272918701,
|
| 402 |
+
"rewards/rejected": -4.211119174957275,
|
| 403 |
+
"step": 220
|
| 404 |
+
},
|
| 405 |
+
{
|
| 406 |
+
"epoch": 1.7490494296577945,
|
| 407 |
+
"grad_norm": 17.75951807278943,
|
| 408 |
+
"learning_rate": 2.2003573712085455e-07,
|
| 409 |
+
"logits/chosen": 0.32914024591445923,
|
| 410 |
+
"logits/rejected": 3.113135576248169,
|
| 411 |
+
"logps/chosen": -353.28631591796875,
|
| 412 |
+
"logps/rejected": -597.5838012695312,
|
| 413 |
+
"loss": 0.1802,
|
| 414 |
+
"rewards/accuracies": 0.981249988079071,
|
| 415 |
+
"rewards/chosen": -1.1312899589538574,
|
| 416 |
+
"rewards/margins": 2.7755885124206543,
|
| 417 |
+
"rewards/rejected": -3.906878709793091,
|
| 418 |
+
"step": 230
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"epoch": 1.8250950570342206,
|
| 422 |
+
"grad_norm": 21.2237904198368,
|
| 423 |
+
"learning_rate": 1.980946230499431e-07,
|
| 424 |
+
"logits/chosen": 0.5761924982070923,
|
| 425 |
+
"logits/rejected": 3.6894748210906982,
|
| 426 |
+
"logps/chosen": -369.281982421875,
|
| 427 |
+
"logps/rejected": -639.9312133789062,
|
| 428 |
+
"loss": 0.1863,
|
| 429 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 430 |
+
"rewards/chosen": -1.4418867826461792,
|
| 431 |
+
"rewards/margins": 2.867917776107788,
|
| 432 |
+
"rewards/rejected": -4.309804439544678,
|
| 433 |
+
"step": 240
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 1.9011406844106464,
|
| 437 |
+
"grad_norm": 25.460606683183755,
|
| 438 |
+
"learning_rate": 1.7656435209470376e-07,
|
| 439 |
+
"logits/chosen": 1.0737273693084717,
|
| 440 |
+
"logits/rejected": 3.9221127033233643,
|
| 441 |
+
"logps/chosen": -389.84417724609375,
|
| 442 |
+
"logps/rejected": -633.5592041015625,
|
| 443 |
+
"loss": 0.1899,
|
| 444 |
+
"rewards/accuracies": 0.9375,
|
| 445 |
+
"rewards/chosen": -1.4279615879058838,
|
| 446 |
+
"rewards/margins": 2.770998954772949,
|
| 447 |
+
"rewards/rejected": -4.198960304260254,
|
| 448 |
+
"step": 250
|
| 449 |
+
},
|
| 450 |
+
{
|
| 451 |
+
"epoch": 1.9011406844106464,
|
| 452 |
+
"eval_logits/chosen": 1.2524254322052002,
|
| 453 |
+
"eval_logits/rejected": 2.571428060531616,
|
| 454 |
+
"eval_logps/chosen": -438.79302978515625,
|
| 455 |
+
"eval_logps/rejected": -556.13232421875,
|
| 456 |
+
"eval_loss": 0.5391203165054321,
|
| 457 |
+
"eval_rewards/accuracies": 0.7456896305084229,
|
| 458 |
+
"eval_rewards/chosen": -1.7883172035217285,
|
| 459 |
+
"eval_rewards/margins": 1.2902345657348633,
|
| 460 |
+
"eval_rewards/rejected": -3.078551769256592,
|
| 461 |
+
"eval_runtime": 96.6165,
|
| 462 |
+
"eval_samples_per_second": 18.951,
|
| 463 |
+
"eval_steps_per_second": 0.3,
|
| 464 |
+
"step": 250
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 1.9771863117870723,
|
| 468 |
