Model save
Browse files
README.md
ADDED
|
@@ -0,0 +1,82 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: alignment-handbook/zephyr-7b-sft-full
|
| 3 |
+
library_name: peft
|
| 4 |
+
license: apache-2.0
|
| 5 |
+
tags:
|
| 6 |
+
- trl
|
| 7 |
+
- dpo
|
| 8 |
+
- generated_from_trainer
|
| 9 |
+
model-index:
|
| 10 |
+
- name: zephyr-dpop-qlora-uf6k-5e-6
|
| 11 |
+
results: []
|
| 12 |
+
---
|
| 13 |
+
|
| 14 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 15 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 16 |
+
|
| 17 |
+
# zephyr-dpop-qlora-uf6k-5e-6
|
| 18 |
+
|
| 19 |
+
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.
|
| 20 |
+
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 0.6879
|
| 22 |
+
- Positive Losses: 0.0941
|
| 23 |
+
- Dpo Losses: 0.6679
|
| 24 |
+
- Rewards/chosen: 0.1483
|
| 25 |
+
- Rewards/rejected: 0.0942
|
| 26 |
+
- Rewards/accuracies: 0.6900
|
| 27 |
+
- Rewards/margins: 0.0541
|
| 28 |
+
- Rewards/margins Max: 0.2146
|
| 29 |
+
- Rewards/margins Min: -0.0945
|
| 30 |
+
- Rewards/margins Std: 0.1028
|
| 31 |
+
- Logps/rejected: -249.1596
|
| 32 |
+
- Logps/chosen: -269.7678
|
| 33 |
+
- Logits/rejected: -2.7464
|
| 34 |
+
- Logits/chosen: -2.7837
|
| 35 |
+
|
| 36 |
+
## Model description
|
| 37 |
+
|
| 38 |
+
More information needed
|
| 39 |
+
|
| 40 |
+
## Intended uses & limitations
|
| 41 |
+
|
| 42 |
+
More information needed
|
| 43 |
+
|
| 44 |
+
## Training and evaluation data
|
| 45 |
+
|
| 46 |
+
More information needed
|
| 47 |
+
|
| 48 |
+
## Training procedure
|
| 49 |
+
|
| 50 |
+
### Training hyperparameters
|
| 51 |
+
|
| 52 |
+
The following hyperparameters were used during training:
|
| 53 |
+
- learning_rate: 5e-06
|
| 54 |
+
- train_batch_size: 4
|
| 55 |
+
- eval_batch_size: 8
|
| 56 |
+
- seed: 42
|
| 57 |
+
- distributed_type: multi-GPU
|
| 58 |
+
- num_devices: 2
|
| 59 |
+
- gradient_accumulation_steps: 2
|
| 60 |
+
- total_train_batch_size: 16
|
| 61 |
+
- total_eval_batch_size: 16
|
| 62 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 63 |
+
- lr_scheduler_type: cosine
|
| 64 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 65 |
+
- num_epochs: 1
|
| 66 |
+
|
| 67 |
+
### Training results
|
| 68 |
+
|
| 69 |
+
| Training Loss | Epoch | Step | Validation Loss | Positive Losses | Dpo Losses | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Rewards/margins Max | Rewards/margins Min | Rewards/margins Std | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 70 |
+
|:-------------:|:-----:|:----:|:---------------:|:---------------:|:----------:|:--------------:|:----------------:|:------------------:|:---------------:|:-------------------:|:-------------------:|:-------------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 71 |
+
| 0.6896 | 0.3 | 100 | 0.6927 | 0.0589 | 0.6767 | 0.1278 | 0.0930 | 0.6710 | 0.0348 | 0.1572 | -0.0767 | 0.0771 | -249.2807 | -271.8154 | -2.7452 | -2.7829 |
|
| 72 |
+
| 0.6919 | 0.61 | 200 | 0.6864 | 0.0952 | 0.6689 | 0.1438 | 0.0918 | 0.6830 | 0.0521 | 0.2115 | -0.0943 | 0.1012 | -249.3993 | -270.2086 | -2.7481 | -2.7856 |
|
| 73 |
+
| 0.6774 | 0.91 | 300 | 0.6879 | 0.0941 | 0.6679 | 0.1483 | 0.0942 | 0.6900 | 0.0541 | 0.2146 | -0.0945 | 0.1028 | -249.1596 | -269.7678 | -2.7464 | -2.7837 |
|
| 74 |
+
|
| 75 |
+
|
| 76 |
+
### Framework versions
|
| 77 |
+
|
| 78 |
+
- PEFT 0.7.1
|
| 79 |
+
- Transformers 4.39.0.dev0
|
| 80 |
+
- Pytorch 2.1.2+cu121
|
| 81 |
+
- Datasets 2.14.6
|
| 82 |
+
- Tokenizers 0.15.2
|
adapter_model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 671150064
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:084a8a0e00a7f6c92d47bf419fe0d26f355d6f36d67f34470cd42eb064428d84
|
| 3 |
size 671150064
|
all_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.6874516553791823,
|
| 4 |
+
"train_runtime": 3895.4317,
|
| 5 |
+
"train_samples": 5263,
|
| 6 |
+
"train_samples_per_second": 1.351,
|
| 7 |
+
"train_steps_per_second": 0.084
|
| 8 |
+
}
|
runs/Jul25_08-24-43_notebook-deployment-48-7d9b6c99-khd85/events.out.tfevents.1721895977.notebook-deployment-48-7d9b6c99-khd85.3153931.0
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1a28df16c988343ea46d30e2b7003b5c759a3543b2a96670546714a00b9b3350
|
| 3 |
+
size 41417
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.6874516553791823,
|
| 4 |
+
"train_runtime": 3895.4317,
|
| 5 |
+
"train_samples": 5263,
|
| 6 |
+
"train_samples_per_second": 1.351,
|
| 7 |
+
"train_steps_per_second": 0.084
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,753 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 1.0,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 329,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"dpo_losses": 0.6931471824645996,
