Commit
•
9a6d085
1
Parent(s):
00cc72a
Model save
Browse files- README.md +75 -0
- all_results.json +9 -0
- generation_config.json +6 -0
- train_results.json +9 -0
- trainer_state.json +826 -0
README.md
ADDED
|
@@ -0,0 +1,75 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: transformers
|
| 3 |
+
tags:
|
| 4 |
+
- trl
|
| 5 |
+
- dpo
|
| 6 |
+
- generated_from_trainer
|
| 7 |
+
model-index:
|
| 8 |
+
- name: zephyr-7b-dpo-full
|
| 9 |
+
results: []
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 13 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 14 |
+
|
| 15 |
+
# zephyr-7b-dpo-full
|
| 16 |
+
|
| 17 |
+
This model was trained from scratch on an unknown dataset.
|
| 18 |
+
It achieves the following results on the evaluation set:
|
| 19 |
+
- Loss: 0.5225
|
| 20 |
+
- Rewards/chosen: -0.5814
|
| 21 |
+
- Rewards/rejected: -1.3885
|
| 22 |
+
- Rewards/accuracies: 0.7773
|
| 23 |
+
- Rewards/margins: 0.8071
|
| 24 |
+
- Logps/rejected: -403.6498
|
| 25 |
+
- Logps/chosen: -323.4799
|
| 26 |
+
- Logits/rejected: 1.6879
|
| 27 |
+
- Logits/chosen: 1.2050
|
| 28 |
+
|
| 29 |
+
## Model description
|
| 30 |
+
|
| 31 |
+
More information needed
|
| 32 |
+
|
| 33 |
+
## Intended uses & limitations
|
| 34 |
+
|
| 35 |
+
More information needed
|
| 36 |
+
|
| 37 |
+
## Training and evaluation data
|
| 38 |
+
|
| 39 |
+
More information needed
|
| 40 |
+
|
| 41 |
+
## Training procedure
|
| 42 |
+
|
| 43 |
+
### Training hyperparameters
|
| 44 |
+
|
| 45 |
+
The following hyperparameters were used during training:
|
| 46 |
+
- learning_rate: 5e-07
|
| 47 |
+
- train_batch_size: 8
|
| 48 |
+
- eval_batch_size: 8
|
| 49 |
+
- seed: 42
|
| 50 |
+
- distributed_type: multi-GPU
|
| 51 |
+
- num_devices: 8
|
| 52 |
+
- gradient_accumulation_steps: 2
|
| 53 |
+
- total_train_batch_size: 128
|
| 54 |
+
- total_eval_batch_size: 64
|
| 55 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 56 |
+
- lr_scheduler_type: cosine
|
| 57 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 58 |
+
- num_epochs: 1
|
| 59 |
+
|
| 60 |
+
### Training results
|
| 61 |
+
|
| 62 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 63 |
+
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 64 |
+
| 0.5999 | 0.2092 | 100 | 0.5966 | -0.3739 | -0.7450 | 0.7109 | 0.3711 | -339.3055 | -302.7284 | -1.6763 | -1.7949 |
|
| 65 |
+
| 0.5606 | 0.4184 | 200 | 0.5555 | -0.6215 | -1.2285 | 0.7461 | 0.6069 | -387.6506 | -327.4932 | 1.5409 | 1.1102 |
|
| 66 |
+
| 0.5211 | 0.6276 | 300 | 0.5284 | -0.6372 | -1.3932 | 0.7617 | 0.7559 | -404.1211 | -329.0649 | 1.2079 | 0.7847 |
|
| 67 |
+
| 0.5189 | 0.8368 | 400 | 0.5225 | -0.5814 | -1.3885 | 0.7773 | 0.8071 | -403.6498 | -323.4799 | 1.6879 | 1.2050 |
|
| 68 |
+
|
| 69 |
+
|
| 70 |
+
### Framework versions
|
| 71 |
+
|
| 72 |
+
- Transformers 4.45.2
|
| 73 |
+
- Pytorch 2.1.0
|
| 74 |
+
- Datasets 3.0.1
|
| 75 |
+
- Tokenizers 0.20.1
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5574632548886862,
|
| 5 |
+
"train_runtime": 4290.8569,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 14.248,
|
| 8 |
+
"train_steps_per_second": 0.111
|
| 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": "4.45.2"
|
| 6 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5574632548886862,
|
| 5 |
+
"train_runtime": 4290.8569,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 14.248,
|
| 8 |
+
"train_steps_per_second": 0.111
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,826 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 1.0,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 478,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.0020920502092050207,
|
| 13 |
+
"grad_norm": 9.102261763693875,
|
| 14 |
+
"learning_rate": 1.0416666666666666e-08,
|
| 15 |
+
"logits/chosen": -2.598843574523926,
|
| 16 |
+
"logits/rejected": -2.5780863761901855,
|
| 17 |
+
"logps/chosen": -269.870849609375,
|
| 18 |
+
"logps/rejected": -361.686767578125,
|
| 19 |
+
"loss": 0.6931,
|
| 20 |
+
"rewards/accuracies": 0.0,
|
| 21 |
+
"rewards/chosen": 0.0,
|
| 22 |
+
"rewards/margins": 0.0,
|
| 23 |
+
"rewards/rejected": 0.0,
|
| 24 |
+
"step": 1
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.02092050209205021,
|
| 28 |
+
"grad_norm": 9.071970452275611,
|
| 29 |
+
"learning_rate": 1.0416666666666667e-07,
|
| 30 |
+
"logits/chosen": -2.4106454849243164,
|
| 31 |
+
"logits/rejected": -2.3727071285247803,
|
| 32 |
+
"logps/chosen": -269.8708801269531,
|
| 33 |
+
"logps/rejected": -254.9169158935547,
|
| 34 |
+
"loss": 0.6932,
|
| 35 |
+
"rewards/accuracies": 0.4236111044883728,
|
| 36 |
+
"rewards/chosen": 3.945813296013512e-05,
|
| 37 |
+
