Jason Lee
commited on
Commit
•
3a97bb8
1
Parent(s):
995493b
Model save
Browse files- README.md +74 -0
- all_results.json +9 -0
- generation_config.json +7 -0
- train_results.json +9 -0
- trainer_state.json +826 -0
README.md
ADDED
|
@@ -0,0 +1,74 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
tags:
|
| 3 |
+
- trl
|
| 4 |
+
- dpo
|
| 5 |
+
- generated_from_trainer
|
| 6 |
+
model-index:
|
| 7 |
+
- name: tinyllama-chat-mine-dpo
|
| 8 |
+
results: []
|
| 9 |
+
---
|
| 10 |
+
|
| 11 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 12 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 13 |
+
|
| 14 |
+
# tinyllama-chat-mine-dpo
|
| 15 |
+
|
| 16 |
+
This model was trained from scratch on an unknown dataset.
|
| 17 |
+
It achieves the following results on the evaluation set:
|
| 18 |
+
- Loss: 0.6240
|
| 19 |
+
- Rewards/chosen: -0.7007
|
| 20 |
+
- Rewards/rejected: -0.9684
|
| 21 |
+
- Rewards/accuracies: 0.6825
|
| 22 |
+
- Rewards/margins: 0.2677
|
| 23 |
+
- Logps/rejected: -395.3808
|
| 24 |
+
- Logps/chosen: -412.6868
|
| 25 |
+
- Logits/rejected: -2.7107
|
| 26 |
+
- Logits/chosen: -2.7399
|
| 27 |
+
|
| 28 |
+
## Model description
|
| 29 |
+
|
| 30 |
+
More information needed
|
| 31 |
+
|
| 32 |
+
## Intended uses & limitations
|
| 33 |
+
|
| 34 |
+
More information needed
|
| 35 |
+
|
| 36 |
+
## Training and evaluation data
|
| 37 |
+
|
| 38 |
+
More information needed
|
| 39 |
+
|
| 40 |
+
## Training procedure
|
| 41 |
+
|
| 42 |
+
### Training hyperparameters
|
| 43 |
+
|
| 44 |
+
The following hyperparameters were used during training:
|
| 45 |
+
- learning_rate: 5e-07
|
| 46 |
+
- train_batch_size: 8
|
| 47 |
+
- eval_batch_size: 8
|
| 48 |
+
- seed: 42
|
| 49 |
+
- distributed_type: multi-GPU
|
| 50 |
+
- num_devices: 4
|
| 51 |
+
- gradient_accumulation_steps: 4
|
| 52 |
+
- total_train_batch_size: 128
|
| 53 |
+
- total_eval_batch_size: 32
|
| 54 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 55 |
+
- lr_scheduler_type: cosine
|
| 56 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 57 |
+
- num_epochs: 1
|
| 58 |
+
|
| 59 |
+
### Training results
|
| 60 |
+
|
| 61 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 62 |
+
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 63 |
+
| 0.6685 | 0.2093 | 100 | 0.6694 | -0.1299 | -0.1943 | 0.6528 | 0.0644 | -317.9682 | -355.6027 | -2.9255 | -2.9507 |
|
| 64 |
+
| 0.642 | 0.4186 | 200 | 0.6407 | -0.4273 | -0.6175 | 0.6726 | 0.1902 | -360.2873 | -385.3470 | -2.7926 | -2.8208 |
|
| 65 |
+
| 0.6285 | 0.6279 | 300 | 0.6331 | -0.4723 | -0.6951 | 0.6647 | 0.2228 | -368.0482 | -389.8438 | -2.7731 | -2.8012 |
|
| 66 |
+
| 0.6222 | 0.8373 | 400 | 0.6240 | -0.7007 | -0.9684 | 0.6825 | 0.2677 | -395.3808 | -412.6868 | -2.7107 | -2.7399 |
|
| 67 |
+
|
| 68 |
+
|
| 69 |
+
### Framework versions
|
| 70 |
+
|
| 71 |
+
- Transformers 4.43.3
|
| 72 |
+
- Pytorch 2.1.2
|
| 73 |
+
- Datasets 2.20.0
|
| 74 |
+
- Tokenizers 0.19.1
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9984301412872841,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.6428392008415558,
|
| 5 |
+
"train_runtime": 6631.0546,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 9.219,
|
| 8 |
+
"train_steps_per_second": 0.072
|
| 9 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token_id": 1,
|
| 3 |
+
"eos_token_id": 2,
|
| 4 |
+
"max_length": 2048,
|
| 5 |
+
"pad_token_id": 0,
|
| 6 |
+
"transformers_version": "4.43.3"
|
| 7 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9984301412872841,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.6428392008415558,
|
| 5 |
+
"train_runtime": 6631.0546,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 9.219,
|
| 8 |
+
"train_steps_per_second": 0.072
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,826 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.9984301412872841,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 477,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.0020931449502878076,
|
| 13 |
+
"grad_norm": 2.741797685623169,
|
| 14 |
+
"learning_rate": 1.0416666666666666e-08,
|
| 15 |
+
"logits/chosen": -3.1097278594970703,
|
| 16 |
+
"logits/rejected": -3.080122470855713,
|
| 17 |
+
"logps/chosen": -389.3681640625,
|
| 18 |
+
"logps/rejected": -352.00482177734375,
|
| 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": 0.020931449502878074,
|
| 28 |
+
"grad_norm": 2.746737480163574,
|
| 29 |
+
"learning_rate": 1.0416666666666667e-07,
|
| 30 |
+
"logits/chosen": -2.9152116775512695,
|
| 31 |
+
"logits/rejected": -2.905562400817871,
|
| 32 |
+
"logps/chosen": -323.6693420410156,
|
| 33 |
+