|
| 13 |
+
"epoch": 0.0,
|
| 14 |
+
"grad_norm": 2.094792348544482,
|
| 15 |
+
"learning_rate": 1.5151515151515152e-07,
|
| 16 |
+
"logits/chosen": -2.6820077896118164,
|
| 17 |
+
"logits/rejected": -2.6930205821990967,
|
| 18 |
+
"logps/chosen": -281.2528381347656,
|
| 19 |
+
"logps/rejected": -258.0622253417969,
|
| 20 |
+
"loss": 0.6931,
|
| 21 |
+
"positive_losses": 0.0,
|
| 22 |
+
"rewards/accuracies": 0.0,
|
| 23 |
+
"rewards/chosen": 0.0,
|
| 24 |
+
"rewards/margins": 0.0,
|
| 25 |
+
"rewards/margins_max": 0.0,
|
| 26 |
+
"rewards/margins_min": 0.0,
|
| 27 |
+
"rewards/margins_std": 0.0,
|
| 28 |
+
"rewards/rejected": 0.0,
|
| 29 |
+
"step": 1
|
| 30 |
+
},
|
| 31 |
+
{
|
| 32 |
+
"dpo_losses": 0.6931204795837402,
|
| 33 |
+
"epoch": 0.03,
|
| 34 |
+
"grad_norm": 7.907274986440123,
|
| 35 |
+
"learning_rate": 1.5151515151515152e-06,
|
| 36 |
+
"logits/chosen": -2.7687692642211914,
|
| 37 |
+
"logits/rejected": -2.7543087005615234,
|
| 38 |
+
"logps/chosen": -284.4098815917969,
|
| 39 |
+
"logps/rejected": -249.6673126220703,
|
| 40 |
+
"loss": 0.6967,
|
| 41 |
+
"positive_losses": 0.03447924554347992,
|
| 42 |
+
"rewards/accuracies": 0.4583333432674408,
|
| 43 |
+
"rewards/chosen": 0.0019425028003752232,
|
| 44 |
+
"rewards/margins": 5.710232289857231e-05,
|
| 45 |
+
"rewards/margins_max": 0.003333130618557334,
|
| 46 |
+
"rewards/margins_min": -0.0030696813482791185,
|
| 47 |
+
"rewards/margins_std": 0.002927461639046669,
|
| 48 |
+
"rewards/rejected": 0.0018854006193578243,
|
| 49 |
+
"step": 10
|
| 50 |
+
},
|
| 51 |
+
{
|
| 52 |
+
"dpo_losses": 0.6927222609519958,
|
| 53 |
+
"epoch": 0.06,
|
| 54 |
+
"grad_norm": 1.9785585238747578,
|
| 55 |
+
"learning_rate": 3.0303030303030305e-06,
|
| 56 |
+
"logits/chosen": -2.8341968059539795,
|
| 57 |
+
"logits/rejected": -2.7815167903900146,
|
| 58 |
+
"logps/chosen": -289.8644714355469,
|
| 59 |
+
"logps/rejected": -268.81475830078125,
|
| 60 |
+
"loss": 0.6935,
|
| 61 |
+
"positive_losses": 0.00822601281106472,
|
| 62 |
+
"rewards/accuracies": 0.5,
|
| 63 |
+
"rewards/chosen": 0.016672641038894653,
|
| 64 |
+
"rewards/margins": 0.0008590976940467954,
|
| 65 |
+
"rewards/margins_max": 0.007572791539132595,
|
| 66 |
+
"rewards/margins_min": -0.004768169019371271,
|
| 67 |
+
"rewards/margins_std": 0.0055806804448366165,
|
| 68 |
+
"rewards/rejected": 0.015813542529940605,
|
| 69 |
+
"step": 20
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"dpo_losses": 0.6912546157836914,
|
| 73 |
+
"epoch": 0.09,
|
| 74 |
+
"grad_norm": 9.993422824831308,
|
| 75 |
+
"learning_rate": 4.5454545454545455e-06,
|
| 76 |
+
"logits/chosen": -2.86173152923584,
|
| 77 |
+
"logits/rejected": -2.814188241958618,
|
| 78 |
+
"logps/chosen": -255.656494140625,
|
| 79 |
+
"logps/rejected": -224.1510467529297,
|
| 80 |
+
"loss": 0.692,
|
| 81 |
+
"positive_losses": 0.00018234252638649195,
|
| 82 |
+
"rewards/accuracies": 0.5625,
|
| 83 |
+
"rewards/chosen": 0.03588816151022911,
|
| 84 |
+
"rewards/margins": 0.00383453955873847,
|
| 85 |
+
"rewards/margins_max": 0.017018133774399757,
|
| 86 |
+
"rewards/margins_min": -0.008637982420623302,
|
| 87 |
+
"rewards/margins_std": 0.011648830957710743,
|
| 88 |
+
"rewards/rejected": 0.0320536233484745,
|
| 89 |
+
"step": 30
|
| 90 |
+
},
|
| 91 |
+
{
|
| 92 |
+
"dpo_losses": 0.6852566003799438,
|
| 93 |
+
"epoch": 0.12,
|
| 94 |
+
"grad_norm": 7.551494566668885,
|
| 95 |
+
"learning_rate": 4.993103596812269e-06,
|
| 96 |
+
"logits/chosen": -2.8286097049713135,
|
| 97 |
+
"logits/rejected": -2.7643656730651855,
|
| 98 |
+
"logps/chosen": -311.00653076171875,
|
| 99 |
+
"logps/rejected": -219.75357055664062,
|
| 100 |
+
"loss": 0.6928,
|
| 101 |
+
"positive_losses": 0.0,
|
| 102 |
+
"rewards/accuracies": 0.6875,
|
| 103 |
+
"rewards/chosen": 0.06535681337118149,
|
| 104 |
+
"rewards/margins": 0.016044817864894867,
|
| 105 |
+
"rewards/margins_max": 0.04139421135187149,
|
| 106 |
+
"rewards/margins_min": -0.006711454130709171,
|
| 107 |
+
"rewards/margins_std": 0.021322522312402725,
|
| 108 |
+
"rewards/rejected": 0.049311988055706024,
|
| 109 |
+
"step": 40
|
| 110 |
+
},
|
| 111 |
+
{
|
| 112 |
+
"dpo_losses": 0.6857399940490723,
|
| 113 |
+
"epoch": 0.15,
|
| 114 |
+
"grad_norm": 7.613114395779342,
|
| 115 |
+
"learning_rate": 4.95941685833271e-06,
|
| 116 |
+
"logits/chosen": -2.7872977256774902,
|
| 117 |
+
"logits/rejected": -2.8009393215179443,
|
| 118 |
+
"logps/chosen": -235.7885284423828,
|
| 119 |
+
"logps/rejected": -274.2627258300781,
|
| 120 |
+
"loss": 0.6875,
|
| 121 |
+
