Model save
Browse files- README.md +87 -0
- adapter_model.safetensors +1 -1
- all_results.json +8 -0
- runs/Jul26_16-00-37_node12/events.out.tfevents.1721977340.node12.2124571.0 +2 -2
- train_results.json +8 -0
- trainer_state.json +2199 -0
README.md
ADDED
|
@@ -0,0 +1,87 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
library_name: peft
|
| 4 |
+
tags:
|
| 5 |
+
- trl
|
| 6 |
+
- dpo
|
| 7 |
+
- generated_from_trainer
|
| 8 |
+
base_model: alignment-handbook/zephyr-7b-sft-full
|
| 9 |
+
model-index:
|
| 10 |
+
- name: zephyr-7b-dpop-uf6k-qlora-5e-7-epoch3
|
| 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-7b-dpop-uf6k-qlora-5e-7-epoch3
|
| 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: 0.6878
|
| 22 |
+
- Positive Losses: 0.0651
|
| 23 |
+
- Dpo Losses: 0.6793
|
| 24 |
+
- Rewards/chosen: 0.0802
|
| 25 |
+
- Rewards/rejected: 0.0515
|
| 26 |
+
- Rewards/accuracies: 0.6706
|
| 27 |
+
- Rewards/margins: 0.0287
|
| 28 |
+
- Rewards/margins Max: 0.0914
|
| 29 |
+
- Rewards/margins Min: -0.0317
|
| 30 |
+
- Rewards/margins Std: 0.0550
|
| 31 |
+
- Logps/rejected: -254.0345
|
| 32 |
+
- Logps/chosen: -277.2037
|
| 33 |
+
- Logits/rejected: -2.8028
|
| 34 |
+
- Logits/chosen: -2.8477
|
| 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: 2
|
| 55 |
+
- eval_batch_size: 4
|
| 56 |
+
- seed: 42
|
| 57 |
+
- distributed_type: multi-GPU
|
| 58 |
+
- num_devices: 8
|
| 59 |
+
- total_train_batch_size: 16
|
| 60 |
+
- total_eval_batch_size: 32
|
| 61 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 62 |
+
- lr_scheduler_type: cosine
|
| 63 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 64 |
+
- num_epochs: 3
|
| 65 |
+
|
| 66 |
+
### Training results
|
| 67 |
+
|
| 68 |
+
| 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 |
|
| 69 |
+
|:-------------:|:-----:|:----:|:---------------:|:---------------:|:----------:|:--------------:|:----------------:|:------------------:|:---------------:|:-------------------:|:-------------------:|:-------------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 70 |
+
| 0.6932 | 0.3 | 100 | 0.6930 | 0.0052 | 0.6925 | 0.0096 | 0.0083 | 0.5794 | 0.0012 | 0.0066 | -0.0040 | 0.0047 | -258.3495 | -284.2642 | -2.8141 | -2.8594 |
|
| 71 |
+
| 0.6919 | 0.61 | 200 | 0.6915 | 0.0119 | 0.6904 | 0.0242 | 0.0187 | 0.6667 | 0.0055 | 0.0195 | -0.0083 | 0.0124 | -257.3141 | -282.7999 | -2.8133 | -2.8583 |
|
| 72 |
+
| 0.6903 | 0.91 | 300 | 0.6899 | 0.0165 | 0.6876 | 0.0395 | 0.0283 | 0.6667 | 0.0112 | 0.0379 | -0.0143 | 0.0232 | -256.3544 | -281.2695 | -2.8086 | -2.8537 |
|
| 73 |
+
| 0.6832 | 1.22 | 400 | 0.6892 | 0.0304 | 0.6847 | 0.0525 | 0.0351 | 0.7024 | 0.0174 | 0.0557 | -0.0196 | 0.0337 | -255.6741 | -279.9755 | -2.8057 | -2.8507 |
|
| 74 |
+
| 0.6776 | 1.52 | 500 | 0.6884 | 0.0444 | 0.6825 | 0.0647 | 0.0427 | 0.6905 | 0.0220 | 0.0710 | -0.0256 | 0.0433 | -254.9144 | -278.7508 | -2.8047 | -2.8495 |
|
| 75 |
+
| 0.677 | 1.82 | 600 | 0.6873 | 0.0459 | 0.6811 | 0.0769 | 0.0519 | 0.6825 | 0.0250 | 0.0803 | -0.0280 | 0.0484 | -253.9932 | -277.5360 | -2.8047 | -2.8494 |
|
| 76 |
+
| 0.6796 | 2.13 | 700 | 0.6872 | 0.0548 | 0.6800 | 0.0798 | 0.0526 | 0.6825 | 0.0272 | 0.0865 | -0.0298 | 0.0521 | -253.9202 | -277.2366 | -2.8026 | -2.8477 |
|
| 77 |
+
| 0.6778 | 2.43 | 800 | 0.6875 | 0.0604 | 0.6795 | 0.0800 | 0.0518 | 0.6825 | 0.0282 | 0.0897 | -0.0307 | 0.0540 | -254.0074 | -277.2222 | -2.8024 | -2.8474 |
|
| 78 |
+
| 0.6739 | 2.74 | 900 | 0.6878 | 0.0651 | 0.6793 | 0.0802 | 0.0515 | 0.6706 | 0.0287 | 0.0914 | -0.0317 | 0.0550 | -254.0345 | -277.2037 | -2.8028 | -2.8477 |
|
| 79 |
+
|
| 80 |
+
|
| 81 |
+
### Framework versions
|
| 82 |
+
|
| 83 |
+
- PEFT 0.7.1
|
| 84 |
+
- Transformers 4.39.0.dev0
|
| 85 |
+
- Pytorch 2.1.2+cu121
|
| 86 |
+
- Datasets 2.14.6
|
| 87 |
+
- 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:f5339b4ca5429745d030c94b880df53f7a2dce8452865089487c30dfda548446
|
| 3 |
size 671150064
|
all_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"train_loss": 0.684076373458754,
|
| 4 |
+
"train_runtime": 8695.656,
|
| 5 |
+
"train_samples": 5263,
|
| 6 |
+
"train_samples_per_second": 1.816,
|
| 7 |
+
"train_steps_per_second": 0.114
|
| 8 |
+
}
|
runs/Jul26_16-00-37_node12/events.out.tfevents.1721977340.node12.2124571.0
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 sha256:5ead509b20ac1f0a2d35b60a20627942180d6a666ed314eac43291aa66daa60d
|
| 3 |
+
size 113332
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"train_loss": 0.684076373458754,
|
| 4 |
+
"train_runtime": 8695.656,
|
| 5 |
+
"train_samples": 5263,
|
| 6 |
+
"train_samples_per_second": 1.816,
|
| 7 |
+
"train_steps_per_second": 0.114
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,2199 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 3.0,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 987,
|
| 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": 2.1006424602315215,
|
| 15 |
+
"learning_rate": 5.050505050505051e-09,
|
| 16 |
+
"logits/chosen": -2.589059352874756,
|
| 17 |
+
"logits/rejected": -2.6418559551239014,
|
| 18 |
+
"logps/chosen": -212.42770385742188,
|
| 19 |
+
"logps/rejected": -135.43048095703125,
|
| 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": 0.6925780773162842,
|
| 33 |
+
"epoch": 0.03,
|
| 34 |
+
"grad_norm": 12.811305441645697,
|
| 35 |
+
"learning_rate": 5.05050505050505e-08,
|
| 36 |
+
"logits/chosen": -2.977125644683838,
|
| 37 |
+
"logits/rejected": -2.981342315673828,
|
| 38 |
+
"logps/chosen": -426.8561096191406,
|
| 39 |
+
"logps/rejected": -265.4306640625,
|
| 40 |
+
"loss": 0.6957,
|
| 41 |
+
"positive_losses": 0.01630179025232792,
|
| 42 |
+
"rewards/accuracies": 0.3888888955116272,
|
| 43 |
+
"rewards/chosen": 0.0003335709043312818,
|
| 44 |
+
"rewards/margins": 0.0011399724753573537,
|
| 45 |
+
"rewards/margins_max": 0.002273110207170248,
|
| 46 |
+
"rewards/margins_min": 6.834672603872605e-06,
|
| 47 |
+
"rewards/margins_std": 0.0016024988144636154,
|
| 48 |
+
"rewards/rejected": -0.0008064016001299024,
|
| 49 |
+
"step": 10
|
| 50 |
+
},
|
| 51 |
+
{
|
| 52 |
+
"dpo_losses": 0.6931944489479065,
|
| 53 |
+
"epoch": 0.06,
|
| 54 |
+
"grad_norm": 17.598175196981984,
|
| 55 |
+
"learning_rate": 1.01010101010101e-07,
|
| 56 |
+
"logits/chosen": -2.992300271987915,
|
| 57 |
+
"logits/rejected": -2.8757376670837402,
|
| 58 |
+
"logps/chosen": -376.1276550292969,
|
| 59 |
+
"logps/rejected": -280.2401428222656,
|
| 60 |
+
"loss": 0.7005,
|
| 61 |
+
"positive_losses": 0.06999759376049042,
|
| 62 |
+
"rewards/accuracies": 0.5,
|
| 63 |
+
"rewards/chosen": 0.0003719997184816748,
|
| 64 |
+
"rewards/margins": -9.284308907808736e-05,
|
| 65 |
+
"rewards/margins_max": 0.0014239997835829854,
|
| 66 |
+
"rewards/margins_min": -0.0016096860636025667,
|
| 67 |
+
"rewards/margins_std": 0.0021451397333294153,
|
| 68 |
+
"rewards/rejected": 0.000464842829387635,
|
| 69 |
+
"step": 20
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"dpo_losses": 0.6936646699905396,
|
| 73 |
+
"epoch": 0.09,
|
| 74 |
+
"grad_norm": 16.284910292127993,
|
| 75 |
+
"learning_rate": 1.5151515151515152e-07,
|
| 76 |
+
"logits/chosen": -2.8964314460754395,
|
| 77 |
+
"logits/rejected": -2.9219629764556885,
|
| 78 |
+
"logps/chosen": -204.34725952148438,
|
| 79 |
+
"logps/rejected": -278.52960205078125,
|
| 80 |
+
"loss": 0.698,
|
| 81 |
+
"positive_losses": 0.042078591883182526,
|
| 82 |
+
"rewards/accuracies": 0.3499999940395355,
|
| 83 |
+
"rewards/chosen": 0.0006226692348718643,
|
| 84 |
+
"rewards/margins": -0.0010327090276405215,
|
| 85 |
+
"rewards/margins_max": 0.0002273331192554906,
|
| 86 |
+
"rewards/margins_min": -0.002292751334607601,
|
| 87 |
+
"rewards/margins_std": 0.001781968749128282,
|
| 88 |
+
"rewards/rejected": 0.001655378146097064,
|
| 89 |
+
"step": 30
|
| 90 |
+
},
|
| 91 |
+
{
|
| 92 |
+
"dpo_losses": 0.6926015615463257,
|
| 93 |
+
"epoch": 0.12,
|
| 94 |
+
"grad_norm": 7.600951496593335,
|
| 95 |
+
"learning_rate": 2.02020202020202e-07,
|
| 96 |
+
"logits/chosen": -2.841174602508545,
|
| 97 |
+
"logits/rejected": -2.759981632232666,
|
| 98 |
+
"logps/chosen": -320.43280029296875,
|
| 99 |
+
"logps/rejected": -249.27462768554688,
|
| 100 |
+
"loss": 0.6944,
|
| 101 |
+
"positive_losses": 0.0006469726795330644,
|
| 102 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 103 |
+
"rewards/chosen": 0.0031745159067213535,
|
| 104 |
+
"rewards/margins": 0.0010936097241938114,
|
| 105 |
+
"rewards/margins_max": 0.0026608791667968035,
|
| 106 |
+
"rewards/margins_min": -0.0004736593982670456,
|
| 107 |
+
"rewards/margins_std": 0.0022164531983435154,
|
| 108 |
+
"rewards/rejected": 0.0020809059496968985,
|
| 109 |
+
"step": 40
|
| 110 |
+
},
|
| 111 |
+
{
|
| 112 |
+
"dpo_losses": 0.6930263638496399,
|
| 113 |
+
"epoch": 0.15,
|
| 114 |
+
"grad_norm": 9.105946335609067,
|
| 115 |
+
"learning_rate": 2.525252525252525e-07,
|
| 116 |
+
"logits/chosen": -2.829671859741211,
|
| 117 |
+
"logits/rejected": -2.848930835723877,
|
| 118 |
+
"logps/chosen": -227.5389404296875,
|
| 119 |
+
"logps/rejected": -240.80026245117188,
|
| 120 |
+
"loss": 0.694,
|
| 121 |
+
"positive_losses": 0.0,
|
| 122 |
+
"rewards/accuracies": 0.5,
|
| 123 |
+
"rewards/chosen": 0.004696831107139587,
|
| 124 |
+
"rewards/margins": 0.00024281651712954044,
|
| 125 |
+
"rewards/margins_max": 0.0018194876611232758,
|
| 126 |
+
"rewards/margins_min": -0.0013338546268641949,
|
| 127 |
+
"rewards/margins_std": 0.0022297496907413006,
|
| 128 |
+
"rewards/rejected": 0.004454014357179403,
|
| 129 |
+
"step": 50
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"dpo_losses": 0.6927760243415833,
|
| 133 |
+
"epoch": 0.18,
|
| 134 |
+
"grad_norm": 1.9982033646829467,
|
| 135 |
+
"learning_rate": 3.0303030303030305e-07,
|
| 136 |
+
"logits/chosen": -2.807051420211792,
|
| 137 |
+
"logits/rejected": -2.722658157348633,
|
| 138 |
+
"logps/chosen": -333.5701599121094,
|
| 139 |
+
"logps/rejected": -296.24591064453125,
|
| 140 |
+
"loss": 0.6935,
|
| 141 |
+
"positive_losses": 0.01206359826028347,
|
| 142 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 143 |
+
"rewards/chosen": 0.007529796566814184,
|
| 144 |
+
"rewards/margins": 0.0007501507643610239,
|
| 145 |
+
"rewards/margins_max": 0.004166206810623407,
|
| 146 |
+
"rewards/margins_min": -0.002665905049070716,
|
| 147 |
+
"rewards/margins_std": 0.004831031896173954,
|
| 148 |
+
"rewards/rejected": 0.006779646035283804,
|
| 149 |
+
"step": 60
|
| 150 |
+
},
|
| 151 |
+
{
|
| 152 |
+
"dpo_losses": 0.692486584186554,
|
| 153 |
+
"epoch": 0.21,
|
| 154 |
+
"grad_norm": 9.583384726566226,
|
| 155 |
+
"learning_rate": 3.535353535353535e-07,
|
| 156 |
+
"logits/chosen": -2.6647205352783203,
|
| 157 |
+
"logits/rejected": -2.6472456455230713,
|
| 158 |
+
"logps/chosen": -305.70159912109375,
|
| 159 |
+
"logps/rejected": -241.12240600585938,
|
| 160 |
+
"loss": 0.6938,
|
| 161 |
+
"positive_losses": 0.006031799130141735,
|
| 162 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 163 |
+
"rewards/chosen": 0.007261182181537151,
|
| 164 |
+
"rewards/margins": 0.0013255949597805738,
|
| 165 |
+
"rewards/margins_max": 0.003496175166219473,
|
| 166 |
+
"rewards/margins_min": -0.0008449858287349343,
|
| 167 |
+
"rewards/margins_std": 0.0030696645844727755,
|
| 168 |
+
"rewards/rejected": 0.0059355879202485085,
|
| 169 |
+
"step": 70
|
| 170 |
+
},
|
| 171 |
+
{
|
| 172 |
+
"dpo_losses": 0.6925214529037476,
|
| 173 |
+
"epoch": 0.24,
|
| 174 |
+
"grad_norm": 2.10104329669103,
|
| 175 |
+
"learning_rate": 4.04040404040404e-07,
|
| 176 |
+
"logits/chosen": -2.965867280960083,
|
| 177 |
+
"logits/rejected": -2.8706986904144287,
|
| 178 |
+
"logps/chosen": -324.9981689453125,
|
| 179 |
+
"logps/rejected": -331.00262451171875,
|
| 180 |
+
"loss": 0.6938,
|
| 181 |
+
"positive_losses": 0.0,
|
| 182 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 183 |
+
"rewards/chosen": 0.009126158431172371,
|
| 184 |
+
"rewards/margins": 0.0012557277223095298,
|
| 185 |
+
"rewards/margins_max": 0.0036045038141310215,
|
| 186 |
+
"rewards/margins_min": -0.0010930480202659965,
|
| 187 |
+
"rewards/margins_std": 0.0033216706942766905,
|
| 188 |
+
"rewards/rejected": 0.007870429195463657,
|
| 189 |
+
"step": 80
|
| 190 |
+
},
|
| 191 |
+
{
|
| 192 |
+
"dpo_losses": 0.6932989358901978,
|
| 193 |
+
"epoch": 0.27,
|
| 194 |
+
"grad_norm": 1.9836450871378364,
|
| 195 |
+
"learning_rate": 4.545454545454545e-07,
|
| 196 |
+
"logits/chosen": -2.7672338485717773,
|
| 197 |
+
"logits/rejected": -2.6813530921936035,
|
| 198 |
+
"logps/chosen": -280.62030029296875,
|
| 199 |
+
"logps/rejected": -277.8973388671875,
|
| 200 |
+
"loss": 0.6933,
|
| 201 |
+
"positive_losses": 0.0020202635787427425,
|
