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-uf-ours-5e-7-epoch1
|
| 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-uf-ours-5e-7-epoch1
|
| 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: 0.7024
|
| 22 |
+
- Positive Losses: 0.1592
|
| 23 |
+
- Dpo Losses: 0.6845
|
| 24 |
+
- Rewards/chosen: 0.0528
|
| 25 |
+
- Rewards/rejected: 0.0346
|
| 26 |
+
- Rewards/accuracies: 0.5940
|
| 27 |
+
- Rewards/margins: 0.0182
|
| 28 |
+
- Rewards/margins Max: 0.1072
|
| 29 |
+
- Rewards/margins Min: -0.0591
|
| 30 |
+
- Rewards/margins Std: 0.0555
|
| 31 |
+
- Logps/rejected: -255.1217
|
| 32 |
+
- Logps/chosen: -279.3168
|
| 33 |
+
- Logits/rejected: -2.7478
|
| 34 |
+
- Logits/chosen: -2.7859
|
| 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-07
|
| 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 |
+
| 0.6834 | 0.28 | 100 | 0.6932 | 0.0200 | 0.6905 | 0.0248 | 0.0194 | 0.5930 | 0.0054 | 0.0341 | -0.0197 | 0.0178 | -256.6423 | -282.1181 | -2.7657 | -2.8044 |
|
| 72 |
+
| 0.6629 | 0.56 | 200 | 0.6977 | 0.1042 | 0.6860 | 0.0485 | 0.0335 | 0.5980 | 0.0149 | 0.0881 | -0.0489 | 0.0456 | -255.2263 | -279.7464 | -2.7492 | -2.7879 |
|
| 73 |
+
| 0.6479 | 0.85 | 300 | 0.7024 | 0.1592 | 0.6845 | 0.0528 | 0.0346 | 0.5940 | 0.0182 | 0.1072 | -0.0591 | 0.0555 | -255.1217 | -279.3168 | -2.7478 | -2.7859 |
|
| 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 sha256:34e4ce3dec547099b6f9e5a3335ef132d4adf3cd5ebf4394c1549ebfa9198205
|
| 3 |
size 671150064
|
all_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.6688590748209349,
|
| 4 |
+
"train_runtime": 4310.2257,
|
| 5 |
+
"train_samples": 5678,
|
| 6 |
+
"train_samples_per_second": 1.317,
|
| 7 |
+
"train_steps_per_second": 0.082
|
| 8 |
+
}
|
runs/Jul29_11-02-19_notebook-deployment-48-7d9b6c99-khd85/events.out.tfevents.1722251039.notebook-deployment-48-7d9b6c99-khd85.3446418.0
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 sha256:07c6ac1ab8e249cd2d511cfabed056ef1e54156e2a8475a8221445210ee799d9
|
| 3 |
+
size 44442
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.6688590748209349,
|
| 4 |
+
"train_runtime": 4310.2257,
|
| 5 |
+
"train_samples": 5678,
|
| 6 |
+
"train_samples_per_second": 1.317,
|
| 7 |
+
"train_steps_per_second": 0.082
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,813 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 1.0,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 355,
|
| 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": 1.6020024881522184,
|
| 15 |
+
"learning_rate": 1.3888888888888887e-08,
|
| 16 |
+
"logits/chosen": -2.861618995666504,
|
| 17 |
+
"logits/rejected": -2.8205904960632324,
|
| 18 |
+
"logps/chosen": -271.06011962890625,
|
| 19 |
+
"logps/rejected": -211.1704559326172,
|
| 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": 0.6933339834213257,
|
| 33 |
+
"epoch": 0.03,
|
| 34 |
+
"grad_norm": 17.749396515379228,
|
| 35 |
+
"learning_rate": 1.3888888888888888e-07,
|
| 36 |
+
"logits/chosen": -2.8337996006011963,
|
| 37 |
+
"logits/rejected": -2.7912583351135254,
|
| 38 |
+
"logps/chosen": -325.06781005859375,
|
| 39 |
+
"logps/rejected": -274.9460754394531,
|
| 40 |
+
"loss": 0.6995,
|
| 41 |
+
"positive_losses": 0.07389768213033676,
|
| 42 |
+
"rewards/accuracies": 0.3888888955116272,
|
| 43 |
+
"rewards/chosen": -0.00017488877347204834,
|
| 44 |
+
"rewards/margins": -0.0003703224065247923,
|
| 45 |
+
"rewards/margins_max": 0.002196074463427067,
|
| 46 |
+
"rewards/margins_min": -0.003493131836876273,
|
| 47 |
+
"rewards/margins_std": 0.0025361739099025726,
|
| 48 |
+
"rewards/rejected": 0.00019543366215657443,
|
| 49 |
+
"step": 10
|
| 50 |
+
},
|
| 51 |
+
{
|
| 52 |
+
"dpo_losses": 0.6930112838745117,
|
| 53 |
+
"epoch": 0.06,
|
| 54 |
+
"grad_norm": 16.42860549070251,
|
| 55 |
+
"learning_rate": 2.7777777777777776e-07,
|
| 56 |
+
"logits/chosen": -2.7250771522521973,
|
| 57 |
+
"logits/rejected": -2.7066714763641357,
|
| 58 |
+
"logps/chosen": -293.7167663574219,
|
| 59 |
+
"logps/rejected": -215.6971893310547,
|
| 60 |
+
"loss": 0.6969,
|
| 61 |
+
"positive_losses": 0.0511661060154438,
|
| 62 |
+
"rewards/accuracies": 0.512499988079071,
|
| 63 |
+
"rewards/chosen": 0.0010916258906945586,
|
| 64 |
+
"rewards/margins": 0.00027377792866900563,
|
| 65 |
+
"rewards/margins_max": 0.0035606161691248417,
|
| 66 |
+
"rewards/margins_min": -0.0026067704893648624,
|
| 67 |
+
"rewards/margins_std": 0.002710341941565275,
|
| 68 |
+
"rewards/rejected": 0.0008178479038178921,
|
| 69 |
+
"step": 20
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"dpo_losses": 0.692296028137207,
|
| 73 |
+
"epoch": 0.08,
|
| 74 |
+