"positive_losses": 0.015755081549286842,
|
| 122 |
+
"rewards/accuracies": 0.6875,
|
| 123 |
+
"rewards/chosen": 0.07146342098712921,
|
| 124 |
+
"rewards/margins": 0.0152506772428751,
|
| 125 |
+
"rewards/margins_max": 0.058862246572971344,
|
| 126 |
+
"rewards/margins_min": -0.022146200761198997,
|
| 127 |
+
"rewards/margins_std": 0.03581809252500534,
|
| 128 |
+
"rewards/rejected": 0.056212734431028366,
|
| 129 |
+
"step": 50
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"dpo_losses": 0.680141806602478,
|
| 133 |
+
"epoch": 0.18,
|
| 134 |
+
"grad_norm": 7.505695469277918,
|
| 135 |
+
"learning_rate": 4.898051734555676e-06,
|
| 136 |
+
"logits/chosen": -2.8290228843688965,
|
| 137 |
+
"logits/rejected": -2.839033603668213,
|
| 138 |
+
"logps/chosen": -312.344482421875,
|
| 139 |
+
"logps/rejected": -276.32183837890625,
|
| 140 |
+
"loss": 0.6846,
|
| 141 |
+
"positive_losses": 0.05971794202923775,
|
| 142 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 143 |
+
"rewards/chosen": 0.0957661122083664,
|
| 144 |
+
"rewards/margins": 0.027223944664001465,
|
| 145 |
+
"rewards/margins_max": 0.09758423268795013,
|
| 146 |
+
"rewards/margins_min": -0.04277171194553375,
|
| 147 |
+
"rewards/margins_std": 0.061675649136304855,
|
| 148 |
+
"rewards/rejected": 0.06854216754436493,
|
| 149 |
+
"step": 60
|
| 150 |
+
},
|
| 151 |
+
{
|
| 152 |
+
"dpo_losses": 0.6815682649612427,
|
| 153 |
+
"epoch": 0.21,
|
| 154 |
+
"grad_norm": 1.5360756176524522,
|
| 155 |
+
"learning_rate": 4.809698831278217e-06,
|
| 156 |
+
"logits/chosen": -2.736055374145508,
|
| 157 |
+
"logits/rejected": -2.7235770225524902,
|
| 158 |
+
"logps/chosen": -257.50408935546875,
|
| 159 |
+
"logps/rejected": -239.8359375,
|
| 160 |
+
"loss": 0.694,
|
| 161 |
+
"positive_losses": 0.1467689573764801,
|
| 162 |
+
"rewards/accuracies": 0.637499988079071,
|
| 163 |
+
"rewards/chosen": 0.08973346650600433,
|
| 164 |
+
"rewards/margins": 0.024347994476556778,
|
| 165 |
+
"rewards/margins_max": 0.0899825394153595,
|
| 166 |
+
"rewards/margins_min": -0.03823970630764961,
|
| 167 |
+
"rewards/margins_std": 0.05780817195773125,
|
| 168 |
+
"rewards/rejected": 0.06538547575473785,
|
| 169 |
+
"step": 70
|
| 170 |
+
},
|
| 171 |
+
{
|
| 172 |
+
"dpo_losses": 0.6777443885803223,
|
| 173 |
+
"epoch": 0.24,
|
| 174 |
+
"grad_norm": 7.6098783040686255,
|
| 175 |
+
"learning_rate": 4.695352475952706e-06,
|
| 176 |
+
"logits/chosen": -2.825540542602539,
|
| 177 |
+
"logits/rejected": -2.8001182079315186,
|
| 178 |
+
"logps/chosen": -272.2143859863281,
|
| 179 |
+
"logps/rejected": -268.1064147949219,
|
| 180 |
+
"loss": 0.6843,
|
| 181 |
+
"positive_losses": 0.019883345812559128,
|
| 182 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 183 |
+
"rewards/chosen": 0.10059104114770889,
|
| 184 |
+
"rewards/margins": 0.03249071165919304,
|
| 185 |
+
"rewards/margins_max": 0.10248656570911407,
|
| 186 |
+
"rewards/margins_min": -0.04477204009890556,
|
| 187 |
+
"rewards/margins_std": 0.06633640825748444,
|
| 188 |
+
"rewards/rejected": 0.06810032576322556,
|
| 189 |
+
"step": 80
|
| 190 |
+
},
|
| 191 |
+
{
|
| 192 |
+
"dpo_losses": 0.6827337741851807,
|
| 193 |
+
"epoch": 0.27,
|
| 194 |
+
"grad_norm": 13.032468304298272,
|
| 195 |
+
"learning_rate": 4.556299527482029e-06,
|
| 196 |
+
"logits/chosen": -2.7769341468811035,
|
| 197 |
+
"logits/rejected": -2.7250094413757324,
|
| 198 |
+
"logps/chosen": -285.997314453125,
|
| 199 |
+
"logps/rejected": -284.3042297363281,
|
| 200 |
+
"loss": 0.6924,
|
| 201 |
+
"positive_losses": 0.0755772590637207,
|
| 202 |
+
"rewards/accuracies": 0.612500011920929,
|
| 203 |
+
"rewards/chosen": 0.09554663300514221,
|
| 204 |
+
"rewards/margins": 0.022927606478333473,
|
| 205 |
+
"rewards/margins_max": 0.11027027666568756,
|
| 206 |
+
"rewards/margins_min": -0.06920486688613892,
|
| 207 |
+
"rewards/margins_std": 0.0804494172334671,
|
| 208 |
+
"rewards/rejected": 0.07261903584003448,
|
| 209 |
+
"step": 90
|
| 210 |
+
},
|
| 211 |
+
{
|
| 212 |
+
"dpo_losses": 0.6703774333000183,
|
| 213 |
+
"epoch": 0.3,
|
| 214 |
+
"grad_norm": 1.6704382025698337,
|
| 215 |
+
"learning_rate": 4.394104893853007e-06,
|
| 216 |
+
"logits/chosen": -2.8705239295959473,
|
| 217 |
+
"logits/rejected": -2.832496166229248,
|
| 218 |
+
"logps/chosen": -262.93939208984375,
|
| 219 |
+
"logps/rejected": -251.376953125,
|
| 220 |
+
"loss": 0.6896,
|
| 221 |
+
"positive_losses": 0.10124854743480682,
|
| 222 |
+
"rewards/accuracies": 0.762499988079071,
|
| 223 |
+
"rewards/chosen": 0.10522875934839249,
|
| 224 |
+
"rewards/margins": 0.04796325042843819,
|
| 225 |
+
"rewards/margins_max": 0.13638350367546082,