| 202 |
+
"rewards/accuracies": 0.550000011920929,
|
| 203 |
+
"rewards/chosen": 0.007793852128088474,
|
| 204 |
+
"rewards/margins": -0.00029996776720508933,
|
| 205 |
+
"rewards/margins_max": 0.0021644725929945707,
|
| 206 |
+
"rewards/margins_min": -0.0027644080109894276,
|
| 207 |
+
"rewards/margins_std": 0.003485244931653142,
|
| 208 |
+
"rewards/rejected": 0.008093819953501225,
|
| 209 |
+
"step": 90
|
| 210 |
+
},
|
| 211 |
+
{
|
| 212 |
+
"dpo_losses": 0.6919553279876709,
|
| 213 |
+
"epoch": 0.3,
|
| 214 |
+
"grad_norm": 1.713342030808654,
|
| 215 |
+
"learning_rate": 4.99998435471878e-07,
|
| 216 |
+
"logits/chosen": -2.971386432647705,
|
| 217 |
+
"logits/rejected": -2.915938377380371,
|
| 218 |
+
"logps/chosen": -269.40509033203125,
|
| 219 |
+
"logps/rejected": -256.8268737792969,
|
| 220 |
+
"loss": 0.6932,
|
| 221 |
+
"positive_losses": 0.0,
|
| 222 |
+
"rewards/accuracies": 0.75,
|
| 223 |
+
"rewards/chosen": 0.008469906635582447,
|
| 224 |
+
"rewards/margins": 0.002389331813901663,
|
| 225 |
+
"rewards/margins_max": 0.004788221325725317,
|
| 226 |
+
"rewards/margins_min": -9.557651537761558e-06,
|
| 227 |
+
"rewards/margins_std": 0.0033925422467291355,
|
| 228 |
+
"rewards/rejected": 0.006080574356019497,
|
| 229 |
+
"step": 100
|
| 230 |
+
},
|
| 231 |
+
{
|
| 232 |
+
"epoch": 0.3,
|
| 233 |
+
"eval_dpo_losses": 0.6925311088562012,
|
| 234 |
+
"eval_logits/chosen": -2.8593878746032715,
|
| 235 |
+
"eval_logits/rejected": -2.8141348361968994,
|
| 236 |
+
"eval_logps/chosen": -284.26416015625,
|
| 237 |
+
"eval_logps/rejected": -258.3495178222656,
|
| 238 |
+
"eval_loss": 0.6929529905319214,
|
| 239 |
+
"eval_positive_losses": 0.0051858690567314625,
|
| 240 |
+
"eval_rewards/accuracies": 0.579365074634552,
|
| 241 |
+
"eval_rewards/chosen": 0.009570538066327572,
|
| 242 |
+
"eval_rewards/margins": 0.0012392366770654917,
|
| 243 |
+
"eval_rewards/margins_max": 0.006621083710342646,
|
| 244 |
+
"eval_rewards/margins_min": -0.003959333524107933,
|
| 245 |
+
"eval_rewards/margins_std": 0.004715622402727604,
|
| 246 |
+
"eval_rewards/rejected": 0.00833130069077015,
|
| 247 |
+
"eval_runtime": 282.6491,
|
| 248 |
+
"eval_samples_per_second": 7.076,
|
| 249 |
+
"eval_steps_per_second": 0.223,
|
| 250 |
+
"step": 100
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"dpo_losses": 0.6927061676979065,
|
| 254 |
+
"epoch": 0.33,
|
| 255 |
+
"grad_norm": 2.1416302847338655,
|
| 256 |
+
"learning_rate": 4.998107157902955e-07,
|
| 257 |
+
"logits/chosen": -2.841651201248169,
|
| 258 |
+
"logits/rejected": -2.74540638923645,
|
| 259 |
+
"logps/chosen": -310.16998291015625,
|
| 260 |
+
"logps/rejected": -254.25537109375,
|
| 261 |
+
"loss": 0.6931,
|
| 262 |
+
"positive_losses": 0.009991454891860485,
|
| 263 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 264 |
+
"rewards/chosen": 0.009014981798827648,
|
| 265 |
+
"rewards/margins": 0.0008873892948031425,
|
| 266 |
+
"rewards/margins_max": 0.0038191110361367464,
|
| 267 |
+
"rewards/margins_min": -0.0020443324465304613,
|
| 268 |
+
"rewards/margins_std": 0.004146079998463392,
|
| 269 |
+
"rewards/rejected": 0.008127592504024506,
|
| 270 |
+
"step": 110
|
| 271 |
+
},
|
| 272 |
+
{
|
| 273 |
+
"dpo_losses": 0.6933146715164185,
|
| 274 |
+
"epoch": 0.36,
|
| 275 |
+
"grad_norm": 2.01446405141137,
|
| 276 |
+
"learning_rate": 4.993103596812268e-07,
|
| 277 |
+
"logits/chosen": -2.7760932445526123,
|
| 278 |
+
"logits/rejected": -2.853368043899536,
|
| 279 |
+
"logps/chosen": -241.74014282226562,
|
| 280 |
+
"logps/rejected": -251.635009765625,
|
| 281 |
+
"loss": 0.6928,
|
| 282 |
+
"positive_losses": 0.0,
|
| 283 |
+
"rewards/accuracies": 0.5,
|
| 284 |
+
"rewards/chosen": 0.011283518746495247,
|
| 285 |
+
"rewards/margins": -0.00032261465094052255,
|
| 286 |
+
"rewards/margins_max": 0.0035319500602781773,
|
| 287 |
+
"rewards/margins_min": -0.00417717918753624,
|
| 288 |
+
"rewards/margins_std": 0.00545117724686861,
|
| 289 |
+
"rewards/rejected": 0.011606132611632347,
|
| 290 |
+
"step": 120
|
| 291 |
+
},
|
| 292 |
+
{
|
| 293 |
+
"dpo_losses": 0.6912622451782227,
|
| 294 |
+
"epoch": 0.4,
|
| 295 |
+
"grad_norm": 1.6295192671904115,
|
| 296 |
+
"learning_rate": 4.984979933369709e-07,
|
| 297 |
+
"logits/chosen": -2.9324135780334473,
|
| 298 |
+
"logits/rejected": -2.867285966873169,
|
| 299 |
+
"logps/chosen": -442.0322265625,
|
| 300 |
+
"logps/rejected": -269.4384765625,
|
| 301 |
+
"loss": 0.6928,
|
| 302 |
+
"positive_losses": 0.0,
|
| 303 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 304 |
+
"rewards/chosen": 0.01560944877564907,
|
| 305 |
+
"rewards/margins": 0.0037853687535971403,
|
| 306 |
+
"rewards/margins_max": 0.007000140845775604,
|
| 307 |
+
"rewards/margins_min": 0.0005705966614186764,
|
| 308 |
+
"rewards/margins_std": 0.004546374548226595,
|
| 309 |
+
"rewards/rejected": 0.011824080720543861,
|
| 310 |
+
"step": 130
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"dpo_losses": 0.6920586824417114,
|
| 314 |
+
"epoch": 0.43,
|
| 315 |
+
"grad_norm": 1.9106724203537326,
|
| 316 |
+
"learning_rate": 4.973746334285336e-07,
|
| 317 |
+
"logits/chosen": -2.803515911102295,
|
| 318 |
+
"logits/rejected": -2.8981337547302246,
|
| 319 |
+
"logps/chosen": -340.9748229980469,
|
| 320 |
+
"logps/rejected": -325.1402282714844,
|
| 321 |
+
"loss": 0.6917,
|
| 322 |
+
"positive_losses": 0.0,
|
| 323 |
+
"rewards/accuracies": 0.5,
|
| 324 |
+
"rewards/chosen": 0.01538917887955904,
|
| 325 |
+
"rewards/margins": 0.002193502150475979,
|
| 326 |
+
"rewards/margins_max": 0.005477132275700569,
|
| 327 |
+
"rewards/margins_min": -0.001090128207579255,
|
| 328 |
+
"rewards/margins_std": 0.004643755033612251,
|
| 329 |
+
"rewards/rejected": 0.013195676729083061,
|
| 330 |
+
"step": 140
|
| 331 |
+
},
|
| 332 |
+
{
|
| 333 |
+
"dpo_losses": 0.6921483278274536,
|
| 334 |
+
"epoch": 0.46,
|
| 335 |
+
"grad_norm": 1.4481594604817483,
|
| 336 |
+
"learning_rate": 4.959416858332709e-07,
|
| 337 |
+
"logits/chosen": -2.855631113052368,
|
| 338 |
+
"logits/rejected": -2.820399045944214,
|
| 339 |
+
"logps/chosen": -271.1258544921875,
|
| 340 |
+
"logps/rejected": -282.1231689453125,
|
| 341 |
+
"loss": 0.6932,
|
| 342 |
+
"positive_losses": 0.0,
|
| 343 |
+
"rewards/accuracies": 0.699999988079071,
|
| 344 |
+
"rewards/chosen": 0.014782702550292015,
|
| 345 |
+
"rewards/margins": 0.002010155702009797,
|
| 346 |
+
"rewards/margins_max": 0.00527102081105113,
|
| 347 |
+
"rewards/margins_min": -0.001250709407031536,
|
| 348 |
+
"rewards/margins_std": 0.004611559212207794,
|
| 349 |
+
"rewards/rejected": 0.012772548012435436,
|
| 350 |
+
"step": 150
|
| 351 |
+
},
|
| 352 |
+
{
|
| 353 |
+
"dpo_losses": 0.6921746730804443,
|
| 354 |
+
"epoch": 0.49,
|
| 355 |
+
"grad_norm": 2.122964894274462,
|
| 356 |
+
"learning_rate": 4.942009438754421e-07,
|
| 357 |
+
"logits/chosen": -2.8902671337127686,
|
| 358 |
+
"logits/rejected": -2.875211238861084,
|
| 359 |
+
"logps/chosen": -210.3759765625,
|
| 360 |
+
"logps/rejected": -202.38551330566406,
|
| 361 |
+
"loss": 0.6939,
|
| 362 |
+
"positive_losses": 0.0,
|
| 363 |
+
"rewards/accuracies": 0.44999998807907104,
|
| 364 |
+
"rewards/chosen": 0.014994035474956036,
|
| 365 |
+
"rewards/margins": 0.0019599986262619495,
|
| 366 |
+
"rewards/margins_max": 0.006691869348287582,
|
| 367 |
+
"rewards/margins_min": -0.0027718725614249706,
|
| 368 |
+
"rewards/margins_std": 0.006691875867545605,
|
| 369 |
+
"rewards/rejected": 0.013034040108323097,
|
| 370 |
+
"step": 160
|
| 371 |
+
},
|
| 372 |
+
{
|
| 373 |
+
"dpo_losses": 0.690701961517334,
|
| 374 |
+
"epoch": 0.52,
|
| 375 |
+
"grad_norm": 1.585863985509441,
|
| 376 |
+
"learning_rate": 4.921545860818778e-07,
|
| 377 |
+
"logits/chosen": -2.8698291778564453,
|
| 378 |
+
"logits/rejected": -2.769212484359741,
|
| 379 |
+
"logps/chosen": -274.9241943359375,
|
| 380 |
+
"logps/rejected": -276.3659973144531,
|
| 381 |
+
"loss": 0.6909,
|
| 382 |
+
"positive_losses": 0.0,
|
| 383 |
+
"rewards/accuracies": 0.75,
|
| 384 |
+
"rewards/chosen": 0.019051264971494675,
|
| 385 |
+
"rewards/margins": 0.004921003244817257,
|
| 386 |
+
"rewards/margins_max": 0.009086159989237785,
|
| 387 |
+
"rewards/margins_min": 0.0007558457436971366,
|
| 388 |
+
"rewards/margins_std": 0.005890422035008669,
|
| 389 |
+
"rewards/rejected": 0.014130261726677418,
|
| 390 |
+
"step": 170
|
| 391 |
+
},
|
| 392 |
+
{
|
| 393 |
+
"dpo_losses": 0.692548394203186,
|
| 394 |
+
"epoch": 0.55,
|
| 395 |
+
"grad_norm": 8.258786695285371,
|
| 396 |
+
"learning_rate": 4.898051734555674e-07,
|
| 397 |
+
"logits/chosen": -2.8515067100524902,
|
| 398 |
+
"logits/rejected": -2.865828037261963,
|
| 399 |
+
"logps/chosen": -193.68429565429688,
|
| 400 |
+
"logps/rejected": -217.70962524414062,
|
| 401 |
+
"loss": 0.6921,
|
| 402 |
+
"positive_losses": 0.0,
|
| 403 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 404 |
+
"rewards/chosen": 0.016250569373369217,
|
| 405 |
+
"rewards/margins": 0.0012078474974259734,
|
| 406 |
+
"rewards/margins_max": 0.0032145425211638212,
|
| 407 |
+
"rewards/margins_min": -0.0007988469442352653,
|
| 408 |
+
"rewards/margins_std": 0.002837894717231393,
|
| 409 |
+
"rewards/rejected": 0.01504272036254406,
|
| 410 |
+
"step": 180
|
| 411 |
+
},
|
| 412 |
+
{
|
| 413 |
+
"dpo_losses": 0.6910722255706787,
|
| 414 |
+
"epoch": 0.58,
|
| 415 |
+
"grad_norm": 9.556509896855202,
|
| 416 |
+
"learning_rate": 4.871556462705816e-07,
|
| 417 |
+
"logits/chosen": -2.663051128387451,
|
| 418 |
+
"logits/rejected": -2.8675954341888428,
|
| 419 |
+
"logps/chosen": -174.3250732421875,
|
| 420 |
+
"logps/rejected": -239.2481231689453,
|
| 421 |
+
"loss": 0.692,
|
| 422 |
+
"positive_losses": 0.0,
|
| 423 |
+
"rewards/accuracies": 0.5,
|
| 424 |
+
"rewards/chosen": 0.020512599498033524,
|
| 425 |
+
"rewards/margins": 0.004181156400591135,
|
| 426 |
+
"rewards/margins_max": 0.009443456307053566,
|
| 427 |
+
"rewards/margins_min": -0.0010811425745487213,
|
| 428 |
+
"rewards/margins_std": 0.007442014757543802,
|
| 429 |
+
"rewards/rejected": 0.016331443563103676,
|
| 430 |
+
"step": 190
|
| 431 |
+
},
|
| 432 |
+
{
|
| 433 |
+
"dpo_losses": 0.6924042701721191,
|
| 434 |
+
"epoch": 0.61,
|
| 435 |
+
"grad_norm": 2.1686233219670368,
|
| 436 |
+
"learning_rate": 4.842093203923387e-07,
|
| 437 |
+
"logits/chosen": -2.906074047088623,
|
| 438 |
+
"logits/rejected": -2.8612585067749023,
|
| 439 |
+
"logps/chosen": -219.8831787109375,
|
| 440 |
+
"logps/rejected": -310.6004638671875,
|
| 441 |
+
"loss": 0.6919,
|
| 442 |
+
"positive_losses": 0.0378265380859375,
|
| 443 |
+
"rewards/accuracies": 0.5,
|
| 444 |
+
"rewards/chosen": 0.02058028057217598,
|
| 445 |
+
"rewards/margins": 0.0015535615384578705,
|
| 446 |
+
"rewards/margins_max": 0.009957343339920044,
|
| 447 |
+
"rewards/margins_min": -0.006850218866020441,
|
| 448 |
+
"rewards/margins_std": 0.011884740553796291,
|
| 449 |
+
"rewards/rejected": 0.01902671717107296,
|
| 450 |
+
"step": 200
|
| 451 |
+
},
|
| 452 |
+
{
|
| 453 |
+
"epoch": 0.61,
|
| 454 |
+
"eval_dpo_losses": 0.6904077529907227,
|
| 455 |
+
"eval_logits/chosen": -2.858304977416992,
|
| 456 |
+
"eval_logits/rejected": -2.813277244567871,
|
| 457 |
+
"eval_logps/chosen": -282.7998962402344,
|
| 458 |
+
"eval_logps/rejected": -257.31414794921875,
|
| 459 |
+
"eval_loss": 0.6915329694747925,
|
| 460 |
+
"eval_positive_losses": 0.01193891279399395,
|
| 461 |
+
"eval_rewards/accuracies": 0.6666666865348816,
|
| 462 |
+
"eval_rewards/chosen": 0.024213315919041634,
|
| 463 |
+
"eval_rewards/margins": 0.0055284094996750355,
|
| 464 |
+
"eval_rewards/margins_max": 0.019484752789139748,
|
| 465 |
+
"eval_rewards/margins_min": -0.008329018950462341,
|
| 466 |
+
"eval_rewards/margins_std": 0.012371117249131203,
|
| 467 |
+
"eval_rewards/rejected": 0.018684905022382736,
|
| 468 |
+
"eval_runtime": 282.1184,
|
| 469 |
+
"eval_samples_per_second": 7.089,
|
| 470 |
+
"eval_steps_per_second": 0.223,
|
| 471 |
+
"step": 200
|
| 472 |
+
},
|
| 473 |
+
{
|
| 474 |
+
"dpo_losses": 0.6858643293380737,
|
| 475 |
+
"epoch": 0.64,
|
| 476 |
+
"grad_norm": 1.962735267678132,
|
| 477 |
+
"learning_rate": 4.809698831278217e-07,
|
| 478 |
+
"logits/chosen": -2.9653470516204834,
|
| 479 |
+
"logits/rejected": -2.9085001945495605,
|
| 480 |
+
"logps/chosen": -375.43804931640625,
|
| 481 |
+
"logps/rejected": -293.70599365234375,
|
| 482 |
+
"loss": 0.6913,
|
| 483 |
+
"positive_losses": 0.0,
|
| 484 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 485 |
+
"rewards/chosen": 0.032657068222761154,
|
| 486 |
+
"rewards/margins": 0.014679856598377228,
|
| 487 |
+
"rewards/margins_max": 0.026673411950469017,
|
| 488 |
+
"rewards/margins_min": 0.002686301711946726,