"rewards/margins": -0.00021239757188595831,
|
| 38 |
+
"rewards/rejected": 0.0002518556430004537,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.04184100418410042,
|
| 43 |
+
"grad_norm": 10.775263417671352,
|
| 44 |
+
"learning_rate": 2.0833333333333333e-07,
|
| 45 |
+
"logits/chosen": -2.4717860221862793,
|
| 46 |
+
"logits/rejected": -2.407602310180664,
|
| 47 |
+
"logps/chosen": -284.3338928222656,
|
| 48 |
+
"logps/rejected": -297.76251220703125,
|
| 49 |
+
"loss": 0.6925,
|
| 50 |
+
"rewards/accuracies": 0.5249999761581421,
|
| 51 |
+
"rewards/chosen": 0.004285941831767559,
|
| 52 |
+
"rewards/margins": 0.0006065635825507343,
|
| 53 |
+
"rewards/rejected": 0.00367937795817852,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.06276150627615062,
|
| 58 |
+
"grad_norm": 7.642238787668073,
|
| 59 |
+
"learning_rate": 3.1249999999999997e-07,
|
| 60 |
+
"logits/chosen": -2.5130527019500732,
|
| 61 |
+
"logits/rejected": -2.4189674854278564,
|
| 62 |
+
"logps/chosen": -302.9613952636719,
|
| 63 |
+
"logps/rejected": -263.80633544921875,
|
| 64 |
+
"loss": 0.6894,
|
| 65 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 66 |
+
"rewards/chosen": 0.024250635877251625,
|
| 67 |
+
"rewards/margins": 0.0054460749961435795,
|
| 68 |
+
"rewards/rejected": 0.018804559484124184,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.08368200836820083,
|
| 73 |
+
"grad_norm": 7.769592061468885,
|
| 74 |
+
"learning_rate": 4.1666666666666667e-07,
|
| 75 |
+
"logits/chosen": -2.4025068283081055,
|
| 76 |
+
"logits/rejected": -2.3595290184020996,
|
| 77 |
+
"logps/chosen": -268.82928466796875,
|
| 78 |
+
"logps/rejected": -243.6490936279297,
|
| 79 |
+
"loss": 0.6827,
|
| 80 |
+
"rewards/accuracies": 0.737500011920929,
|
| 81 |
+
"rewards/chosen": 0.050604142248630524,
|
| 82 |
+
"rewards/margins": 0.028540844097733498,
|
| 83 |
+
"rewards/rejected": 0.022063296288251877,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.10460251046025104,
|
| 88 |
+
"grad_norm": 7.836365186344556,
|
| 89 |
+
"learning_rate": 4.999733114418725e-07,
|
| 90 |
+
"logits/chosen": -2.3352060317993164,
|
| 91 |
+
"logits/rejected": -2.3020994663238525,
|
| 92 |
+
"logps/chosen": -283.0472717285156,
|
| 93 |
+
"logps/rejected": -297.33740234375,
|
| 94 |
+
"loss": 0.6733,
|
| 95 |
+
"rewards/accuracies": 0.637499988079071,
|
| 96 |
+
"rewards/chosen": 0.014853554777801037,
|
| 97 |
+
"rewards/margins": 0.029812458902597427,
|
| 98 |
+
"rewards/rejected": -0.014958900399506092,
|
| 99 |
+
"step": 50
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.12552301255230125,
|
| 103 |
+
"grad_norm": 15.966435105260237,
|
| 104 |
+
"learning_rate": 4.990398100856366e-07,
|
| 105 |
+
"logits/chosen": -2.396578311920166,
|
| 106 |
+
"logits/rejected": -2.329003095626831,
|
| 107 |
+
"logps/chosen": -274.9017333984375,
|
| 108 |
+
"logps/rejected": -308.89105224609375,
|
| 109 |
+
"loss": 0.6527,
|
| 110 |
+
"rewards/accuracies": 0.699999988079071,
|
| 111 |
+
"rewards/chosen": 0.003318754257634282,
|
| 112 |
+
"rewards/margins": 0.09059403091669083,
|
| 113 |
+
"rewards/rejected": -0.08727528154850006,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.14644351464435146,
|
| 118 |
+
"grad_norm": 9.243487328642475,
|
| 119 |
+
"learning_rate": 4.967775735898179e-07,
|
| 120 |
+
"logits/chosen": -2.244589328765869,
|
| 121 |
+
"logits/rejected": -2.24755859375,
|
| 122 |
+
"logps/chosen": -278.61004638671875,
|
| 123 |
+
"logps/rejected": -280.5500183105469,
|
| 124 |
+
"loss": 0.6348,
|
| 125 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 126 |
+
"rewards/chosen": -0.08607344329357147,
|
| 127 |
+
"rewards/margins": 0.11931805312633514,
|
| 128 |
+
"rewards/rejected": -0.20539149641990662,
|
| 129 |
+
"step": 70
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.16736401673640167,
|
| 133 |
+
"grad_norm": 14.953704290427632,
|
| 134 |
+
"learning_rate": 4.931986719649298e-07,
|
| 135 |
+
"logits/chosen": -2.3784019947052,
|
| 136 |
+
"logits/rejected": -2.2930545806884766,
|
| 137 |
+
"logps/chosen": -330.93719482421875,
|
| 138 |
+
"logps/rejected": -288.47344970703125,
|
| 139 |
+
"loss": 0.6192,
|
| 140 |
+
"rewards/accuracies": 0.668749988079071,
|
| 141 |
+
"rewards/chosen": -0.03501243516802788,
|
| 142 |
+
"rewards/margins": 0.18967404961585999,
|
| 143 |
+
"rewards/rejected": -0.22468650341033936,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.18828451882845187,
|
| 148 |
+
"grad_norm": 15.867488807966232,
|
| 149 |
+
"learning_rate": 4.883222001996351e-07,
|
| 150 |
+
"logits/chosen": -2.0707411766052246,
|
| 151 |
+
"logits/rejected": -1.9940516948699951,
|
| 152 |
+
"logps/chosen": -288.53009033203125,