|
| 226 |
+
"rewards/margins_min": -0.02367212623357773,
|
| 227 |
+
"rewards/margins_std": 0.07211659848690033,
|
| 228 |
+
"rewards/rejected": 0.0572655126452446,
|
| 229 |
+
"step": 100
|
| 230 |
+
},
|
| 231 |
+
{
|
| 232 |
+
"epoch": 0.3,
|
| 233 |
+
"eval_dpo_losses": 0.6766990423202515,
|
| 234 |
+
"eval_logits/chosen": -2.782886028289795,
|
| 235 |
+
"eval_logits/rejected": -2.7451796531677246,
|
| 236 |
+
"eval_logps/chosen": -271.81536865234375,
|
| 237 |
+
"eval_logps/rejected": -249.28070068359375,
|
| 238 |
+
"eval_loss": 0.692669689655304,
|
| 239 |
+
"eval_positive_losses": 0.05891977623105049,
|
| 240 |
+
"eval_rewards/accuracies": 0.6710000038146973,
|
| 241 |
+
"eval_rewards/chosen": 0.12778052687644958,
|
| 242 |
+
"eval_rewards/margins": 0.03479834645986557,
|
| 243 |
+
"eval_rewards/margins_max": 0.15719929337501526,
|
| 244 |
+
"eval_rewards/margins_min": -0.0766773670911789,
|
| 245 |
+
"eval_rewards/margins_std": 0.07707200944423676,
|
| 246 |
+
"eval_rewards/rejected": 0.09298218786716461,
|
| 247 |
+
"eval_runtime": 429.6713,
|
| 248 |
+
"eval_samples_per_second": 4.655,
|
| 249 |
+
"eval_steps_per_second": 0.291,
|
| 250 |
+
"step": 100
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"dpo_losses": 0.669114887714386,
|
| 254 |
+
"epoch": 0.33,
|
| 255 |
+
"grad_norm": 2.0588703005951716,
|
| 256 |
+
"learning_rate": 4.210593920593201e-06,
|
| 257 |
+
"logits/chosen": -2.732736110687256,
|
| 258 |
+
"logits/rejected": -2.717144727706909,
|
| 259 |
+
"logps/chosen": -297.5173645019531,
|
| 260 |
+
"logps/rejected": -227.2909698486328,
|
| 261 |
+
"loss": 0.6716,
|
| 262 |
+
"positive_losses": 0.005372238345444202,
|
| 263 |
+
"rewards/accuracies": 0.6875,
|
| 264 |
+
"rewards/chosen": 0.14238855242729187,
|
| 265 |
+
"rewards/margins": 0.06108074635267258,
|
| 266 |
+
"rewards/margins_max": 0.21817930042743683,
|
| 267 |
+
"rewards/margins_min": -0.03649354726076126,
|
| 268 |
+
"rewards/margins_std": 0.11923160403966904,
|
| 269 |
+
"rewards/rejected": 0.08130781352519989,
|
| 270 |
+
"step": 110
|
| 271 |
+
},
|
| 272 |
+
{
|
| 273 |
+
"dpo_losses": 0.6759502291679382,
|
| 274 |
+
"epoch": 0.36,
|
| 275 |
+
"grad_norm": 2.114750885236866,
|
| 276 |
+
"learning_rate": 4.007831848252212e-06,
|
| 277 |
+
"logits/chosen": -2.71893572807312,
|
| 278 |
+
"logits/rejected": -2.7199361324310303,
|
| 279 |
+
"logps/chosen": -312.4023132324219,
|
| 280 |
+
"logps/rejected": -266.9503479003906,
|
| 281 |
+
"loss": 0.6841,
|
| 282 |
+
"positive_losses": 0.1982906311750412,
|
| 283 |
+
"rewards/accuracies": 0.637499988079071,
|
| 284 |
+
"rewards/chosen": 0.11162187904119492,
|
| 285 |
+
"rewards/margins": 0.03713056445121765,
|
| 286 |
+
"rewards/margins_max": 0.13481460511684418,
|
| 287 |
+
"rewards/margins_min": -0.06108905002474785,
|
| 288 |
+
"rewards/margins_std": 0.08896367251873016,
|
| 289 |
+
"rewards/rejected": 0.07449130713939667,
|
| 290 |
+
"step": 120
|
| 291 |
+
},
|
| 292 |
+
{
|
| 293 |
+
"dpo_losses": 0.6827476024627686,
|
| 294 |
+
"epoch": 0.4,
|
| 295 |
+
"grad_norm": 1.6378871780311928,
|
| 296 |
+
"learning_rate": 3.7881005700938635e-06,
|
| 297 |
+
"logits/chosen": -2.7933266162872314,
|
| 298 |
+
"logits/rejected": -2.803476333618164,
|
| 299 |
+
"logps/chosen": -254.1282196044922,
|
| 300 |
+
"logps/rejected": -224.1819610595703,
|
| 301 |
+
"loss": 0.6872,
|
| 302 |
+
"positive_losses": 0.06588058173656464,
|
| 303 |
+
"rewards/accuracies": 0.612500011920929,
|
| 304 |
+
"rewards/chosen": 0.12265358120203018,
|
| 305 |
+
"rewards/margins": 0.023296961560845375,
|
| 306 |
+
"rewards/margins_max": 0.1208866611123085,
|
| 307 |
+
"rewards/margins_min": -0.07595702260732651,
|
| 308 |
+
"rewards/margins_std": 0.08657418191432953,
|
| 309 |
+
"rewards/rejected": 0.09935661405324936,
|
| 310 |
+
"step": 130
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"dpo_losses": 0.6661409735679626,
|
| 314 |
+
"epoch": 0.43,
|
| 315 |
+
"grad_norm": 1.861594662390886,
|
| 316 |
+
"learning_rate": 3.553872951569236e-06,
|
| 317 |
+
"logits/chosen": -2.7491869926452637,
|
| 318 |
+
"logits/rejected": -2.7360832691192627,
|
| 319 |
+
"logps/chosen": -280.62664794921875,
|
| 320 |
+
"logps/rejected": -262.3188781738281,
|
| 321 |
+
"loss": 0.6745,
|
| 322 |
+
"positive_losses": 0.11735143512487411,
|
| 323 |
+
"rewards/accuracies": 0.737500011920929,
|
| 324 |
+
"rewards/chosen": 0.13766564428806305,
|
| 325 |
+
"rewards/margins": 0.05706212669610977,
|
| 326 |
+
"rewards/margins_max": 0.1553352326154709,
|
| 327 |
+
"rewards/margins_min": -0.027615731582045555,
|
| 328 |
+
"rewards/margins_std": 0.08170460909605026,