|
| 489 |
+
"rewards/margins_std": 0.01696144789457321,
|
| 490 |
+
"rewards/rejected": 0.017977211624383926,
|
| 491 |
+
"step": 210
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"dpo_losses": 0.6917377710342407,
|
| 495 |
+
"epoch": 0.67,
|
| 496 |
+
"grad_norm": 1.9665123469481351,
|
| 497 |
+
"learning_rate": 4.774413886109386e-07,
|
| 498 |
+
"logits/chosen": -2.9416866302490234,
|
| 499 |
+
"logits/rejected": -2.848280429840088,
|
| 500 |
+
"logps/chosen": -281.5489807128906,
|
| 501 |
+
"logps/rejected": -228.3254852294922,
|
| 502 |
+
"loss": 0.6904,
|
| 503 |
+
"positive_losses": 0.0,
|
| 504 |
+
"rewards/accuracies": 0.550000011920929,
|
| 505 |
+
"rewards/chosen": 0.024796273559331894,
|
| 506 |
+
"rewards/margins": 0.002851649885997176,
|
| 507 |
+
"rewards/margins_max": 0.00997900776565075,
|
| 508 |
+
"rewards/margins_min": -0.004275707062333822,
|
| 509 |
+
"rewards/margins_std": 0.010079605504870415,
|
| 510 |
+
"rewards/rejected": 0.021944623440504074,
|
| 511 |
+
"step": 220
|
| 512 |
+
},
|
| 513 |
+
{
|
| 514 |
+
"dpo_losses": 0.6900176405906677,
|
| 515 |
+
"epoch": 0.7,
|
| 516 |
+
"grad_norm": 8.719769818128936,
|
| 517 |
+
"learning_rate": 4.736282527288008e-07,
|
| 518 |
+
"logits/chosen": -2.9049994945526123,
|
| 519 |
+
"logits/rejected": -2.930379629135132,
|
| 520 |
+
"logps/chosen": -204.6458740234375,
|
| 521 |
+
"logps/rejected": -201.35501098632812,
|
| 522 |
+
"loss": 0.6932,
|
| 523 |
+
"positive_losses": 0.0,
|
| 524 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 525 |
+
"rewards/chosen": 0.02477971278131008,
|
| 526 |
+
"rewards/margins": 0.006342197768390179,
|
| 527 |
+
"rewards/margins_max": 0.015214154496788979,
|
| 528 |
+
"rewards/margins_min": -0.0025297594256699085,
|
| 529 |
+
"rewards/margins_std": 0.012546841986477375,
|
| 530 |
+
"rewards/rejected": 0.018437515944242477,
|
| 531 |
+
"step": 230
|
| 532 |
+
},
|
| 533 |
+
{
|
| 534 |
+
"dpo_losses": 0.6900538802146912,
|
| 535 |
+
"epoch": 0.73,
|
| 536 |
+
"grad_norm": 2.0583669693189552,
|
| 537 |
+
"learning_rate": 4.6953524759527053e-07,
|
| 538 |
+
"logits/chosen": -3.0629992485046387,
|
| 539 |
+
"logits/rejected": -2.9669101238250732,
|
| 540 |
+
"logps/chosen": -341.70391845703125,
|
| 541 |
+
"logps/rejected": -327.5718078613281,
|
| 542 |
+
"loss": 0.6911,
|
| 543 |
+
"positive_losses": 0.0,
|
| 544 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 545 |
+
"rewards/chosen": 0.029439503327012062,
|
| 546 |
+
"rewards/margins": 0.006293328013271093,
|
| 547 |
+
"rewards/margins_max": 0.019671425223350525,
|
| 548 |
+
"rewards/margins_min": -0.0070847696624696255,
|
| 549 |
+
"rewards/margins_std": 0.018919486552476883,
|
| 550 |
+
"rewards/rejected": 0.023146172985434532,
|
| 551 |
+
"step": 240
|
| 552 |
+
},
|
| 553 |
+
{
|
| 554 |
+
"dpo_losses": 0.6918362379074097,
|
| 555 |
+
"epoch": 0.76,
|
| 556 |
+
"grad_norm": 1.9864876382070107,
|
| 557 |
+
"learning_rate": 4.6516749557869185e-07,
|
| 558 |
+
"logits/chosen": -2.9660696983337402,
|
| 559 |
+
"logits/rejected": -2.871572494506836,
|
| 560 |
+
"logps/chosen": -272.2618408203125,
|
| 561 |
+
"logps/rejected": -243.72860717773438,
|
| 562 |
+
"loss": 0.6926,
|
| 563 |
+
"positive_losses": 0.0,
|
| 564 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 565 |
+
"rewards/chosen": 0.030278582125902176,
|
| 566 |
+
"rewards/margins": 0.002722159493714571,
|
| 567 |
+
"rewards/margins_max": 0.014538141898810863,
|
| 568 |
+
"rewards/margins_min": -0.009093823842704296,
|
| 569 |
+
"rewards/margins_std": 0.016710322350263596,
|
| 570 |
+
"rewards/rejected": 0.027556419372558594,
|
| 571 |
+
"step": 250
|
| 572 |
+
},
|
| 573 |
+
{
|
| 574 |
+
"dpo_losses": 0.6899689435958862,
|
| 575 |
+
"epoch": 0.79,
|
| 576 |
+
"grad_norm": 2.12838016274278,
|
| 577 |
+
"learning_rate": 4.6053046289128194e-07,
|
| 578 |
+
"logits/chosen": -2.956109046936035,
|
| 579 |
+
"logits/rejected": -2.9426538944244385,
|
| 580 |
+
"logps/chosen": -241.18115234375,
|
| 581 |
+
"logps/rejected": -204.5635528564453,
|
| 582 |
+
"loss": 0.6896,
|
| 583 |
+
"positive_losses": 0.00806427001953125,
|
| 584 |
+
"rewards/accuracies": 0.550000011920929,
|
| 585 |
+
"rewards/chosen": 0.029516180977225304,
|
| 586 |
+
"rewards/margins": 0.006419041194021702,
|
| 587 |
+
"rewards/margins_max": 0.014670031145215034,
|
| 588 |
+
"rewards/margins_min": -0.0018319508526474237,
|
| 589 |
+
"rewards/margins_std": 0.011668664403259754,
|
| 590 |
+
"rewards/rejected": 0.023097142577171326,
|
| 591 |
+
"step": 260
|
| 592 |
+
},
|
| 593 |
+
{
|
| 594 |
+
"dpo_losses": 0.6896145343780518,
|
| 595 |
+
"epoch": 0.82,
|
| 596 |
+
"grad_norm": 2.012252233192398,
|
| 597 |
+
"learning_rate": 4.5562995274820283e-07,
|
| 598 |
+
"logits/chosen": -2.819646120071411,
|
| 599 |
+
"logits/rejected": -2.845240354537964,
|
| 600 |
+
"logps/chosen": -307.28582763671875,
|
| 601 |
+
"logps/rejected": -313.50897216796875,
|
| 602 |
+
"loss": 0.6913,
|
| 603 |
+
"positive_losses": 0.06691131740808487,
|
| 604 |
+
"rewards/accuracies": 0.550000011920929,
|
| 605 |
+
"rewards/chosen": 0.025536488741636276,
|
| 606 |
+
"rewards/margins": 0.007166997995227575,
|
| 607 |
+
"rewards/margins_max": 0.01585778035223484,
|
| 608 |
+
"rewards/margins_min": -0.0015237818006426096,
|
| 609 |
+
"rewards/margins_std": 0.012290618382394314,
|
| 610 |
+
"rewards/rejected": 0.018369488418102264,
|
| 611 |
+
"step": 270
|
| 612 |
+
},
|
| 613 |
+
{
|
| 614 |
+
"dpo_losses": 0.6887096166610718,
|
| 615 |
+
"epoch": 0.85,
|
| 616 |
+
"grad_norm": 2.181123977999317,
|
| 617 |
+
"learning_rate": 4.5047209810487705e-07,
|
| 618 |
+
"logits/chosen": -2.971086025238037,
|
| 619 |
+
"logits/rejected": -2.9251911640167236,
|
| 620 |
+
"logps/chosen": -351.91864013671875,
|
| 621 |
+
"logps/rejected": -323.68658447265625,
|
| 622 |
+
"loss": 0.6897,
|
| 623 |
+
"positive_losses": 0.03092041052877903,
|
| 624 |
+
"rewards/accuracies": 0.699999988079071,
|
| 625 |
+
"rewards/chosen": 0.04124215617775917,
|
| 626 |
+
"rewards/margins": 0.009057796560227871,
|
| 627 |
+
"rewards/margins_max": 0.017794614657759666,
|
| 628 |
+
"rewards/margins_min": 0.00032097622170113027,
|
| 629 |
+
"rewards/margins_std": 0.012355729006230831,
|
| 630 |
+
"rewards/rejected": 0.032184358686208725,
|
| 631 |
+
"step": 280
|
| 632 |
+
},
|
| 633 |
+
{
|
| 634 |
+
"dpo_losses": 0.6910039186477661,
|
| 635 |
+
"epoch": 0.88,
|
| 636 |
+
"grad_norm": 1.7264365586137154,
|
| 637 |
+
"learning_rate": 4.450633539816351e-07,
|
| 638 |
+
"logits/chosen": -2.669238567352295,
|
| 639 |
+
"logits/rejected": -2.6185073852539062,
|
| 640 |
+
"logps/chosen": -229.40542602539062,
|
| 641 |
+
"logps/rejected": -268.8614501953125,
|
| 642 |
+
"loss": 0.6898,
|
| 643 |
+
"positive_losses": 0.0014183043967932463,
|
| 644 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 645 |
+
"rewards/chosen": 0.029668454080820084,
|
| 646 |
+
"rewards/margins": 0.00436769612133503,
|
| 647 |
+
"rewards/margins_max": 0.01580682210624218,
|
| 648 |
+
"rewards/margins_min": -0.007071429397910833,
|
| 649 |
+
"rewards/margins_std": 0.01617736741900444,
|
| 650 |
+
"rewards/rejected": 0.025300756096839905,
|
| 651 |
+
"step": 290
|
| 652 |
+
},
|
| 653 |
+
{
|
| 654 |
+
"dpo_losses": 0.688437819480896,
|
| 655 |
+
"epoch": 0.91,
|
| 656 |
+
"grad_norm": 7.040817600804112,
|
| 657 |
+
"learning_rate": 4.394104893853007e-07,
|
| 658 |
+
"logits/chosen": -3.0286426544189453,
|
| 659 |
+
"logits/rejected": -2.9285566806793213,
|
| 660 |
+
"logps/chosen": -266.41265869140625,
|
| 661 |
+
"logps/rejected": -276.48675537109375,
|
| 662 |
+
"loss": 0.6903,
|
| 663 |
+
"positive_losses": 0.08214263617992401,
|
| 664 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 665 |
+
"rewards/chosen": 0.037517085671424866,
|
| 666 |
+
"rewards/margins": 0.009616566821932793,
|
| 667 |
+
"rewards/margins_max": 0.027836177498102188,
|
| 668 |
+
"rewards/margins_min": -0.008603041991591454,
|
| 669 |
+
"rewards/margins_std": 0.025766421109437943,
|
| 670 |
+
"rewards/rejected": 0.027900516986846924,
|
| 671 |
+
"step": 300
|
| 672 |
+
},
|
| 673 |
+
{
|
| 674 |
+
"epoch": 0.91,
|
| 675 |
+
"eval_dpo_losses": 0.6876209378242493,
|
| 676 |
+
"eval_logits/chosen": -2.853738784790039,
|
| 677 |
+
"eval_logits/rejected": -2.8086090087890625,
|
| 678 |
+
"eval_logps/chosen": -281.2695007324219,
|
| 679 |
+
"eval_logps/rejected": -256.35443115234375,
|
| 680 |
+
"eval_loss": 0.6898627877235413,
|
| 681 |
+
"eval_positive_losses": 0.016462840139865875,
|
| 682 |
+
"eval_rewards/accuracies": 0.6666666865348816,
|
| 683 |
+
"eval_rewards/chosen": 0.03951713442802429,
|
| 684 |
+
"eval_rewards/margins": 0.011234736070036888,
|
| 685 |
+
"eval_rewards/margins_max": 0.0378512442111969,
|
| 686 |
+
"eval_rewards/margins_min": -0.014283931814134121,
|
| 687 |
+
"eval_rewards/margins_std": 0.023247137665748596,
|
| 688 |
+
"eval_rewards/rejected": 0.028282402083277702,
|
| 689 |
+
"eval_runtime": 281.1088,
|
| 690 |
+
"eval_samples_per_second": 7.115,
|
| 691 |
+
"eval_steps_per_second": 0.224,
|
| 692 |
+
"step": 300
|
| 693 |
+
},
|
| 694 |
+
{
|
| 695 |
+
"dpo_losses": 0.6858936548233032,
|
| 696 |
+
"epoch": 0.94,
|
| 697 |
+
"grad_norm": 1.534301986044828,
|
| 698 |
+
"learning_rate": 4.33520578837825e-07,
|
| 699 |
+
"logits/chosen": -2.7751994132995605,
|
| 700 |
+
"logits/rejected": -2.767688035964966,
|
| 701 |
+
"logps/chosen": -286.08197021484375,
|
| 702 |
+
"logps/rejected": -240.5784912109375,
|
| 703 |
+
"loss": 0.6894,
|
| 704 |
+
"positive_losses": 0.0,
|
| 705 |
+
"rewards/accuracies": 0.699999988079071,
|
| 706 |
+
"rewards/chosen": 0.04037940502166748,
|
| 707 |
+
"rewards/margins": 0.014671111479401588,
|
| 708 |
+
"rewards/margins_max": 0.025094876065850258,
|
| 709 |
+
"rewards/margins_min": 0.004247347824275494,
|
| 710 |
+
"rewards/margins_std": 0.014741426333785057,
|
| 711 |
+
"rewards/rejected": 0.02570829726755619,
|
| 712 |
+
"step": 310
|
| 713 |
+
},
|
| 714 |
+
{
|
| 715 |
+
"dpo_losses": 0.6850165128707886,
|
| 716 |
+
"epoch": 0.97,
|
| 717 |
+
"grad_norm": 1.9190817282141555,
|
| 718 |
+
"learning_rate": 4.2740099352256954e-07,
|
| 719 |
+
"logits/chosen": -2.9025392532348633,
|
| 720 |
+
"logits/rejected": -2.907529592514038,
|
| 721 |
+
"logps/chosen": -202.9761505126953,
|
| 722 |
+
"logps/rejected": -240.7790985107422,
|
| 723 |
+
"loss": 0.6868,
|
| 724 |
+
"positive_losses": 0.0,
|
| 725 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 726 |
+
"rewards/chosen": 0.04233746603131294,
|
| 727 |
+
"rewards/margins": 0.016391225159168243,
|
| 728 |
+
"rewards/margins_max": 0.023528441786766052,
|
| 729 |
+
"rewards/margins_min": 0.00925400760024786,
|
| 730 |
+
"rewards/margins_std": 0.010093549266457558,
|
| 731 |
+
"rewards/rejected": 0.02594623900949955,
|
| 732 |
+
"step": 320
|
| 733 |
+
},
|
| 734 |
+
{
|
| 735 |
+
"dpo_losses": 0.6829380989074707,
|
| 736 |
+
"epoch": 1.0,
|
| 737 |
+
"grad_norm": 2.017378170300957,
|
| 738 |
+
"learning_rate": 4.2105939205932005e-07,
|
| 739 |
+
"logits/chosen": -2.9684784412384033,
|
| 740 |
+
"logits/rejected": -2.8465845584869385,
|
| 741 |
+
"logps/chosen": -384.70379638671875,
|
| 742 |
+
"logps/rejected": -371.2561950683594,
|
| 743 |
+
"loss": 0.6893,
|
| 744 |
+
"positive_losses": 0.0,
|
| 745 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 746 |
+
"rewards/chosen": 0.056094102561473846,
|
| 747 |
+
"rewards/margins": 0.02067440189421177,
|
| 748 |
+
"rewards/margins_max": 0.03426554799079895,
|
| 749 |
+
"rewards/margins_min": 0.007083254866302013,
|
| 750 |
+
"rewards/margins_std": 0.019220784306526184,
|
| 751 |
+
"rewards/rejected": 0.03541969880461693,
|
| 752 |
+
"step": 330
|
| 753 |
+
},
|
| 754 |
+
{
|
| 755 |
+
"dpo_losses": 0.6854172348976135,
|
| 756 |
+
"epoch": 1.03,
|
| 757 |
+
"grad_norm": 2.002276198207298,
|
| 758 |
+
"learning_rate": 4.145037109195757e-07,
|
| 759 |
+
"logits/chosen": -2.9032974243164062,
|
| 760 |
+
"logits/rejected": -2.9131686687469482,
|
| 761 |
+
"logps/chosen": -265.29217529296875,
|
| 762 |
+
"logps/rejected": -192.96426391601562,
|
| 763 |
+
"loss": 0.6866,
|
| 764 |
+
"positive_losses": 0.01656951941549778,
|
| 765 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 766 |
+
"rewards/chosen": 0.044899724423885345,
|
| 767 |
+
"rewards/margins": 0.01578436978161335,
|
| 768 |
+
"rewards/margins_max": 0.03546427935361862,
|
| 769 |
+
"rewards/margins_min": -0.003895541187375784,
|
| 770 |
+
"rewards/margins_std": 0.02783159911632538,
|
| 771 |
+
"rewards/rejected": 0.029115352779626846,
|
| 772 |
+
"step": 340
|
| 773 |
+
},
|
| 774 |
+
{
|
| 775 |
+
"dpo_losses": 0.6908459067344666,
|
| 776 |
+
"epoch": 1.06,