|
| 153 |
+
"logps/rejected": -311.29840087890625,
|
| 154 |
+
"loss": 0.6004,
|
| 155 |
+
"rewards/accuracies": 0.768750011920929,
|
| 156 |
+
"rewards/chosen": -0.2625655233860016,
|
| 157 |
+
"rewards/margins": 0.3136993646621704,
|
| 158 |
+
"rewards/rejected": -0.5762647986412048,
|
| 159 |
+
"step": 90
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.20920502092050208,
|
| 163 |
+
"grad_norm": 13.25483251042058,
|
| 164 |
+
"learning_rate": 4.821741763807186e-07,
|
| 165 |
+
"logits/chosen": -1.913400411605835,
|
| 166 |
+
"logits/rejected": -1.7971922159194946,
|
| 167 |
+
"logps/chosen": -292.3438415527344,
|
| 168 |
+
"logps/rejected": -316.77410888671875,
|
| 169 |
+
"loss": 0.5999,
|
| 170 |
+
"rewards/accuracies": 0.6875,
|
| 171 |
+
"rewards/chosen": -0.2575649917125702,
|
| 172 |
+
"rewards/margins": 0.323893278837204,
|
| 173 |
+
"rewards/rejected": -0.5814582705497742,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 0.20920502092050208,
|
| 178 |
+
"eval_logits/chosen": -1.794881820678711,
|
| 179 |
+
"eval_logits/rejected": -1.6762588024139404,
|
| 180 |
+
"eval_logps/chosen": -302.7283935546875,
|
| 181 |
+
"eval_logps/rejected": -339.30548095703125,
|
| 182 |
+
"eval_loss": 0.5966345071792603,
|
| 183 |
+
"eval_rewards/accuracies": 0.7109375,
|
| 184 |
+
"eval_rewards/chosen": -0.37388285994529724,
|
| 185 |
+
"eval_rewards/margins": 0.371126264333725,
|
| 186 |
+
"eval_rewards/rejected": -0.7450091242790222,
|
| 187 |
+
"eval_runtime": 53.5348,
|
| 188 |
+
"eval_samples_per_second": 37.359,
|
| 189 |
+
"eval_steps_per_second": 0.598,
|
| 190 |
+
"step": 100
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 0.2301255230125523,
|
| 194 |
+
"grad_norm": 15.2172693582291,
|
| 195 |
+
"learning_rate": 4.747874028753375e-07,
|
| 196 |
+
"logits/chosen": -1.628220796585083,
|
| 197 |
+
"logits/rejected": -1.3428412675857544,
|
| 198 |
+
"logps/chosen": -351.1053466796875,
|
| 199 |
+
"logps/rejected": -339.97393798828125,
|
| 200 |
+
"loss": 0.5893,
|
| 201 |
+
"rewards/accuracies": 0.699999988079071,
|
| 202 |
+
"rewards/chosen": -0.41141095757484436,
|
| 203 |
+
"rewards/margins": 0.3850555717945099,
|
| 204 |
+
"rewards/rejected": -0.7964665293693542,
|
| 205 |
+
"step": 110
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.2510460251046025,
|
| 209 |
+
"grad_norm": 15.83081907349326,
|
| 210 |
+
"learning_rate": 4.662012913161997e-07,
|
| 211 |
+
"logits/chosen": -0.7602194547653198,
|
| 212 |
+
"logits/rejected": -0.5894412994384766,
|
| 213 |
+
"logps/chosen": -322.77508544921875,
|
| 214 |
+
"logps/rejected": -345.4307861328125,
|
| 215 |
+
"loss": 0.5779,
|
| 216 |
+
"rewards/accuracies": 0.75,
|
| 217 |
+
"rewards/chosen": -0.3538314998149872,
|
| 218 |
+
"rewards/margins": 0.45046257972717285,
|
| 219 |
+
"rewards/rejected": -0.8042941093444824,
|
| 220 |
+
"step": 120
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.2719665271966527,
|
| 224 |
+
"grad_norm": 15.055402351260229,
|
| 225 |
+
"learning_rate": 4.5646165232345103e-07,
|
| 226 |
+
"logits/chosen": -0.357295423746109,
|
| 227 |
+
"logits/rejected": -0.0887320414185524,
|
| 228 |
+
"logps/chosen": -332.0826110839844,
|
| 229 |
+
"logps/rejected": -368.31256103515625,
|
| 230 |
+
"loss": 0.549,
|
| 231 |
+
"rewards/accuracies": 0.75,
|
| 232 |
+
"rewards/chosen": -0.441201388835907,
|
| 233 |
+
"rewards/margins": 0.5332251787185669,
|
| 234 |
+
"rewards/rejected": -0.9744265675544739,
|
| 235 |
+
"step": 130
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.2928870292887029,
|
| 239 |
+
"grad_norm": 18.351529760698416,
|
| 240 |
+
"learning_rate": 4.456204510851956e-07,
|
| 241 |
+
"logits/chosen": -0.05173628777265549,
|
| 242 |
+
"logits/rejected": 0.2410668432712555,
|
| 243 |
+
"logps/chosen": -356.73333740234375,
|
| 244 |
+
"logps/rejected": -390.5004577636719,
|
| 245 |
+
"loss": 0.5555,
|
| 246 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 247 |
+
"rewards/chosen": -0.37852978706359863,
|
| 248 |
+
"rewards/margins": 0.49712079763412476,
|
| 249 |
+
"rewards/rejected": -0.8756505846977234,
|
| 250 |
+
"step": 140
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 0.3138075313807531,
|
| 254 |
+
"grad_norm": 22.63597752112208,
|
| 255 |
+
"learning_rate": 4.337355301007335e-07,
|
| 256 |
+
"logits/chosen": 0.6089886426925659,
|
| 257 |
+
"logits/rejected": 0.9238691329956055,
|
| 258 |
+
"logps/chosen": -365.7376403808594,
|
| 259 |
+
"logps/rejected": -402.1335144042969,
|
| 260 |
+
"loss": 0.5568,
|
| 261 |
+
"rewards/accuracies": 0.706250011920929,
|
| 262 |
+
"rewards/chosen": -0.7808293104171753,
|
| 263 |
+
"rewards/margins": 0.449969619512558,