|
| 329 |
+
"rewards/rejected": 0.08060351014137268,
|
| 330 |
+
"step": 140
|
| 331 |
+
},
|
| 332 |
+
{
|
| 333 |
+
"dpo_losses": 0.6667992472648621,
|
| 334 |
+
"epoch": 0.46,
|
| 335 |
+
"grad_norm": 1.3867667176452936,
|
| 336 |
+
"learning_rate": 3.307785000580313e-06,
|
| 337 |
+
"logits/chosen": -2.8069396018981934,
|
| 338 |
+
"logits/rejected": -2.788170099258423,
|
| 339 |
+
"logps/chosen": -256.98443603515625,
|
| 340 |
+
"logps/rejected": -237.07907104492188,
|
| 341 |
+
"loss": 0.6927,
|
| 342 |
+
"positive_losses": 0.10638203471899033,
|
| 343 |
+
"rewards/accuracies": 0.762499988079071,
|
| 344 |
+
"rewards/chosen": 0.1365698128938675,
|
| 345 |
+
"rewards/margins": 0.05615685135126114,
|
| 346 |
+
"rewards/margins_max": 0.1702376753091812,
|
| 347 |
+
"rewards/margins_min": -0.057074159383773804,
|
| 348 |
+
"rewards/margins_std": 0.099973663687706,
|
| 349 |
+
"rewards/rejected": 0.08041295409202576,
|
| 350 |
+
"step": 150
|
| 351 |
+
},
|
| 352 |
+
{
|
| 353 |
+
"dpo_losses": 0.678627610206604,
|
| 354 |
+
"epoch": 0.49,
|
| 355 |
+
"grad_norm": 2.064113448248422,
|
| 356 |
+
"learning_rate": 3.052606201731325e-06,
|
| 357 |
+
"logits/chosen": -2.7706875801086426,
|
| 358 |
+
"logits/rejected": -2.7537920475006104,
|
| 359 |
+
"logps/chosen": -240.9077606201172,
|
| 360 |
+
"logps/rejected": -229.16552734375,
|
| 361 |
+
"loss": 0.6928,
|
| 362 |
+
"positive_losses": 0.03613028675317764,
|
| 363 |
+
"rewards/accuracies": 0.574999988079071,
|
| 364 |
+
"rewards/chosen": 0.14596374332904816,
|
| 365 |
+
"rewards/margins": 0.03164920210838318,
|
| 366 |
+
"rewards/margins_max": 0.13097994029521942,
|
| 367 |
+
"rewards/margins_min": -0.0587020106613636,
|
| 368 |
+
"rewards/margins_std": 0.08477158099412918,
|
| 369 |
+
"rewards/rejected": 0.11431453377008438,
|
| 370 |
+
"step": 160
|
| 371 |
+
},
|
| 372 |
+
{
|
| 373 |
+
"dpo_losses": 0.6667446494102478,
|
| 374 |
+
"epoch": 0.52,
|
| 375 |
+
"grad_norm": 1.6001303619683518,
|
| 376 |
+
"learning_rate": 2.7912083484274266e-06,
|
| 377 |
+
"logits/chosen": -2.7891182899475098,
|
| 378 |
+
"logits/rejected": -2.7103054523468018,
|
| 379 |
+
"logps/chosen": -287.3363342285156,
|
| 380 |
+
"logps/rejected": -262.26116943359375,
|
| 381 |
+
"loss": 0.67,
|
| 382 |
+
"positive_losses": 0.011825943365693092,
|
| 383 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 384 |
+
"rewards/chosen": 0.16382232308387756,
|
| 385 |
+
"rewards/margins": 0.05722727254033089,
|
| 386 |
+
"rewards/margins_max": 0.15611840784549713,
|
| 387 |
+
"rewards/margins_min": -0.05285083130002022,
|
| 388 |
+
"rewards/margins_std": 0.09372057765722275,
|
| 389 |
+
"rewards/rejected": 0.10659502446651459,
|
| 390 |
+
"step": 170
|
| 391 |
+
},
|
| 392 |
+
{
|
| 393 |
+
"dpo_losses": 0.674170196056366,
|
| 394 |
+
"epoch": 0.55,
|
| 395 |
+
"grad_norm": 9.426032570856917,
|
| 396 |
+
"learning_rate": 2.526533223585641e-06,
|
| 397 |
+
"logits/chosen": -2.815913438796997,
|
| 398 |
+
"logits/rejected": -2.7526583671569824,
|
| 399 |
+
"logps/chosen": -241.7661590576172,
|
| 400 |
+
"logps/rejected": -218.5923614501953,
|
| 401 |
+
"loss": 0.6845,
|
| 402 |
+
"positive_losses": 0.0364038459956646,
|
| 403 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 404 |
+
"rewards/chosen": 0.1436757743358612,
|
| 405 |
+
"rewards/margins": 0.040739208459854126,
|
| 406 |
+
"rewards/margins_max": 0.14018966257572174,
|
| 407 |
+
"rewards/margins_min": -0.03748547285795212,
|
| 408 |
+
"rewards/margins_std": 0.07952649146318436,
|
| 409 |
+
"rewards/rejected": 0.10293656587600708,
|
| 410 |
+
"step": 180
|
| 411 |
+
},
|
| 412 |
+
{
|
| 413 |
+
"dpo_losses": 0.6727156639099121,
|
| 414 |
+
"epoch": 0.58,
|
| 415 |
+
"grad_norm": 9.828820911992421,
|
| 416 |
+
"learning_rate": 2.2615594926807554e-06,
|
| 417 |
+
"logits/chosen": -2.754612922668457,
|
| 418 |
+
"logits/rejected": -2.739302158355713,
|
| 419 |
+
"logps/chosen": -284.7493896484375,
|
| 420 |
+
"logps/rejected": -259.38885498046875,
|
| 421 |
+
"loss": 0.7021,
|
| 422 |
+
"positive_losses": 0.08002634346485138,
|
| 423 |
+
"rewards/accuracies": 0.675000011920929,
|
| 424 |
+
"rewards/chosen": 0.1583152711391449,
|
| 425 |
+
"rewards/margins": 0.04616940766572952,
|
| 426 |
+
"rewards/margins_max": 0.19369181990623474,
|
| 427 |
+
"rewards/margins_min": -0.09055180847644806,
|
| 428 |
+
"rewards/margins_std": 0.12469442188739777,
|
| 429 |
+
"rewards/rejected": 0.11214585602283478,
|
| 430 |
+
"step": 190
|
| 431 |
+
},
|
| 432 |
+
{
|
| 433 |
+
"dpo_losses": 0.6642587184906006,
|
| 434 |
+
"epoch": 0.61,
|
| 435 |
+
"grad_norm": 12.372285795702703,