|
| 777 |
+
"grad_norm": 1.847200301127122,
|
| 778 |
+
"learning_rate": 4.077421544941079e-07,
|
| 779 |
+
"logits/chosen": -2.8273587226867676,
|
| 780 |
+
"logits/rejected": -2.8343966007232666,
|
| 781 |
+
"logps/chosen": -225.12521362304688,
|
| 782 |
+
"logps/rejected": -211.3616180419922,
|
| 783 |
+
"loss": 0.6867,
|
| 784 |
+
"positive_losses": 0.089324951171875,
|
| 785 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 786 |
+
"rewards/chosen": 0.038552217185497284,
|
| 787 |
+
"rewards/margins": 0.004919026046991348,
|
| 788 |
+
"rewards/margins_max": 0.021458707749843597,
|
| 789 |
+
"rewards/margins_min": -0.011620650999248028,
|
| 790 |
+
"rewards/margins_std": 0.0233906377106905,
|
| 791 |
+
"rewards/rejected": 0.03363319858908653,
|
| 792 |
+
"step": 350
|
| 793 |
+
},
|
| 794 |
+
{
|
| 795 |
+
"dpo_losses": 0.6812031865119934,
|
| 796 |
+
"epoch": 1.09,
|
| 797 |
+
"grad_norm": 1.9045252266933426,
|
| 798 |
+
"learning_rate": 4.0078318482522114e-07,
|
| 799 |
+
"logits/chosen": -2.8550143241882324,
|
| 800 |
+
"logits/rejected": -2.851294755935669,
|
| 801 |
+
"logps/chosen": -236.28811645507812,
|
| 802 |
+
"logps/rejected": -254.1648406982422,
|
| 803 |
+
"loss": 0.6854,
|
| 804 |
+
"positive_losses": 0.0,
|
| 805 |
+
"rewards/accuracies": 0.699999988079071,
|
| 806 |
+
"rewards/chosen": 0.05590778589248657,
|
| 807 |
+
"rewards/margins": 0.024477163329720497,
|
| 808 |
+
"rewards/margins_max": 0.046524226665496826,
|
| 809 |
+
"rewards/margins_min": 0.0024301104713231325,
|
| 810 |
+
"rewards/margins_std": 0.031179243698716164,
|
| 811 |
+
"rewards/rejected": 0.031430620700120926,
|
| 812 |
+
"step": 360
|
| 813 |
+
},
|
| 814 |
+
{
|
| 815 |
+
"dpo_losses": 0.6793321371078491,
|
| 816 |
+
"epoch": 1.12,
|
| 817 |
+
"grad_norm": 2.2246778576418613,
|
| 818 |
+
"learning_rate": 3.9363551101656356e-07,
|
| 819 |
+
"logits/chosen": -2.8856711387634277,
|
| 820 |
+
"logits/rejected": -2.7337353229522705,
|
| 821 |
+
"logps/chosen": -360.27142333984375,
|
| 822 |
+
"logps/rejected": -211.5453338623047,
|
| 823 |
+
"loss": 0.6838,
|
| 824 |
+
"positive_losses": 0.0,
|
| 825 |
+
"rewards/accuracies": 0.75,
|
| 826 |
+
"rewards/chosen": 0.05444517731666565,
|
| 827 |
+
"rewards/margins": 0.02811727486550808,
|
| 828 |
+
"rewards/margins_max": 0.04928036779165268,
|
| 829 |
+
"rewards/margins_min": 0.006954180542379618,
|
| 830 |
+
"rewards/margins_std": 0.029929134994745255,
|
| 831 |
+
"rewards/rejected": 0.02632790245115757,
|
| 832 |
+
"step": 370
|
| 833 |
+
},
|
| 834 |
+
{
|
| 835 |
+
"dpo_losses": 0.6877564787864685,
|
| 836 |
+
"epoch": 1.16,
|
| 837 |
+
"grad_norm": 2.093111994313669,
|
| 838 |
+
"learning_rate": 3.8630807833374277e-07,
|
| 839 |
+
"logits/chosen": -2.9553494453430176,
|
| 840 |
+
"logits/rejected": -2.8332180976867676,
|
| 841 |
+
"logps/chosen": -270.87994384765625,
|
| 842 |
+
"logps/rejected": -317.43060302734375,
|
| 843 |
+
"loss": 0.6851,
|
| 844 |
+
"positive_losses": 0.0,
|
| 845 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 846 |
+
"rewards/chosen": 0.052814941853284836,
|
| 847 |
+
"rewards/margins": 0.011218393221497536,
|
| 848 |
+
"rewards/margins_max": 0.03882121667265892,
|
| 849 |
+
"rewards/margins_min": -0.01638443022966385,
|
| 850 |
+
"rewards/margins_std": 0.039036285132169724,
|
| 851 |
+
"rewards/rejected": 0.04159654304385185,
|
| 852 |
+
"step": 380
|
| 853 |
+
},
|
| 854 |
+
{
|
| 855 |
+
"dpo_losses": 0.6842631101608276,
|
| 856 |
+
"epoch": 1.19,
|
| 857 |
+
"grad_norm": 1.9478894060100704,
|
| 858 |
+
"learning_rate": 3.7881005700938627e-07,
|
| 859 |
+
"logits/chosen": -2.883104085922241,
|
| 860 |
+
"logits/rejected": -2.9553048610687256,
|
| 861 |
+
"logps/chosen": -286.1315002441406,
|
| 862 |
+
"logps/rejected": -273.46533203125,
|
| 863 |
+
"loss": 0.6845,
|
| 864 |
+
"positive_losses": 0.0,
|
| 865 |
+
"rewards/accuracies": 0.699999988079071,
|
| 866 |
+
"rewards/chosen": 0.061849962919950485,
|
| 867 |
+
"rewards/margins": 0.01832859218120575,
|
| 868 |
+
"rewards/margins_max": 0.04364844784140587,
|
| 869 |
+
"rewards/margins_min": -0.006991264875978231,
|
| 870 |
+
"rewards/margins_std": 0.03580768406391144,
|
| 871 |
+
"rewards/rejected": 0.043521374464035034,
|
| 872 |
+
"step": 390
|
| 873 |
+
},
|
| 874 |
+
{
|
| 875 |
+
"dpo_losses": 0.6858700513839722,
|
| 876 |
+
"epoch": 1.22,
|
| 877 |
+
"grad_norm": 2.027188274912056,
|
| 878 |
+
"learning_rate": 3.711508307666567e-07,
|
| 879 |
+
"logits/chosen": -2.8832974433898926,
|
| 880 |
+
"logits/rejected": -2.911393404006958,
|
| 881 |
+
"logps/chosen": -202.76405334472656,
|
| 882 |
+
"logps/rejected": -219.851318359375,
|
| 883 |
+
"loss": 0.6832,
|
| 884 |
+
"positive_losses": 0.0,
|
| 885 |
+
"rewards/accuracies": 0.699999988079071,
|
| 886 |
+
"rewards/chosen": 0.053437333554029465,
|
| 887 |
+
"rewards/margins": 0.014844384975731373,
|
| 888 |
+
"rewards/margins_max": 0.03350861743092537,
|
| 889 |
+
"rewards/margins_min": -0.0038198456168174744,
|
| 890 |
+
"rewards/margins_std": 0.02639520727097988,
|
| 891 |
+
"rewards/rejected": 0.038592949509620667,
|
| 892 |
+
"step": 400
|
| 893 |
+
},
|
| 894 |
+
{
|
| 895 |
+
"epoch": 1.22,
|
| 896 |
+
"eval_dpo_losses": 0.6846545934677124,
|
| 897 |
+
"eval_logits/chosen": -2.850677490234375,
|
| 898 |
+
"eval_logits/rejected": -2.8056881427764893,
|
| 899 |
+
"eval_logps/chosen": -279.9754638671875,
|
| 900 |
+
"eval_logps/rejected": -255.6741485595703,
|
| 901 |
+
"eval_loss": 0.6891601085662842,
|
| 902 |
+
"eval_positive_losses": 0.030417578294873238,
|
| 903 |
+
"eval_rewards/accuracies": 0.7023809552192688,
|
| 904 |
+
"eval_rewards/chosen": 0.052457116544246674,
|
| 905 |
+
"eval_rewards/margins": 0.01737220212817192,
|
| 906 |
+
"eval_rewards/margins_max": 0.05574695020914078,
|
| 907 |
+
"eval_rewards/margins_min": -0.019588936120271683,
|
| 908 |
+
"eval_rewards/margins_std": 0.03368803858757019,
|
| 909 |
+
"eval_rewards/rejected": 0.03508492186665535,
|
| 910 |
+
"eval_runtime": 281.5128,
|
| 911 |
+
"eval_samples_per_second": 7.104,
|
| 912 |
+
"eval_steps_per_second": 0.224,
|
| 913 |
+
"step": 400
|
| 914 |
+
},
|
| 915 |
+
{
|
| 916 |
+
"dpo_losses": 0.6754962205886841,
|
| 917 |
+
"epoch": 1.25,
|
| 918 |
+
"grad_norm": 1.8539817875046798,
|
| 919 |
+
"learning_rate": 3.6333998507558596e-07,
|
| 920 |
+
"logits/chosen": -2.7948994636535645,
|
| 921 |
+
"logits/rejected": -2.838921070098877,
|
| 922 |
+
"logps/chosen": -243.80331420898438,
|
| 923 |
+
"logps/rejected": -267.6720275878906,
|
| 924 |
+
"loss": 0.6841,
|
| 925 |
+
"positive_losses": 0.0,
|
| 926 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 927 |
+
"rewards/chosen": 0.061250943690538406,
|
| 928 |
+
"rewards/margins": 0.03602340444922447,
|
| 929 |
+
"rewards/margins_max": 0.05897093936800957,
|
| 930 |
+
"rewards/margins_min": 0.013075867667794228,
|
| 931 |
+
"rewards/margins_std": 0.032452717423439026,
|
| 932 |
+
"rewards/rejected": 0.025227541103959084,
|
| 933 |
+
"step": 410
|
| 934 |
+
},
|
| 935 |
+
{
|
| 936 |
+
"dpo_losses": 0.6807809472084045,
|
| 937 |
+
"epoch": 1.28,
|
| 938 |
+
"grad_norm": 6.364237058452619,
|
| 939 |
+
"learning_rate": 3.5538729515692354e-07,
|
| 940 |
+
"logits/chosen": -2.943338394165039,
|
| 941 |
+
"logits/rejected": -2.7816109657287598,
|
| 942 |
+
"logps/chosen": -335.1874084472656,
|
| 943 |
+
"logps/rejected": -279.122802734375,
|
| 944 |
+
"loss": 0.6895,
|
| 945 |
+
"positive_losses": 0.0,
|
| 946 |
+
"rewards/accuracies": 0.699999988079071,
|
| 947 |
+
"rewards/chosen": 0.06752719730138779,
|
| 948 |
+
"rewards/margins": 0.025191819295287132,
|
| 949 |
+
"rewards/margins_max": 0.04893922060728073,
|
| 950 |
+
"rewards/margins_min": 0.001444423571228981,
|
| 951 |
+
"rewards/margins_std": 0.03358389437198639,
|
| 952 |
+
"rewards/rejected": 0.04233537241816521,
|
| 953 |
+
"step": 420
|
| 954 |
+
},
|
| 955 |
+
{
|
| 956 |
+
"dpo_losses": 0.6775410175323486,
|
| 957 |
+
"epoch": 1.31,
|
| 958 |
+
"grad_norm": 2.1288430860755887,
|
| 959 |
+
"learning_rate": 3.4730271374851453e-07,
|
| 960 |
+
"logits/chosen": -2.8643321990966797,
|
| 961 |
+
"logits/rejected": -2.8559885025024414,
|
| 962 |
+
"logps/chosen": -335.872802734375,
|
| 963 |
+
"logps/rejected": -249.48501586914062,
|
| 964 |
+
"loss": 0.6787,
|
| 965 |
+
"positive_losses": 0.0,
|
| 966 |
+
"rewards/accuracies": 0.699999988079071,
|
| 967 |
+
"rewards/chosen": 0.06267213821411133,
|
| 968 |
+
"rewards/margins": 0.03193260729312897,
|
| 969 |
+
"rewards/margins_max": 0.05750267580151558,
|
| 970 |
+
"rewards/margins_min": 0.006362536456435919,
|
| 971 |
+
"rewards/margins_std": 0.03616153448820114,
|
| 972 |
+
"rewards/rejected": 0.03073953464627266,
|
| 973 |
+
"step": 430
|
| 974 |
+
},
|
| 975 |
+
{
|
| 976 |
+
"dpo_losses": 0.6760178208351135,
|
| 977 |
+
"epoch": 1.34,
|
| 978 |
+
"grad_norm": 2.302278973535758,
|
| 979 |
+
"learning_rate": 3.390963586495142e-07,
|
| 980 |
+
"logits/chosen": -2.7527740001678467,
|
| 981 |
+
"logits/rejected": -2.7448365688323975,
|
| 982 |
+
"logps/chosen": -291.7049560546875,
|
| 983 |
+
"logps/rejected": -197.5590362548828,
|
| 984 |
+
"loss": 0.6836,
|
| 985 |
+
"positive_losses": 0.0,
|
| 986 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 987 |
+
"rewards/chosen": 0.06701546162366867,
|
| 988 |
+
"rewards/margins": 0.03498576581478119,
|
| 989 |
+
"rewards/margins_max": 0.05517752841114998,
|
| 990 |
+
"rewards/margins_min": 0.01479400135576725,
|
| 991 |
+
"rewards/margins_std": 0.02855546772480011,
|
| 992 |
+
"rewards/rejected": 0.03202969953417778,
|
| 993 |
+
"step": 440
|
| 994 |
+
},
|
| 995 |
+
{
|
| 996 |
+
"dpo_losses": 0.6835552453994751,
|
| 997 |
+
"epoch": 1.37,
|
| 998 |
+
"grad_norm": 7.531724083885525,
|
| 999 |
+
"learning_rate": 3.3077850005803125e-07,
|
| 1000 |
+
"logits/chosen": -3.047640562057495,
|
| 1001 |
+
"logits/rejected": -2.9399123191833496,
|
| 1002 |
+
"logps/chosen": -273.14666748046875,
|
| 1003 |
+
"logps/rejected": -232.124267578125,
|
| 1004 |
+
"loss": 0.6811,
|
| 1005 |
+
"positive_losses": 0.0,
|
| 1006 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1007 |
+
"rewards/chosen": 0.05829363316297531,
|
| 1008 |
+
"rewards/margins": 0.019574137404561043,
|
| 1009 |
+
"rewards/margins_max": 0.040946681052446365,
|
| 1010 |
+
"rewards/margins_min": -0.001798404031433165,
|
| 1011 |
+
"rewards/margins_std": 0.03022533655166626,
|
| 1012 |
+
"rewards/rejected": 0.03871949762105942,
|
| 1013 |
+
"step": 450
|
| 1014 |
+
},
|
| 1015 |
+
{
|
| 1016 |
+
"dpo_losses": 0.6802719831466675,
|
| 1017 |
+
"epoch": 1.4,
|
| 1018 |
+
"grad_norm": 2.1783384024286963,
|
| 1019 |
+
"learning_rate": 3.2235954771804384e-07,
|
| 1020 |
+
"logits/chosen": -2.899862051010132,
|
| 1021 |
+
"logits/rejected": -2.8675050735473633,
|
| 1022 |
+
"logps/chosen": -288.083740234375,
|
| 1023 |
+
"logps/rejected": -268.29876708984375,
|
| 1024 |
+
"loss": 0.683,
|
| 1025 |
+
"positive_losses": 0.0,
|
| 1026 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1027 |
+
"rewards/chosen": 0.0625755786895752,
|
| 1028 |
+
"rewards/margins": 0.026135718449950218,
|
| 1029 |
+
"rewards/margins_max": 0.04849334806203842,
|
| 1030 |
+
"rewards/margins_min": 0.0037780862767249346,
|
| 1031 |
+
"rewards/margins_std": 0.03161846846342087,
|
| 1032 |
+
"rewards/rejected": 0.03643985837697983,
|
| 1033 |
+
"step": 460
|
| 1034 |
+
},
|
| 1035 |
+
{
|
| 1036 |
+
"dpo_losses": 0.6933653354644775,
|
| 1037 |
+
"epoch": 1.43,
|
| 1038 |
+
"grad_norm": 6.179682870277185,
|
| 1039 |
+
"learning_rate": 3.138500378916752e-07,
|
| 1040 |
+
"logits/chosen": -2.8621251583099365,
|
| 1041 |
+
"logits/rejected": -2.8821539878845215,
|
| 1042 |
+
"logps/chosen": -181.8573455810547,
|
| 1043 |
+
"logps/rejected": -222.8794708251953,
|
| 1044 |
+
"loss": 0.6844,
|
| 1045 |
+
"positive_losses": 0.0033210753463208675,
|
| 1046 |
+
"rewards/accuracies": 0.44999998807907104,
|
| 1047 |
+
"rewards/chosen": 0.04994913935661316,
|
| 1048 |
+
"rewards/margins": -5.03659236983367e-07,
|
| 1049 |
+
"rewards/margins_max": 0.01866455376148224,
|
| 1050 |
+
"rewards/margins_min": -0.018665555864572525,
|
| 1051 |
+
"rewards/margins_std": 0.0263963732868433,
|
| 1052 |
+
"rewards/rejected": 0.04994964599609375,
|
| 1053 |
+
"step": 470
|
| 1054 |
+
},
|
| 1055 |
+
{
|
| 1056 |
+
"dpo_losses": 0.6902952790260315,
|
| 1057 |
+
"epoch": 1.46,
|
| 1058 |
+
"grad_norm": 2.096436783641514,
|
| 1059 |
+
"learning_rate": 3.0526062017313247e-07,
|
| 1060 |
+
"logits/chosen": -2.8722407817840576,
|
| 1061 |
+
"logits/rejected": -2.9027137756347656,
|
| 1062 |
+
"logps/chosen": -245.9871826171875,
|
| 1063 |
+
"logps/rejected": -332.47283935546875,
|
| 1064 |