"logps/rejected": -297.4331359863281,
|
| 34 |
+
"loss": 0.6932,
|
| 35 |
+
"rewards/accuracies": 0.4201388955116272,
|
| 36 |
+
"rewards/chosen": -0.00030070680077187717,
|
| 37 |
+
"rewards/margins": -0.00017917003424372524,
|
| 38 |
+
"rewards/rejected": -0.00012153676652815193,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.04186289900575615,
|
| 43 |
+
"grad_norm": 2.543320655822754,
|
| 44 |
+
"learning_rate": 2.0833333333333333e-07,
|
| 45 |
+
"logits/chosen": -2.9843814373016357,
|
| 46 |
+
"logits/rejected": -2.9781064987182617,
|
| 47 |
+
"logps/chosen": -335.30413818359375,
|
| 48 |
+
"logps/rejected": -316.4299621582031,
|
| 49 |
+
"loss": 0.6931,
|
| 50 |
+
"rewards/accuracies": 0.5218750238418579,
|
| 51 |
+
"rewards/chosen": 0.0004041799402330071,
|
| 52 |
+
"rewards/margins": 0.0005227966466918588,
|
| 53 |
+
"rewards/rejected": -0.0001186166555271484,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.06279434850863422,
|
| 58 |
+
"grad_norm": 2.5893585681915283,
|
| 59 |
+
"learning_rate": 3.1249999999999997e-07,
|
| 60 |
+
"logits/chosen": -2.997720241546631,
|
| 61 |
+
"logits/rejected": -2.9677913188934326,
|
| 62 |
+
"logps/chosen": -345.15240478515625,
|
| 63 |
+
"logps/rejected": -295.96966552734375,
|
| 64 |
+
"loss": 0.6928,
|
| 65 |
+
"rewards/accuracies": 0.528124988079071,
|
| 66 |
+
"rewards/chosen": -0.0004641309496946633,
|
| 67 |
+
"rewards/margins": 0.000602933403570205,
|
| 68 |
+
"rewards/rejected": -0.0010670643532648683,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.0837257980115123,
|
| 73 |
+
"grad_norm": 2.6134979724884033,
|
| 74 |
+
"learning_rate": 4.1666666666666667e-07,
|
| 75 |
+
"logits/chosen": -2.9335551261901855,
|
| 76 |
+
"logits/rejected": -2.9302399158477783,
|
| 77 |
+
"logps/chosen": -321.5934143066406,
|
| 78 |
+
"logps/rejected": -289.6072998046875,
|
| 79 |
+
"loss": 0.6919,
|
| 80 |
+
"rewards/accuracies": 0.637499988079071,
|
| 81 |
+
"rewards/chosen": -0.0015825815498828888,
|
| 82 |
+
"rewards/margins": 0.0028794072568416595,
|
| 83 |
+
"rewards/rejected": -0.004461988341063261,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.10465724751439037,
|
| 88 |
+
"grad_norm": 2.6024513244628906,
|
| 89 |
+
"learning_rate": 4.999731868769026e-07,
|
| 90 |
+
"logits/chosen": -2.9198384284973145,
|
| 91 |
+
"logits/rejected": -2.8993093967437744,
|
| 92 |
+
"logps/chosen": -328.54644775390625,
|
| 93 |
+
"logps/rejected": -309.58087158203125,
|
| 94 |
+
"loss": 0.6901,
|
| 95 |
+
"rewards/accuracies": 0.5562499761581421,
|
| 96 |
+
"rewards/chosen": -0.005653353873640299,
|
| 97 |
+
"rewards/margins": 0.004262409172952175,
|
| 98 |
+
"rewards/rejected": -0.009915763512253761,
|
| 99 |
+
"step": 50
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.12558869701726844,
|
| 103 |
+
"grad_norm": 2.767859697341919,
|
| 104 |
+
"learning_rate": 4.990353313429303e-07,
|
| 105 |
+
"logits/chosen": -2.982860565185547,
|
| 106 |
+
"logits/rejected": -2.969365119934082,
|
| 107 |
+
"logps/chosen": -307.7032470703125,
|
| 108 |
+
"logps/rejected": -304.3933410644531,
|
| 109 |
+
"loss": 0.6874,
|
| 110 |
+
"rewards/accuracies": 0.578125,
|
| 111 |
+
"rewards/chosen": -0.015346085652709007,
|
| 112 |
+
"rewards/margins": 0.010624411515891552,
|
| 113 |
+
"rewards/rejected": -0.025970498099923134,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.14652014652014653,
|
| 118 |
+
"grad_norm": 2.7651147842407227,
|
| 119 |
+
"learning_rate": 4.967625656594781e-07,
|
| 120 |
+
"logits/chosen": -2.9431705474853516,
|
| 121 |
+
"logits/rejected": -2.953648090362549,
|
| 122 |
+
"logps/chosen": -346.7774658203125,
|
| 123 |
+
"logps/rejected": -312.93865966796875,
|
| 124 |
+
"loss": 0.6826,
|
| 125 |
+
"rewards/accuracies": 0.653124988079071,
|
| 126 |
+
"rewards/chosen": -0.04352644830942154,
|
| 127 |
+
"rewards/margins": 0.024388128891587257,
|
| 128 |
+
"rewards/rejected": -0.06791457533836365,
|
| 129 |
+
"step": 70
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.1674515960230246,
|
| 133 |
+
"grad_norm": 2.939892053604126,
|
| 134 |
+
"learning_rate": 4.93167072587771e-07,
|
| 135 |
+
"logits/chosen": -2.973877429962158,
|
| 136 |
+
"logits/rejected": -2.9495885372161865,
|
| 137 |
+
"logps/chosen": -362.0614929199219,
|
| 138 |
+
"logps/rejected": -301.11322021484375,
|
| 139 |
+
"loss": 0.6771,
|
| 140 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 141 |
+
"rewards/chosen": -0.08665237575769424,
|
| 142 |
+
"rewards/margins": 0.0347968153655529,
|
| 143 |
+
"rewards/rejected": -0.12144919484853745,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.18838304552590268,
|
| 148 |