+
"grad_norm": 28.509116337681686,
|
| 469 |
+
"learning_rate": 1.5561534135101884e-07,
|
| 470 |
+
"logits/chosen": 1.822516679763794,
|
| 471 |
+
"logits/rejected": 4.293866157531738,
|
| 472 |
+
"logps/chosen": -417.22503662109375,
|
| 473 |
+
"logps/rejected": -664.3339233398438,
|
| 474 |
+
"loss": 0.1979,
|
| 475 |
+
"rewards/accuracies": 0.9375,
|
| 476 |
+
"rewards/chosen": -1.6485140323638916,
|
| 477 |
+
"rewards/margins": 2.746192455291748,
|
| 478 |
+
"rewards/rejected": -4.394707202911377,
|
| 479 |
+
"step": 260
|
| 480 |
+
},
|
| 481 |
+
{
|
| 482 |
+
"epoch": 2.053231939163498,
|
| 483 |
+
"grad_norm": 9.761431912302397,
|
| 484 |
+
"learning_rate": 1.3541340710517546e-07,
|
| 485 |
+
"logits/chosen": 1.401188611984253,
|
| 486 |
+
"logits/rejected": 4.360350608825684,
|
| 487 |
+
"logps/chosen": -416.37481689453125,
|
| 488 |
+
"logps/rejected": -680.0263671875,
|
| 489 |
+
"loss": 0.1217,
|
| 490 |
+
"rewards/accuracies": 0.9375,
|
| 491 |
+
"rewards/chosen": -1.4166698455810547,
|
| 492 |
+
"rewards/margins": 2.9713897705078125,
|
| 493 |
+
"rewards/rejected": -4.388059616088867,
|
| 494 |
+
"step": 270
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 2.129277566539924,
|
| 498 |
+
"grad_norm": 19.479793936953538,
|
| 499 |
+
"learning_rate": 1.1611845235944143e-07,
|
| 500 |
+
"logits/chosen": 2.3020176887512207,
|
| 501 |
+
"logits/rejected": 5.1446309089660645,
|
| 502 |
+
"logps/chosen": -407.40936279296875,
|
| 503 |
+
"logps/rejected": -718.0748901367188,
|
| 504 |
+
"loss": 0.1082,
|
| 505 |
+
"rewards/accuracies": 0.981249988079071,
|
| 506 |
+
"rewards/chosen": -1.5115373134613037,
|
| 507 |
+
"rewards/margins": 3.4099037647247314,
|
| 508 |
+
"rewards/rejected": -4.921441555023193,
|
| 509 |
+
"step": 280
|
| 510 |
+
},
|
| 511 |
+
{
|
| 512 |
+
"epoch": 2.20532319391635,
|
| 513 |
+
"grad_norm": 12.421604139145975,
|
| 514 |
+
"learning_rate": 9.788320116265892e-08,
|
| 515 |
+
"logits/chosen": 1.9953467845916748,
|
| 516 |
+
"logits/rejected": 5.533808708190918,
|
| 517 |
+
"logps/chosen": -459.0770568847656,
|
| 518 |
+
"logps/rejected": -714.7064208984375,
|
| 519 |
+
"loss": 0.112,
|
| 520 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 521 |
+
"rewards/chosen": -1.560991883277893,
|
| 522 |
+
"rewards/margins": 3.3094687461853027,
|
| 523 |
+
"rewards/rejected": -4.8704609870910645,
|
| 524 |
+
"step": 290
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"epoch": 2.2813688212927756,
|
| 528 |
+
"grad_norm": 17.541719244774725,
|
| 529 |
+
"learning_rate": 8.085198976392124e-08,
|
| 530 |
+
"logits/chosen": 2.661414384841919,
|
| 531 |
+
"logits/rejected": 5.309275150299072,
|
| 532 |
+
"logps/chosen": -396.86163330078125,
|
| 533 |
+
"logps/rejected": -682.7183837890625,