|
| 436 |
+
"learning_rate": 1.9992691817133025e-06,
|
| 437 |
+
"logits/chosen": -2.760648727416992,
|
| 438 |
+
"logits/rejected": -2.7311196327209473,
|
| 439 |
+
"logps/chosen": -267.6399230957031,
|
| 440 |
+
"logps/rejected": -280.528076171875,
|
| 441 |
+
"loss": 0.6919,
|
| 442 |
+
"positive_losses": 0.2932780385017395,
|
| 443 |
+
"rewards/accuracies": 0.6875,
|
| 444 |
+
"rewards/chosen": 0.13960692286491394,
|
| 445 |
+
"rewards/margins": 0.06209836155176163,
|
| 446 |
+
"rewards/margins_max": 0.1877867430448532,
|
| 447 |
+
"rewards/margins_min": -0.06410057097673416,
|
| 448 |
+
"rewards/margins_std": 0.11066991090774536,
|
| 449 |
+
"rewards/rejected": 0.07750856131315231,
|
| 450 |
+
"step": 200
|
| 451 |
+
},
|
| 452 |
+
{
|
| 453 |
+
"epoch": 0.61,
|
| 454 |
+
"eval_dpo_losses": 0.6688500046730042,
|
| 455 |
+
"eval_logits/chosen": -2.785562753677368,
|
| 456 |
+
"eval_logits/rejected": -2.7480554580688477,
|
| 457 |
+
"eval_logps/chosen": -270.2085876464844,
|
| 458 |
+
"eval_logps/rejected": -249.3992919921875,
|
| 459 |
+
"eval_loss": 0.6864475011825562,
|
| 460 |
+
"eval_positive_losses": 0.09515316039323807,
|
| 461 |
+
"eval_rewards/accuracies": 0.6830000281333923,
|
| 462 |
+
"eval_rewards/chosen": 0.1438482403755188,
|
| 463 |
+
"eval_rewards/margins": 0.052051883190870285,
|
| 464 |
+
"eval_rewards/margins_max": 0.21146319806575775,
|
| 465 |
+
"eval_rewards/margins_min": -0.09425032883882523,
|
| 466 |
+
"eval_rewards/margins_std": 0.10122820734977722,
|
| 467 |
+
"eval_rewards/rejected": 0.09179634600877762,
|
| 468 |
+
"eval_runtime": 428.7542,
|
| 469 |
+
"eval_samples_per_second": 4.665,
|
| 470 |
+
"eval_steps_per_second": 0.292,
|
| 471 |
+
"step": 200
|
| 472 |
+
},
|
| 473 |
+
{
|
| 474 |
+
"dpo_losses": 0.6639125943183899,
|
| 475 |
+
"epoch": 0.64,
|
| 476 |
+
"grad_norm": 5.938345116150278,
|
| 477 |
+
"learning_rate": 1.742614117358029e-06,
|
| 478 |
+
"logits/chosen": -2.7786309719085693,
|
| 479 |
+
"logits/rejected": -2.737117290496826,
|
| 480 |
+
"logps/chosen": -291.00030517578125,
|
| 481 |
+
"logps/rejected": -280.262451171875,
|
| 482 |
+
"loss": 0.6787,
|
| 483 |
+
"positive_losses": 0.15116862952709198,
|
| 484 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 485 |
+
"rewards/chosen": 0.14316870272159576,
|
| 486 |
+
"rewards/margins": 0.06285777688026428,
|
| 487 |
+
"rewards/margins_max": 0.1971181035041809,
|
| 488 |
+
"rewards/margins_min": -0.05821852758526802,
|
| 489 |
+
"rewards/margins_std": 0.11372953653335571,
|
| 490 |
+
"rewards/rejected": 0.08031092584133148,
|
| 491 |
+
"step": 210
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"dpo_losses": 0.6715448498725891,
|
| 495 |
+
"epoch": 0.67,
|
| 496 |
+
"grad_norm": 1.9113410130667947,
|
| 497 |
+
"learning_rate": 1.4944827069769125e-06,
|
| 498 |
+
"logits/chosen": -2.8290793895721436,
|
| 499 |
+
"logits/rejected": -2.8052849769592285,
|
| 500 |
+
"logps/chosen": -296.83233642578125,
|
| 501 |
+
"logps/rejected": -254.8493194580078,
|
| 502 |
+
"loss": 0.6738,
|
| 503 |
+
"positive_losses": 0.0,
|
| 504 |
+
"rewards/accuracies": 0.675000011920929,
|
| 505 |
+
"rewards/chosen": 0.16140839457511902,
|
| 506 |
+
"rewards/margins": 0.04584597796201706,
|
| 507 |
+
"rewards/margins_max": 0.14889702200889587,
|
| 508 |
+
"rewards/margins_min": -0.04799387976527214,
|
| 509 |
+
"rewards/margins_std": 0.08770614862442017,
|
| 510 |
+
"rewards/rejected": 0.11556243896484375,
|
| 511 |
+
"step": 220
|
| 512 |
+
},
|
| 513 |
+
{
|
| 514 |
+
"dpo_losses": 0.6808522939682007,
|
| 515 |
+
"epoch": 0.7,
|
| 516 |
+
"grad_norm": 14.916786804283728,
|
| 517 |
+
"learning_rate": 1.257667432355893e-06,
|
| 518 |
+
"logits/chosen": -2.79703950881958,
|
| 519 |
+
"logits/rejected": -2.8196463584899902,
|
| 520 |
+
"logps/chosen": -274.06671142578125,
|
| 521 |
+
"logps/rejected": -251.3921661376953,
|
| 522 |
+
"loss": 0.6919,
|
| 523 |
+
"positive_losses": 0.1536567509174347,
|
| 524 |
+
"rewards/accuracies": 0.550000011920929,
|
| 525 |
+
"rewards/chosen": 0.1457880586385727,
|
| 526 |
+
"rewards/margins": 0.032118141651153564,
|
| 527 |
+
"rewards/margins_max": 0.19181016087532043,
|
| 528 |
+
"rewards/margins_min": -0.09802166372537613,
|
| 529 |
+
"rewards/margins_std": 0.13066089153289795,
|
| 530 |
+
"rewards/rejected": 0.11366990953683853,
|
| 531 |
+
"step": 230
|
| 532 |
+
},
|
| 533 |
+
{
|
| 534 |
+
"dpo_losses": 0.6678847074508667,
|
| 535 |
+
"epoch": 0.73,
|
| 536 |
+
"grad_norm": 2.0373215532852718,
|
| 537 |
+
"learning_rate": 1.0348334229922677e-06,
|
| 538 |
+
"logits/chosen": -2.854497194290161,
|
| 539 |
+