"grad_norm": 8.629940663439823,
|
| 75 |
+
"learning_rate": 4.1666666666666667e-07,
|
| 76 |
+
"logits/chosen": -2.8196403980255127,
|
| 77 |
+
"logits/rejected": -2.750633716583252,
|
| 78 |
+
"logps/chosen": -303.18206787109375,
|
| 79 |
+
"logps/rejected": -232.16091918945312,
|
| 80 |
+
"loss": 0.6932,
|
| 81 |
+
"positive_losses": 0.0034084320068359375,
|
| 82 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 83 |
+
"rewards/chosen": 0.005238121375441551,
|
| 84 |
+
"rewards/margins": 0.0017049547750502825,
|
| 85 |
+
"rewards/margins_max": 0.0041273366659879684,
|
| 86 |
+
"rewards/margins_min": -0.0010792845860123634,
|
| 87 |
+
"rewards/margins_std": 0.0023812309373170137,
|
| 88 |
+
"rewards/rejected": 0.003533167066052556,
|
| 89 |
+
"step": 30
|
| 90 |
+
},
|
| 91 |
+
{
|
| 92 |
+
"dpo_losses": 0.6920384764671326,
|
| 93 |
+
"epoch": 0.11,
|
| 94 |
+
"grad_norm": 1.7378063453273105,
|
| 95 |
+
"learning_rate": 4.998060489154965e-07,
|
| 96 |
+
"logits/chosen": -2.8426527976989746,
|
| 97 |
+
"logits/rejected": -2.7624073028564453,
|
| 98 |
+
"logps/chosen": -276.6007385253906,
|
| 99 |
+
"logps/rejected": -225.0730438232422,
|
| 100 |
+
"loss": 0.6922,
|
| 101 |
+
"positive_losses": 0.0007575989002361894,
|
| 102 |
+
"rewards/accuracies": 0.637499988079071,
|
| 103 |
+
"rewards/chosen": 0.008838965557515621,
|
| 104 |
+
"rewards/margins": 0.0022226297296583652,
|
| 105 |
+
"rewards/margins_max": 0.006442135665565729,
|
| 106 |
+
"rewards/margins_min": -0.0014907930744811893,
|
| 107 |
+
"rewards/margins_std": 0.0035486381966620684,
|
| 108 |
+
"rewards/rejected": 0.006616336293518543,
|
| 109 |
+
"step": 40
|
| 110 |
+
},
|
| 111 |
+
{
|
| 112 |
+
"dpo_losses": 0.6905555725097656,
|
| 113 |
+
"epoch": 0.14,
|
| 114 |
+
"grad_norm": 9.354247049667382,
|
| 115 |
+
"learning_rate": 4.976275538042932e-07,
|
| 116 |
+
"logits/chosen": -2.810286283493042,
|
| 117 |
+
"logits/rejected": -2.7382943630218506,
|
| 118 |
+
"logps/chosen": -274.2152404785156,
|
| 119 |
+
"logps/rejected": -233.04257202148438,
|
| 120 |
+
"loss": 0.6905,
|
| 121 |
+
"positive_losses": 0.0,
|
| 122 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 123 |
+
"rewards/chosen": 0.013549095019698143,
|
| 124 |
+
"rewards/margins": 0.005201784893870354,
|
| 125 |
+
"rewards/margins_max": 0.0127005185931921,
|
| 126 |
+
"rewards/margins_min": -0.0008777756011113524,
|
| 127 |
+
"rewards/margins_std": 0.006243027281016111,
|
| 128 |
+
"rewards/rejected": 0.008347309194505215,
|
| 129 |
+
"step": 50
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"dpo_losses": 0.6887227892875671,
|
| 133 |
+
"epoch": 0.17,
|
| 134 |
+
"grad_norm": 2.4332730738698984,
|
| 135 |
+
"learning_rate": 4.930493069997119e-07,
|
| 136 |
+
"logits/chosen": -2.749894142150879,
|
| 137 |
+
"logits/rejected": -2.7108871936798096,
|
| 138 |
+
"logps/chosen": -311.87884521484375,
|
| 139 |
+
"logps/rejected": -261.8995361328125,
|
| 140 |
+
"loss": 0.6896,
|
| 141 |
+
"positive_losses": 0.0,
|
| 142 |
+
"rewards/accuracies": 0.875,
|
| 143 |
+
"rewards/chosen": 0.017607735469937325,
|
| 144 |
+
"rewards/margins": 0.008888162672519684,
|
| 145 |
+
"rewards/margins_max": 0.017459740862250328,
|
| 146 |
+
"rewards/margins_min": 0.0007526646950282156,
|
| 147 |
+
"rewards/margins_std": 0.0074041737243533134,
|
| 148 |
+
"rewards/rejected": 0.008719570934772491,
|
| 149 |
+
"step": 60
|
| 150 |
+
},
|
| 151 |
+
{
|
| 152 |
+
"dpo_losses": 0.6875914931297302,
|
| 153 |
+
"epoch": 0.2,
|
| 154 |
+
"grad_norm": 1.7936785587345936,
|
| 155 |
+
"learning_rate": 4.861156761634013e-07,
|
| 156 |
+
"logits/chosen": -2.7838871479034424,
|
| 157 |
+
"logits/rejected": -2.7218618392944336,
|
| 158 |
+
"logps/chosen": -320.61077880859375,
|
| 159 |
+
"logps/rejected": -234.509033203125,
|
| 160 |
+
"loss": 0.6877,
|
| 161 |
+
"positive_losses": 0.0053993226028978825,
|
| 162 |
+
"rewards/accuracies": 0.862500011920929,
|
| 163 |
+
"rewards/chosen": 0.022504538297653198,
|
| 164 |
+
"rewards/margins": 0.01117872167378664,
|
| 165 |
+
"rewards/margins_max": 0.024569755420088768,
|
| 166 |
+
"rewards/margins_min": 0.00035735839628614485,
|
| 167 |
+
"rewards/margins_std": 0.0108140017837286,
|
| 168 |
+
"rewards/rejected": 0.011325814761221409,
|
| 169 |
+
"step": 70
|
| 170 |
+
},
|
| 171 |
+
{
|
| 172 |
+
"dpo_losses": 0.6856142282485962,
|
| 173 |
+
"epoch": 0.23,
|
| 174 |
+
"grad_norm": 1.8616136810198376,
|
| 175 |
+
"learning_rate": 4.768938549177392e-07,
|
| 176 |
+
"logits/chosen": -2.8332719802856445,
|
| 177 |
+
"logits/rejected": -2.7764153480529785,
|
| 178 |
+
"logps/chosen": -318.5867614746094,
|
| 179 |
+