+
"grad_norm": 2.9581210613250732,
|
| 149 |
+
"learning_rate": 4.882681251368548e-07,
|
| 150 |
+
"logits/chosen": -2.9634270668029785,
|
| 151 |
+
"logits/rejected": -2.9433810710906982,
|
| 152 |
+
"logps/chosen": -323.24359130859375,
|
| 153 |
+
"logps/rejected": -298.01763916015625,
|
| 154 |
+
"loss": 0.6728,
|
| 155 |
+
"rewards/accuracies": 0.625,
|
| 156 |
+
"rewards/chosen": -0.13328225910663605,
|
| 157 |
+
"rewards/margins": 0.03973008692264557,
|
| 158 |
+
"rewards/rejected": -0.17301234602928162,
|
| 159 |
+
"step": 90
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.20931449502878074,
|
| 163 |
+
"grad_norm": 3.1116480827331543,
|
| 164 |
+
"learning_rate": 4.820919832540181e-07,
|
| 165 |
+
"logits/chosen": -2.9381461143493652,
|
| 166 |
+
"logits/rejected": -2.9422552585601807,
|
| 167 |
+
"logps/chosen": -342.9717102050781,
|
| 168 |
+
"logps/rejected": -310.58758544921875,
|
| 169 |
+
"loss": 0.6685,
|
| 170 |
+
"rewards/accuracies": 0.637499988079071,
|
| 171 |
+
"rewards/chosen": -0.14851507544517517,
|
| 172 |
+
"rewards/margins": 0.057429663836956024,
|
| 173 |
+
"rewards/rejected": -0.2059447318315506,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 0.20931449502878074,
|
| 178 |
+
"eval_logits/chosen": -2.950716972351074,
|
| 179 |
+
"eval_logits/rejected": -2.9254934787750244,
|
| 180 |
+
"eval_logps/chosen": -355.60272216796875,
|
| 181 |
+
"eval_logps/rejected": -317.9681701660156,
|
| 182 |
+
"eval_loss": 0.6694281101226807,
|
| 183 |
+
"eval_rewards/accuracies": 0.6527777910232544,
|
| 184 |
+
"eval_rewards/chosen": -0.1298944652080536,
|
| 185 |
+
"eval_rewards/margins": 0.06442055851221085,
|
| 186 |
+
"eval_rewards/rejected": -0.19431501626968384,
|
| 187 |
+
"eval_runtime": 86.0849,
|
| 188 |
+
"eval_samples_per_second": 23.233,
|
| 189 |
+
"eval_steps_per_second": 0.732,
|
| 190 |
+
"step": 100
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 0.2302459445316588,
|
| 194 |
+
"grad_norm": 3.7738163471221924,
|
| 195 |
+
"learning_rate": 4.7467175306295647e-07,
|
| 196 |
+
"logits/chosen": -2.926741123199463,
|
| 197 |
+
"logits/rejected": -2.913757562637329,
|
| 198 |
+
"logps/chosen": -356.9674377441406,
|
| 199 |
+
"logps/rejected": -310.5869140625,
|
| 200 |
+
"loss": 0.6697,
|
| 201 |
+
"rewards/accuracies": 0.606249988079071,
|
| 202 |
+
"rewards/chosen": -0.14377950131893158,
|
| 203 |
+
"rewards/margins": 0.06109999865293503,
|
| 204 |
+
"rewards/rejected": -0.204879492521286,
|
| 205 |
+
"step": 110
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.25117739403453687,
|
| 209 |
+
"grad_norm": 3.45920467376709,
|
| 210 |
+
"learning_rate": 4.6604720940421207e-07,
|
| 211 |
+
"logits/chosen": -2.854685068130493,
|
| 212 |
+
"logits/rejected": -2.851743459701538,
|
| 213 |
+
"logps/chosen": -349.94189453125,
|
| 214 |
+
"logps/rejected": -327.46820068359375,
|
| 215 |
+
"loss": 0.6557,
|
| 216 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 217 |
+
"rewards/chosen": -0.174159973859787,
|
| 218 |
+
"rewards/margins": 0.08476099371910095,
|
| 219 |
+
"rewards/rejected": -0.25892096757888794,
|
| 220 |
+
"step": 120
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.272108843537415,
|
| 224 |
+
"grad_norm": 4.213115692138672,
|
| 225 |
+
"learning_rate": 4.5626458262912735e-07,
|
| 226 |
+
"logits/chosen": -2.8723671436309814,
|
| 227 |
+
"logits/rejected": -2.8682103157043457,
|
| 228 |
+
"logps/chosen": -364.5023498535156,
|
| 229 |
+
"logps/rejected": -344.17413330078125,
|
| 230 |
+
"loss": 0.65,
|
| 231 |
+
"rewards/accuracies": 0.640625,
|
| 232 |
+
"rewards/chosen": -0.26356348395347595,
|
| 233 |
+
"rewards/margins": 0.08755739033222198,
|
| 234 |
+
"rewards/rejected": -0.35112088918685913,
|
| 235 |
+
"step": 130
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.29304029304029305,
|
| 239 |
+
"grad_norm": 4.351049900054932,
|
| 240 |
+
"learning_rate": 4.453763107901675e-07,
|
| 241 |
+
"logits/chosen": -2.8395862579345703,
|
| 242 |
+
"logits/rejected": -2.8383350372314453,
|
| 243 |
+
"logps/chosen": -386.54754638671875,
|
| 244 |
+
"logps/rejected": -353.8211975097656,
|
| 245 |
+
"loss": 0.6576,
|
| 246 |
+
"rewards/accuracies": 0.643750011920929,
|
| 247 |
+
"rewards/chosen": -0.30667954683303833,
|
| 248 |
+
"rewards/margins": 0.11848233640193939,
|
| 249 |
+
"rewards/rejected": -0.4251618981361389,
|
| 250 |
+
"step": 140
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 0.3139717425431711,
|
| 254 |
+
"grad_norm": 5.293145656585693,
|
| 255 |
+
"learning_rate": 4.3344075855595097e-07,
|
| 256 |
+
"logits/chosen": -2.837963581085205,
|
| 257 |
+