|
| 264 |
+
"rewards/rejected": -1.2307989597320557,
|
| 265 |
+
"step": 150
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.33472803347280333,
|
| 269 |
+
"grad_norm": 20.03187629205258,
|
| 270 |
+
"learning_rate": 4.2087030056579986e-07,
|
| 271 |
+
"logits/chosen": 0.2681460976600647,
|
| 272 |
+
"logits/rejected": 0.93768310546875,
|
| 273 |
+
"logps/chosen": -325.05755615234375,
|
| 274 |
+
"logps/rejected": -369.85015869140625,
|
| 275 |
+
"loss": 0.5599,
|
| 276 |
+
"rewards/accuracies": 0.75,
|
| 277 |
+
"rewards/chosen": -0.4498773515224457,
|
| 278 |
+
"rewards/margins": 0.700762927532196,
|
| 279 |
+
"rewards/rejected": -1.1506402492523193,
|
| 280 |
+
"step": 160
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 0.35564853556485354,
|
| 284 |
+
"grad_norm": 25.495361660969547,
|
| 285 |
+
"learning_rate": 4.070934040463998e-07,
|
| 286 |
+
"logits/chosen": 0.556589663028717,
|
| 287 |
+
"logits/rejected": 0.9764513969421387,
|
| 288 |
+
"logps/chosen": -326.57281494140625,
|
| 289 |
+
"logps/rejected": -349.85894775390625,
|
| 290 |
+
"loss": 0.5672,
|
| 291 |
+
"rewards/accuracies": 0.65625,
|
| 292 |
+
"rewards/chosen": -0.7536547780036926,
|
| 293 |
+
"rewards/margins": 0.4030492901802063,
|
| 294 |
+
"rewards/rejected": -1.1567041873931885,
|
| 295 |
+
"step": 170
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 0.37656903765690375,
|
| 299 |
+
"grad_norm": 24.47540687241319,
|
| 300 |
+
"learning_rate": 3.9247834624635404e-07,
|
| 301 |
+
"logits/chosen": 1.273604393005371,
|
| 302 |
+
"logits/rejected": 1.7295089960098267,
|
| 303 |
+
"logps/chosen": -344.38751220703125,
|
| 304 |
+
"logps/rejected": -365.9759521484375,
|
| 305 |
+
"loss": 0.5254,
|
| 306 |
+
"rewards/accuracies": 0.71875,
|
| 307 |
+
"rewards/chosen": -0.8270662426948547,
|
| 308 |
+
"rewards/margins": 0.5501136779785156,
|
| 309 |
+
"rewards/rejected": -1.3771798610687256,
|
| 310 |
+
"step": 180
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.39748953974895396,
|
| 314 |
+
"grad_norm": 22.91459265715999,
|
| 315 |
+
"learning_rate": 3.7710310482256523e-07,
|
| 316 |
+
"logits/chosen": 1.0068578720092773,
|
| 317 |
+
"logits/rejected": 1.2909595966339111,
|
| 318 |
+
"logps/chosen": -328.5067138671875,
|
| 319 |
+
"logps/rejected": -372.14324951171875,
|
| 320 |
+
"loss": 0.5444,
|
| 321 |
+
"rewards/accuracies": 0.706250011920929,
|
| 322 |
+
"rewards/chosen": -0.5530312061309814,
|
| 323 |
+
"rewards/margins": 0.5563492178916931,
|
| 324 |
+
"rewards/rejected": -1.1093804836273193,
|
| 325 |
+
"step": 190
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 0.41841004184100417,
|
| 329 |
+
"grad_norm": 36.86200995874469,
|
| 330 |
+
"learning_rate": 3.610497133404795e-07,
|
| 331 |
+
"logits/chosen": 1.4111734628677368,
|
| 332 |
+
"logits/rejected": 1.7813835144042969,
|
| 333 |
+
"logps/chosen": -336.32794189453125,
|
| 334 |
+
"logps/rejected": -386.16827392578125,
|
| 335 |
+
"loss": 0.5606,
|
| 336 |
+
"rewards/accuracies": 0.762499988079071,
|
| 337 |
+
"rewards/chosen": -0.7106108069419861,
|
| 338 |
+
"rewards/margins": 0.5927971601486206,
|
| 339 |
+
"rewards/rejected": -1.303407907485962,
|
| 340 |
+
"step": 200
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 0.41841004184100417,
|
| 344 |
+
"eval_logits/chosen": 1.1101882457733154,
|
| 345 |
+
"eval_logits/rejected": 1.5408776998519897,
|
| 346 |
+
"eval_logps/chosen": -327.4931640625,
|
| 347 |
+
"eval_logps/rejected": -387.65057373046875,
|
| 348 |
+
"eval_loss": 0.5555490851402283,
|
| 349 |
+
"eval_rewards/accuracies": 0.74609375,
|
| 350 |
+
"eval_rewards/chosen": -0.6215302348136902,
|
| 351 |
+
"eval_rewards/margins": 0.606930136680603,
|
| 352 |
+
"eval_rewards/rejected": -1.228460431098938,
|
| 353 |
+
"eval_runtime": 53.4692,
|
| 354 |
+
"eval_samples_per_second": 37.405,
|
| 355 |
+
"eval_steps_per_second": 0.598,
|
| 356 |
+
"step": 200
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 0.4393305439330544,
|
| 360 |
+
"grad_norm": 19.538681466466336,
|
| 361 |
+
"learning_rate": 3.4440382358952115e-07,
|
| 362 |
+
"logits/chosen": 1.2745749950408936,
|
| 363 |
+
"logits/rejected": 1.9181947708129883,
|
| 364 |
+
"logps/chosen": -358.096435546875,
|
| 365 |
+
"logps/rejected": -373.4177551269531,
|
| 366 |
+
"loss": 0.5641,
|
| 367 |
+
"rewards/accuracies": 0.75,
|
| 368 |
+
"rewards/chosen": -0.5784404873847961,
|
| 369 |
+
"rewards/margins": 0.6223729252815247,
|
| 370 |
+
"rewards/rejected": -1.2008132934570312,
|
| 371 |
+
"step": 210
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 0.4602510460251046,
|
| 375 |
+
"grad_norm": 20.86939524596463,
|
| 376 |