"logps/rejected": -285.2149963378906,
|
| 180 |
+
"loss": 0.6869,
|
| 181 |
+
"positive_losses": 0.006173324771225452,
|
| 182 |
+
"rewards/accuracies": 0.800000011920929,
|
| 183 |
+
"rewards/chosen": 0.026837289333343506,
|
| 184 |
+
"rewards/margins": 0.015200009569525719,
|
| 185 |
+
"rewards/margins_max": 0.03601212427020073,
|
| 186 |
+
"rewards/margins_min": -0.00037241075187921524,
|
| 187 |
+
"rewards/margins_std": 0.01664547622203827,
|
| 188 |
+
"rewards/rejected": 0.011637277901172638,
|
| 189 |
+
"step": 80
|
| 190 |
+
},
|
| 191 |
+
{
|
| 192 |
+
"dpo_losses": 0.6834368705749512,
|
| 193 |
+
"epoch": 0.25,
|
| 194 |
+
"grad_norm": 2.576979966636063,
|
| 195 |
+
"learning_rate": 4.654732116743193e-07,
|
| 196 |
+
"logits/chosen": -2.7666513919830322,
|
| 197 |
+
"logits/rejected": -2.724118709564209,
|
| 198 |
+
"logps/chosen": -272.710693359375,
|
| 199 |
+
"logps/rejected": -197.66629028320312,
|
| 200 |
+
"loss": 0.6833,
|
| 201 |
+
"positive_losses": 0.0,
|
| 202 |
+
"rewards/accuracies": 0.862500011920929,
|
| 203 |
+
"rewards/chosen": 0.03284075856208801,
|
| 204 |
+
"rewards/margins": 0.01963154971599579,
|
| 205 |
+
"rewards/margins_max": 0.04466867819428444,
|
| 206 |
+
"rewards/margins_min": 0.0018520345911383629,
|
| 207 |
+
"rewards/margins_std": 0.020031433552503586,
|
| 208 |
+
"rewards/rejected": 0.013209208846092224,
|
| 209 |
+
"step": 90
|
| 210 |
+
},
|
| 211 |
+
{
|
| 212 |
+
"dpo_losses": 0.6846113801002502,
|
| 213 |
+
"epoch": 0.28,
|
| 214 |
+
"grad_norm": 1.9012962136053702,
|
| 215 |
+
"learning_rate": 4.519644235671752e-07,
|
| 216 |
+
"logits/chosen": -2.8058769702911377,
|
| 217 |
+
"logits/rejected": -2.7677180767059326,
|
| 218 |
+
"logps/chosen": -278.3069763183594,
|
| 219 |
+
"logps/rejected": -260.80572509765625,
|
| 220 |
+
"loss": 0.6834,
|
| 221 |
+
"positive_losses": 0.00875930767506361,
|
| 222 |
+
"rewards/accuracies": 0.8125,
|
| 223 |
+
"rewards/chosen": 0.031047578901052475,
|
| 224 |
+
"rewards/margins": 0.01723078265786171,
|
| 225 |
+
"rewards/margins_max": 0.03883099928498268,
|
| 226 |
+
"rewards/margins_min": 0.00033125150366686285,
|
| 227 |
+
"rewards/margins_std": 0.017453256994485855,
|
| 228 |
+
"rewards/rejected": 0.013816798105835915,
|
| 229 |
+
"step": 100
|
| 230 |
+
},
|
| 231 |
+
{
|
| 232 |
+
"epoch": 0.28,
|
| 233 |
+
"eval_dpo_losses": 0.6905006766319275,
|
| 234 |
+
"eval_logits/chosen": -2.804429292678833,
|
| 235 |
+
"eval_logits/rejected": -2.7657294273376465,
|
| 236 |
+
"eval_logps/chosen": -282.1181335449219,
|
| 237 |
+
"eval_logps/rejected": -256.64227294921875,
|
| 238 |
+
"eval_loss": 0.6931570768356323,
|
| 239 |
+
"eval_positive_losses": 0.02003120444715023,
|
| 240 |
+
"eval_rewards/accuracies": 0.5929999947547913,
|
| 241 |
+
"eval_rewards/chosen": 0.024752924218773842,
|
| 242 |
+
"eval_rewards/margins": 0.005386360455304384,
|
| 243 |
+
"eval_rewards/margins_max": 0.03411807864904404,
|
| 244 |
+
"eval_rewards/margins_min": -0.019733905792236328,
|
| 245 |
+
"eval_rewards/margins_std": 0.017771249637007713,
|
| 246 |
+
"eval_rewards/rejected": 0.019366566091775894,
|
| 247 |
+
"eval_runtime": 428.432,
|
| 248 |
+
"eval_samples_per_second": 4.668,
|
| 249 |
+
"eval_steps_per_second": 0.292,
|
| 250 |
+
"step": 100
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"dpo_losses": 0.676801860332489,
|
| 254 |
+
"epoch": 0.31,
|
| 255 |
+
"grad_norm": 2.3464200895802714,
|
| 256 |
+
"learning_rate": 4.364984038837727e-07,
|
| 257 |
+
"logits/chosen": -2.87202787399292,
|
| 258 |
+
"logits/rejected": -2.7757174968719482,
|
| 259 |
+
"logps/chosen": -369.87457275390625,
|
| 260 |
+
"logps/rejected": -285.02069091796875,
|
| 261 |
+
"loss": 0.6771,
|
| 262 |
+
"positive_losses": 0.0,
|
| 263 |
+
"rewards/accuracies": 0.9375,
|
| 264 |
+
"rewards/chosen": 0.0473114475607872,
|
| 265 |
+
"rewards/margins": 0.03318057209253311,
|
| 266 |
+
"rewards/margins_max": 0.06624362617731094,
|
| 267 |
+
"rewards/margins_min": 0.005891180131584406,
|
| 268 |
+
"rewards/margins_std": 0.026871800422668457,
|
| 269 |
+
"rewards/rejected": 0.014130881056189537,
|
| 270 |
+
"step": 110
|
| 271 |
+
},
|
| 272 |
+
{
|
| 273 |
+
"dpo_losses": 0.6757510900497437,
|
| 274 |
+
"epoch": 0.34,
|
| 275 |
+
"grad_norm": 2.0654592381160293,
|
| 276 |
+
"learning_rate": 4.1922503338800447e-07,
|
| 277 |
+
"logits/chosen": -2.8409907817840576,
|
| 278 |
+
"logits/rejected": -2.7864270210266113,
|
| 279 |
+
"logps/chosen": -343.3132629394531,
|
| 280 |
+
"logps/rejected": -264.656982421875,
|
| 281 |
+
"loss": 0.6761,
|
| 282 |
+
"positive_losses": 0.00038242340087890625,
|