"logits/rejected": -2.8277411460876465,
|
| 258 |
+
"logps/chosen": -363.89532470703125,
|
| 259 |
+
"logps/rejected": -334.4164733886719,
|
| 260 |
+
"loss": 0.6562,
|
| 261 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 262 |
+
"rewards/chosen": -0.3559018671512604,
|
| 263 |
+
"rewards/margins": 0.0798763707280159,
|
| 264 |
+
"rewards/rejected": -0.43577829003334045,
|
| 265 |
+
"step": 150
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.3349031920460492,
|
| 269 |
+
"grad_norm": 5.583098411560059,
|
| 270 |
+
"learning_rate": 4.2052190435769554e-07,
|
| 271 |
+
"logits/chosen": -2.8631508350372314,
|
| 272 |
+
"logits/rejected": -2.8266890048980713,
|
| 273 |
+
"logps/chosen": -359.99200439453125,
|
| 274 |
+
"logps/rejected": -320.49151611328125,
|
| 275 |
+
"loss": 0.6453,
|
| 276 |
+
"rewards/accuracies": 0.628125011920929,
|
| 277 |
+
"rewards/chosen": -0.34305575489997864,
|
| 278 |
+
"rewards/margins": 0.13357409834861755,
|
| 279 |
+
"rewards/rejected": -0.4766298234462738,
|
| 280 |
+
"step": 160
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 0.35583464154892724,
|
| 284 |
+
"grad_norm": 5.565986633300781,
|
| 285 |
+
"learning_rate": 4.0668899744407567e-07,
|
| 286 |
+
"logits/chosen": -2.800814628601074,
|
| 287 |
+
"logits/rejected": -2.7905514240264893,
|
| 288 |
+
"logps/chosen": -349.87725830078125,
|
| 289 |
+
"logps/rejected": -323.04150390625,
|
| 290 |
+
"loss": 0.6516,
|
| 291 |
+
"rewards/accuracies": 0.606249988079071,
|
| 292 |
+
"rewards/chosen": -0.4009243845939636,
|
| 293 |
+
"rewards/margins": 0.12320031225681305,
|
| 294 |
+
"rewards/rejected": -0.5241247415542603,
|
| 295 |
+
"step": 170
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 0.37676609105180536,
|
| 299 |
+
"grad_norm": 6.890281677246094,
|
| 300 |
+
"learning_rate": 3.920161866827889e-07,
|
| 301 |
+
"logits/chosen": -2.8282012939453125,
|
| 302 |
+
"logits/rejected": -2.8248977661132812,
|
| 303 |
+
"logps/chosen": -382.04364013671875,
|
| 304 |
+
"logps/rejected": -350.888427734375,
|
| 305 |
+
"loss": 0.64,
|
| 306 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 307 |
+
"rewards/chosen": -0.4164988100528717,
|
| 308 |
+
"rewards/margins": 0.15513385832309723,
|
| 309 |
+
"rewards/rejected": -0.5716326832771301,
|
| 310 |
+
"step": 180
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.3976975405546834,
|
| 314 |
+
"grad_norm": 5.489590167999268,
|
| 315 |
+
"learning_rate": 3.765821230985757e-07,
|
| 316 |
+
"logits/chosen": -2.8130276203155518,
|
| 317 |
+
"logits/rejected": -2.7892467975616455,
|
| 318 |
+
"logps/chosen": -350.63873291015625,
|
| 319 |
+
"logps/rejected": -333.55718994140625,
|
| 320 |
+
"loss": 0.6408,
|
| 321 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 322 |
+
"rewards/chosen": -0.3725319802761078,
|
| 323 |
+
"rewards/margins": 0.13300345838069916,
|
| 324 |
+
"rewards/rejected": -0.5055354833602905,
|
| 325 |
+
"step": 190
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 0.4186289900575615,
|
| 329 |
+
"grad_norm": 6.967275142669678,
|
| 330 |
+
"learning_rate": 3.604695382782159e-07,
|
| 331 |
+
"logits/chosen": -2.787351608276367,
|
| 332 |
+
"logits/rejected": -2.789077043533325,
|
| 333 |
+
"logps/chosen": -370.17864990234375,
|
| 334 |
+
"logps/rejected": -361.43988037109375,
|
| 335 |
+
"loss": 0.642,
|
| 336 |
+
"rewards/accuracies": 0.637499988079071,
|
| 337 |
+
"rewards/chosen": -0.38552746176719666,
|
| 338 |
+
"rewards/margins": 0.16969361901283264,
|
| 339 |
+
"rewards/rejected": -0.5552210807800293,
|
| 340 |
+
"step": 200
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 0.4186289900575615,
|
| 344 |
+
"eval_logits/chosen": -2.8207905292510986,
|
| 345 |
+
"eval_logits/rejected": -2.7925562858581543,
|
| 346 |
+
"eval_logps/chosen": -385.3469543457031,
|
| 347 |
+
"eval_logps/rejected": -360.2872619628906,
|
| 348 |
+
"eval_loss": 0.6406751871109009,
|
| 349 |
+
"eval_rewards/accuracies": 0.6726190447807312,
|
| 350 |
+
"eval_rewards/chosen": -0.4273369312286377,
|
| 351 |
+
"eval_rewards/margins": 0.19016937911510468,
|
| 352 |
+
"eval_rewards/rejected": -0.6175063252449036,
|
| 353 |
+
"eval_runtime": 86.2297,
|
| 354 |
+
"eval_samples_per_second": 23.194,
|
| 355 |
+
"eval_steps_per_second": 0.731,
|
| 356 |
+
"step": 200
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 0.43956043956043955,
|
| 360 |
+
"grad_norm": 8.497602462768555,
|
| 361 |
+
"learning_rate": 3.4376480090239047e-07,
|
| 362 |
+
"logits/chosen": -2.7831473350524902,
|
| 363 |
+
"logits/rejected": -2.7467095851898193,
|
| 364 |
+
"logps/chosen": -387.8429870605469,
|
| 365 |
+