|
| 534 |
+
"loss": 0.1083,
|
| 535 |
+
"rewards/accuracies": 0.956250011920929,
|
| 536 |
+
"rewards/chosen": -1.661924958229065,
|
| 537 |
+
"rewards/margins": 3.159823179244995,
|
| 538 |
+
"rewards/rejected": -4.821747779846191,
|
| 539 |
+
"step": 300
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 2.2813688212927756,
|
| 543 |
+
"eval_logits/chosen": 2.2903313636779785,
|
| 544 |
+
"eval_logits/rejected": 3.6920409202575684,
|
| 545 |
+
"eval_logps/chosen": -470.052734375,
|
| 546 |
+
"eval_logps/rejected": -601.4120483398438,
|
| 547 |
+
"eval_loss": 0.558447539806366,
|
| 548 |
+
"eval_rewards/accuracies": 0.7629310488700867,
|
| 549 |
+
"eval_rewards/chosen": -2.100914239883423,
|
| 550 |
+
"eval_rewards/margins": 1.4304344654083252,
|
| 551 |
+
"eval_rewards/rejected": -3.531348466873169,
|
| 552 |
+
"eval_runtime": 94.198,
|
| 553 |
+
"eval_samples_per_second": 19.438,
|
| 554 |
+
"eval_steps_per_second": 0.308,
|
| 555 |
+
"step": 300
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 2.3574144486692017,
|
| 559 |
+
"grad_norm": 11.598664016084259,
|
| 560 |
+
"learning_rate": 6.515962415763369e-08,
|
| 561 |
+
"logits/chosen": 2.566627025604248,
|
| 562 |
+
"logits/rejected": 5.178008079528809,
|
| 563 |
+
"logps/chosen": -364.78521728515625,
|
| 564 |
+
"logps/rejected": -709.1622314453125,
|
| 565 |
+
"loss": 0.0941,
|
| 566 |
+
"rewards/accuracies": 0.987500011920929,
|
| 567 |
+
"rewards/chosen": -1.6015697717666626,
|
| 568 |
+
"rewards/margins": 3.526782989501953,
|
| 569 |
+
"rewards/rejected": -5.128352165222168,
|
| 570 |
+
"step": 310
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"epoch": 2.4334600760456273,
|
| 574 |
+
"grad_norm": 22.496239920192775,
|
| 575 |
+
"learning_rate": 5.093031306275308e-08,
|
| 576 |
+
"logits/chosen": 2.157808303833008,
|
| 577 |
+
"logits/rejected": 5.751224040985107,
|
| 578 |
+
"logps/chosen": -454.247314453125,
|
| 579 |
+
"logps/rejected": -807.82275390625,
|
| 580 |
+
"loss": 0.1048,
|
| 581 |
+
"rewards/accuracies": 0.987500011920929,
|
| 582 |
+
"rewards/chosen": -1.797298789024353,
|
| 583 |
+
"rewards/margins": 3.7231411933898926,
|
| 584 |
+
"rewards/rejected": -5.520440101623535,
|
| 585 |
+
"step": 320
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 2.5095057034220534,
|
| 589 |
+
"grad_norm": 13.016382898289478,
|
| 590 |
+
"learning_rate": 3.827668478192578e-08,
|
| 591 |
+
"logits/chosen": 2.29127836227417,
|
| 592 |
+
"logits/rejected": 5.964070796966553,
|
| 593 |
+
"logps/chosen": -475.0154724121094,
|
| 594 |
+
"logps/rejected": -776.1507568359375,
|
| 595 |
+
"loss": 0.0961,
|
| 596 |
+
"rewards/accuracies": 0.96875,
|
| 597 |
+
"rewards/chosen": -1.8038787841796875,
|
| 598 |
+
