Model save
Browse files
README.md
ADDED
@@ -0,0 +1,82 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
base_model: alignment-handbook/zephyr-7b-sft-full
|
3 |
+
library_name: peft
|
4 |
+
license: apache-2.0
|
5 |
+
tags:
|
6 |
+
- trl
|
7 |
+
- dpo
|
8 |
+
- generated_from_trainer
|
9 |
+
model-index:
|
10 |
+
- name: zephyr-dpop-qlora-uf-ours-5e-6-epoch1
|
11 |
+
results: []
|
12 |
+
---
|
13 |
+
|
14 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
15 |
+
should probably proofread and complete it, then remove this comment. -->
|
16 |
+
|
17 |
+
# zephyr-dpop-qlora-uf-ours-5e-6-epoch1
|
18 |
+
|
19 |
+
This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full) on the None dataset.
|
20 |
+
It achieves the following results on the evaluation set:
|
21 |
+
- Loss: 1.6048
|
22 |
+
- Positive Losses: 8.7062
|
23 |
+
- Dpo Losses: 0.6448
|
24 |
+
- Rewards/chosen: 0.0014
|
25 |
+
- Rewards/rejected: -0.1274
|
26 |
+
- Rewards/accuracies: 0.6470
|
27 |
+
- Rewards/margins: 0.1288
|
28 |
+
- Rewards/margins Max: 0.6375
|
29 |
+
- Rewards/margins Min: -0.3461
|
30 |
+
- Rewards/margins Std: 0.3295
|
31 |
+
- Logps/rejected: -271.3224
|
32 |
+
- Logps/chosen: -284.4528
|
33 |
+
- Logits/rejected: -2.6733
|
34 |
+
- Logits/chosen: -2.7153
|
35 |
+
|
36 |
+
## Model description
|
37 |
+
|
38 |
+
More information needed
|
39 |
+
|
40 |
+
## Intended uses & limitations
|
41 |
+
|
42 |
+
More information needed
|
43 |
+
|
44 |
+
## Training and evaluation data
|
45 |
+
|
46 |
+
More information needed
|
47 |
+
|
48 |
+
## Training procedure
|
49 |
+
|
50 |
+
### Training hyperparameters
|
51 |
+
|
52 |
+
The following hyperparameters were used during training:
|
53 |
+
- learning_rate: 5e-06
|
54 |
+
- train_batch_size: 4
|
55 |
+
- eval_batch_size: 8
|
56 |
+
- seed: 42
|
57 |
+
- distributed_type: multi-GPU
|
58 |
+
- num_devices: 2
|
59 |
+
- gradient_accumulation_steps: 2
|
60 |
+
- total_train_batch_size: 16
|
61 |
+
- total_eval_batch_size: 16
|
62 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
63 |
+
- lr_scheduler_type: cosine
|
64 |
+
- lr_scheduler_warmup_ratio: 0.1
|
65 |
+
- num_epochs: 1
|
66 |
+
|
67 |
+
### Training results
|
68 |
+
|
69 |
+
| Training Loss | Epoch | Step | Validation Loss | Positive Losses | Dpo Losses | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Rewards/margins Max | Rewards/margins Min | Rewards/margins Std | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
70 |
+
|:-------------:|:-----:|:----:|:---------------:|:---------------:|:----------:|:--------------:|:----------------:|:------------------:|:---------------:|:-------------------:|:-------------------:|:-------------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
71 |
+
| 0.6232 | 0.28 | 100 | 1.1413 | 4.2610 | 0.6656 | 0.0429 | -0.0252 | 0.6230 | 0.0680 | 0.4086 | -0.2281 | 0.2094 | -261.0972 | -280.3080 | -2.6361 | -2.6726 |
|
72 |
+
| 0.5625 | 0.56 | 200 | 1.7186 | 9.6677 | 0.6469 | -0.0183 | -0.1426 | 0.6420 | 0.1243 | 0.6362 | -0.3433 | 0.3277 | -272.8399 | -286.4236 | -2.6380 | -2.6780 |
|
73 |
+