| 283 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 284 |
+
"rewards/chosen": 0.05072961002588272,
|
| 285 |
+
"rewards/margins": 0.035357214510440826,
|
| 286 |
+
"rewards/margins_max": 0.0718097984790802,
|
| 287 |
+
"rewards/margins_min": 0.007976613938808441,
|
| 288 |
+
"rewards/margins_std": 0.029289904981851578,
|
| 289 |
+
"rewards/rejected": 0.015372401103377342,
|
| 290 |
+
"step": 120
|
| 291 |
+
},
|
| 292 |
+
{
|
| 293 |
+
"dpo_losses": 0.672697901725769,
|
| 294 |
+
"epoch": 0.37,
|
| 295 |
+
"grad_norm": 1.805644520581111,
|
| 296 |
+
"learning_rate": 4.003117078299021e-07,
|
| 297 |
+
"logits/chosen": -2.833576202392578,
|
| 298 |
+
"logits/rejected": -2.7511653900146484,
|
| 299 |
+
"logps/chosen": -376.5670471191406,
|
| 300 |
+
"logps/rejected": -298.7974548339844,
|
| 301 |
+
"loss": 0.6728,
|
| 302 |
+
"positive_losses": 0.007812881842255592,
|
| 303 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 304 |
+
"rewards/chosen": 0.05797192454338074,
|
| 305 |
+
"rewards/margins": 0.041637517511844635,
|
| 306 |
+
"rewards/margins_max": 0.07995648682117462,
|
| 307 |
+
"rewards/margins_min": 0.005986917298287153,
|
| 308 |
+
"rewards/margins_std": 0.03285466879606247,
|
| 309 |
+
"rewards/rejected": 0.016334403306245804,
|
| 310 |
+
"step": 130
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"dpo_losses": 0.6723438501358032,
|
| 314 |
+
"epoch": 0.39,
|
| 315 |
+
"grad_norm": 1.4837668528318801,
|
| 316 |
+
"learning_rate": 3.799417157181075e-07,
|
| 317 |
+
"logits/chosen": -2.7876131534576416,
|
| 318 |
+
"logits/rejected": -2.748243808746338,
|
| 319 |
+
"logps/chosen": -312.4587707519531,
|
| 320 |
+
"logps/rejected": -268.16253662109375,
|
| 321 |
+
"loss": 0.6749,
|
| 322 |
+
"positive_losses": 0.0,
|
| 323 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 324 |
+
"rewards/chosen": 0.059552956372499466,
|
| 325 |
+
"rewards/margins": 0.04268346354365349,
|
| 326 |
+
"rewards/margins_max": 0.09346815198659897,
|
| 327 |
+
"rewards/margins_min": 0.005906702019274235,
|
| 328 |
+
"rewards/margins_std": 0.04082999378442764,
|
| 329 |
+
"rewards/rejected": 0.01686949096620083,
|
| 330 |
+
"step": 140
|
| 331 |
+
},
|
| 332 |
+
{
|
| 333 |
+
"dpo_losses": 0.6723035573959351,
|
| 334 |
+
"epoch": 0.42,
|
| 335 |
+
"grad_norm": 1.9734696023947083,
|
| 336 |
+
"learning_rate": 3.583124620760659e-07,
|
| 337 |
+
"logits/chosen": -2.7900147438049316,
|
| 338 |
+
"logits/rejected": -2.7500596046447754,
|
| 339 |
+
"logps/chosen": -282.7308654785156,
|
| 340 |
+
"logps/rejected": -211.3845672607422,
|
| 341 |
+
"loss": 0.6711,
|
| 342 |
+
"positive_losses": 0.0,
|
| 343 |
+
"rewards/accuracies": 0.925000011920929,
|
| 344 |
+
"rewards/chosen": 0.05954126641154289,
|
| 345 |
+
"rewards/margins": 0.04255921393632889,
|
| 346 |
+
"rewards/margins_max": 0.08835546672344208,
|
| 347 |
+
"rewards/margins_min": 0.005340501666069031,
|
| 348 |
+
"rewards/margins_std": 0.037935130298137665,
|
| 349 |
+
"rewards/rejected": 0.016982052475214005,
|
| 350 |
+
"step": 150
|
| 351 |
+
},
|
| 352 |
+
{
|
| 353 |
+
"dpo_losses": 0.669468343257904,
|
| 354 |
+
"epoch": 0.45,
|
| 355 |
+
"grad_norm": 1.6641482478870189,
|
| 356 |
+
"learning_rate": 3.356335553954679e-07,
|
| 357 |
+
"logits/chosen": -2.768503189086914,
|
| 358 |
+
"logits/rejected": -2.7139506340026855,
|
| 359 |
+
"logps/chosen": -293.73223876953125,
|
| 360 |
+
"logps/rejected": -232.4127655029297,
|
| 361 |
+
"loss": 0.6692,
|
| 362 |
+
"positive_losses": 0.010938262566924095,
|
| 363 |
+
"rewards/accuracies": 0.925000011920929,
|
| 364 |
+
"rewards/chosen": 0.06814642250537872,
|
| 365 |
+
"rewards/margins": 0.04834110662341118,
|
| 366 |
+
"rewards/margins_max": 0.09320978820323944,
|
| 367 |
+
"rewards/margins_min": 0.00712673831731081,
|
| 368 |
+
"rewards/margins_std": 0.038901425898075104,
|
| 369 |
+
"rewards/rejected": 0.019805317744612694,
|
| 370 |
+
"step": 160
|
| 371 |
+
},
|
| 372 |
+
{
|
| 373 |
+
"dpo_losses": 0.6628905534744263,
|
| 374 |
+
"epoch": 0.48,
|
| 375 |
+
"grad_norm": 1.8761085412198597,
|
| 376 |
+
"learning_rate": 3.121247763262235e-07,
|
| 377 |
+
"logits/chosen": -2.8206450939178467,
|
| 378 |
+
"logits/rejected": -2.762035846710205,
|
| 379 |
+
"logps/chosen": -321.90386962890625,
|
| 380 |
+
"logps/rejected": -294.7594299316406,
|
| 381 |
+
"loss": 0.6649,
|
| 382 |
+
"positive_losses": 0.0,
|
| 383 |
+
"rewards/accuracies": 0.887499988079071,
|
| 384 |
+
"rewards/chosen": 0.07717735320329666,
|
| 385 |
+
"rewards/margins": 0.062114737927913666,
|
| 386 |
+
"rewards/margins_max": 0.11694135516881943,
|
| 387 |
+
"rewards/margins_min": 0.011941976845264435,