"logits/rejected": -2.8086347579956055,
|
| 540 |
+
"logps/chosen": -276.0729064941406,
|
| 541 |
+
"logps/rejected": -265.6085510253906,
|
| 542 |
+
"loss": 0.689,
|
| 543 |
+
"positive_losses": 0.08907928317785263,
|
| 544 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 545 |
+
"rewards/chosen": 0.14944827556610107,
|
| 546 |
+
"rewards/margins": 0.05373426154255867,
|
| 547 |
+
"rewards/margins_max": 0.16546599566936493,
|
| 548 |
+
"rewards/margins_min": -0.03850485011935234,
|
| 549 |
+
"rewards/margins_std": 0.093926340341568,
|
| 550 |
+
"rewards/rejected": 0.0957140102982521,
|
| 551 |
+
"step": 240
|
| 552 |
+
},
|
| 553 |
+
{
|
| 554 |
+
"dpo_losses": 0.6652002334594727,
|
| 555 |
+
"epoch": 0.76,
|
| 556 |
+
"grad_norm": 2.0611778414132793,
|
| 557 |
+
"learning_rate": 8.284884626103165e-07,
|
| 558 |
+
"logits/chosen": -2.7938308715820312,
|
| 559 |
+
"logits/rejected": -2.762302875518799,
|
| 560 |
+
"logps/chosen": -285.65093994140625,
|
| 561 |
+
"logps/rejected": -295.174560546875,
|
| 562 |
+
"loss": 0.7577,
|
| 563 |
+
"positive_losses": 0.03922748565673828,
|
| 564 |
+
"rewards/accuracies": 0.637499988079071,
|
| 565 |
+
"rewards/chosen": 0.15438929200172424,
|
| 566 |
+
"rewards/margins": 0.06061772257089615,
|
| 567 |
+
"rewards/margins_max": 0.1899864226579666,
|
| 568 |
+
"rewards/margins_min": -0.060655929148197174,
|
| 569 |
+
"rewards/margins_std": 0.11546076834201813,
|
| 570 |
+
"rewards/rejected": 0.09377159178256989,
|
| 571 |
+
"step": 250
|
| 572 |
+
},
|
| 573 |
+
{
|
| 574 |
+
"dpo_losses": 0.6641849279403687,
|
| 575 |
+
"epoch": 0.79,
|
| 576 |
+
"grad_norm": 5.534968963232783,
|
| 577 |
+
"learning_rate": 6.409547664531734e-07,
|
| 578 |
+
"logits/chosen": -2.817701816558838,
|
| 579 |
+
"logits/rejected": -2.7872586250305176,
|
| 580 |
+
"logps/chosen": -316.77471923828125,
|
| 581 |
+
"logps/rejected": -301.3828125,
|
| 582 |
+
"loss": 0.6791,
|
| 583 |
+
"positive_losses": 0.059366799890995026,
|
| 584 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 585 |
+
"rewards/chosen": 0.17207232117652893,
|
| 586 |
+
"rewards/margins": 0.060843367129564285,
|
| 587 |
+
"rewards/margins_max": 0.1557174026966095,
|
| 588 |
+
"rewards/margins_min": -0.024733999744057655,
|
| 589 |
+
"rewards/margins_std": 0.0783897191286087,
|
| 590 |
+
"rewards/rejected": 0.11122896522283554,
|
| 591 |
+
"step": 260
|
| 592 |
+
},
|
| 593 |
+
{
|
| 594 |
+
"dpo_losses": 0.6656020879745483,
|
| 595 |
+
"epoch": 0.82,
|
| 596 |
+
"grad_norm": 2.074871528049425,
|
| 597 |
+
"learning_rate": 4.743428469705336e-07,
|
| 598 |
+
"logits/chosen": -2.7688679695129395,
|
| 599 |
+
"logits/rejected": -2.767603874206543,
|
| 600 |
+
"logps/chosen": -289.3965759277344,
|
| 601 |
+
"logps/rejected": -299.40924072265625,
|
| 602 |
+
"loss": 0.6782,
|
| 603 |
+
"positive_losses": 0.13356761634349823,
|
| 604 |
+
"rewards/accuracies": 0.637499988079071,
|
| 605 |
+
"rewards/chosen": 0.14411255717277527,
|
| 606 |
+
"rewards/margins": 0.058389417827129364,
|
| 607 |
+
"rewards/margins_max": 0.1643931269645691,
|
| 608 |
+
"rewards/margins_min": -0.037889931350946426,
|
| 609 |
+
"rewards/margins_std": 0.09290634840726852,
|
| 610 |
+
"rewards/rejected": 0.0857231393456459,
|
| 611 |
+
"step": 270
|
| 612 |
+
},
|
| 613 |
+
{
|
| 614 |
+
"dpo_losses": 0.6564372181892395,
|
| 615 |
+
"epoch": 0.85,
|
| 616 |
+
"grad_norm": 2.142298246340979,
|
| 617 |
+
"learning_rate": 3.3052776201888266e-07,
|
| 618 |
+
"logits/chosen": -2.818432569503784,
|
| 619 |
+
"logits/rejected": -2.8038887977600098,
|
| 620 |
+
"logps/chosen": -308.4718933105469,
|
| 621 |
+
"logps/rejected": -260.4856262207031,
|
| 622 |
+
"loss": 0.6756,
|
| 623 |
+
"positive_losses": 0.07436485588550568,
|
| 624 |
+
"rewards/accuracies": 0.762499988079071,
|
| 625 |
+
"rewards/chosen": 0.17093774676322937,
|
| 626 |
+
"rewards/margins": 0.07805446535348892,
|
| 627 |
+
"rewards/margins_max": 0.18430791795253754,
|
| 628 |
+
"rewards/margins_min": -0.037783779203891754,
|
| 629 |
+
"rewards/margins_std": 0.09794415533542633,
|
| 630 |
+
"rewards/rejected": 0.09288326650857925,
|
| 631 |
+
"step": 280
|
| 632 |
+
},
|
| 633 |
+
{
|
| 634 |
+
"dpo_losses": 0.668868899345398,
|
| 635 |
+
"epoch": 0.88,
|
| 636 |
+
"grad_norm": 8.396243354371068,
|
| 637 |
+
"learning_rate": 2.111280128780638e-07,
|
| 638 |
+
"logits/chosen": -2.756545305252075,
|
| 639 |
+
"logits/rejected": -2.7231669425964355,
|
| 640 |
+
"logps/chosen": -259.9716491699219,
|
| 641 |
+
"logps/rejected": -236.48770141601562,
|
| 642 |
+
"loss": 0.684,
|
| 643 |
+
"positive_losses": 0.07783536612987518,
|
| 644 |
+
"rewards/accuracies": 0.675000011920929,