+
"loss": 0.6832,
|
| 1065 |
+
"positive_losses": 0.0,
|
| 1066 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 1067 |
+
"rewards/chosen": 0.0516946017742157,
|
| 1068 |
+
"rewards/margins": 0.006326904986053705,
|
| 1069 |
+
"rewards/margins_max": 0.03602270036935806,
|
| 1070 |
+
"rewards/margins_min": -0.023368891328573227,
|
| 1071 |
+
"rewards/margins_std": 0.041996195912361145,
|
| 1072 |
+
"rewards/rejected": 0.04536769539117813,
|
| 1073 |
+
"step": 480
|
| 1074 |
+
},
|
| 1075 |
+
{
|
| 1076 |
+
"dpo_losses": 0.6820996999740601,
|
| 1077 |
+
"epoch": 1.49,
|
| 1078 |
+
"grad_norm": 2.214654285326634,
|
| 1079 |
+
"learning_rate": 2.966020441608114e-07,
|
| 1080 |
+
"logits/chosen": -2.940601110458374,
|
| 1081 |
+
"logits/rejected": -2.950160026550293,
|
| 1082 |
+
"logps/chosen": -408.1592712402344,
|
| 1083 |
+
"logps/rejected": -367.73052978515625,
|
| 1084 |
+
"loss": 0.683,
|
| 1085 |
+
"positive_losses": 0.10776062309741974,
|
| 1086 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1087 |
+
"rewards/chosen": 0.06694754213094711,
|
| 1088 |
+
"rewards/margins": 0.02292126789689064,
|
| 1089 |
+
"rewards/margins_max": 0.04947842285037041,
|
| 1090 |
+
"rewards/margins_min": -0.0036358858924359083,
|
| 1091 |
+
"rewards/margins_std": 0.037557490170001984,
|
| 1092 |
+
"rewards/rejected": 0.04402626305818558,
|
| 1093 |
+
"step": 490
|
| 1094 |
+
},
|
| 1095 |
+
{
|
| 1096 |
+
"dpo_losses": 0.6780664324760437,
|
| 1097 |
+
"epoch": 1.52,
|
| 1098 |
+
"grad_norm": 2.3489902153022464,
|
| 1099 |
+
"learning_rate": 2.8788514600424653e-07,
|
| 1100 |
+
"logits/chosen": -2.9777138233184814,
|
| 1101 |
+
"logits/rejected": -2.8717598915100098,
|
| 1102 |
+
"logps/chosen": -271.792236328125,
|
| 1103 |
+
"logps/rejected": -217.39794921875,
|
| 1104 |
+
"loss": 0.6776,
|
| 1105 |
+
"positive_losses": 0.04692535474896431,
|
| 1106 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1107 |
+
"rewards/chosen": 0.07212454080581665,
|
| 1108 |
+
"rewards/margins": 0.031092515215277672,
|
| 1109 |
+
"rewards/margins_max": 0.05294063687324524,
|
| 1110 |
+
"rewards/margins_min": 0.009244391694664955,
|
| 1111 |
+
"rewards/margins_std": 0.03089791163802147,
|
| 1112 |
+
"rewards/rejected": 0.04103202372789383,
|
| 1113 |
+
"step": 500
|
| 1114 |
+
},
|
| 1115 |
+
{
|
| 1116 |
+
"epoch": 1.52,
|
| 1117 |
+
"eval_dpo_losses": 0.6824542880058289,
|
| 1118 |
+
"eval_logits/chosen": -2.8494935035705566,
|
| 1119 |
+
"eval_logits/rejected": -2.8047306537628174,
|
| 1120 |
+
"eval_logps/chosen": -278.7507629394531,
|
| 1121 |
+
"eval_logps/rejected": -254.91441345214844,
|
| 1122 |
+
"eval_loss": 0.6883726716041565,
|
| 1123 |
+
"eval_positive_losses": 0.04435727000236511,
|
| 1124 |
+
"eval_rewards/accuracies": 0.6904761791229248,
|
| 1125 |
+
"eval_rewards/chosen": 0.06470431387424469,
|
| 1126 |
+
"eval_rewards/margins": 0.02202197164297104,
|
| 1127 |
+
"eval_rewards/margins_max": 0.07098579406738281,
|
| 1128 |
+
"eval_rewards/margins_min": -0.025560587644577026,
|
| 1129 |
+
"eval_rewards/margins_std": 0.04328176751732826,
|
| 1130 |
+
"eval_rewards/rejected": 0.042682331055402756,
|
| 1131 |
+
"eval_runtime": 282.2305,
|
| 1132 |
+
"eval_samples_per_second": 7.086,
|
| 1133 |
+
"eval_steps_per_second": 0.223,
|
| 1134 |
+
"step": 500
|
| 1135 |
+
},
|
| 1136 |
+
{
|
| 1137 |
+
"dpo_losses": 0.684786319732666,
|
| 1138 |
+
"epoch": 1.55,
|
| 1139 |
+
"grad_norm": 1.8216734943221224,
|
| 1140 |
+
"learning_rate": 2.791208348427426e-07,
|
| 1141 |
+
"logits/chosen": -2.806025981903076,
|
| 1142 |
+
"logits/rejected": -2.7332119941711426,
|
| 1143 |
+
"logps/chosen": -350.2364501953125,
|
| 1144 |
+
"logps/rejected": -264.30572509765625,
|
| 1145 |
+
"loss": 0.6831,
|
| 1146 |
+
"positive_losses": 0.04189910739660263,
|
| 1147 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 1148 |
+
"rewards/chosen": 0.05516824126243591,
|
| 1149 |
+
"rewards/margins": 0.017182758077979088,
|
| 1150 |
+
"rewards/margins_max": 0.042931947857141495,
|
| 1151 |
+
"rewards/margins_min": -0.008566427044570446,
|
| 1152 |
+
"rewards/margins_std": 0.036414846777915955,
|
| 1153 |
+
"rewards/rejected": 0.037985485047101974,
|
| 1154 |
+
"step": 510
|
| 1155 |
+
},
|
| 1156 |
+
{
|
| 1157 |
+
"dpo_losses": 0.6818202137947083,
|
| 1158 |
+
"epoch": 1.58,
|
| 1159 |
+
"grad_norm": 7.65808980206578,
|
| 1160 |
+
"learning_rate": 2.703200791526611e-07,
|
| 1161 |
+
"logits/chosen": -3.0046448707580566,
|
| 1162 |
+
"logits/rejected": -2.982851505279541,
|
| 1163 |
+
"logps/chosen": -296.64495849609375,
|
| 1164 |
+
"logps/rejected": -249.4301300048828,
|
| 1165 |
+
"loss": 0.6837,
|
| 1166 |
+
"positive_losses": 0.0,
|
| 1167 |
+
"rewards/accuracies": 0.550000011920929,
|
| 1168 |
+
"rewards/chosen": 0.07770957797765732,
|
| 1169 |
+
"rewards/margins": 0.02354852482676506,
|
| 1170 |
+
"rewards/margins_max": 0.04674399644136429,
|
| 1171 |
+
"rewards/margins_min": 0.000353056937456131,
|
| 1172 |
+
"rewards/margins_std": 0.03280334919691086,
|
| 1173 |
+
"rewards/rejected": 0.05416105315089226,
|
| 1174 |
+
"step": 520
|
| 1175 |
+
},
|
| 1176 |
+
{
|
| 1177 |
+
"dpo_losses": 0.6779841184616089,
|
| 1178 |
+
"epoch": 1.61,
|
| 1179 |
+
"grad_norm": 2.0053415493661566,
|
| 1180 |
+
"learning_rate": 2.6149389302044487e-07,
|
| 1181 |
+
"logits/chosen": -2.849656105041504,
|
| 1182 |
+
"logits/rejected": -2.743305206298828,
|
| 1183 |
+
"logps/chosen": -389.1894836425781,
|
| 1184 |
+
"logps/rejected": -333.93048095703125,
|
| 1185 |
+
"loss": 0.6807,
|
| 1186 |
+
"positive_losses": 0.0,
|
| 1187 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1188 |
+
"rewards/chosen": 0.08877243846654892,
|
| 1189 |
+
"rewards/margins": 0.031258475035429,
|
| 1190 |
+
"rewards/margins_max": 0.05623454600572586,
|
| 1191 |
+
"rewards/margins_min": 0.006282416172325611,
|
| 1192 |
+
"rewards/margins_std": 0.03532148897647858,
|
| 1193 |
+
"rewards/rejected": 0.05751396343111992,
|
| 1194 |
+
"step": 530
|
| 1195 |
+
},
|
| 1196 |
+
{
|
| 1197 |
+
"dpo_losses": 0.6694897413253784,
|
| 1198 |
+
"epoch": 1.64,
|
| 1199 |
+
"grad_norm": 1.8763346054855339,
|
| 1200 |
+
"learning_rate": 2.526533223585641e-07,
|
| 1201 |
+
"logits/chosen": -2.7003915309906006,
|
| 1202 |
+
"logits/rejected": -2.768725872039795,
|
| 1203 |
+
"logps/chosen": -374.7425231933594,
|
| 1204 |
+
"logps/rejected": -326.8890075683594,
|
| 1205 |
+
"loss": 0.6778,
|
| 1206 |
+
"positive_losses": 0.0,
|
| 1207 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1208 |
+
"rewards/chosen": 0.08555268496274948,
|
| 1209 |
+
"rewards/margins": 0.048603277653455734,
|
| 1210 |
+
"rewards/margins_max": 0.07389990985393524,
|
| 1211 |
+
"rewards/margins_min": 0.023306647315621376,
|
| 1212 |
+
"rewards/margins_std": 0.0357748381793499,
|
| 1213 |
+
"rewards/rejected": 0.03694940358400345,
|
| 1214 |
+
"step": 540
|
| 1215 |
+
},
|
| 1216 |
+
{
|
| 1217 |
+
"dpo_losses": 0.6740890741348267,
|
| 1218 |
+
"epoch": 1.67,
|
| 1219 |
+
"grad_norm": 7.77121120757971,
|
| 1220 |
+
"learning_rate": 2.438094310816307e-07,
|
| 1221 |
+
"logits/chosen": -2.7915353775024414,
|
| 1222 |
+
"logits/rejected": -2.8357605934143066,
|
| 1223 |
+
"logps/chosen": -232.6016387939453,
|
| 1224 |
+
"logps/rejected": -294.88299560546875,
|
| 1225 |
+
"loss": 0.6826,
|
| 1226 |
+
"positive_losses": 0.0,
|
| 1227 |
+
"rewards/accuracies": 0.75,
|
| 1228 |
+
"rewards/chosen": 0.06982794404029846,
|
| 1229 |
+
"rewards/margins": 0.03929734230041504,
|
| 1230 |
+
"rewards/margins_max": 0.07744838297367096,
|
| 1231 |
+
"rewards/margins_min": 0.0011462997645139694,
|
| 1232 |
+
"rewards/margins_std": 0.05395371839404106,
|
| 1233 |
+
"rewards/rejected": 0.030530598014593124,
|
| 1234 |
+
"step": 550
|
| 1235 |
+
},
|
| 1236 |
+
{
|
| 1237 |
+
"dpo_losses": 0.6748778820037842,
|
| 1238 |
+
"epoch": 1.7,
|
| 1239 |
+
"grad_norm": 2.150634086609646,
|
| 1240 |
+
"learning_rate": 2.3497328725998406e-07,
|
| 1241 |
+
"logits/chosen": -2.8007655143737793,
|
| 1242 |
+
"logits/rejected": -2.8195149898529053,
|
| 1243 |
+
"logps/chosen": -272.33734130859375,
|
| 1244 |
+
"logps/rejected": -247.6636199951172,
|
| 1245 |
+
"loss": 0.6794,
|
| 1246 |
+
"positive_losses": 0.18440398573875427,
|
| 1247 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1248 |
+
"rewards/chosen": 0.08441298454999924,
|
| 1249 |
+
"rewards/margins": 0.03730101138353348,
|
| 1250 |
+
"rewards/margins_max": 0.05858795717358589,
|
| 1251 |
+
"rewards/margins_min": 0.016014060005545616,
|
| 1252 |
+
"rewards/margins_std": 0.030104290693998337,
|
| 1253 |
+
"rewards/rejected": 0.047111984342336655,
|
| 1254 |
+
"step": 560
|
| 1255 |
+
},
|
| 1256 |
+
{
|
| 1257 |
+
"dpo_losses": 0.6834501028060913,
|
| 1258 |
+
"epoch": 1.73,
|
| 1259 |
+
"grad_norm": 1.9329413652442144,
|
| 1260 |
+
"learning_rate": 2.261559492680755e-07,
|
| 1261 |
+
"logits/chosen": -3.00966215133667,
|
| 1262 |
+
"logits/rejected": -2.9343342781066895,
|
| 1263 |
+
"logps/chosen": -364.2731018066406,
|
| 1264 |
+
"logps/rejected": -270.2425537109375,
|
| 1265 |
+
"loss": 0.6883,
|
| 1266 |
+
"positive_losses": 0.03923187404870987,
|
| 1267 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 1268 |
+
"rewards/chosen": 0.07380800694227219,
|
| 1269 |
+
"rewards/margins": 0.020400431007146835,
|
| 1270 |
+
"rewards/margins_max": 0.051741547882556915,
|
| 1271 |
+
"rewards/margins_min": -0.010940684005618095,
|
| 1272 |
+
"rewards/margins_std": 0.04432303458452225,
|
| 1273 |
+
"rewards/rejected": 0.05340757966041565,
|
| 1274 |
+
"step": 570
|
| 1275 |
+
},
|
| 1276 |
+
{
|
| 1277 |
+
"dpo_losses": 0.683830738067627,
|
| 1278 |
+
"epoch": 1.76,
|
| 1279 |
+
"grad_norm": 6.498857250997875,
|
| 1280 |
+
"learning_rate": 2.1736845194498716e-07,
|
| 1281 |
+
"logits/chosen": -2.8374483585357666,
|
| 1282 |
+
"logits/rejected": -2.778815507888794,
|
| 1283 |
+
"logps/chosen": -294.1668395996094,
|
| 1284 |
+
"logps/rejected": -277.2710876464844,
|
| 1285 |
+
"loss": 0.6838,
|
| 1286 |
+
"positive_losses": 0.0,
|
| 1287 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1288 |
+
"rewards/chosen": 0.0776522159576416,
|
| 1289 |
+
"rewards/margins": 0.018975799903273582,
|
| 1290 |
+
"rewards/margins_max": 0.03656230866909027,
|
| 1291 |
+
"rewards/margins_min": 0.0013892930001020432,
|
| 1292 |
+
"rewards/margins_std": 0.02487107738852501,
|
| 1293 |
+
"rewards/rejected": 0.05867641046643257,
|
| 1294 |
+
"step": 580
|
| 1295 |
+
},
|
| 1296 |
+
{
|
| 1297 |
+
"dpo_losses": 0.6692390441894531,
|
| 1298 |
+
"epoch": 1.79,
|
| 1299 |
+
"grad_norm": 14.442968348535452,
|
| 1300 |
+
"learning_rate": 2.0862179278440504e-07,
|
| 1301 |
+
"logits/chosen": -2.752941608428955,
|
| 1302 |
+
"logits/rejected": -2.595552921295166,
|
| 1303 |
+
"logps/chosen": -326.27947998046875,
|
| 1304 |
+
"logps/rejected": -363.86175537109375,
|
| 1305 |
+
"loss": 0.6826,
|
| 1306 |
+
"positive_losses": 0.0,
|
| 1307 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1308 |
+
"rewards/chosen": 0.09450069069862366,
|
| 1309 |
+
"rewards/margins": 0.04905655235052109,
|
| 1310 |
+
"rewards/margins_max": 0.08510641753673553,
|
| 1311 |
+
"rewards/margins_min": 0.013006689958274364,
|
| 1312 |
+
"rewards/margins_std": 0.05098220705986023,
|
| 1313 |
+
"rewards/rejected": 0.04544413462281227,
|
| 1314 |
+
"step": 590
|
| 1315 |
+
},
|
| 1316 |
+
{
|
| 1317 |
+
"dpo_losses": 0.6739311218261719,
|
| 1318 |
+
"epoch": 1.82,
|
| 1319 |
+
"grad_norm": 1.8571868330601113,
|
| 1320 |
+
"learning_rate": 1.9992691817133024e-07,
|
| 1321 |
+
"logits/chosen": -2.814953327178955,
|
| 1322 |
+
"logits/rejected": -2.798966884613037,
|
| 1323 |
+
"logps/chosen": -228.36764526367188,
|
| 1324 |
+
"logps/rejected": -220.4230499267578,
|
| 1325 |
+
"loss": 0.677,
|
| 1326 |
+
"positive_losses": 0.0,
|
| 1327 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1328 |
+
"rewards/chosen": 0.07805188000202179,
|
| 1329 |
+
"rewards/margins": 0.03935372084379196,
|
| 1330 |
+
"rewards/margins_max": 0.06514634191989899,
|
| 1331 |
+
"rewards/margins_min": 0.013561112806200981,
|
| 1332 |
+
"rewards/margins_std": 0.0364762619137764,
|
| 1333 |
+
"rewards/rejected": 0.03869815915822983,
|
| 1334 |
+
"step": 600
|
| 1335 |
+
},
|
| 1336 |
+
{
|
| 1337 |
+
"epoch": 1.82,
|
| 1338 |
+
"eval_dpo_losses": 0.6810709238052368,
|
| 1339 |
+
"eval_logits/chosen": -2.8493857383728027,
|
| 1340 |
+
"eval_logits/rejected": -2.804689407348633,
|
| 1341 |
+
"eval_logps/chosen": -277.5360412597656,
|
| 1342 |
+
"eval_logps/rejected": -253.9931640625,
|
| 1343 |
+
"eval_loss": 0.6872838139533997,
|
| 1344 |
+
"eval_positive_losses": 0.04589047655463219,
|
| 1345 |
+