"logps/rejected": -351.0397644042969,
|
| 366 |
+
"loss": 0.6471,
|
| 367 |
+
"rewards/accuracies": 0.637499988079071,
|
| 368 |
+
"rewards/chosen": -0.5111196041107178,
|
| 369 |
+
"rewards/margins": 0.1388254463672638,
|
| 370 |
+
"rewards/rejected": -0.649945080280304,
|
| 371 |
+
"step": 210
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 0.4604918890633176,
|
| 375 |
+
"grad_norm": 6.599562168121338,
|
| 376 |
+
"learning_rate": 3.265574537815398e-07,
|
| 377 |
+
"logits/chosen": -2.7829642295837402,
|
| 378 |
+
"logits/rejected": -2.7764458656311035,
|
| 379 |
+
"logps/chosen": -379.44464111328125,
|
| 380 |
+
"logps/rejected": -354.76092529296875,
|
| 381 |
+
"loss": 0.6423,
|
| 382 |
+
"rewards/accuracies": 0.637499988079071,
|
| 383 |
+
"rewards/chosen": -0.49129724502563477,
|
| 384 |
+
"rewards/margins": 0.16036386787891388,
|
| 385 |
+
"rewards/rejected": -0.6516611576080322,
|
| 386 |
+
"step": 220
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 0.48142333856619574,
|
| 390 |
+
"grad_norm": 6.966579914093018,
|
| 391 |
+
"learning_rate": 3.0893973387735683e-07,
|
| 392 |
+
"logits/chosen": -2.7712459564208984,
|
| 393 |
+
"logits/rejected": -2.760551929473877,
|
| 394 |
+
"logps/chosen": -370.9303283691406,
|
| 395 |
+
"logps/rejected": -359.6304626464844,
|
| 396 |
+
"loss": 0.6429,
|
| 397 |
+
"rewards/accuracies": 0.637499988079071,
|
| 398 |
+
"rewards/chosen": -0.436328649520874,
|
| 399 |
+
"rewards/margins": 0.17412248253822327,
|
| 400 |
+
"rewards/rejected": -0.6104511022567749,
|
| 401 |
+
"step": 230
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 0.5023547880690737,
|
| 405 |
+
"grad_norm": 6.429652690887451,
|
| 406 |
+
"learning_rate": 2.910060778827554e-07,
|
| 407 |
+
"logits/chosen": -2.7925572395324707,
|
| 408 |
+
"logits/rejected": -2.7672150135040283,
|
| 409 |
+
"logps/chosen": -370.3599548339844,
|
| 410 |
+
"logps/rejected": -349.8984680175781,
|
| 411 |
+
"loss": 0.6292,
|
| 412 |
+
"rewards/accuracies": 0.671875,
|
| 413 |
+
"rewards/chosen": -0.4170507490634918,
|
| 414 |
+
"rewards/margins": 0.2068098783493042,
|
| 415 |
+
"rewards/rejected": -0.6238606572151184,
|
| 416 |
+
"step": 240
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 0.5232862375719518,
|
| 420 |
+
"grad_norm": 6.210602283477783,
|
| 421 |
+
"learning_rate": 2.7285261601056697e-07,
|
| 422 |
+
"logits/chosen": -2.791151285171509,
|
| 423 |
+
"logits/rejected": -2.761967897415161,
|
| 424 |
+
"logps/chosen": -386.1891784667969,
|
| 425 |
+
"logps/rejected": -342.8139343261719,
|
| 426 |
+
"loss": 0.6327,
|
| 427 |
+
"rewards/accuracies": 0.6781250238418579,
|
| 428 |
+
"rewards/chosen": -0.42713984847068787,
|
| 429 |
+
"rewards/margins": 0.22475413978099823,
|
| 430 |
+
"rewards/rejected": -0.6518939137458801,
|
| 431 |
+
"step": 250
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 0.54421768707483,
|
| 435 |
+
"grad_norm": 6.485888481140137,
|
| 436 |
+
"learning_rate": 2.5457665670441937e-07,
|
| 437 |
+
"logits/chosen": -2.8333914279937744,
|
| 438 |
+
"logits/rejected": -2.8468918800354004,
|
| 439 |
+
"logps/chosen": -381.37078857421875,
|
| 440 |
+
"logps/rejected": -365.30963134765625,
|
| 441 |
+
"loss": 0.6332,
|
| 442 |
+
"rewards/accuracies": 0.684374988079071,
|
| 443 |
+
"rewards/chosen": -0.45365315675735474,
|
| 444 |
+
"rewards/margins": 0.20745711028575897,
|
| 445 |
+
"rewards/rejected": -0.6611102223396301,
|
| 446 |
+
"step": 260
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 0.565149136577708,
|
| 450 |
+
"grad_norm": 7.725765705108643,
|
| 451 |
+
"learning_rate": 2.3627616503391812e-07,
|
| 452 |
+
"logits/chosen": -2.8207736015319824,
|
| 453 |
+
"logits/rejected": -2.8081822395324707,
|
| 454 |
+
"logps/chosen": -386.81976318359375,
|
| 455 |
+
"logps/rejected": -361.7131652832031,
|
| 456 |
+
"loss": 0.6314,
|
| 457 |
+
"rewards/accuracies": 0.659375011920929,
|
| 458 |
+
"rewards/chosen": -0.44246116280555725,
|
| 459 |
+
"rewards/margins": 0.18473473191261292,
|
| 460 |
+
"rewards/rejected": -0.6271958351135254,
|
| 461 |
+
"step": 270
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.5860805860805861,
|
| 465 |
+
"grad_norm": 7.315251350402832,
|
| 466 |
+
"learning_rate": 2.1804923757009882e-07,
|
| 467 |
+
"logits/chosen": -2.7474558353424072,
|
| 468 |
+
"logits/rejected": -2.7452392578125,
|
| 469 |
+
"logps/chosen": -373.4273681640625,
|
| 470 |
+
"logps/rejected": -346.6417541503906,
|
| 471 |
+
"loss": 0.6341,
|
| 472 |
+
"rewards/accuracies": 0.671875,
|
| 473 |
+
"rewards/chosen": -0.4729071259498596,
|
| 474 |
+
"rewards/margins": 0.2166450470685959,
|
| 475 |
+