|
| 388 |
+
"rewards/margins_std": 0.0472017265856266,
|
| 389 |
+
"rewards/rejected": 0.01506261806935072,
|
| 390 |
+
"step": 170
|
| 391 |
+
},
|
| 392 |
+
{
|
| 393 |
+
"dpo_losses": 0.6624481678009033,
|
| 394 |
+
"epoch": 0.51,
|
| 395 |
+
"grad_norm": 1.9655992636640798,
|
| 396 |
+
"learning_rate": 2.880139477883347e-07,
|
| 397 |
+
"logits/chosen": -2.8146770000457764,
|
| 398 |
+
"logits/rejected": -2.726132869720459,
|
| 399 |
+
"logps/chosen": -323.855712890625,
|
| 400 |
+
"logps/rejected": -292.4509582519531,
|
| 401 |
+
"loss": 0.6619,
|
| 402 |
+
"positive_losses": 0.0,
|
| 403 |
+
"rewards/accuracies": 0.9375,
|
| 404 |
+
"rewards/chosen": 0.08045096695423126,
|
| 405 |
+
"rewards/margins": 0.06311116367578506,
|
| 406 |
+
"rewards/margins_max": 0.1282820701599121,
|
| 407 |
+
"rewards/margins_min": 0.017489684745669365,
|
| 408 |
+
"rewards/margins_std": 0.049531489610672,
|
| 409 |
+
"rewards/rejected": 0.0173397995531559,
|
| 410 |
+
"step": 180
|
| 411 |
+
},
|
| 412 |
+
{
|
| 413 |
+
"dpo_losses": 0.6590775847434998,
|
| 414 |
+
"epoch": 0.54,
|
| 415 |
+
"grad_norm": 2.3566449266948037,
|
| 416 |
+
"learning_rate": 2.635347271463544e-07,
|
| 417 |
+
"logits/chosen": -2.776195526123047,
|
| 418 |
+
"logits/rejected": -2.683864116668701,
|
| 419 |
+
"logps/chosen": -313.2172546386719,
|
| 420 |
+
"logps/rejected": -235.8129119873047,
|
| 421 |
+
"loss": 0.6608,
|
| 422 |
+
"positive_losses": 0.006581115536391735,
|
| 423 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 424 |
+
"rewards/chosen": 0.0874362364411354,
|
| 425 |
+
"rewards/margins": 0.07068866491317749,
|
| 426 |
+
"rewards/margins_max": 0.1461794078350067,
|
| 427 |
+
"rewards/margins_min": 0.012810202315449715,
|
| 428 |
+
"rewards/margins_std": 0.06181079149246216,
|
| 429 |
+
"rewards/rejected": 0.01674756594002247,
|
| 430 |
+
"step": 190
|
| 431 |
+
},
|
| 432 |
+
{
|
| 433 |
+
"dpo_losses": 0.659989058971405,
|
| 434 |
+
"epoch": 0.56,
|
| 435 |
+
"grad_norm": 8.082919972131291,
|
| 436 |
+
"learning_rate": 2.3892434184240534e-07,
|
| 437 |
+
"logits/chosen": -2.851855516433716,
|
| 438 |
+
"logits/rejected": -2.7749085426330566,
|
| 439 |
+
"logps/chosen": -330.7215881347656,
|
| 440 |
+
"logps/rejected": -263.6206359863281,
|
| 441 |
+
"loss": 0.6629,
|
| 442 |
+
"positive_losses": 0.010635947808623314,
|
| 443 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 444 |
+
"rewards/chosen": 0.08811243623495102,
|
| 445 |
+
"rewards/margins": 0.06834487617015839,
|
| 446 |
+
"rewards/margins_max": 0.13839691877365112,
|
| 447 |
+
"rewards/margins_min": 0.01415681280195713,
|
| 448 |
+
"rewards/margins_std": 0.05431779474020004,
|
| 449 |
+
"rewards/rejected": 0.01976756379008293,
|
| 450 |
+
"step": 200
|
| 451 |
+
},
|
| 452 |
+
{
|
| 453 |
+
"epoch": 0.56,
|
| 454 |
+
"eval_dpo_losses": 0.6859865784645081,
|
| 455 |
+
"eval_logits/chosen": -2.787888765335083,
|
| 456 |
+
"eval_logits/rejected": -2.7491748332977295,
|
| 457 |
+
"eval_logps/chosen": -279.74639892578125,
|
| 458 |
+
"eval_logps/rejected": -255.2262725830078,
|
| 459 |
+
"eval_loss": 0.6976820230484009,
|
| 460 |
+
"eval_positive_losses": 0.10420288890600204,
|
| 461 |
+
"eval_rewards/accuracies": 0.5979999899864197,
|
| 462 |
+
"eval_rewards/chosen": 0.04846998676657677,
|
| 463 |
+
"eval_rewards/margins": 0.01494329422712326,
|
| 464 |
+
"eval_rewards/margins_max": 0.08812851458787918,
|
| 465 |
+
"eval_rewards/margins_min": -0.04890156909823418,
|
| 466 |
+
"eval_rewards/margins_std": 0.04564943537116051,
|
| 467 |
+
"eval_rewards/rejected": 0.033526696264743805,
|
| 468 |
+
"eval_runtime": 427.8302,
|
| 469 |
+
"eval_samples_per_second": 4.675,
|
| 470 |
+
"eval_steps_per_second": 0.292,
|
| 471 |
+
"step": 200
|
| 472 |
+
},
|
| 473 |
+
{
|
| 474 |
+
"dpo_losses": 0.6545363068580627,
|
| 475 |
+
"epoch": 0.59,
|
| 476 |
+
"grad_norm": 1.9636450688854714,
|
| 477 |
+
"learning_rate": 2.1442129043167873e-07,
|
| 478 |
+
"logits/chosen": -2.754328966140747,
|
| 479 |
+
"logits/rejected": -2.7271111011505127,
|
| 480 |
+
"logps/chosen": -310.51593017578125,
|
| 481 |
+
"logps/rejected": -254.80386352539062,
|
| 482 |
+
"loss": 0.6579,
|
| 483 |
+
"positive_losses": 0.019413376227021217,
|
| 484 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 485 |
+
"rewards/chosen": 0.096683569252491,
|
| 486 |
+
"rewards/margins": 0.08047457039356232,
|
| 487 |
+
"rewards/margins_max": 0.16606834530830383,
|
| 488 |
+
"rewards/margins_min": 0.009150387719273567,
|
| 489 |
+
"rewards/margins_std": 0.07150334119796753,
|
| 490 |
+
"rewards/rejected": 0.01620900072157383,
|
| 491 |
+
"step": 210