+
"learning_rate": 3.272542485937368e-07,
|
| 377 |
+
"logits/chosen": 0.8960593342781067,
|
| 378 |
+
"logits/rejected": 1.4320571422576904,
|
| 379 |
+
"logps/chosen": -332.91864013671875,
|
| 380 |
+
"logps/rejected": -364.5188293457031,
|
| 381 |
+
"loss": 0.5474,
|
| 382 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 383 |
+
"rewards/chosen": -0.5551973581314087,
|
| 384 |
+
"rewards/margins": 0.6443191170692444,
|
| 385 |
+
"rewards/rejected": -1.1995164155960083,
|
| 386 |
+
"step": 220
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 0.4811715481171548,
|
| 390 |
+
"grad_norm": 19.099059937166764,
|
| 391 |
+
"learning_rate": 3.096924887558854e-07,
|
| 392 |
+
"logits/chosen": 0.6132632493972778,
|
| 393 |
+
"logits/rejected": 0.9618092775344849,
|
| 394 |
+
"logps/chosen": -323.7577209472656,
|
| 395 |
+
"logps/rejected": -385.98565673828125,
|
| 396 |
+
"loss": 0.552,
|
| 397 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 398 |
+
"rewards/chosen": -0.7150660753250122,
|
| 399 |
+
"rewards/margins": 0.6639611721038818,
|
| 400 |
+
"rewards/rejected": -1.3790271282196045,
|
| 401 |
+
"step": 230
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 0.502092050209205,
|
| 405 |
+
"grad_norm": 17.891063709919703,
|
| 406 |
+
"learning_rate": 2.9181224366319943e-07,
|
| 407 |
+
"logits/chosen": 0.4524998664855957,
|
| 408 |
+
"logits/rejected": 0.8361547589302063,
|
| 409 |
+
"logps/chosen": -321.6592102050781,
|
| 410 |
+
"logps/rejected": -369.956298828125,
|
| 411 |
+
"loss": 0.5196,
|
| 412 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 413 |
+
"rewards/chosen": -0.551439642906189,
|
| 414 |
+
"rewards/margins": 0.6203877925872803,
|
| 415 |
+
"rewards/rejected": -1.1718274354934692,
|
| 416 |
+
"step": 240
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 0.5230125523012552,
|
| 420 |
+
"grad_norm": 19.34076101834313,
|
| 421 |
+
"learning_rate": 2.7370891215954565e-07,
|
| 422 |
+
"logits/chosen": 1.31077241897583,
|
| 423 |
+
"logits/rejected": 1.824054479598999,
|
| 424 |
+
"logps/chosen": -356.10235595703125,
|
| 425 |
+
"logps/rejected": -388.8703918457031,
|
| 426 |
+
"loss": 0.5268,
|
| 427 |
+
"rewards/accuracies": 0.768750011920929,
|
| 428 |
+
"rewards/chosen": -0.48947277665138245,
|
| 429 |
+
"rewards/margins": 0.7497965097427368,
|
| 430 |
+
"rewards/rejected": -1.2392691373825073,
|
| 431 |
+
"step": 250
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 0.5439330543933054,
|
| 435 |
+
"grad_norm": 26.294415278038993,
|
| 436 |
+
"learning_rate": 2.55479083351317e-07,
|
| 437 |
+
"logits/chosen": 1.545461893081665,
|
| 438 |
+
"logits/rejected": 2.077069044113159,
|
| 439 |
+
"logps/chosen": -352.08587646484375,
|
| 440 |
+
"logps/rejected": -382.9310607910156,
|
| 441 |
+
"loss": 0.5298,
|
| 442 |
+
"rewards/accuracies": 0.768750011920929,
|
| 443 |
+
"rewards/chosen": -0.5431629419326782,
|
| 444 |
+
"rewards/margins": 0.7132429480552673,
|
| 445 |
+
"rewards/rejected": -1.2564058303833008,
|
| 446 |
+
"step": 260
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 0.5648535564853556,
|
| 450 |
+
"grad_norm": 16.790572679540336,
|
| 451 |
+
"learning_rate": 2.3722002126275822e-07,
|
| 452 |
+
"logits/chosen": 1.6584622859954834,
|
| 453 |
+
"logits/rejected": 2.1149163246154785,
|
| 454 |
+
"logps/chosen": -342.0724182128906,
|
| 455 |
+
"logps/rejected": -371.71588134765625,
|
| 456 |
+
"loss": 0.5367,
|
| 457 |
+
"rewards/accuracies": 0.668749988079071,
|
| 458 |
+
"rewards/chosen": -0.615090548992157,
|
| 459 |
+
"rewards/margins": 0.5203320980072021,
|
| 460 |
+
"rewards/rejected": -1.1354224681854248,
|
| 461 |
+
"step": 270
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.5857740585774058,
|
| 465 |
+
"grad_norm": 24.772008946191377,
|
| 466 |
+
"learning_rate": 2.19029145890313e-07,
|
| 467 |
+
"logits/chosen": 1.7721284627914429,
|
| 468 |
+
"logits/rejected": 2.3066000938415527,
|
| 469 |
+
"logps/chosen": -326.43603515625,
|
| 470 |
+
"logps/rejected": -370.3632507324219,
|
| 471 |
+
"loss": 0.5442,
|
| 472 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 473 |
+
"rewards/chosen": -0.5788434147834778,
|
| 474 |
+
"rewards/margins": 0.684018075466156,
|
| 475 |
+
"rewards/rejected": -1.2628614902496338,
|
| 476 |
+
"step": 280
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 0.606694560669456,
|
| 480 |
+
"grad_norm": 19.443360004007754,
|
| 481 |
+
"learning_rate": 2.0100351342479216e-07,
|
| 482 |
+
"logits/chosen": 1.2106627225875854,
|
| 483 |
+
"logits/rejected": 1.4834706783294678,
|
| 484 |
+
"logps/chosen": -333.8368225097656,
|
| 485 |
+
"logps/rejected": -384.6605224609375,