"rewards/rejected": -0.6895521879196167,
|
| 476 |
+
"step": 280
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 0.6070120355834642,
|
| 480 |
+
"grad_norm": 7.481057643890381,
|
| 481 |
+
"learning_rate": 1.9999357655598891e-07,
|
| 482 |
+
"logits/chosen": -2.8017992973327637,
|
| 483 |
+
"logits/rejected": -2.7807507514953613,
|
| 484 |
+
"logps/chosen": -372.48126220703125,
|
| 485 |
+
"logps/rejected": -341.9537048339844,
|
| 486 |
+
"loss": 0.6264,
|
| 487 |
+
"rewards/accuracies": 0.675000011920929,
|
| 488 |
+
"rewards/chosen": -0.5083416700363159,
|
| 489 |
+
"rewards/margins": 0.176277756690979,
|
| 490 |
+
"rewards/rejected": -0.6846194863319397,
|
| 491 |
+
"step": 290
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.6279434850863422,
|
| 495 |
+
"grad_norm": 8.147472381591797,
|
| 496 |
+
"learning_rate": 1.8220596619089573e-07,
|
| 497 |
+
"logits/chosen": -2.7440390586853027,
|
| 498 |
+
"logits/rejected": -2.717979907989502,
|
| 499 |
+
"logps/chosen": -375.93035888671875,
|
| 500 |
+
"logps/rejected": -361.93121337890625,
|
| 501 |
+
"loss": 0.6285,
|
| 502 |
+
"rewards/accuracies": 0.625,
|
| 503 |
+
"rewards/chosen": -0.4719668924808502,
|
| 504 |
+
"rewards/margins": 0.17040875554084778,
|
| 505 |
+
"rewards/rejected": -0.642375648021698,
|
| 506 |
+
"step": 300
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 0.6279434850863422,
|
| 510 |
+
"eval_logits/chosen": -2.801197052001953,
|
| 511 |
+
"eval_logits/rejected": -2.773117780685425,
|
| 512 |
+
"eval_logps/chosen": -389.8437805175781,
|
| 513 |
+
"eval_logps/rejected": -368.0482177734375,
|
| 514 |
+
"eval_loss": 0.6331161856651306,
|
| 515 |
+
"eval_rewards/accuracies": 0.6646825671195984,
|
| 516 |
+
"eval_rewards/chosen": -0.4723050594329834,
|
| 517 |
+
"eval_rewards/margins": 0.22281017899513245,
|
| 518 |
+
"eval_rewards/rejected": -0.6951152086257935,
|
| 519 |
+
"eval_runtime": 85.9334,
|
| 520 |
+
"eval_samples_per_second": 23.274,
|
| 521 |
+
"eval_steps_per_second": 0.733,
|
| 522 |
+
"step": 300
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 0.6488749345892203,
|
| 526 |
+
"grad_norm": 11.20168399810791,
|
| 527 |
+
"learning_rate": 1.647817538357072e-07,
|
| 528 |
+
"logits/chosen": -2.736398935317993,
|
| 529 |
+
"logits/rejected": -2.738858461380005,
|
| 530 |
+
"logps/chosen": -391.0700378417969,
|
| 531 |
+
"logps/rejected": -347.84356689453125,
|
| 532 |
+
"loss": 0.622,
|
| 533 |
+
"rewards/accuracies": 0.675000011920929,
|
| 534 |
+
"rewards/chosen": -0.4913768768310547,
|
| 535 |
+
"rewards/margins": 0.23215806484222412,
|
| 536 |
+
"rewards/rejected": -0.723534882068634,
|
| 537 |
+
"step": 310
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 0.6698063840920984,
|
| 541 |
+
"grad_norm": 7.335338592529297,
|
| 542 |
+
"learning_rate": 1.478143389201113e-07,
|
| 543 |
+
"logits/chosen": -2.7694268226623535,
|
| 544 |
+
"logits/rejected": -2.74495267868042,
|
| 545 |
+
"logps/chosen": -382.05419921875,
|
| 546 |
+
"logps/rejected": -351.9586486816406,
|
| 547 |
+
"loss": 0.6179,
|
| 548 |
+
"rewards/accuracies": 0.637499988079071,
|
| 549 |
+
"rewards/chosen": -0.5782755017280579,
|
| 550 |
+
"rewards/margins": 0.24550530314445496,
|
| 551 |
+
"rewards/rejected": -0.8237808346748352,
|
| 552 |
+
"step": 320
|
| 553 |
+
},
|
| 554 |
+
{
|
| 555 |
+
"epoch": 0.6907378335949764,
|
| 556 |
+
"grad_norm": 9.174397468566895,
|
| 557 |
+
"learning_rate": 1.3139467229135998e-07,
|
| 558 |
+
"logits/chosen": -2.7542152404785156,
|
| 559 |
+
"logits/rejected": -2.74013352394104,
|
| 560 |
+
"logps/chosen": -391.40618896484375,
|
| 561 |
+
"logps/rejected": -386.9635314941406,
|
| 562 |
+
"loss": 0.63,
|
| 563 |
+
"rewards/accuracies": 0.609375,
|
| 564 |
+
"rewards/chosen": -0.6302947998046875,
|
| 565 |
+
"rewards/margins": 0.17986619472503662,
|
| 566 |
+
"rewards/rejected": -0.8101609945297241,
|
| 567 |
+
"step": 330
|
| 568 |
+
},
|
| 569 |
+
{
|
| 570 |
+
"epoch": 0.7116692830978545,
|
| 571 |
+
"grad_norm": 8.58338451385498,
|
| 572 |
+
"learning_rate": 1.1561076868822755e-07,
|
| 573 |
+
"logits/chosen": -2.75496768951416,
|
| 574 |
+
"logits/rejected": -2.7243494987487793,
|
| 575 |
+
"logps/chosen": -427.29315185546875,
|
| 576 |
+
"logps/rejected": -422.3783264160156,
|
| 577 |
+
"loss": 0.6206,
|
| 578 |
+
"rewards/accuracies": 0.659375011920929,
|
| 579 |
+
"rewards/chosen": -0.6146808862686157,
|
| 580 |
+
"rewards/margins": 0.28181296586990356,
|
| 581 |
+
"rewards/rejected": -0.8964937925338745,
|
| 582 |
+
"step": 340
|
| 583 |
+
},
|
| 584 |
+
{
|
| 585 |
+
"epoch": 0.7326007326007326,
|
| 586 |
+