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"dpo_losses": 0.6517983675003052,
|
| 495 |
+
"epoch": 0.62,
|
| 496 |
+
"grad_norm": 1.8542376801111211,
|
| 497 |
+
"learning_rate": 1.9026303129961048e-07,
|
| 498 |
+
"logits/chosen": -2.85148024559021,
|
| 499 |
+
"logits/rejected": -2.75597882270813,
|
| 500 |
+
"logps/chosen": -343.08294677734375,
|
| 501 |
+
"logps/rejected": -272.4956359863281,
|
| 502 |
+
"loss": 0.6555,
|
| 503 |
+
"positive_losses": 0.0,
|
| 504 |
+
"rewards/accuracies": 0.949999988079071,
|
| 505 |
+
"rewards/chosen": 0.1037331372499466,
|
| 506 |
+
"rewards/margins": 0.08541466295719147,
|
| 507 |
+
"rewards/margins_max": 0.1512610912322998,
|
| 508 |
+
"rewards/margins_min": 0.02069752663373947,
|
| 509 |
+
"rewards/margins_std": 0.05890106409788132,
|
| 510 |
+
"rewards/rejected": 0.018318474292755127,
|
| 511 |
+
"step": 220
|
| 512 |
+
},
|
| 513 |
+
{
|
| 514 |
+
"dpo_losses": 0.6518235206604004,
|
| 515 |
+
"epoch": 0.65,
|
| 516 |
+
"grad_norm": 1.9608270986414384,
|
| 517 |
+
"learning_rate": 1.6668368145931396e-07,
|
| 518 |
+
"logits/chosen": -2.8535895347595215,
|
| 519 |
+
"logits/rejected": -2.7599148750305176,
|
| 520 |
+
"logps/chosen": -361.3349609375,
|
| 521 |
+
"logps/rejected": -260.3254089355469,
|
| 522 |
+
"loss": 0.6498,
|
| 523 |
+
"positive_losses": 0.010654067620635033,
|
| 524 |
+
"rewards/accuracies": 0.925000011920929,
|
| 525 |
+
"rewards/chosen": 0.10340724140405655,
|
| 526 |
+
"rewards/margins": 0.08572699129581451,
|
| 527 |
+
"rewards/margins_max": 0.16569408774375916,
|
| 528 |
+
"rewards/margins_min": 0.01854148879647255,
|
| 529 |
+
"rewards/margins_std": 0.06613387167453766,
|
| 530 |
+
"rewards/rejected": 0.017680250108242035,
|
| 531 |
+
"step": 230
|
| 532 |
+
},
|
| 533 |
+
{
|
| 534 |
+
"dpo_losses": 0.6590827703475952,
|
| 535 |
+
"epoch": 0.68,
|
| 536 |
+
"grad_norm": 1.5969745701990838,
|
| 537 |
+
"learning_rate": 1.4391174773015834e-07,
|
| 538 |
+
"logits/chosen": -2.8000075817108154,
|
| 539 |
+
"logits/rejected": -2.729769229888916,
|
| 540 |
+
"logps/chosen": -316.0329284667969,
|
| 541 |
+
"logps/rejected": -281.4412841796875,
|
| 542 |
+
"loss": 0.6592,
|
| 543 |
+
"positive_losses": 0.009823227301239967,
|
| 544 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 545 |
+
"rewards/chosen": 0.0876411646604538,
|
| 546 |
+
"rewards/margins": 0.07030560076236725,
|
| 547 |
+
"rewards/margins_max": 0.14278724789619446,
|
| 548 |
+
"rewards/margins_min": 0.013438734225928783,
|
| 549 |
+
"rewards/margins_std": 0.05923761799931526,
|
| 550 |
+
"rewards/rejected": 0.017335567623376846,
|
| 551 |
+
"step": 240
|
| 552 |
+
},
|
| 553 |
+
{
|
| 554 |
+
"dpo_losses": 0.6565552949905396,
|
| 555 |
+
"epoch": 0.7,
|
| 556 |
+
"grad_norm": 1.9164626662561326,
|
| 557 |
+
"learning_rate": 1.2216791228457775e-07,
|
| 558 |
+
"logits/chosen": -2.8003289699554443,
|
| 559 |
+
"logits/rejected": -2.7232441902160645,
|
| 560 |
+
"logps/chosen": -302.2171325683594,
|
| 561 |
+
"logps/rejected": -249.5670928955078,
|
| 562 |
+
"loss": 0.661,
|
| 563 |
+
"positive_losses": 0.009924506768584251,
|
| 564 |
+
"rewards/accuracies": 0.949999988079071,
|
| 565 |
+
"rewards/chosen": 0.090640127658844,
|
| 566 |
+
"rewards/margins": 0.07586830854415894,
|
| 567 |
+
"rewards/margins_max": 0.15460054576396942,
|
| 568 |
+
"rewards/margins_min": 0.012039058841764927,
|
| 569 |
+
"rewards/margins_std": 0.06554323434829712,
|
| 570 |
+
"rewards/rejected": 0.014771823771297932,
|
| 571 |
+
"step": 250
|
| 572 |
+
},
|
| 573 |
+
{
|
| 574 |
+
"dpo_losses": 0.6541475057601929,
|
| 575 |
+
"epoch": 0.73,
|
| 576 |
+
"grad_norm": 1.7167350749240156,
|
| 577 |
+
"learning_rate": 1.0166289402331391e-07,
|
| 578 |
+
"logits/chosen": -2.8486251831054688,
|
| 579 |
+
"logits/rejected": -2.7656264305114746,
|
| 580 |
+
"logps/chosen": -282.35260009765625,
|
| 581 |
+
"logps/rejected": -257.30596923828125,
|
| 582 |
+
"loss": 0.6505,
|
| 583 |
+
"positive_losses": 0.0,
|
| 584 |
+
"rewards/accuracies": 0.949999988079071,
|
| 585 |
+
"rewards/chosen": 0.09740927070379257,
|
| 586 |
+
"rewards/margins": 0.08115915954113007,
|
| 587 |
+
"rewards/margins_max": 0.16602402925491333,
|
| 588 |
+
"rewards/margins_min": 0.016714217141270638,
|
| 589 |
+
"rewards/margins_std": 0.06957190483808517,
|
| 590 |
+
"rewards/rejected": 0.016250116750597954,
|
| 591 |
+
"step": 260
|
| 592 |
+
},
|
| 593 |
+
{
|
| 594 |
+
"dpo_losses": 0.6572530269622803,
|
| 595 |
+
"epoch": 0.76,
|
| 596 |
+
"grad_norm": 2.017681285366355,
|
| 597 |
+
"learning_rate": 8.259540650444734e-08,
|
| 598 |
+
"logits/chosen": -2.7961807250976562,
|