|
| 486 |
+
"loss": 0.5396,
|
| 487 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 488 |
+
"rewards/chosen": -0.8036863207817078,
|
| 489 |
+
"rewards/margins": 0.5939573645591736,
|
| 490 |
+
"rewards/rejected": -1.397643804550171,
|
| 491 |
+
"step": 290
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.6276150627615062,
|
| 495 |
+
"grad_norm": 24.510951465909894,
|
| 496 |
+
"learning_rate": 1.8323929841460178e-07,
|
| 497 |
+
"logits/chosen": 0.9158282279968262,
|
| 498 |
+
"logits/rejected": 1.5702682733535767,
|
| 499 |
+
"logps/chosen": -382.8146057128906,
|
| 500 |
+
"logps/rejected": -400.11474609375,
|
| 501 |
+
"loss": 0.5211,
|
| 502 |
+
"rewards/accuracies": 0.6875,
|
| 503 |
+
"rewards/chosen": -0.7668856382369995,
|
| 504 |
+
"rewards/margins": 0.6154419183731079,
|
| 505 |
+
"rewards/rejected": -1.382327675819397,
|
| 506 |
+
"step": 300
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 0.6276150627615062,
|
| 510 |
+
"eval_logits/chosen": 0.7847145199775696,
|
| 511 |
+
"eval_logits/rejected": 1.2079328298568726,
|
| 512 |
+
"eval_logps/chosen": -329.06488037109375,
|
| 513 |
+
"eval_logps/rejected": -404.1211242675781,
|
| 514 |
+
"eval_loss": 0.5284228324890137,
|
| 515 |
+
"eval_rewards/accuracies": 0.76171875,
|
| 516 |
+
"eval_rewards/chosen": -0.6372475028038025,
|
| 517 |
+
"eval_rewards/margins": 0.7559184432029724,
|
| 518 |
+
"eval_rewards/rejected": -1.393165946006775,
|
| 519 |
+
"eval_runtime": 53.4499,
|
| 520 |
+
"eval_samples_per_second": 37.418,
|
| 521 |
+
"eval_steps_per_second": 0.599,
|
| 522 |
+
"step": 300
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 0.6485355648535565,
|
| 526 |
+
"grad_norm": 21.649958560460085,
|
| 527 |
+
"learning_rate": 1.6583128063291573e-07,
|
| 528 |
+
"logits/chosen": 0.9612062573432922,
|
| 529 |
+
"logits/rejected": 1.3238331079483032,
|
| 530 |
+
"logps/chosen": -379.32904052734375,
|
| 531 |
+
"logps/rejected": -410.33685302734375,
|
| 532 |
+
"loss": 0.5135,
|
| 533 |
+
"rewards/accuracies": 0.731249988079071,
|
| 534 |
+
"rewards/chosen": -0.7942469120025635,
|
| 535 |
+
"rewards/margins": 0.6349788904190063,
|
| 536 |
+
"rewards/rejected": -1.4292259216308594,
|
| 537 |
+
"step": 310
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 0.6694560669456067,
|
| 541 |
+
"grad_norm": 21.732265089014117,
|
| 542 |
+
"learning_rate": 1.488723393865766e-07,
|
| 543 |
+
"logits/chosen": 0.9882587194442749,
|
| 544 |
+
"logits/rejected": 1.442718744277954,
|
| 545 |
+
"logps/chosen": -375.4773254394531,
|
| 546 |
+
"logps/rejected": -389.82733154296875,
|
| 547 |
+
"loss": 0.506,
|
| 548 |
+
"rewards/accuracies": 0.75,
|
| 549 |
+
"rewards/chosen": -0.7661019563674927,
|
| 550 |
+
"rewards/margins": 0.6393279433250427,
|
| 551 |
+
"rewards/rejected": -1.4054299592971802,
|
| 552 |
+
"step": 320
|
| 553 |
+
},
|
| 554 |
+
{
|
| 555 |
+
"epoch": 0.6903765690376569,
|
| 556 |
+
"grad_norm": 25.995557260279178,
|
| 557 |
+
"learning_rate": 1.3245295796480788e-07,
|
| 558 |
+
"logits/chosen": 0.8205612897872925,
|
| 559 |
+
"logits/rejected": 1.306123971939087,
|
| 560 |
+
"logps/chosen": -333.33074951171875,
|
| 561 |
+
"logps/rejected": -386.80072021484375,
|
| 562 |
+
"loss": 0.5192,
|
| 563 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 564 |
+
"rewards/chosen": -0.5908290147781372,
|
| 565 |
+
"rewards/margins": 0.5889186859130859,
|
| 566 |
+
"rewards/rejected": -1.1797475814819336,
|
| 567 |
+
"step": 330
|
| 568 |
+
},
|
| 569 |
+
{
|
| 570 |
+
"epoch": 0.7112970711297071,
|
| 571 |
+
"grad_norm": 34.06994351015907,
|
| 572 |
+
"learning_rate": 1.1666074087171627e-07,
|
| 573 |
+
"logits/chosen": 0.8795354962348938,
|
| 574 |
+
"logits/rejected": 1.3036845922470093,
|
| 575 |
+
"logps/chosen": -349.6972961425781,
|
| 576 |
+
"logps/rejected": -425.7892150878906,
|
| 577 |
+
"loss": 0.5208,
|
| 578 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 579 |
+
"rewards/chosen": -0.5422636270523071,
|
| 580 |
+
"rewards/margins": 0.8381665945053101,
|
| 581 |
+
"rewards/rejected": -1.3804302215576172,
|
| 582 |
+
"step": 340
|
| 583 |
+
},
|
| 584 |
+
{
|
| 585 |
+
"epoch": 0.7322175732217573,
|
| 586 |
+
"grad_norm": 21.28540346436605,
|
| 587 |
+
"learning_rate": 1.0157994641835734e-07,
|
| 588 |
+
"logits/chosen": 1.0639841556549072,
|
| 589 |
+
"logits/rejected": 1.6034936904907227,
|
| 590 |
+
"logps/chosen": -329.88690185546875,
|
| 591 |
+
"logps/rejected": -385.4267578125,
|
| 592 |
+
"loss": 0.4889,
|
| 593 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 594 |
+
"rewards/chosen": -0.6864292621612549,
|
| 595 |
+
"rewards/margins": 0.7708432674407959,
|
| 596 |