"eval_rewards/accuracies": 0.682539701461792,
|
| 1346 |
+
"eval_rewards/chosen": 0.07685134559869766,
|
| 1347 |
+
"eval_rewards/margins": 0.02495652064681053,
|
| 1348 |
+
"eval_rewards/margins_max": 0.08033440262079239,
|
| 1349 |
+
"eval_rewards/margins_min": -0.02796892076730728,
|
| 1350 |
+
"eval_rewards/margins_std": 0.04838396981358528,
|
| 1351 |
+
"eval_rewards/rejected": 0.05189482495188713,
|
| 1352 |
+
"eval_runtime": 282.2305,
|
| 1353 |
+
"eval_samples_per_second": 7.086,
|
| 1354 |
+
"eval_steps_per_second": 0.223,
|
| 1355 |
+
"step": 600
|
| 1356 |
+
},
|
| 1357 |
+
{
|
| 1358 |
+
"dpo_losses": 0.6791930794715881,
|
| 1359 |
+
"epoch": 1.85,
|
| 1360 |
+
"grad_norm": 2.108916508331074,
|
| 1361 |
+
"learning_rate": 1.912947096827513e-07,
|
| 1362 |
+
"logits/chosen": -2.977818250656128,
|
| 1363 |
+
"logits/rejected": -2.8510375022888184,
|
| 1364 |
+
"logps/chosen": -303.4037170410156,
|
| 1365 |
+
"logps/rejected": -240.5347137451172,
|
| 1366 |
+
"loss": 0.6825,
|
| 1367 |
+
"positive_losses": 0.0,
|
| 1368 |
+
"rewards/accuracies": 0.75,
|
| 1369 |
+
"rewards/chosen": 0.08365954458713531,
|
| 1370 |
+
"rewards/margins": 0.028567474335432053,
|
| 1371 |
+
"rewards/margins_max": 0.05369936302304268,
|
| 1372 |
+
"rewards/margins_min": 0.003435590071603656,
|
| 1373 |
+
"rewards/margins_std": 0.03554185479879379,
|
| 1374 |
+
"rewards/rejected": 0.05509205535054207,
|
| 1375 |
+
"step": 610
|
| 1376 |
+
},
|
| 1377 |
+
{
|
| 1378 |
+
"dpo_losses": 0.6858159303665161,
|
| 1379 |
+
"epoch": 1.88,
|
| 1380 |
+
"grad_norm": 2.0427241279215775,
|
| 1381 |
+
"learning_rate": 1.8273597046942452e-07,
|
| 1382 |
+
"logits/chosen": -2.8828835487365723,
|
| 1383 |
+
"logits/rejected": -2.7817318439483643,
|
| 1384 |
+
"logps/chosen": -252.5178985595703,
|
| 1385 |
+
"logps/rejected": -227.63070678710938,
|
| 1386 |
+
"loss": 0.6838,
|
| 1387 |
+
"positive_losses": 0.16164550185203552,
|
| 1388 |
+
"rewards/accuracies": 0.550000011920929,
|
| 1389 |
+
"rewards/chosen": 0.06277048587799072,
|
| 1390 |
+
"rewards/margins": 0.015206271782517433,
|
| 1391 |
+
"rewards/margins_max": 0.04130411148071289,
|
| 1392 |
+
"rewards/margins_min": -0.010891567915678024,
|
| 1393 |
+
"rewards/margins_std": 0.03690791875123978,
|
| 1394 |
+
"rewards/rejected": 0.04756421223282814,
|
| 1395 |
+
"step": 620
|
| 1396 |
+
},
|
| 1397 |
+
{
|
| 1398 |
+
"dpo_losses": 0.6795656681060791,
|
| 1399 |
+
"epoch": 1.91,
|
| 1400 |
+
"grad_norm": 4.79445166991863,
|
| 1401 |
+
"learning_rate": 1.742614117358029e-07,
|
| 1402 |
+
"logits/chosen": -2.863675594329834,
|
| 1403 |
+
"logits/rejected": -2.8280770778656006,
|
| 1404 |
+
"logps/chosen": -238.7898712158203,
|
| 1405 |
+
"logps/rejected": -213.51846313476562,
|
| 1406 |
+
"loss": 0.6827,
|
| 1407 |
+
"positive_losses": 0.0,
|
| 1408 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 1409 |
+
"rewards/chosen": 0.06867232173681259,
|
| 1410 |
+
"rewards/margins": 0.028171828016638756,
|
| 1411 |
+
"rewards/margins_max": 0.06020760536193848,
|
| 1412 |
+
"rewards/margins_min": -0.0038639488629996777,
|
| 1413 |
+
"rewards/margins_std": 0.04530543461441994,
|
| 1414 |
+
"rewards/rejected": 0.040500491857528687,
|
| 1415 |
+
"step": 630
|
| 1416 |
+
},
|
| 1417 |
+
{
|
| 1418 |
+
"dpo_losses": 0.6790474057197571,
|
| 1419 |
+
"epoch": 1.95,
|
| 1420 |
+
"grad_norm": 1.9164560727509983,
|
| 1421 |
+
"learning_rate": 1.6588163933503686e-07,
|
| 1422 |
+
"logits/chosen": -2.7663025856018066,
|
| 1423 |
+
"logits/rejected": -2.7693681716918945,
|
| 1424 |
+
"logps/chosen": -180.65650939941406,
|
| 1425 |
+
"logps/rejected": -231.96517944335938,
|
| 1426 |
+
"loss": 0.6845,
|
| 1427 |
+
"positive_losses": 0.0,
|
| 1428 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1429 |
+
"rewards/chosen": 0.05268232896924019,
|
| 1430 |
+
"rewards/margins": 0.028681308031082153,
|
| 1431 |
+
"rewards/margins_max": 0.053858645260334015,
|
| 1432 |
+
"rewards/margins_min": 0.0035039703361690044,
|
| 1433 |
+
"rewards/margins_std": 0.035606130957603455,
|
| 1434 |
+
"rewards/rejected": 0.02400101348757744,
|
| 1435 |
+
"step": 640
|
| 1436 |
+
},
|
| 1437 |
+
{
|
| 1438 |
+
"dpo_losses": 0.682746946811676,
|
| 1439 |
+
"epoch": 1.98,
|
| 1440 |
+
"grad_norm": 2.1910688443112587,
|
| 1441 |
+
"learning_rate": 1.5760714049581947e-07,
|
| 1442 |
+
"logits/chosen": -2.835636615753174,
|
| 1443 |
+
"logits/rejected": -2.823740243911743,
|
| 1444 |
+
"logps/chosen": -217.59878540039062,
|
| 1445 |
+
"logps/rejected": -202.2057342529297,
|
| 1446 |
+
"loss": 0.6825,
|
| 1447 |
+
"positive_losses": 0.016841506585478783,
|
| 1448 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 1449 |
+
"rewards/chosen": 0.05880022048950195,
|
| 1450 |
+
"rewards/margins": 0.021725308150053024,
|
| 1451 |
+
"rewards/margins_max": 0.05336114764213562,
|
| 1452 |
+
"rewards/margins_min": -0.009910527616739273,
|
| 1453 |
+
"rewards/margins_std": 0.04473983123898506,
|
| 1454 |
+
"rewards/rejected": 0.03707490488886833,
|
| 1455 |
+
"step": 650
|
| 1456 |
+
},
|
| 1457 |
+
{
|
| 1458 |
+
"dpo_losses": 0.6814676523208618,
|
| 1459 |
+
"epoch": 2.01,
|
| 1460 |
+
"grad_norm": 1.922230509309539,
|
| 1461 |
+
"learning_rate": 1.4944827069769122e-07,
|
| 1462 |
+
"logits/chosen": -3.023953914642334,
|
| 1463 |
+
"logits/rejected": -2.9771876335144043,
|
| 1464 |
+
"logps/chosen": -332.7305603027344,
|
| 1465 |
+
"logps/rejected": -338.9562072753906,
|
| 1466 |
+
"loss": 0.6765,
|
| 1467 |
+
"positive_losses": 0.052425384521484375,
|
| 1468 |
+
"rewards/accuracies": 0.75,
|
| 1469 |
+
"rewards/chosen": 0.054904378950595856,
|
| 1470 |
+
"rewards/margins": 0.02420732006430626,
|
| 1471 |
+
"rewards/margins_max": 0.05272083729505539,
|
| 1472 |
+
"rewards/margins_min": -0.004306197632104158,
|
| 1473 |
+
"rewards/margins_std": 0.04032420367002487,
|
| 1474 |
+
"rewards/rejected": 0.03069705329835415,
|
| 1475 |
+
"step": 660
|
| 1476 |
+
},
|
| 1477 |
+
{
|
| 1478 |
+
"dpo_losses": 0.6808332204818726,
|
| 1479 |
+
"epoch": 2.04,
|
| 1480 |
+
"grad_norm": 2.128169444960246,
|
| 1481 |
+
"learning_rate": 1.4141524071122656e-07,
|
| 1482 |
+
"logits/chosen": -2.90728497505188,
|
| 1483 |
+
"logits/rejected": -2.9327125549316406,
|
| 1484 |
+
"logps/chosen": -294.28411865234375,
|
| 1485 |
+
"logps/rejected": -306.5546569824219,
|
| 1486 |
+
"loss": 0.6781,
|
| 1487 |
+
"positive_losses": 0.0,
|
| 1488 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1489 |
+
"rewards/chosen": 0.08569607883691788,
|
| 1490 |
+
"rewards/margins": 0.026000851765275,
|
| 1491 |
+
"rewards/margins_max": 0.06097273901104927,
|
| 1492 |
+
"rewards/margins_min": -0.008971036411821842,
|
| 1493 |
+
"rewards/margins_std": 0.049457717686891556,
|
| 1494 |
+
"rewards/rejected": 0.059695225208997726,
|
| 1495 |
+
"step": 670
|
| 1496 |
+
},
|
| 1497 |
+
{
|
| 1498 |
+
"dpo_losses": 0.6652864217758179,
|
| 1499 |
+
"epoch": 2.07,
|
| 1500 |
+
"grad_norm": 2.0596328226644616,
|
| 1501 |
+
"learning_rate": 1.3351810381932387e-07,
|
| 1502 |
+
"logits/chosen": -2.879633665084839,
|
| 1503 |
+
"logits/rejected": -2.866821765899658,
|
| 1504 |
+
"logps/chosen": -277.6556091308594,
|
| 1505 |
+
"logps/rejected": -225.43563842773438,
|
| 1506 |
+
"loss": 0.6728,
|
| 1507 |
+
"positive_losses": 0.06625213474035263,
|
| 1508 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1509 |
+
"rewards/chosen": 0.09781384468078613,
|
| 1510 |
+
"rewards/margins": 0.05752427130937576,
|
| 1511 |
+
"rewards/margins_max": 0.09321505576372147,
|
| 1512 |
+
"rewards/margins_min": 0.02183348871767521,
|
| 1513 |
+
"rewards/margins_std": 0.0504743866622448,
|
| 1514 |
+
"rewards/rejected": 0.04028957709670067,
|
| 1515 |
+
"step": 680
|
| 1516 |
+
},
|
| 1517 |
+
{
|
| 1518 |
+
"dpo_losses": 0.6790655255317688,
|
| 1519 |
+
"epoch": 2.1,
|
| 1520 |
+
"grad_norm": 2.366475517541238,
|
| 1521 |
+
"learning_rate": 1.2576674323558928e-07,
|
| 1522 |
+
"logits/chosen": -2.954852819442749,
|
| 1523 |
+
"logits/rejected": -2.885873794555664,
|
| 1524 |
+
"logps/chosen": -260.78973388671875,
|
| 1525 |
+
"logps/rejected": -258.0441589355469,
|
| 1526 |
+
"loss": 0.6782,
|
| 1527 |
+
"positive_losses": 0.0,
|
| 1528 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1529 |
+
"rewards/chosen": 0.08070608973503113,
|
| 1530 |
+
"rewards/margins": 0.02996811270713806,
|
| 1531 |
+
"rewards/margins_max": 0.07714684307575226,
|
| 1532 |
+
"rewards/margins_min": -0.017210613936185837,
|
| 1533 |
+
"rewards/margins_std": 0.06672079861164093,
|
| 1534 |
+
"rewards/rejected": 0.050737977027893066,
|
| 1535 |
+
"step": 690
|
| 1536 |
+
},
|
| 1537 |
+
{
|
| 1538 |
+
"dpo_losses": 0.677202582359314,
|
| 1539 |
+
"epoch": 2.13,
|
| 1540 |
+
"grad_norm": 2.333653791310998,
|
| 1541 |
+
"learning_rate": 1.1817085973556152e-07,
|
| 1542 |
+
"logits/chosen": -2.869096279144287,
|
| 1543 |
+
"logits/rejected": -2.842957019805908,
|
| 1544 |
+
"logps/chosen": -240.80905151367188,
|
| 1545 |
+
"logps/rejected": -281.9169006347656,
|
| 1546 |
+
"loss": 0.6796,
|
| 1547 |
+
"positive_losses": 0.0,
|
| 1548 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1549 |
+
"rewards/chosen": 0.08999605476856232,
|
| 1550 |
+
"rewards/margins": 0.032623276114463806,
|
| 1551 |
+
"rewards/margins_max": 0.05816454440355301,
|
| 1552 |
+
"rewards/margins_min": 0.007082013878971338,
|
| 1553 |
+
"rewards/margins_std": 0.03612080588936806,
|
| 1554 |
+
"rewards/rejected": 0.057372767478227615,
|
| 1555 |
+
"step": 700
|
| 1556 |
+
},
|
| 1557 |
+
{
|
| 1558 |
+
"epoch": 2.13,
|
| 1559 |
+
"eval_dpo_losses": 0.6800086498260498,
|
| 1560 |
+
"eval_logits/chosen": -2.8477256298065186,
|
| 1561 |
+
"eval_logits/rejected": -2.802640676498413,
|
| 1562 |
+
"eval_logps/chosen": -277.2366027832031,
|
| 1563 |
+
"eval_logps/rejected": -253.92018127441406,
|
| 1564 |
+
"eval_loss": 0.687153697013855,
|
| 1565 |
+
"eval_positive_losses": 0.05479403957724571,
|
| 1566 |
+
"eval_rewards/accuracies": 0.682539701461792,
|
| 1567 |
+
"eval_rewards/chosen": 0.0798461064696312,
|
| 1568 |
+
"eval_rewards/margins": 0.0272214412689209,
|
| 1569 |
+
"eval_rewards/margins_max": 0.08652784675359726,
|
| 1570 |
+
"eval_rewards/margins_min": -0.029835335910320282,
|
| 1571 |
+
"eval_rewards/margins_std": 0.05211076885461807,
|
| 1572 |
+
"eval_rewards/rejected": 0.0526246652007103,
|
| 1573 |
+
"eval_runtime": 281.0737,
|
| 1574 |
+
"eval_samples_per_second": 7.116,
|
| 1575 |
+
"eval_steps_per_second": 0.224,
|
| 1576 |
+
"step": 700
|
| 1577 |
+
},
|
| 1578 |
+
{
|
| 1579 |
+
"dpo_losses": 0.6692663431167603,
|
| 1580 |
+
"epoch": 2.16,
|
| 1581 |
+
"grad_norm": 1.7167790990262888,
|
| 1582 |
+
"learning_rate": 1.1073995951625664e-07,
|
| 1583 |
+
"logits/chosen": -2.9118266105651855,
|
| 1584 |
+
"logits/rejected": -2.8686835765838623,
|
| 1585 |
+
"logps/chosen": -400.4477844238281,
|
| 1586 |
+
"logps/rejected": -264.7125549316406,
|
| 1587 |
+
"loss": 0.677,
|
| 1588 |
+
"positive_losses": 0.0,
|
| 1589 |
+
"rewards/accuracies": 0.75,
|
| 1590 |
+
"rewards/chosen": 0.10832663625478745,
|
| 1591 |
+
"rewards/margins": 0.0492747537791729,
|
| 1592 |
+
"rewards/margins_max": 0.08400574326515198,
|
| 1593 |
+
"rewards/margins_min": 0.014543762430548668,
|
| 1594 |
+
"rewards/margins_std": 0.049117036163806915,
|
| 1595 |
+
"rewards/rejected": 0.05905189365148544,
|
| 1596 |
+
"step": 710
|
| 1597 |
+
},
|
| 1598 |
+
{
|
| 1599 |
+
"dpo_losses": 0.670403778553009,
|
| 1600 |
+
"epoch": 2.19,
|
| 1601 |
+
"grad_norm": 1.8644942647733735,
|
| 1602 |
+
"learning_rate": 1.0348334229922676e-07,
|
| 1603 |
+
"logits/chosen": -2.9610977172851562,
|
| 1604 |
+
"logits/rejected": -2.86533784866333,
|
| 1605 |
+
"logps/chosen": -368.39862060546875,
|
| 1606 |
+
"logps/rejected": -269.9642028808594,
|
| 1607 |
+
"loss": 0.6754,
|
| 1608 |
+
"positive_losses": 0.0,
|
| 1609 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1610 |
+
"rewards/chosen": 0.08919496089220047,
|
| 1611 |
+
"rewards/margins": 0.04688384756445885,
|
| 1612 |
+
"rewards/margins_max": 0.08247358351945877,
|
| 1613 |
+
"rewards/margins_min": 0.01129411906003952,
|
| 1614 |
+
"rewards/margins_std": 0.0503314845263958,
|
| 1615 |
+
"rewards/rejected": 0.04231110215187073,
|
| 1616 |
+
"step": 720
|
| 1617 |
+
},
|
| 1618 |
+
{
|
| 1619 |
+
"dpo_losses": 0.6769901514053345,
|
| 1620 |
+
"epoch": 2.22,
|
| 1621 |
+
"grad_norm": 2.180859431848076,
|
| 1622 |
+
"learning_rate": 9.6410089692021e-08,
|
| 1623 |
+
"logits/chosen": -2.9250197410583496,
|
| 1624 |
+
"logits/rejected": -2.9387307167053223,
|
| 1625 |
+
"logps/chosen": -361.2002868652344,
|
| 1626 |
+
"logps/rejected": -294.02313232421875,
|
| 1627 |
+
"loss": 0.6754,
|
| 1628 |
+