"grad_norm": 7.745279788970947,
|
| 587 |
+
"learning_rate": 1.0054723495346482e-07,
|
| 588 |
+
"logits/chosen": -2.7472786903381348,
|
| 589 |
+
"logits/rejected": -2.727072238922119,
|
| 590 |
+
"logps/chosen": -380.8338623046875,
|
| 591 |
+
"logps/rejected": -356.40875244140625,
|
| 592 |
+
"loss": 0.6116,
|
| 593 |
+
"rewards/accuracies": 0.6656249761581421,
|
| 594 |
+
"rewards/chosen": -0.6483052968978882,
|
| 595 |
+
"rewards/margins": 0.22691123187541962,
|
| 596 |
+
"rewards/rejected": -0.8752166628837585,
|
| 597 |
+
"step": 350
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"epoch": 0.7535321821036107,
|
| 601 |
+
"grad_norm": 8.188222885131836,
|
| 602 |
+
"learning_rate": 8.628481651367875e-08,
|
| 603 |
+
"logits/chosen": -2.7281575202941895,
|
| 604 |
+
"logits/rejected": -2.714717388153076,
|
| 605 |
+
"logps/chosen": -399.79217529296875,
|
| 606 |
+
"logps/rejected": -374.2989807128906,
|
| 607 |
+
"loss": 0.6313,
|
| 608 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 609 |
+
"rewards/chosen": -0.7062140107154846,
|
| 610 |
+
"rewards/margins": 0.22601504623889923,
|
| 611 |
+
"rewards/rejected": -0.9322290420532227,
|
| 612 |
+
"step": 360
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"epoch": 0.7744636316064888,
|
| 616 |
+
"grad_norm": 7.2593889236450195,
|
| 617 |
+
"learning_rate": 7.289996455765748e-08,
|
| 618 |
+
"logits/chosen": -2.753688097000122,
|
| 619 |
+
"logits/rejected": -2.7259750366210938,
|
| 620 |
+
"logps/chosen": -383.2185974121094,
|
| 621 |
+
"logps/rejected": -367.9438781738281,
|
| 622 |
+
"loss": 0.6152,
|
| 623 |
+
"rewards/accuracies": 0.637499988079071,
|
| 624 |
+
"rewards/chosen": -0.7225135564804077,
|
| 625 |
+
"rewards/margins": 0.21898405253887177,
|
| 626 |
+
"rewards/rejected": -0.9414976239204407,
|
| 627 |
+
"step": 370
|
| 628 |
+
},
|
| 629 |
+
{
|
| 630 |
+
"epoch": 0.7953950811093669,
|
| 631 |
+
"grad_norm": 7.146636009216309,
|
| 632 |
+
"learning_rate": 6.046442623320145e-08,
|
| 633 |
+
"logits/chosen": -2.6859023571014404,
|
| 634 |
+
"logits/rejected": -2.695509433746338,
|
| 635 |
+
"logps/chosen": -363.49493408203125,
|
| 636 |
+
"logps/rejected": -377.5242004394531,
|
| 637 |
+
"loss": 0.6159,
|
| 638 |
+
"rewards/accuracies": 0.6656249761581421,
|
| 639 |
+
"rewards/chosen": -0.7377739548683167,
|
| 640 |
+
"rewards/margins": 0.25747013092041016,
|
| 641 |
+
"rewards/rejected": -0.995244026184082,
|
| 642 |
+
"step": 380
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 0.8163265306122449,
|
| 646 |
+
"grad_norm": 7.638937473297119,
|
| 647 |
+
"learning_rate": 4.904486005914027e-08,
|
| 648 |
+
"logits/chosen": -2.6865615844726562,
|
| 649 |
+
"logits/rejected": -2.6732540130615234,
|
| 650 |
+
"logps/chosen": -433.0126037597656,
|
| 651 |
+
"logps/rejected": -432.28448486328125,
|
| 652 |
+
"loss": 0.6203,
|
| 653 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 654 |
+
"rewards/chosen": -0.6989966034889221,
|
| 655 |
+
"rewards/margins": 0.23467639088630676,
|
| 656 |
+
"rewards/rejected": -0.9336729049682617,
|
| 657 |
+
"step": 390
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 0.837257980115123,
|
| 661 |
+
"grad_norm": 8.645991325378418,
|
| 662 |
+
"learning_rate": 3.8702478614051345e-08,
|
| 663 |
+
"logits/chosen": -2.6967921257019043,
|
| 664 |
+
"logits/rejected": -2.6760878562927246,
|
| 665 |
+
"logps/chosen": -396.04595947265625,
|
| 666 |
+
"logps/rejected": -385.61810302734375,
|
| 667 |
+
"loss": 0.6222,
|
| 668 |
+
"rewards/accuracies": 0.6781250238418579,
|
| 669 |
+
"rewards/chosen": -0.674105167388916,
|
| 670 |
+
"rewards/margins": 0.25409311056137085,
|
| 671 |
+
"rewards/rejected": -0.9281982183456421,
|
| 672 |
+
"step": 400
|
| 673 |
+
},
|
| 674 |
+
{
|
| 675 |
+
"epoch": 0.837257980115123,
|
| 676 |
+
"eval_logits/chosen": -2.7399401664733887,
|
| 677 |
+
"eval_logits/rejected": -2.7107229232788086,
|
| 678 |
+
"eval_logps/chosen": -412.686767578125,
|
| 679 |
+
"eval_logps/rejected": -395.38079833984375,
|
| 680 |
+
"eval_loss": 0.6240472197532654,
|
| 681 |
+
"eval_rewards/accuracies": 0.682539701461792,
|
| 682 |
+
"eval_rewards/chosen": -0.7007347345352173,
|
| 683 |
+
"eval_rewards/margins": 0.2677067816257477,
|
| 684 |
+
"eval_rewards/rejected": -0.9684414863586426,
|
| 685 |
+
"eval_runtime": 86.0349,
|
| 686 |
+
"eval_samples_per_second": 23.246,
|
| 687 |
+
"eval_steps_per_second": 0.732,
|
| 688 |
+
"step": 400
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"epoch": 0.858189429618001,
|
| 692 |
+
"grad_norm": 8.63017749786377,
|
| 693 |
+
"learning_rate": 2.9492720416985e-08,
|
| 694 |
+