| 599 |
+
"logits/rejected": -2.743107318878174,
|
| 600 |
+
"logps/chosen": -299.6520690917969,
|
| 601 |
+
"logps/rejected": -260.46929931640625,
|
| 602 |
+
"loss": 0.6604,
|
| 603 |
+
"positive_losses": 0.046803951263427734,
|
| 604 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 605 |
+
"rewards/chosen": 0.09463498741388321,
|
| 606 |
+
"rewards/margins": 0.07433197647333145,
|
| 607 |
+
"rewards/margins_max": 0.14396773278713226,
|
| 608 |
+
"rewards/margins_min": 0.020581519231200218,
|
| 609 |
+
"rewards/margins_std": 0.05590103194117546,
|
| 610 |
+
"rewards/rejected": 0.020303016528487206,
|
| 611 |
+
"step": 270
|
| 612 |
+
},
|
| 613 |
+
{
|
| 614 |
+
"dpo_losses": 0.655379593372345,
|
| 615 |
+
"epoch": 0.79,
|
| 616 |
+
"grad_norm": 1.734695130092448,
|
| 617 |
+
"learning_rate": 6.515023221586721e-08,
|
| 618 |
+
"logits/chosen": -2.7713265419006348,
|
| 619 |
+
"logits/rejected": -2.7317543029785156,
|
| 620 |
+
"logps/chosen": -293.76214599609375,
|
| 621 |
+
"logps/rejected": -270.7189025878906,
|
| 622 |
+
"loss": 0.6575,
|
| 623 |
+
"positive_losses": 0.04788818210363388,
|
| 624 |
+
"rewards/accuracies": 0.949999988079071,
|
| 625 |
+
"rewards/chosen": 0.09694816917181015,
|
| 626 |
+
"rewards/margins": 0.07816567271947861,
|
| 627 |
+
"rewards/margins_max": 0.14518895745277405,
|
| 628 |
+
"rewards/margins_min": 0.019240325316786766,
|
| 629 |
+
"rewards/margins_std": 0.057953737676143646,
|
| 630 |
+
"rewards/rejected": 0.018782509490847588,
|
| 631 |
+
"step": 280
|
| 632 |
+
},
|
| 633 |
+
{
|
| 634 |
+
"dpo_losses": 0.6595104336738586,
|
| 635 |
+
"epoch": 0.82,
|
| 636 |
+
"grad_norm": 1.7562807996771055,
|
| 637 |
+
"learning_rate": 4.949643185335287e-08,
|
| 638 |
+
"logits/chosen": -2.7762978076934814,
|
| 639 |
+
"logits/rejected": -2.7266323566436768,
|
| 640 |
+
"logps/chosen": -278.39996337890625,
|
| 641 |
+
"logps/rejected": -262.4858093261719,
|
| 642 |
+
"loss": 0.6646,
|
| 643 |
+
"positive_losses": 0.06580867618322372,
|
| 644 |
+
"rewards/accuracies": 0.875,
|
| 645 |
+
"rewards/chosen": 0.08516252040863037,
|
| 646 |
+
"rewards/margins": 0.06983821839094162,
|
| 647 |
+
"rewards/margins_max": 0.15399742126464844,
|
| 648 |
+
"rewards/margins_min": 0.005857336334884167,
|
| 649 |
+
"rewards/margins_std": 0.06812156736850739,
|
| 650 |
+
"rewards/rejected": 0.015324289910495281,
|
| 651 |
+
"step": 290
|
| 652 |
+
},
|
| 653 |
+
{
|
| 654 |
+
"dpo_losses": 0.6477023363113403,
|
| 655 |
+
"epoch": 0.85,
|
| 656 |
+
"grad_norm": 1.9324048018602529,
|
| 657 |
+
"learning_rate": 3.578570595810274e-08,
|
| 658 |
+
"logits/chosen": -2.8429884910583496,
|
| 659 |
+
"logits/rejected": -2.765357732772827,
|
| 660 |
+
"logps/chosen": -335.2912292480469,
|
| 661 |
+
"logps/rejected": -287.69110107421875,
|
| 662 |
+
"loss": 0.6479,
|
| 663 |
+
"positive_losses": 0.0,
|
| 664 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 665 |
+
"rewards/chosen": 0.1122179627418518,
|
| 666 |
+
"rewards/margins": 0.09479531645774841,
|
| 667 |
+
"rewards/margins_max": 0.18769413232803345,
|
| 668 |
+
"rewards/margins_min": 0.015501707792282104,
|
| 669 |
+
"rewards/margins_std": 0.07714836299419403,
|
| 670 |
+
"rewards/rejected": 0.017422644421458244,
|
| 671 |
+
"step": 300
|
| 672 |
+
},
|
| 673 |
+
{
|
| 674 |
+
"epoch": 0.85,
|
| 675 |
+
"eval_dpo_losses": 0.6845089793205261,
|
| 676 |
+
"eval_logits/chosen": -2.785942554473877,
|
| 677 |
+
"eval_logits/rejected": -2.7478015422821045,
|
| 678 |
+
"eval_logps/chosen": -279.3167724609375,
|
| 679 |
+
"eval_logps/rejected": -255.1217498779297,
|
| 680 |
+
"eval_loss": 0.702374279499054,
|
| 681 |
+
"eval_positive_losses": 0.15923255681991577,
|
| 682 |
+
"eval_rewards/accuracies": 0.593999981880188,
|
| 683 |
+
"eval_rewards/chosen": 0.05276622623205185,
|
| 684 |
+
"eval_rewards/margins": 0.01819423958659172,
|
| 685 |
+
"eval_rewards/margins_max": 0.10716991871595383,
|
| 686 |
+
"eval_rewards/margins_min": -0.059129390865564346,
|
| 687 |
+
"eval_rewards/margins_std": 0.055548474192619324,
|
| 688 |
+
"eval_rewards/rejected": 0.03457198664546013,
|
| 689 |
+
"eval_runtime": 428.1446,
|
| 690 |
+
"eval_samples_per_second": 4.671,
|
| 691 |
+
"eval_steps_per_second": 0.292,
|
| 692 |
+
"step": 300
|
| 693 |
+
},
|
| 694 |
+
{
|
| 695 |
+
"dpo_losses": 0.6494620442390442,
|
| 696 |
+
"epoch": 0.87,
|
| 697 |
+
"grad_norm": 1.7775417343369382,
|
| 698 |
+
"learning_rate": 2.415092479103503e-08,
|
| 699 |
+
"logits/chosen": -2.8399946689605713,
|
| 700 |
+
"logits/rejected": -2.7444241046905518,
|
| 701 |
+
"logps/chosen": -287.1294250488281,
|
| 702 |
+
"logps/rejected": -212.641845703125,