+
"rewards/rejected": -1.4572727680206299,
|
| 597 |
+
"step": 350
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"epoch": 0.7531380753138075,
|
| 601 |
+
"grad_norm": 23.66473954551423,
|
| 602 |
+
"learning_rate": 8.729103716819111e-08,
|
| 603 |
+
"logits/chosen": 1.199178695678711,
|
| 604 |
+
"logits/rejected": 1.9472402334213257,
|
| 605 |
+
"logps/chosen": -380.9803161621094,
|
| 606 |
+
"logps/rejected": -419.5332946777344,
|
| 607 |
+
"loss": 0.5316,
|
| 608 |
+
"rewards/accuracies": 0.768750011920929,
|
| 609 |
+
"rewards/chosen": -0.7623031735420227,
|
| 610 |
+
"rewards/margins": 0.7938345670700073,
|
| 611 |
+
"rewards/rejected": -1.5561376810073853,
|
| 612 |
+
"step": 360
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"epoch": 0.7740585774058577,
|
| 616 |
+
"grad_norm": 22.814657698551834,
|
| 617 |
+
"learning_rate": 7.387025063449081e-08,
|
| 618 |
+
"logits/chosen": 1.5637246370315552,
|
| 619 |
+
"logits/rejected": 2.090482234954834,
|
| 620 |
+
"logps/chosen": -364.23126220703125,
|
| 621 |
+
"logps/rejected": -389.804931640625,
|
| 622 |
+
"loss": 0.5165,
|
| 623 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 624 |
+
"rewards/chosen": -0.8549477458000183,
|
| 625 |
+
"rewards/margins": 0.6235242486000061,
|
| 626 |
+
"rewards/rejected": -1.4784719944000244,
|
| 627 |
+
"step": 370
|
| 628 |
+
},
|
| 629 |
+
{
|
| 630 |
+
"epoch": 0.7949790794979079,
|
| 631 |
+
"grad_norm": 24.019268622369378,
|
| 632 |
+
"learning_rate": 6.138919252022435e-08,
|
| 633 |
+
"logits/chosen": 1.690513253211975,
|
| 634 |
+
"logits/rejected": 1.896419882774353,
|
| 635 |
+
"logps/chosen": -330.74395751953125,
|
| 636 |
+
"logps/rejected": -429.8720703125,
|
| 637 |
+
"loss": 0.5175,
|
| 638 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 639 |
+
"rewards/chosen": -0.8654612302780151,
|
| 640 |
+
"rewards/margins": 0.7997757196426392,
|
| 641 |
+
"rewards/rejected": -1.6652368307113647,
|
| 642 |
+
"step": 380
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 0.8158995815899581,
|
| 646 |
+
"grad_norm": 23.872152876539662,
|
| 647 |
+
"learning_rate": 4.991445467064689e-08,
|
| 648 |
+
"logits/chosen": 1.24801766872406,
|
| 649 |
+
"logits/rejected": 1.5409824848175049,
|
| 650 |
+
"logps/chosen": -374.94500732421875,
|
| 651 |
+
"logps/rejected": -423.5159606933594,
|
| 652 |
+
"loss": 0.5064,
|
| 653 |
+
"rewards/accuracies": 0.731249988079071,
|
| 654 |
+
"rewards/chosen": -0.623989462852478,
|
| 655 |
+
"rewards/margins": 0.7279406785964966,
|
| 656 |
+
"rewards/rejected": -1.351930022239685,
|
| 657 |
+
"step": 390
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 0.8368200836820083,
|
| 661 |
+
"grad_norm": 22.17989276400099,
|
| 662 |
+
"learning_rate": 3.9507259776993954e-08,
|
| 663 |
+
"logits/chosen": 1.4913052320480347,
|
| 664 |
+
"logits/rejected": 1.9780304431915283,
|
| 665 |
+
"logps/chosen": -344.193359375,
|
| 666 |
+
"logps/rejected": -412.5020446777344,
|
| 667 |
+
"loss": 0.5189,
|
| 668 |
+
"rewards/accuracies": 0.762499988079071,
|
| 669 |
+
"rewards/chosen": -0.6954695582389832,
|
| 670 |
+
"rewards/margins": 0.7355377078056335,
|
| 671 |
+
"rewards/rejected": -1.4310071468353271,
|
| 672 |
+
"step": 400
|
| 673 |
+
},
|
| 674 |
+
{
|
| 675 |
+
"epoch": 0.8368200836820083,
|
| 676 |
+
"eval_logits/chosen": 1.2049928903579712,
|
| 677 |
+
"eval_logits/rejected": 1.687930941581726,
|
| 678 |
+
"eval_logps/chosen": -323.4798889160156,
|
| 679 |
+
"eval_logps/rejected": -403.6497802734375,
|
| 680 |
+
"eval_loss": 0.5224906206130981,
|
| 681 |
+
"eval_rewards/accuracies": 0.77734375,
|
| 682 |
+
"eval_rewards/chosen": -0.5813976526260376,
|
| 683 |
+
"eval_rewards/margins": 0.8070548176765442,
|
| 684 |
+
"eval_rewards/rejected": -1.3884525299072266,
|
| 685 |
+
"eval_runtime": 53.5131,
|
| 686 |
+
"eval_samples_per_second": 37.374,
|
| 687 |
+
"eval_steps_per_second": 0.598,
|
| 688 |
+
"step": 400
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"epoch": 0.8577405857740585,
|
| 692 |
+
"grad_norm": 24.78416106509579,
|
| 693 |
+
"learning_rate": 3.022313472693447e-08,
|
| 694 |
+
"logits/chosen": 1.401047945022583,
|
| 695 |
+
"logits/rejected": 2.122811794281006,
|
| 696 |
+
"logps/chosen": -370.56951904296875,
|
| 697 |
+
"logps/rejected": -416.39080810546875,
|
| 698 |
+
"loss": 0.5278,
|
| 699 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 700 |
+
"rewards/chosen": -0.6760027408599854,
|
| 701 |
+
"rewards/margins": 0.7805537581443787,
|
| 702 |
+
"rewards/rejected": -1.4565564393997192,
|
| 703 |
+
"step": 410
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 0.8786610878661087,
|
| 707 |
+