"rewards/margins": 3.5154175758361816,
|
| 599 |
+
"rewards/rejected": -5.319295883178711,
|
| 600 |
+
"step": 330
|
| 601 |
+
},
|
| 602 |
+
{
|
| 603 |
+
"epoch": 2.585551330798479,
|
| 604 |
+
"grad_norm": 18.59300730528273,
|
| 605 |
+
"learning_rate": 2.729889572230856e-08,
|
| 606 |
+
"logits/chosen": 2.6335761547088623,
|
| 607 |
+
"logits/rejected": 6.27906608581543,
|
| 608 |
+
"logps/chosen": -456.11676025390625,
|
| 609 |
+
"logps/rejected": -846.2113037109375,
|
| 610 |
+
"loss": 0.0807,
|
| 611 |
+
"rewards/accuracies": 0.987500011920929,
|
| 612 |
+
"rewards/chosen": -2.0658538341522217,
|
| 613 |
+
"rewards/margins": 4.058461666107178,
|
| 614 |
+
"rewards/rejected": -6.1243157386779785,
|
| 615 |
+
"step": 340
|
| 616 |
+
},
|
| 617 |
+
{
|
| 618 |
+
"epoch": 2.661596958174905,
|
| 619 |
+
"grad_norm": 13.300072255332687,
|
| 620 |
+
"learning_rate": 1.8083837634341766e-08,
|
| 621 |
+
"logits/chosen": 2.266738176345825,
|
| 622 |
+
"logits/rejected": 6.4141387939453125,
|
| 623 |
+
"logps/chosen": -500.92901611328125,
|
| 624 |
+
"logps/rejected": -874.1981201171875,
|
| 625 |
+
"loss": 0.0955,
|
| 626 |
+
"rewards/accuracies": 1.0,
|
| 627 |
+
"rewards/chosen": -2.200644016265869,
|
| 628 |
+
"rewards/margins": 4.188169956207275,
|
| 629 |
+
"rewards/rejected": -6.3888139724731445,
|
| 630 |
+
"step": 350
|
| 631 |
+
},
|
| 632 |
+
{
|
| 633 |
+
"epoch": 2.661596958174905,
|
| 634 |
+
"eval_logits/chosen": 2.532045602798462,
|
| 635 |
+
"eval_logits/rejected": 4.256929397583008,
|
| 636 |
+
"eval_logps/chosen": -544.331787109375,
|
| 637 |
+
"eval_logps/rejected": -697.9909057617188,
|
| 638 |
+
"eval_loss": 0.5829063653945923,
|
| 639 |
+
"eval_rewards/accuracies": 0.7629310488700867,
|
| 640 |
+
"eval_rewards/chosen": -2.8437047004699707,
|
| 641 |
+
"eval_rewards/margins": 1.6534322500228882,
|
| 642 |
+
"eval_rewards/rejected": -4.49713659286499,
|
| 643 |
+
"eval_runtime": 93.5502,
|
| 644 |
+
"eval_samples_per_second": 19.572,
|
| 645 |
+
"eval_steps_per_second": 0.31,
|
| 646 |
+
"step": 350
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"epoch": 2.7376425855513307,
|
| 650 |
+
"grad_norm": 14.902459964324052,
|
| 651 |
+
"learning_rate": 1.0704449843359498e-08,
|
| 652 |
+
"logits/chosen": 2.454040050506592,
|
| 653 |
+
"logits/rejected": 6.299553871154785,
|
| 654 |
+
"logps/chosen": -483.86669921875,
|
| 655 |
+
"logps/rejected": -877.189453125,
|
| 656 |
+
"loss": 0.0847,
|
| 657 |
+
"rewards/accuracies": 1.0,
|
| 658 |
+
"rewards/chosen": -2.1329433917999268,
|
| 659 |
+
"rewards/margins": 4.206452369689941,
|
| 660 |
+
"rewards/rejected": -6.3393964767456055,
|
| 661 |
+
"step": 360
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 2.8136882129277567,
|
| 665 |
+
"grad_norm": 15.13819700442962,