"logits/chosen": -2.72432279586792,
|
| 695 |
+
"logits/rejected": -2.709479808807373,
|
| 696 |
+
"logps/chosen": -423.2232971191406,
|
| 697 |
+
"logps/rejected": -398.17779541015625,
|
| 698 |
+
"loss": 0.6347,
|
| 699 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 700 |
+
"rewards/chosen": -0.6851434111595154,
|
| 701 |
+
"rewards/margins": 0.2306232750415802,
|
| 702 |
+
"rewards/rejected": -0.915766716003418,
|
| 703 |
+
"step": 410
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 0.8791208791208791,
|
| 707 |
+
"grad_norm": 8.239731788635254,
|
| 708 |
+
"learning_rate": 2.1464952759020856e-08,
|
| 709 |
+
"logits/chosen": -2.6693649291992188,
|
| 710 |
+
"logits/rejected": -2.672886371612549,
|
| 711 |
+
"logps/chosen": -392.8204040527344,
|
| 712 |
+
"logps/rejected": -393.4344177246094,
|
| 713 |
+
"loss": 0.6118,
|
| 714 |
+
"rewards/accuracies": 0.640625,
|
| 715 |
+
"rewards/chosen": -0.7332924008369446,
|
| 716 |
+
"rewards/margins": 0.24412448704242706,
|
| 717 |
+
"rewards/rejected": -0.9774168133735657,
|
| 718 |
+
"step": 420
|
| 719 |
+
},
|
| 720 |
+
{
|
| 721 |
+
"epoch": 0.9000523286237572,
|
| 722 |
+
"grad_norm": 8.310417175292969,
|
| 723 |
+
"learning_rate": 1.4662207078575684e-08,
|
| 724 |
+
"logits/chosen": -2.6678881645202637,
|
| 725 |
+
"logits/rejected": -2.6358609199523926,
|
| 726 |
+
"logps/chosen": -409.0691833496094,
|
| 727 |
+
"logps/rejected": -372.57183837890625,
|
| 728 |
+
"loss": 0.6136,
|
| 729 |
+
"rewards/accuracies": 0.653124988079071,
|
| 730 |
+
"rewards/chosen": -0.7174139618873596,
|
| 731 |
+
"rewards/margins": 0.21988801658153534,
|
| 732 |
+
"rewards/rejected": -0.9373019933700562,
|
| 733 |
+
"step": 430
|
| 734 |
+
},
|
| 735 |
+
{
|
| 736 |
+
"epoch": 0.9209837781266352,
|
| 737 |
+
"grad_norm": 13.682318687438965,
|
| 738 |
+
"learning_rate": 9.12094829893642e-09,
|
| 739 |
+
"logits/chosen": -2.718200445175171,
|
| 740 |
+
"logits/rejected": -2.689492702484131,
|
| 741 |
+
"logps/chosen": -413.35528564453125,
|
| 742 |
+
"logps/rejected": -406.2598876953125,
|
| 743 |
+
"loss": 0.6068,
|
| 744 |
+
"rewards/accuracies": 0.625,
|
| 745 |
+
"rewards/chosen": -0.7059027552604675,
|
| 746 |
+
"rewards/margins": 0.23966805636882782,
|
| 747 |
+
"rewards/rejected": -0.9455708265304565,
|
| 748 |
+
"step": 440
|
| 749 |
+
},
|
| 750 |
+
{
|
| 751 |
+
"epoch": 0.9419152276295133,
|
| 752 |
+
"grad_norm": 8.156304359436035,
|
| 753 |
+
"learning_rate": 4.8708793644441086e-09,
|
| 754 |
+
"logits/chosen": -2.63850736618042,
|
| 755 |
+
"logits/rejected": -2.6505093574523926,
|
| 756 |
+
"logps/chosen": -413.0054626464844,
|
| 757 |
+
"logps/rejected": -420.00018310546875,
|
| 758 |
+
"loss": 0.6209,
|
| 759 |
+
"rewards/accuracies": 0.671875,
|
| 760 |
+
"rewards/chosen": -0.7106130719184875,
|
| 761 |
+
"rewards/margins": 0.24646687507629395,
|
| 762 |
+
"rewards/rejected": -0.957080066204071,
|
| 763 |
+
"step": 450
|
| 764 |
+
},
|
| 765 |
+
{
|
| 766 |
+
"epoch": 0.9628466771323915,
|
| 767 |
+
"grad_norm": 10.356741905212402,
|
| 768 |
+
"learning_rate": 1.9347820230782295e-09,
|
| 769 |
+
"logits/chosen": -2.698603630065918,
|
| 770 |
+
"logits/rejected": -2.671870231628418,
|
| 771 |
+
"logps/chosen": -374.42828369140625,
|
| 772 |
+
"logps/rejected": -359.28692626953125,
|
| 773 |
+
"loss": 0.6226,
|
| 774 |
+
"rewards/accuracies": 0.643750011920929,
|
| 775 |
+
"rewards/chosen": -0.7496173977851868,
|
| 776 |
+
"rewards/margins": 0.20760098099708557,
|
| 777 |
+
"rewards/rejected": -0.95721834897995,
|
| 778 |
+
"step": 460
|
| 779 |
+
},
|
| 780 |
+
{
|
| 781 |
+
"epoch": 0.9837781266352695,
|
| 782 |
+
"grad_norm": 9.971595764160156,
|
| 783 |
+
"learning_rate": 3.2839470889836627e-10,
|
| 784 |
+
"logits/chosen": -2.7027974128723145,
|
| 785 |
+
"logits/rejected": -2.684480905532837,
|
| 786 |
+
"logps/chosen": -412.5672302246094,
|
| 787 |
+
"logps/rejected": -383.63372802734375,
|
| 788 |
+
"loss": 0.6113,
|
| 789 |
+
"rewards/accuracies": 0.640625,
|
| 790 |
+
"rewards/chosen": -0.7107462882995605,
|
| 791 |
+
"rewards/margins": 0.1883649080991745,
|
| 792 |
+
"rewards/rejected": -0.8991111516952515,
|
| 793 |
+
"step": 470
|
| 794 |
+
},
|
| 795 |
+
{
|
| 796 |
+
"epoch": 0.9984301412872841,
|
| 797 |
+
"step": 477,
|
| 798 |
+
"total_flos": 0.0,
|
| 799 |
+
"train_loss": 0.6428392008415558,
|
| 800 |
+
"train_runtime": 6631.0546,
|
| 801 |
+
"train_samples_per_second": 9.219,
|
| 802 |
+
"train_steps_per_second": 0.072
|
| 803 |
+
}
|
| 804 |
+
],
|
| 805 |
+
"logging_steps": 10,
|
| 806 |
+
"max_steps": 477,
|
| 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 |
+
}
|