"positive_losses": 0.214344784617424,
|
| 1629 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1630 |
+
"rewards/chosen": 0.07332402467727661,
|
| 1631 |
+
"rewards/margins": 0.033169932663440704,
|
| 1632 |
+
"rewards/margins_max": 0.05973232537508011,
|
| 1633 |
+
"rewards/margins_min": 0.0066075399518013,
|
| 1634 |
+
"rewards/margins_std": 0.037564896047115326,
|
| 1635 |
+
"rewards/rejected": 0.04015409201383591,
|
| 1636 |
+
"step": 730
|
| 1637 |
+
},
|
| 1638 |
+
{
|
| 1639 |
+
"dpo_losses": 0.6715595126152039,
|
| 1640 |
+
"epoch": 2.25,
|
| 1641 |
+
"grad_norm": 2.04722557268885,
|
| 1642 |
+
"learning_rate": 8.95290538226148e-08,
|
| 1643 |
+
"logits/chosen": -2.7215945720672607,
|
| 1644 |
+
"logits/rejected": -2.681288957595825,
|
| 1645 |
+
"logps/chosen": -238.29833984375,
|
| 1646 |
+
"logps/rejected": -157.76754760742188,
|
| 1647 |
+
"loss": 0.6755,
|
| 1648 |
+
"positive_losses": 0.0,
|
| 1649 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1650 |
+
"rewards/chosen": 0.08287638425827026,
|
| 1651 |
+
"rewards/margins": 0.04476654529571533,
|
| 1652 |
+
"rewards/margins_max": 0.07331255823373795,
|
| 1653 |
+
"rewards/margins_min": 0.016220541670918465,
|
| 1654 |
+
"rewards/margins_std": 0.040370143949985504,
|
| 1655 |
+
"rewards/rejected": 0.03810984641313553,
|
| 1656 |
+
"step": 740
|
| 1657 |
+
},
|
| 1658 |
+
{
|
| 1659 |
+
"dpo_losses": 0.6766847372055054,
|
| 1660 |
+
"epoch": 2.28,
|
| 1661 |
+
"grad_norm": 7.254140526558107,
|
| 1662 |
+
"learning_rate": 8.284884626103164e-08,
|
| 1663 |
+
"logits/chosen": -2.8504419326782227,
|
| 1664 |
+
"logits/rejected": -2.817267656326294,
|
| 1665 |
+
"logps/chosen": -230.2511749267578,
|
| 1666 |
+
"logps/rejected": -262.9688415527344,
|
| 1667 |
+
"loss": 0.6787,
|
| 1668 |
+
"positive_losses": 0.0,
|
| 1669 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1670 |
+
"rewards/chosen": 0.07979516685009003,
|
| 1671 |
+
"rewards/margins": 0.03387243300676346,
|
| 1672 |
+
"rewards/margins_max": 0.05933348461985588,
|
| 1673 |
+
"rewards/margins_min": 0.008411377668380737,
|
| 1674 |
+
"rewards/margins_std": 0.036007367074489594,
|
| 1675 |
+
"rewards/rejected": 0.04592275246977806,
|
| 1676 |
+
"step": 750
|
| 1677 |
+
},
|
| 1678 |
+
{
|
| 1679 |
+
"dpo_losses": 0.6724633574485779,
|
| 1680 |
+
"epoch": 2.31,
|
| 1681 |
+
"grad_norm": 2.1195303141123993,
|
| 1682 |
+
"learning_rate": 7.637782724202072e-08,
|
| 1683 |
+
"logits/chosen": -2.8214287757873535,
|
| 1684 |
+
"logits/rejected": -2.8338170051574707,
|
| 1685 |
+
"logps/chosen": -247.1202850341797,
|
| 1686 |
+
"logps/rejected": -217.5040740966797,
|
| 1687 |
+
"loss": 0.684,
|
| 1688 |
+
"positive_losses": 0.0,
|
| 1689 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1690 |
+
"rewards/chosen": 0.07794385403394699,
|
| 1691 |
+
"rewards/margins": 0.04256601259112358,
|
| 1692 |
+
"rewards/margins_max": 0.07706017792224884,
|
| 1693 |
+
"rewards/margins_min": 0.008071859367191792,
|
| 1694 |
+
"rewards/margins_std": 0.0487821027636528,
|
| 1695 |
+
"rewards/rejected": 0.03537784516811371,
|
| 1696 |
+
"step": 760
|
| 1697 |
+
},
|
| 1698 |
+
{
|
| 1699 |
+
"dpo_losses": 0.6884227991104126,
|
| 1700 |
+
"epoch": 2.34,
|
| 1701 |
+
"grad_norm": 6.591694443648593,
|
| 1702 |
+
"learning_rate": 7.01240952022801e-08,
|
| 1703 |
+
"logits/chosen": -2.8677477836608887,
|
| 1704 |
+
"logits/rejected": -2.832568883895874,
|
| 1705 |
+
"logps/chosen": -238.2762451171875,
|
| 1706 |
+
"logps/rejected": -235.57272338867188,
|
| 1707 |
+
"loss": 0.6773,
|
| 1708 |
+
"positive_losses": 0.13628844916820526,
|
| 1709 |
+
"rewards/accuracies": 0.550000011920929,
|
| 1710 |
+
"rewards/chosen": 0.06362812221050262,
|
| 1711 |
+
"rewards/margins": 0.009944443590939045,
|
| 1712 |
+
"rewards/margins_max": 0.03966521471738815,
|
| 1713 |
+
"rewards/margins_min": -0.019776327535510063,
|
| 1714 |
+
"rewards/margins_std": 0.04203151911497116,
|
| 1715 |
+
"rewards/rejected": 0.053683679550886154,
|
| 1716 |
+
"step": 770
|
| 1717 |
+
},
|
| 1718 |
+
{
|
| 1719 |
+
"dpo_losses": 0.6836405396461487,
|
| 1720 |
+
"epoch": 2.37,
|
| 1721 |
+
"grad_norm": 2.3289966359685415,
|
| 1722 |
+
"learning_rate": 6.409547664531733e-08,
|
| 1723 |
+
"logits/chosen": -2.867485523223877,
|
| 1724 |
+
"logits/rejected": -2.867396831512451,
|
| 1725 |
+
"logps/chosen": -221.3419189453125,
|
| 1726 |
+
"logps/rejected": -208.03866577148438,
|
| 1727 |
+
"loss": 0.6768,
|
| 1728 |
+
"positive_losses": 0.0,
|
| 1729 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 1730 |
+
"rewards/chosen": 0.06145999953150749,
|
| 1731 |
+
"rewards/margins": 0.019474489614367485,
|
| 1732 |
+
"rewards/margins_max": 0.04479853808879852,
|
| 1733 |
+
"rewards/margins_min": -0.00584955420345068,
|
| 1734 |
+
"rewards/margins_std": 0.03581360727548599,
|
| 1735 |
+
"rewards/rejected": 0.04198550805449486,
|
| 1736 |
+
"step": 780
|
| 1737 |
+
},
|
| 1738 |
+
{
|
| 1739 |
+
"dpo_losses": 0.6730223894119263,
|
| 1740 |
+
"epoch": 2.4,
|
| 1741 |
+
"grad_norm": 2.003874601891291,
|
| 1742 |
+
"learning_rate": 5.829951634663324e-08,
|
| 1743 |
+
"logits/chosen": -2.7790565490722656,
|
| 1744 |
+
"logits/rejected": -2.7391462326049805,
|
| 1745 |
+
"logps/chosen": -245.35067749023438,
|
| 1746 |
+
"logps/rejected": -180.18910217285156,
|
| 1747 |
+
"loss": 0.6751,
|
| 1748 |
+
"positive_losses": 0.0,
|
| 1749 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1750 |
+
"rewards/chosen": 0.08253587782382965,
|
| 1751 |
+
"rewards/margins": 0.04125886410474777,
|
| 1752 |
+
"rewards/margins_max": 0.0634620413184166,
|
| 1753 |
+
"rewards/margins_min": 0.019055673852562904,
|
| 1754 |
+
"rewards/margins_std": 0.03140005096793175,
|
| 1755 |
+
"rewards/rejected": 0.041277021169662476,
|
| 1756 |
+
"step": 790
|
| 1757 |
+
},
|
| 1758 |
+
{
|
| 1759 |
+
"dpo_losses": 0.6658504605293274,
|
| 1760 |
+
"epoch": 2.43,
|
| 1761 |
+
"grad_norm": 2.0012405839420944,
|
| 1762 |
+
"learning_rate": 5.274346791148601e-08,
|
| 1763 |
+
"logits/chosen": -2.9153144359588623,
|
| 1764 |
+
"logits/rejected": -2.84319806098938,
|
| 1765 |
+
"logps/chosen": -366.9844665527344,
|
| 1766 |
+
"logps/rejected": -288.7860412597656,
|
| 1767 |
+
"loss": 0.6778,
|
| 1768 |
+
"positive_losses": 0.0,
|
| 1769 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1770 |
+
"rewards/chosen": 0.08800138533115387,
|
| 1771 |
+
"rewards/margins": 0.05663619562983513,
|
| 1772 |
+
"rewards/margins_max": 0.09046684950590134,
|
| 1773 |
+
"rewards/margins_min": 0.022805536165833473,
|
| 1774 |
+
"rewards/margins_std": 0.04784378036856651,
|
| 1775 |
+
"rewards/rejected": 0.031365178525447845,
|
| 1776 |
+
"step": 800
|
| 1777 |
+
},
|
| 1778 |
+
{
|
| 1779 |
+
"epoch": 2.43,
|
| 1780 |
+
"eval_dpo_losses": 0.6795361638069153,
|
| 1781 |
+
"eval_logits/chosen": -2.8473622798919678,
|
| 1782 |
+
"eval_logits/rejected": -2.8024210929870605,
|
| 1783 |
+
"eval_logps/chosen": -277.22216796875,
|
| 1784 |
+
"eval_logps/rejected": -254.00743103027344,
|
| 1785 |
+
"eval_loss": 0.6875322461128235,
|
| 1786 |
+
"eval_positive_losses": 0.060370072722435,
|
| 1787 |
+
"eval_rewards/accuracies": 0.682539701461792,
|
| 1788 |
+
"eval_rewards/chosen": 0.07999046891927719,
|
| 1789 |
+
"eval_rewards/margins": 0.0282380823045969,
|
| 1790 |
+
"eval_rewards/margins_max": 0.08973579108715057,
|
| 1791 |
+
"eval_rewards/margins_min": -0.030655015259981155,
|
| 1792 |
+
"eval_rewards/margins_std": 0.05399897322058678,
|
| 1793 |
+
"eval_rewards/rejected": 0.05175239220261574,
|
| 1794 |
+
"eval_runtime": 281.3076,
|
| 1795 |
+
"eval_samples_per_second": 7.11,
|
| 1796 |
+
"eval_steps_per_second": 0.224,
|
| 1797 |
+
"step": 800
|
| 1798 |
+
},
|
| 1799 |
+
{
|
| 1800 |
+
"dpo_losses": 0.6727338433265686,
|
| 1801 |
+
"epoch": 2.46,
|
| 1802 |
+
"grad_norm": 7.411310583604656,
|
| 1803 |
+
"learning_rate": 4.743428469705335e-08,
|
| 1804 |
+
"logits/chosen": -2.6556921005249023,
|
| 1805 |
+
"logits/rejected": -2.691716432571411,
|
| 1806 |
+
"logps/chosen": -270.0729064941406,
|
| 1807 |
+
"logps/rejected": -222.0420379638672,
|
| 1808 |
+
"loss": 0.6752,
|
| 1809 |
+
"positive_losses": 0.0,
|
| 1810 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1811 |
+
"rewards/chosen": 0.08230028301477432,
|
| 1812 |
+
"rewards/margins": 0.04306231066584587,
|
| 1813 |
+
"rewards/margins_max": 0.09098808467388153,
|
| 1814 |
+
"rewards/margins_min": -0.004863465670496225,
|
| 1815 |
+
"rewards/margins_std": 0.06777729094028473,
|
| 1816 |
+
"rewards/rejected": 0.03923797979950905,
|
| 1817 |
+
"step": 810
|
| 1818 |
+
},
|
| 1819 |
+
{
|
| 1820 |
+
"dpo_losses": 0.6766294240951538,
|
| 1821 |
+
"epoch": 2.49,
|
| 1822 |
+
"grad_norm": 2.0160351712928146,
|
| 1823 |
+
"learning_rate": 4.2378611110352707e-08,
|
| 1824 |
+
"logits/chosen": -2.8572134971618652,
|
| 1825 |
+
"logits/rejected": -2.9511802196502686,
|
| 1826 |
+
"logps/chosen": -193.8530731201172,
|
| 1827 |
+
"logps/rejected": -267.66522216796875,
|
| 1828 |
+
"loss": 0.6747,
|
| 1829 |
+
"positive_losses": 0.0,
|
| 1830 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 1831 |
+
"rewards/chosen": 0.06872855871915817,
|
| 1832 |
+
"rewards/margins": 0.03442535921931267,
|
| 1833 |
+
"rewards/margins_max": 0.07178723812103271,
|
| 1834 |
+
"rewards/margins_min": -0.0029365241061896086,
|
| 1835 |
+
"rewards/margins_std": 0.05283767729997635,
|
| 1836 |
+
"rewards/rejected": 0.034303199499845505,
|
| 1837 |
+
"step": 820
|
| 1838 |
+
},
|
| 1839 |
+
{
|
| 1840 |
+
"dpo_losses": 0.6828433275222778,
|
| 1841 |
+
"epoch": 2.52,
|
| 1842 |
+
"grad_norm": 2.1498185591320773,
|
| 1843 |
+
"learning_rate": 3.75827742928107e-08,
|
| 1844 |
+
"logits/chosen": -2.9401209354400635,
|
| 1845 |
+
"logits/rejected": -2.8616251945495605,
|
| 1846 |
+
"logps/chosen": -311.86260986328125,
|
| 1847 |
+
"logps/rejected": -264.8041076660156,
|
| 1848 |
+
"loss": 0.6773,
|
| 1849 |
+
"positive_losses": 0.0,
|
| 1850 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 1851 |
+
"rewards/chosen": 0.087630495429039,
|
| 1852 |
+
"rewards/margins": 0.022305091843008995,
|
| 1853 |
+
"rewards/margins_max": 0.07450206577777863,
|
| 1854 |
+
"rewards/margins_min": -0.029891883954405785,
|
| 1855 |
+
"rewards/margins_std": 0.0738176628947258,
|
| 1856 |
+
"rewards/rejected": 0.06532540172338486,
|
| 1857 |
+
"step": 830
|
| 1858 |
+
},
|
| 1859 |
+
{
|
| 1860 |
+
"dpo_losses": 0.6704001426696777,
|
| 1861 |
+
"epoch": 2.55,
|
| 1862 |
+
"grad_norm": 11.453340827604897,
|
| 1863 |
+
"learning_rate": 3.305277620188826e-08,
|
| 1864 |
+
"logits/chosen": -2.805208921432495,
|
| 1865 |
+
"logits/rejected": -2.840792179107666,
|
| 1866 |
+
"logps/chosen": -336.3541564941406,
|
| 1867 |
+
"logps/rejected": -344.42413330078125,
|
| 1868 |
+
"loss": 0.6733,
|
| 1869 |
+
"positive_losses": 0.0,
|
| 1870 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1871 |
+
"rewards/chosen": 0.09611042588949203,
|
| 1872 |
+
"rewards/margins": 0.04694870114326477,
|
| 1873 |
+
"rewards/margins_max": 0.08145497739315033,
|
| 1874 |
+
"rewards/margins_min": 0.012442430481314659,
|
| 1875 |
+
"rewards/margins_std": 0.04879923164844513,
|
| 1876 |
+
"rewards/rejected": 0.04916173219680786,
|
| 1877 |
+
"step": 840
|
| 1878 |
+
},
|
| 1879 |
+
{
|
| 1880 |
+
"dpo_losses": 0.6803882122039795,
|
| 1881 |
+
"epoch": 2.58,
|
| 1882 |
+
"grad_norm": 2.3437007644568086,
|
| 1883 |
+
"learning_rate": 2.879428609967105e-08,
|
| 1884 |
+
"logits/chosen": -2.921419620513916,
|
| 1885 |
+
"logits/rejected": -2.8526148796081543,
|
| 1886 |
+
"logps/chosen": -216.5243377685547,
|
| 1887 |
+
"logps/rejected": -294.4681091308594,
|
| 1888 |
+
"loss": 0.6739,
|
| 1889 |
+
"positive_losses": 0.0,
|
| 1890 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1891 |
+
"rewards/chosen": 0.07691861689090729,
|
| 1892 |
+
"rewards/margins": 0.02676830254495144,
|
| 1893 |
+
"rewards/margins_max": 0.05155067518353462,
|
| 1894 |
+
"rewards/margins_min": 0.001985931070521474,
|
| 1895 |
+
"rewards/margins_std": 0.035047560930252075,
|
| 1896 |
+
"rewards/rejected": 0.0501503124833107,
|
| 1897 |
+
"step": 850
|
| 1898 |
+
},
|
| 1899 |
+
{
|
| 1900 |
+
"dpo_losses": 0.679145872592926,
|
| 1901 |
+
"epoch": 2.61,
|
| 1902 |
+
"grad_norm": 2.249308373857521,
|
| 1903 |
+
"learning_rate": 2.4812633457826666e-08,
|
| 1904 |
+
"logits/chosen": -2.868685007095337,
|
| 1905 |
+
"logits/rejected": -2.853390693664551,
|
| 1906 |
+
"logps/chosen": -259.72357177734375,
|
| 1907 |
+
"logps/rejected": -335.20465087890625,
|
| 1908 |
+
"loss": 0.6762,
|
| 1909 |
+
"positive_losses": 0.0,
|
| 1910 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1911 |
+