|
| 666 |
+
"learning_rate": 5.2191419178871935e-09,
|
| 667 |
+
"logits/chosen": 2.5856270790100098,
|
| 668 |
+
"logits/rejected": 6.600296974182129,
|
| 669 |
+
"logps/chosen": -461.5941467285156,
|
| 670 |
+
"logps/rejected": -831.21240234375,
|
| 671 |
+
"loss": 0.0833,
|
| 672 |
+
"rewards/accuracies": 0.96875,
|
| 673 |
+
"rewards/chosen": -2.348266363143921,
|
| 674 |
+
"rewards/margins": 4.047614574432373,
|
| 675 |
+
"rewards/rejected": -6.395880699157715,
|
| 676 |
+
"step": 370
|
| 677 |
+
},
|
| 678 |
+
{
|
| 679 |
+
"epoch": 2.8897338403041823,
|
| 680 |
+
"grad_norm": 12.919733942557965,
|
| 681 |
+
"learning_rate": 1.6713313443384724e-09,
|
| 682 |
+
"logits/chosen": 2.7355175018310547,
|
| 683 |
+
"logits/rejected": 6.480588436126709,
|
| 684 |
+
"logps/chosen": -487.62274169921875,
|
| 685 |
+
"logps/rejected": -859.5900268554688,
|
| 686 |
+
"loss": 0.0859,
|
| 687 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 688 |
+
"rewards/chosen": -2.41930890083313,
|
| 689 |
+
"rewards/margins": 3.938915967941284,
|
| 690 |
+
"rewards/rejected": -6.358224391937256,
|
| 691 |
+
"step": 380
|
| 692 |
+
},
|
| 693 |
+
{
|
| 694 |
+
"epoch": 2.9657794676806084,
|
| 695 |
+
"grad_norm": 10.73491870552627,
|
| 696 |
+
"learning_rate": 8.909986752470012e-11,
|
| 697 |
+
"logits/chosen": 2.8181276321411133,
|
| 698 |
+
"logits/rejected": 6.899574279785156,
|
| 699 |
+
"logps/chosen": -519.549560546875,
|
| 700 |
+
"logps/rejected": -837.3341674804688,
|
| 701 |
+
"loss": 0.0796,
|
| 702 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 703 |
+
"rewards/chosen": -2.500697612762451,
|
| 704 |
+
"rewards/margins": 3.8103880882263184,
|
| 705 |
+
"rewards/rejected": -6.3110857009887695,
|
| 706 |
+
"step": 390
|
| 707 |
+
},
|
| 708 |
+
{
|
| 709 |
+
"epoch": 2.988593155893536,
|
| 710 |
+
"step": 393,
|
| 711 |
+
"total_flos": 0.0,
|
| 712 |
+
"train_loss": 0.2546291965564699,
|
| 713 |
+
"train_runtime": 10353.4149,
|
| 714 |
+
"train_samples_per_second": 4.868,
|
| 715 |
+
"train_steps_per_second": 0.038
|
| 716 |
+
}
|
| 717 |
+
],
|
| 718 |
+
"logging_steps": 10,
|
| 719 |
+
"max_steps": 393,
|
| 720 |
+
"num_input_tokens_seen": 0,
|
| 721 |
+
"num_train_epochs": 3,
|
| 722 |
+
"save_steps": 100,
|
| 723 |
+
"stateful_callbacks": {
|
| 724 |
+
"TrainerControl": {
|
| 725 |
+
"args": {
|
| 726 |
+
"should_epoch_stop": false,
|
| 727 |
+
"should_evaluate": false,
|
| 728 |
+
"should_log": false,
|
| 729 |
+
"should_save": true,
|
| 730 |
+
"should_training_stop": true
|
| 731 |
+
},
|
| 732 |
+
"attributes": {}
|
| 733 |
+
}
|
| 734 |
+
},
|
| 735 |
+
"total_flos": 0.0,
|
| 736 |
+
"train_batch_size": 8,
|
| 737 |
+
"trial_name": null,
|
| 738 |
+
"trial_params": null
|
| 739 |
+
}
|