|
| 703 |
+
"loss": 0.6483,
|
| 704 |
+
"positive_losses": 0.0,
|
| 705 |
+
"rewards/accuracies": 0.987500011920929,
|
| 706 |
+
"rewards/chosen": 0.10686449706554413,
|
| 707 |
+
"rewards/margins": 0.0910436287522316,
|
| 708 |
+
"rewards/margins_max": 0.18475615978240967,
|
| 709 |
+
"rewards/margins_min": 0.017825862392783165,
|
| 710 |
+
"rewards/margins_std": 0.07425465434789658,
|
| 711 |
+
"rewards/rejected": 0.015820873901247978,
|
| 712 |
+
"step": 310
|
| 713 |
+
},
|
| 714 |
+
{
|
| 715 |
+
"dpo_losses": 0.654121994972229,
|
| 716 |
+
"epoch": 0.9,
|
| 717 |
+
"grad_norm": 1.8615740426477174,
|
| 718 |
+
"learning_rate": 1.4704840690808656e-08,
|
| 719 |
+
"logits/chosen": -2.8066587448120117,
|
| 720 |
+
"logits/rejected": -2.7482457160949707,
|
| 721 |
+
"logps/chosen": -298.34295654296875,
|
| 722 |
+
"logps/rejected": -258.84478759765625,
|
| 723 |
+
"loss": 0.6527,
|
| 724 |
+
"positive_losses": 0.004542541690170765,
|
| 725 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 726 |
+
"rewards/chosen": 0.09930218756198883,
|
| 727 |
+
"rewards/margins": 0.0810491144657135,
|
| 728 |
+
"rewards/margins_max": 0.17170652747154236,
|
| 729 |
+
"rewards/margins_min": 0.012719864025712013,
|
| 730 |
+
"rewards/margins_std": 0.07160074263811111,
|
| 731 |
+
"rewards/rejected": 0.018253061920404434,
|
| 732 |
+
"step": 320
|
| 733 |
+
},
|
| 734 |
+
{
|
| 735 |
+
"dpo_losses": 0.6466237306594849,
|
| 736 |
+
"epoch": 0.93,
|
| 737 |
+
"grad_norm": 1.9427490502829974,
|
| 738 |
+
"learning_rate": 7.538995394063995e-09,
|
| 739 |
+
"logits/chosen": -2.8800158500671387,
|
| 740 |
+
"logits/rejected": -2.7991366386413574,
|
| 741 |
+
"logps/chosen": -343.4425964355469,
|
| 742 |
+
"logps/rejected": -265.7518005371094,
|
| 743 |
+
"loss": 0.6494,
|
| 744 |
+
"positive_losses": 0.0037406920455396175,
|
| 745 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 746 |
+
"rewards/chosen": 0.10877249389886856,
|
| 747 |
+
"rewards/margins": 0.09689434617757797,
|
| 748 |
+
"rewards/margins_max": 0.17881345748901367,
|
| 749 |
+
"rewards/margins_min": 0.022731659933924675,
|
| 750 |
+
"rewards/margins_std": 0.07121957838535309,
|
| 751 |
+
"rewards/rejected": 0.011878135614097118,
|
| 752 |
+
"step": 330
|
| 753 |
+
},
|
| 754 |
+
{
|
| 755 |
+
"dpo_losses": 0.6502379179000854,
|
| 756 |
+
"epoch": 0.96,
|
| 757 |
+
"grad_norm": 1.916115854424015,
|
| 758 |
+
"learning_rate": 2.7228329070159705e-09,
|
| 759 |
+
"logits/chosen": -2.767482280731201,
|
| 760 |
+
"logits/rejected": -2.7154014110565186,
|
| 761 |
+
"logps/chosen": -289.31561279296875,
|
| 762 |
+
"logps/rejected": -249.5873260498047,
|
| 763 |
+
"loss": 0.6484,
|
| 764 |
+
"positive_losses": 0.02227325364947319,
|
| 765 |
+
"rewards/accuracies": 0.925000011920929,
|
| 766 |
+
"rewards/chosen": 0.10529886186122894,
|
| 767 |
+
"rewards/margins": 0.08947577327489853,
|
| 768 |
+
"rewards/margins_max": 0.1924385130405426,
|
| 769 |
+
"rewards/margins_min": 0.021350277587771416,
|
| 770 |
+
"rewards/margins_std": 0.07764019072055817,
|
| 771 |
+
"rewards/rejected": 0.01582309789955616,
|
| 772 |
+
"step": 340
|
| 773 |
+
},
|
| 774 |
+
{
|
| 775 |
+
"dpo_losses": 0.6559463143348694,
|
| 776 |
+
"epoch": 0.99,
|
| 777 |
+
"grad_norm": 5.156600662792089,
|
| 778 |
+
"learning_rate": 3.0302652553296226e-10,
|
| 779 |
+
"logits/chosen": -2.750614881515503,
|
| 780 |
+
"logits/rejected": -2.7069051265716553,
|
| 781 |
+
"logps/chosen": -305.8713073730469,
|
| 782 |
+
"logps/rejected": -285.0591735839844,
|
| 783 |
+
"loss": 0.6565,
|
| 784 |
+
"positive_losses": 0.04594574123620987,
|
| 785 |
+
"rewards/accuracies": 0.9375,
|
| 786 |
+
"rewards/chosen": 0.09087814390659332,
|
| 787 |
+
"rewards/margins": 0.07695063203573227,
|
| 788 |
+
"rewards/margins_max": 0.1514495462179184,
|
| 789 |
+
"rewards/margins_min": 0.016952747479081154,
|
| 790 |
+
"rewards/margins_std": 0.06040460616350174,
|
| 791 |
+
"rewards/rejected": 0.01392750721424818,
|
| 792 |
+
"step": 350
|
| 793 |
+
},
|
| 794 |
+
{
|
| 795 |
+
"epoch": 1.0,
|
| 796 |
+
"step": 355,
|
| 797 |
+
"total_flos": 0.0,
|
| 798 |
+
"train_loss": 0.6688590748209349,
|
| 799 |
+
"train_runtime": 4310.2257,
|
| 800 |
+
"train_samples_per_second": 1.317,
|
| 801 |
+
"train_steps_per_second": 0.082
|
| 802 |
+
}
|
| 803 |
+
],
|
| 804 |
+
"logging_steps": 10,
|
| 805 |
+
"max_steps": 355,
|
| 806 |
+
"num_input_tokens_seen": 0,
|
| 807 |
+
"num_train_epochs": 1,
|
| 808 |
+
"save_steps": 100,
|
| 809 |
+
"total_flos": 0.0,
|
| 810 |
+
"train_batch_size": 4,
|
| 811 |
+
"trial_name": null,
|
| 812 |
+
"trial_params": null
|
| 813 |
+
}
|