"rewards/chosen": 0.0714501366019249,
|
| 1912 |
+
"rewards/margins": 0.028765708208084106,
|
| 1913 |
+
"rewards/margins_max": 0.051595449447631836,
|
| 1914 |
+
"rewards/margins_min": 0.00593597162514925,
|
| 1915 |
+
"rewards/margins_std": 0.03228612244129181,
|
| 1916 |
+
"rewards/rejected": 0.04268442839384079,
|
| 1917 |
+
"step": 860
|
| 1918 |
+
},
|
| 1919 |
+
{
|
| 1920 |
+
"dpo_losses": 0.6712166666984558,
|
| 1921 |
+
"epoch": 2.64,
|
| 1922 |
+
"grad_norm": 1.888865132587614,
|
| 1923 |
+
"learning_rate": 2.1112801287806375e-08,
|
| 1924 |
+
"logits/chosen": -2.803732395172119,
|
| 1925 |
+
"logits/rejected": -2.8339738845825195,
|
| 1926 |
+
"logps/chosen": -336.16632080078125,
|
| 1927 |
+
"logps/rejected": -346.7503662109375,
|
| 1928 |
+
"loss": 0.6767,
|
| 1929 |
+
"positive_losses": 0.0,
|
| 1930 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1931 |
+
"rewards/chosen": 0.10950883477926254,
|
| 1932 |
+
"rewards/margins": 0.04584169760346413,
|
| 1933 |
+
"rewards/margins_max": 0.0953935980796814,
|
| 1934 |
+
"rewards/margins_min": -0.003710202407091856,
|
| 1935 |
+
"rewards/margins_std": 0.07007695734500885,
|
| 1936 |
+
"rewards/rejected": 0.06366713345050812,
|
| 1937 |
+
"step": 870
|
| 1938 |
+
},
|
| 1939 |
+
{
|
| 1940 |
+
"dpo_losses": 0.663690447807312,
|
| 1941 |
+
"epoch": 2.67,
|
| 1942 |
+
"grad_norm": 2.5335259272331125,
|
| 1943 |
+
"learning_rate": 1.7699419904640022e-08,
|
| 1944 |
+
"logits/chosen": -2.864894151687622,
|
| 1945 |
+
"logits/rejected": -2.822705030441284,
|
| 1946 |
+
"logps/chosen": -253.73703002929688,
|
| 1947 |
+
"logps/rejected": -260.36767578125,
|
| 1948 |
+
"loss": 0.6763,
|
| 1949 |
+
"positive_losses": 0.0,
|
| 1950 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1951 |
+
"rewards/chosen": 0.1016063317656517,
|
| 1952 |
+
"rewards/margins": 0.060955654829740524,
|
| 1953 |
+
"rewards/margins_max": 0.10850463807582855,
|
| 1954 |
+
"rewards/margins_min": 0.013406666927039623,
|
| 1955 |
+
"rewards/margins_std": 0.06724441796541214,
|
| 1956 |
+
"rewards/rejected": 0.040650688111782074,
|
| 1957 |
+
"step": 880
|
| 1958 |
+
},
|
| 1959 |
+
{
|
| 1960 |
+
"dpo_losses": 0.6715737581253052,
|
| 1961 |
+
"epoch": 2.71,
|
| 1962 |
+
"grad_norm": 11.524954878911132,
|
| 1963 |
+
"learning_rate": 1.4576761132127912e-08,
|
| 1964 |
+
"logits/chosen": -3.0019936561584473,
|
| 1965 |
+
"logits/rejected": -2.8649494647979736,
|
| 1966 |
+
"logps/chosen": -330.13470458984375,
|
| 1967 |
+
"logps/rejected": -248.55508422851562,
|
| 1968 |
+
"loss": 0.674,
|
| 1969 |
+
"positive_losses": 0.0,
|
| 1970 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1971 |
+
"rewards/chosen": 0.07323987782001495,
|
| 1972 |
+
"rewards/margins": 0.044342849403619766,
|
| 1973 |
+
"rewards/margins_max": 0.0754455104470253,
|
| 1974 |
+
"rewards/margins_min": 0.013240176253020763,
|
| 1975 |
+
"rewards/margins_std": 0.04398581385612488,
|
| 1976 |
+
"rewards/rejected": 0.028897028416395187,
|
| 1977 |
+
"step": 890
|
| 1978 |
+
},
|
| 1979 |
+
{
|
| 1980 |
+
"dpo_losses": 0.6749640107154846,
|
| 1981 |
+
"epoch": 2.74,
|
| 1982 |
+
"grad_norm": 1.8036243560449348,
|
| 1983 |
+
"learning_rate": 1.1748732956682023e-08,
|
| 1984 |
+
"logits/chosen": -2.7051098346710205,
|
| 1985 |
+
"logits/rejected": -2.655970335006714,
|
| 1986 |
+
"logps/chosen": -257.9394836425781,
|
| 1987 |
+
"logps/rejected": -181.5701446533203,
|
| 1988 |
+
"loss": 0.6739,
|
| 1989 |
+
"positive_losses": 0.0,
|
| 1990 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 1991 |
+
"rewards/chosen": 0.07508902996778488,
|
| 1992 |
+
"rewards/margins": 0.03800603747367859,
|
| 1993 |
+
"rewards/margins_max": 0.09310616552829742,
|
| 1994 |
+
"rewards/margins_min": -0.017094094306230545,
|
| 1995 |
+
"rewards/margins_std": 0.07792335003614426,
|
| 1996 |
+
"rewards/rejected": 0.03708299994468689,
|
| 1997 |
+
"step": 900
|
| 1998 |
+
},
|
| 1999 |
+
{
|
| 2000 |
+
"epoch": 2.74,
|
| 2001 |
+
"eval_dpo_losses": 0.679327130317688,
|
| 2002 |
+
"eval_logits/chosen": -2.8477160930633545,
|
| 2003 |
+
"eval_logits/rejected": -2.802847385406494,
|
| 2004 |
+
"eval_logps/chosen": -277.20367431640625,
|
| 2005 |
+
"eval_logps/rejected": -254.03451538085938,
|
| 2006 |
+
"eval_loss": 0.6877949833869934,
|
| 2007 |
+
"eval_positive_losses": 0.06513993442058563,
|
| 2008 |
+
"eval_rewards/accuracies": 0.670634925365448,
|
| 2009 |
+
"eval_rewards/chosen": 0.08017554879188538,
|
| 2010 |
+
"eval_rewards/margins": 0.028694164007902145,
|
| 2011 |
+
"eval_rewards/margins_max": 0.09142320603132248,
|
| 2012 |
+
"eval_rewards/margins_min": -0.03173976391553879,
|
| 2013 |
+
"eval_rewards/margins_std": 0.05502929911017418,
|
| 2014 |
+
"eval_rewards/rejected": 0.05148139223456383,
|
| 2015 |
+
"eval_runtime": 282.246,
|
| 2016 |
+
"eval_samples_per_second": 7.086,
|
| 2017 |
+
"eval_steps_per_second": 0.223,
|
| 2018 |
+
"step": 900
|
| 2019 |
+
},
|
| 2020 |
+
{
|
| 2021 |
+
"dpo_losses": 0.6807221174240112,
|
| 2022 |
+
"epoch": 2.77,
|
| 2023 |
+
"grad_norm": 7.587296620491664,
|
| 2024 |
+
"learning_rate": 9.218874636507556e-09,
|
| 2025 |
+
"logits/chosen": -2.8844008445739746,
|
| 2026 |
+
"logits/rejected": -2.8472609519958496,
|
| 2027 |
+
"logps/chosen": -232.85934448242188,
|
| 2028 |
+
"logps/rejected": -256.0882873535156,
|
| 2029 |
+
"loss": 0.6782,
|
| 2030 |
+
"positive_losses": 0.17994289100170135,
|
| 2031 |
+
"rewards/accuracies": 0.699999988079071,
|
| 2032 |
+
"rewards/chosen": 0.07232363522052765,
|
| 2033 |
+
"rewards/margins": 0.02574983797967434,
|
| 2034 |
+
"rewards/margins_max": 0.06079038232564926,
|
| 2035 |
+
"rewards/margins_min": -0.009290714748203754,
|
| 2036 |
+
"rewards/margins_std": 0.049554821103811264,
|
| 2037 |
+
"rewards/rejected": 0.04657380282878876,
|
| 2038 |
+
"step": 910
|
| 2039 |
+
},
|
| 2040 |
+
{
|
| 2041 |
+
"dpo_losses": 0.674708902835846,
|
| 2042 |
+
"epoch": 2.8,
|
| 2043 |
+
"grad_norm": 2.0209217117752267,
|
| 2044 |
+
"learning_rate": 6.9903522722450645e-09,
|
| 2045 |
+
"logits/chosen": -2.6785788536071777,
|
| 2046 |
+
"logits/rejected": -2.683105945587158,
|
| 2047 |
+
"logps/chosen": -205.6979522705078,
|
| 2048 |
+
"logps/rejected": -194.4645233154297,
|
| 2049 |
+
"loss": 0.6772,
|
| 2050 |
+
"positive_losses": 0.0,
|
| 2051 |
+
"rewards/accuracies": 0.800000011920929,
|
| 2052 |
+
"rewards/chosen": 0.06881477683782578,
|
| 2053 |
+
"rewards/margins": 0.03808169811964035,
|
| 2054 |
+
"rewards/margins_max": 0.06708801537752151,
|
| 2055 |
+
"rewards/margins_min": 0.00907537154853344,
|
| 2056 |
+
"rewards/margins_std": 0.041021138429641724,
|
| 2057 |
+
"rewards/rejected": 0.030733084306120872,
|
| 2058 |
+
"step": 920
|
| 2059 |
+
},
|
| 2060 |
+
{
|
| 2061 |
+
"dpo_losses": 0.6782993078231812,
|
| 2062 |
+
"epoch": 2.83,
|
| 2063 |
+
"grad_norm": 2.1553504283654608,
|
| 2064 |
+
"learning_rate": 5.065954844616721e-09,
|
| 2065 |
+
"logits/chosen": -2.8473258018493652,
|
| 2066 |
+
"logits/rejected": -2.778991222381592,
|
| 2067 |
+
"logps/chosen": -272.9423522949219,
|
| 2068 |
+
"logps/rejected": -323.91290283203125,
|
| 2069 |
+
"loss": 0.6829,
|
| 2070 |
+
"positive_losses": 0.127574160695076,
|
| 2071 |
+
"rewards/accuracies": 0.699999988079071,
|
| 2072 |
+
"rewards/chosen": 0.07656483352184296,
|
| 2073 |
+
"rewards/margins": 0.03158252686262131,
|
| 2074 |
+
"rewards/margins_max": 0.08533596247434616,
|
| 2075 |
+
"rewards/margins_min": -0.022170905023813248,
|
| 2076 |
+
"rewards/margins_std": 0.07601883262395859,
|
| 2077 |
+
"rewards/rejected": 0.04498231038451195,
|
| 2078 |
+
"step": 930
|
| 2079 |
+
},
|
| 2080 |
+
{
|
| 2081 |
+
"dpo_losses": 0.6669495701789856,
|
| 2082 |
+
"epoch": 2.86,
|
| 2083 |
+
"grad_norm": 1.8894258589844486,
|
| 2084 |
+
"learning_rate": 3.4480907240357904e-09,
|
| 2085 |
+
"logits/chosen": -2.886190176010132,
|
| 2086 |
+
"logits/rejected": -2.852233409881592,
|
| 2087 |
+
"logps/chosen": -237.05038452148438,
|
| 2088 |
+
"logps/rejected": -192.5042724609375,
|
| 2089 |
+
"loss": 0.6785,
|
| 2090 |
+
"positive_losses": 0.08552627265453339,
|
| 2091 |
+
"rewards/accuracies": 0.800000011920929,
|
| 2092 |
+
"rewards/chosen": 0.08725001662969589,
|
| 2093 |
+
"rewards/margins": 0.05445738509297371,
|
| 2094 |
+
"rewards/margins_max": 0.08135871589183807,
|
| 2095 |
+
"rewards/margins_min": 0.027556050568819046,
|
| 2096 |
+
"rewards/margins_std": 0.03804423287510872,
|
| 2097 |
+
"rewards/rejected": 0.03279264643788338,
|
| 2098 |
+
"step": 940
|
| 2099 |
+
},
|
| 2100 |
+
{
|
| 2101 |
+
"dpo_losses": 0.6740903854370117,
|
| 2102 |
+
"epoch": 2.89,
|
| 2103 |
+
"grad_norm": 7.272085291531218,
|
| 2104 |
+
"learning_rate": 2.1387846565474044e-09,
|
| 2105 |
+
"logits/chosen": -2.7927393913269043,
|
| 2106 |
+
"logits/rejected": -2.80344295501709,
|
| 2107 |
+
"logps/chosen": -295.1185607910156,
|
| 2108 |
+
"logps/rejected": -264.1958312988281,
|
| 2109 |
+
"loss": 0.6773,
|
| 2110 |
+
"positive_losses": 0.0,
|
| 2111 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 2112 |
+
"rewards/chosen": 0.09159694612026215,
|
| 2113 |
+
"rewards/margins": 0.039398252964019775,
|
| 2114 |
+
"rewards/margins_max": 0.07913588732481003,
|
| 2115 |
+
"rewards/margins_min": -0.00033939722925424576,
|
| 2116 |
+
"rewards/margins_std": 0.056197524070739746,
|
| 2117 |
+
"rewards/rejected": 0.05219869688153267,
|
| 2118 |
+
"step": 950
|
| 2119 |
+
},
|
| 2120 |
+
{
|
| 2121 |
+
"dpo_losses": 0.676175594329834,
|
| 2122 |
+
"epoch": 2.92,
|
| 2123 |
+
"grad_norm": 2.0704055477321535,
|
| 2124 |
+
"learning_rate": 1.1396752298723499e-09,
|
| 2125 |
+
"logits/chosen": -2.6965384483337402,
|
| 2126 |
+
"logits/rejected": -2.7728757858276367,
|
| 2127 |
+
"logps/chosen": -243.5664825439453,
|
| 2128 |
+
"logps/rejected": -246.42904663085938,
|
| 2129 |
+
"loss": 0.6762,
|
| 2130 |
+
"positive_losses": 0.0,
|
| 2131 |
+
"rewards/accuracies": 0.699999988079071,
|
| 2132 |
+
"rewards/chosen": 0.07029319554567337,
|
| 2133 |
+
"rewards/margins": 0.03469633311033249,
|
| 2134 |
+
"rewards/margins_max": 0.0634746104478836,
|
| 2135 |
+
"rewards/margins_min": 0.005918048322200775,
|
| 2136 |
+
"rewards/margins_std": 0.040698640048503876,
|
| 2137 |
+
"rewards/rejected": 0.03559685871005058,
|
| 2138 |
+
"step": 960
|
| 2139 |
+
},
|
| 2140 |
+
{
|
| 2141 |
+
"dpo_losses": 0.6741344332695007,
|
| 2142 |
+
"epoch": 2.95,
|
| 2143 |
+
"grad_norm": 4.375655860786601,
|
| 2144 |
+
"learning_rate": 4.520128227257225e-10,
|
| 2145 |
+
"logits/chosen": -2.9053943157196045,
|
| 2146 |
+
"logits/rejected": -2.903526782989502,
|
| 2147 |
+
"logps/chosen": -324.132568359375,
|
| 2148 |
+
"logps/rejected": -269.60394287109375,
|
| 2149 |
+
"loss": 0.6787,
|
| 2150 |
+
"positive_losses": 0.0,
|
| 2151 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 2152 |
+
"rewards/chosen": 0.10518592596054077,
|
| 2153 |
+
"rewards/margins": 0.04040554165840149,
|
| 2154 |
+
"rewards/margins_max": 0.09129390865564346,
|
| 2155 |
+
"rewards/margins_min": -0.010482819750905037,
|
| 2156 |
+
"rewards/margins_std": 0.07196702063083649,
|
| 2157 |
+
"rewards/rejected": 0.06478038430213928,
|
| 2158 |
+
"step": 970
|
| 2159 |
+
},
|
| 2160 |
+
{
|
| 2161 |
+
"dpo_losses": 0.6782342791557312,
|
| 2162 |
+
"epoch": 2.98,
|
| 2163 |
+
"grad_norm": 2.102961531885413,
|
| 2164 |
+
"learning_rate": 7.665803997633502e-11,
|
| 2165 |
+
"logits/chosen": -2.785641670227051,
|
| 2166 |
+
"logits/rejected": -2.674614667892456,
|
| 2167 |
+
"logps/chosen": -221.5739288330078,
|
| 2168 |
+
"logps/rejected": -235.2669677734375,
|
| 2169 |
+
"loss": 0.6753,
|
| 2170 |
+
"positive_losses": 0.0,
|
| 2171 |
+
"rewards/accuracies": 0.75,
|
| 2172 |
+
"rewards/chosen": 0.06868946552276611,
|
| 2173 |
+
"rewards/margins": 0.030555952340364456,
|
| 2174 |
+
"rewards/margins_max": 0.06132732704281807,
|
| 2175 |
+
"rewards/margins_min": -0.00021542570902965963,
|
| 2176 |
+
"rewards/margins_std": 0.04351729154586792,
|
| 2177 |
+
"rewards/rejected": 0.03813350945711136,
|
| 2178 |
+
"step": 980
|
| 2179 |
+
},
|
| 2180 |
+
{
|
| 2181 |
+
"epoch": 3.0,
|
| 2182 |
+
"step": 987,
|
| 2183 |
+
"total_flos": 0.0,
|
| 2184 |
+
"train_loss": 0.684076373458754,
|
| 2185 |
+
"train_runtime": 8695.656,
|
| 2186 |
+
"train_samples_per_second": 1.816,
|
| 2187 |
+
"train_steps_per_second": 0.114
|
| 2188 |
+
}
|
| 2189 |
+
],
|
| 2190 |
+
"logging_steps": 10,
|
| 2191 |
+
"max_steps": 987,
|
| 2192 |
+
"num_input_tokens_seen": 0,
|
| 2193 |
+
"num_train_epochs": 3,
|
| 2194 |
+
"save_steps": 100,
|
| 2195 |
+
"total_flos": 0.0,
|
| 2196 |
+
"train_batch_size": 2,
|
| 2197 |
+
"trial_name": null,
|
| 2198 |
+
"trial_params": null
|
| 2199 |
+
}
|