"grad_norm": 24.036813785272063,
|
| 708 |
+
"learning_rate": 2.2111614344599684e-08,
|
| 709 |
+
"logits/chosen": 1.3202136754989624,
|
| 710 |
+
"logits/rejected": 1.759892225265503,
|
| 711 |
+
"logps/chosen": -373.802734375,
|
| 712 |
+
"logps/rejected": -416.2435607910156,
|
| 713 |
+
"loss": 0.5136,
|
| 714 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 715 |
+
"rewards/chosen": -0.7081748843193054,
|
| 716 |
+
"rewards/margins": 0.6878335475921631,
|
| 717 |
+
"rewards/rejected": -1.3960084915161133,
|
| 718 |
+
"step": 420
|
| 719 |
+
},
|
| 720 |
+
{
|
| 721 |
+
"epoch": 0.899581589958159,
|
| 722 |
+
"grad_norm": 23.876888782252916,
|
| 723 |
+
"learning_rate": 1.521597710086439e-08,
|
| 724 |
+
"logits/chosen": 1.59340238571167,
|
| 725 |
+
"logits/rejected": 2.011382579803467,
|
| 726 |
+
"logps/chosen": -357.6289978027344,
|
| 727 |
+
"logps/rejected": -414.678466796875,
|
| 728 |
+
"loss": 0.4929,
|
| 729 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 730 |
+
"rewards/chosen": -0.7411682605743408,
|
| 731 |
+
"rewards/margins": 0.7640897631645203,
|
| 732 |
+
"rewards/rejected": -1.5052578449249268,
|
| 733 |
+
"step": 430
|
| 734 |
+
},
|
| 735 |
+
{
|
| 736 |
+
"epoch": 0.9205020920502092,
|
| 737 |
+
"grad_norm": 23.73816215215032,
|
| 738 |
+
"learning_rate": 9.57301420397924e-09,
|
| 739 |
+
"logits/chosen": 1.240863561630249,
|
| 740 |
+
"logits/rejected": 1.8725379705429077,
|
| 741 |
+
"logps/chosen": -354.1866149902344,
|
| 742 |
+
"logps/rejected": -410.48187255859375,
|
| 743 |
+
"loss": 0.5066,
|
| 744 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 745 |
+
"rewards/chosen": -0.6514572501182556,
|
| 746 |
+
"rewards/margins": 0.7045157551765442,
|
| 747 |
+
"rewards/rejected": -1.3559730052947998,
|
| 748 |
+
"step": 440
|
| 749 |
+
},
|
| 750 |
+
{
|
| 751 |
+
"epoch": 0.9414225941422594,
|
| 752 |
+
"grad_norm": 23.69382783496658,
|
| 753 |
+
"learning_rate": 5.212833302556258e-09,
|
| 754 |
+
"logits/chosen": 1.4327576160430908,
|
| 755 |
+
"logits/rejected": 1.8339331150054932,
|
| 756 |
+
"logps/chosen": -374.9490661621094,
|
| 757 |
+
"logps/rejected": -459.17169189453125,
|
| 758 |
+
"loss": 0.51,
|
| 759 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 760 |
+
"rewards/chosen": -0.7898253202438354,
|
| 761 |
+
"rewards/margins": 0.6647788286209106,
|
| 762 |
+
"rewards/rejected": -1.454604148864746,
|
| 763 |
+
"step": 450
|
| 764 |
+
},
|
| 765 |
+
{
|
| 766 |
+
"epoch": 0.9623430962343096,
|
| 767 |
+
"grad_norm": 28.599243740532902,
|
| 768 |
+
"learning_rate": 2.158697848236607e-09,
|
| 769 |
+
"logits/chosen": 1.528529405593872,
|
| 770 |
+
"logits/rejected": 2.007927179336548,
|
| 771 |
+
"logps/chosen": -352.3992004394531,
|
| 772 |
+
"logps/rejected": -387.2112121582031,
|
| 773 |
+
"loss": 0.5217,
|
| 774 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 775 |
+
"rewards/chosen": -0.7364510893821716,
|
| 776 |
+
"rewards/margins": 0.685036838054657,
|
| 777 |
+
"rewards/rejected": -1.4214880466461182,
|
| 778 |
+
"step": 460
|
| 779 |
+
},
|
| 780 |
+
{
|
| 781 |
+
"epoch": 0.9832635983263598,
|
| 782 |
+
"grad_norm": 20.454762644290014,
|
| 783 |
+
"learning_rate": 4.269029751107489e-10,
|
| 784 |
+
"logits/chosen": 1.380399227142334,
|
| 785 |
+
"logits/rejected": 1.8892757892608643,
|
| 786 |
+
"logps/chosen": -350.90447998046875,
|
| 787 |
+
"logps/rejected": -421.0154724121094,
|
| 788 |
+
"loss": 0.5088,
|
| 789 |
+
"rewards/accuracies": 0.737500011920929,
|
| 790 |
+
"rewards/chosen": -0.7164818644523621,
|
| 791 |
+
"rewards/margins": 0.6821088194847107,
|
| 792 |
+
"rewards/rejected": -1.3985908031463623,
|
| 793 |
+
"step": 470
|
| 794 |
+
},
|
| 795 |
+
{
|
| 796 |
+
"epoch": 1.0,
|
| 797 |
+
"step": 478,
|
| 798 |
+
"total_flos": 0.0,
|
| 799 |
+
"train_loss": 0.5574632548886862,
|
| 800 |
+
"train_runtime": 4290.8569,
|
| 801 |
+
"train_samples_per_second": 14.248,
|
| 802 |
+
"train_steps_per_second": 0.111
|
| 803 |
+
}
|
| 804 |
+
],
|
| 805 |
+
"logging_steps": 10,
|
| 806 |
+
"max_steps": 478,
|
| 807 |
+
"num_input_tokens_seen": 0,
|
| 808 |
+
"num_train_epochs": 1,
|
| 809 |
+
"save_steps": 100,
|
| 810 |
+
"stateful_callbacks": {
|
| 811 |
+
"TrainerControl": {
|
| 812 |
+
"args": {
|
| 813 |
+
"should_epoch_stop": false,
|
| 814 |
+
"should_evaluate": false,
|
| 815 |
+
"should_log": false,
|
| 816 |
+
"should_save": true,
|
| 817 |
+
"should_training_stop": true
|
| 818 |
+
},
|
| 819 |
+
"attributes": {}
|
| 820 |
+
}
|
| 821 |
+
},
|
| 822 |
+
"total_flos": 0.0,
|
| 823 |
+
"train_batch_size": 8,
|
| 824 |
+
"trial_name": null,
|
| 825 |
+
"trial_params": null
|
| 826 |
+
}
|