|
| 645 |
+
"rewards/chosen": 0.1421826183795929,
|
| 646 |
+
"rewards/margins": 0.051858507096767426,
|
| 647 |
+
"rewards/margins_max": 0.16274937987327576,
|
| 648 |
+
"rewards/margins_min": -0.04277097061276436,
|
| 649 |
+
"rewards/margins_std": 0.09085315465927124,
|
| 650 |
+
"rewards/rejected": 0.09032410383224487,
|
| 651 |
+
"step": 290
|
| 652 |
+
},
|
| 653 |
+
{
|
| 654 |
+
"dpo_losses": 0.6606593132019043,
|
| 655 |
+
"epoch": 0.91,
|
| 656 |
+
"grad_norm": 7.1470269484352835,
|
| 657 |
+
"learning_rate": 1.1748732956682023e-07,
|
| 658 |
+
"logits/chosen": -2.8557376861572266,
|
| 659 |
+
"logits/rejected": -2.791503667831421,
|
| 660 |
+
"logps/chosen": -309.07489013671875,
|
| 661 |
+
"logps/rejected": -277.8647155761719,
|
| 662 |
+
"loss": 0.6774,
|
| 663 |
+
"positive_losses": 0.09160251915454865,
|
| 664 |
+
"rewards/accuracies": 0.737500011920929,
|
| 665 |
+
"rewards/chosen": 0.14638249576091766,
|
| 666 |
+
"rewards/margins": 0.06913793087005615,
|
| 667 |
+
"rewards/margins_max": 0.1924934983253479,
|
| 668 |
+
"rewards/margins_min": -0.04603328928351402,
|
| 669 |
+
"rewards/margins_std": 0.10587652772665024,
|
| 670 |
+
"rewards/rejected": 0.07724456489086151,
|
| 671 |
+
"step": 300
|
| 672 |
+
},
|
| 673 |
+
{
|
| 674 |
+
"epoch": 0.91,
|
| 675 |
+
"eval_dpo_losses": 0.6679198145866394,
|
| 676 |
+
"eval_logits/chosen": -2.7836546897888184,
|
| 677 |
+
"eval_logits/rejected": -2.746371030807495,
|
| 678 |
+
"eval_logps/chosen": -269.767822265625,
|
| 679 |
+
"eval_logps/rejected": -249.15957641601562,
|
| 680 |
+
"eval_loss": 0.687869131565094,
|
| 681 |
+
"eval_positive_losses": 0.09410960972309113,
|
| 682 |
+
"eval_rewards/accuracies": 0.6899999976158142,
|
| 683 |
+
"eval_rewards/chosen": 0.1482561081647873,
|
| 684 |
+
"eval_rewards/margins": 0.05406271293759346,
|
| 685 |
+
"eval_rewards/margins_max": 0.2146386057138443,
|
| 686 |
+
"eval_rewards/margins_min": -0.0945044532418251,
|
| 687 |
+
"eval_rewards/margins_std": 0.10281499475240707,
|
| 688 |
+
"eval_rewards/rejected": 0.09419341385364532,
|
| 689 |
+
"eval_runtime": 428.9213,
|
| 690 |
+
"eval_samples_per_second": 4.663,
|
| 691 |
+
"eval_steps_per_second": 0.291,
|
| 692 |
+
"step": 300
|
| 693 |
+
},
|
| 694 |
+
{
|
| 695 |
+
"dpo_losses": 0.6655529141426086,
|
| 696 |
+
"epoch": 0.94,
|
| 697 |
+
"grad_norm": 1.6083474443571015,
|
| 698 |
+
"learning_rate": 5.065954844616722e-08,
|
| 699 |
+
"logits/chosen": -2.7985780239105225,
|
| 700 |
+
"logits/rejected": -2.756269931793213,
|
| 701 |
+
"logps/chosen": -261.9189758300781,
|
| 702 |
+
"logps/rejected": -271.715576171875,
|
| 703 |
+
"loss": 0.6852,
|
| 704 |
+
"positive_losses": 0.16939087212085724,
|
| 705 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 706 |
+
"rewards/chosen": 0.1520276516675949,
|
| 707 |
+
"rewards/margins": 0.05855011194944382,
|
| 708 |
+
"rewards/margins_max": 0.17010439932346344,
|
| 709 |
+
"rewards/margins_min": -0.046878743916749954,
|
| 710 |
+
"rewards/margins_std": 0.09575958549976349,
|
| 711 |
+
"rewards/rejected": 0.0934775322675705,
|
| 712 |
+
"step": 310
|
| 713 |
+
},
|
| 714 |
+
{
|
| 715 |
+
"dpo_losses": 0.6655773520469666,
|
| 716 |
+
"epoch": 0.97,
|
| 717 |
+
"grad_norm": 1.8914240259960515,
|
| 718 |
+
"learning_rate": 1.1396752298723501e-08,
|
| 719 |
+
"logits/chosen": -2.8393120765686035,
|
| 720 |
+
"logits/rejected": -2.7888612747192383,
|
| 721 |
+
"logps/chosen": -235.32876586914062,
|
| 722 |
+
"logps/rejected": -249.7720184326172,
|
| 723 |
+
"loss": 0.666,
|
| 724 |
+
"positive_losses": 0.027046585455536842,
|
| 725 |
+
"rewards/accuracies": 0.762499988079071,
|
| 726 |
+
"rewards/chosen": 0.1361643224954605,
|
| 727 |
+
"rewards/margins": 0.05848071724176407,
|
| 728 |
+
"rewards/margins_max": 0.16818095743656158,
|
| 729 |
+
"rewards/margins_min": -0.04760957881808281,
|
| 730 |
+
"rewards/margins_std": 0.09690945595502853,
|
| 731 |
+
"rewards/rejected": 0.07768361270427704,
|
| 732 |
+
"step": 320
|
| 733 |
+
},
|
| 734 |
+
{
|
| 735 |
+
"epoch": 1.0,
|
| 736 |
+
"step": 329,
|
| 737 |
+
"total_flos": 0.0,
|
| 738 |
+
"train_loss": 0.6874516553791823,
|
| 739 |
+
"train_runtime": 3895.4317,
|
| 740 |
+
"train_samples_per_second": 1.351,
|
| 741 |
+
"train_steps_per_second": 0.084
|
| 742 |
+
}
|
| 743 |
+
],
|
| 744 |
+
"logging_steps": 10,
|
| 745 |
+
"max_steps": 329,
|
| 746 |
+
"num_input_tokens_seen": 0,
|
| 747 |
+
"num_train_epochs": 1,
|
| 748 |
+
"save_steps": 100,
|
| 749 |
+
"total_flos": 0.0,
|
| 750 |
+
"train_batch_size": 4,
|
| 751 |
+
"trial_name": null,
|
| 752 |
+
"trial_params": null
|
| 753 |
+
}
|