| 0.4748 | 0.85 | 300 | 1.6048 | 8.7062 | 0.6448 | 0.0014 | -0.1274 | 0.6470 | 0.1288 | 0.6375 | -0.3461 | 0.3295 | -271.3224 | -284.4528 | -2.6733 | -2.7153 |
|
74 |
+
|
75 |
+
|
76 |
+
### Framework versions
|
77 |
+
|
78 |
+
- PEFT 0.7.1
|
79 |
+
- Transformers 4.39.0.dev0
|
80 |
+
- Pytorch 2.1.2+cu121
|
81 |
+
- Datasets 2.14.6
|
82 |
+
- Tokenizers 0.15.2
|
adapter_model.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 671150064
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:157adab1be932a160809b22d2cd46ebc2232e78082ff3d33c796d6ccb4400be7
|
3 |
size 671150064
|
all_results.json
ADDED
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 1.0,
|
3 |
+
"train_loss": 0.5743894765074824,
|
4 |
+
"train_runtime": 4311.1014,
|
5 |
+
"train_samples": 5678,
|
6 |
+
"train_samples_per_second": 1.317,
|
7 |
+
"train_steps_per_second": 0.082
|
8 |
+
}
|
runs/Jul29_11-02-20_notebook-deployment-48-7d9b6c99-khd85/events.out.tfevents.1722251035.notebook-deployment-48-7d9b6c99-khd85.3446409.0
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 sha256:65c29e5ebef7728389473b36ffab0229575c62799e7ab454c0a958d4e15bf729
|
3 |
+
size 44442
|
train_results.json
ADDED
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 1.0,
|
3 |
+
"train_loss": 0.5743894765074824,
|
4 |
+
"train_runtime": 4311.1014,
|
5 |
+
"train_samples": 5678,
|
6 |
+
"train_samples_per_second": 1.317,
|
7 |
+
"train_steps_per_second": 0.082
|
8 |
+
}
|
trainer_state.json
ADDED
@@ -0,0 +1,813 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_metric": null,
|
3 |
+
"best_model_checkpoint": null,
|
4 |
+
"epoch": 1.0,
|
5 |
+
"eval_steps": 100,
|
6 |
+
"global_step": 355,
|
7 |
+
"is_hyper_param_search": false,
|
8 |
+
"is_local_process_zero": true,
|
9 |
+
"is_world_process_zero": true,
|
10 |
+
"log_history": [
|
11 |
+
{
|
12 |
+
"dpo_losses": 0.6931471824645996,
|
13 |
+
"epoch": 0.0,
|
14 |
+
"grad_norm": 1.6018567815095135,
|
15 |
+
"learning_rate": 1.3888888888888888e-07,
|
16 |
+
"logits/chosen": -2.861618995666504,
|
17 |
+
"logits/rejected": -2.8205904960632324,
|
18 |
+
"logps/chosen": -271.06011962890625,
|
19 |
+
"logps/rejected": -211.1704559326172,
|
20 |
+
"loss": 0.6931,
|
21 |
+
"positive_losses": 0.0,
|
22 |
+
"rewards/accuracies": 0.0,
|
23 |
+
"rewards/chosen": 0.0,
|
24 |
+
"rewards/margins": 0.0,
|
25 |
+
"rewards/margins_max": 0.0,
|
26 |
+
"rewards/margins_min": 0.0,
|
27 |
+
"rewards/margins_std": 0.0,
|
28 |
+
"rewards/rejected": 0.0,
|
29 |
+
"step": 1
|
30 |
+
},
|
31 |
+
{
|
32 |
+
"dpo_losses": 0.6928361654281616,
|
33 |
+
"epoch": 0.03,
|
34 |
+
"grad_norm": 14.098492351037597,
|
35 |
+
"learning_rate": 1.3888888888888892e-06,
|
36 |
+
"logits/chosen": -2.8340628147125244,
|
37 |
+
"logits/rejected": -2.7916715145111084,
|
38 |
+
"logps/chosen": -324.87408447265625,
|
39 |
+
"logps/rejected": -274.8518371582031,
|
40 |
+
"loss": 0.6969,
|
41 |
+
"positive_losses": 0.03656284138560295,
|
42 |
+
"rewards/accuracies": 0.5138888955116272,
|
43 |
+
"rewards/chosen": 0.001762597355991602,
|
44 |
+
"rewards/margins": 0.0006246823468245566,
|
45 |
+
"rewards/margins_max": 0.0034460597671568394,
|
46 |
+
"rewards/margins_min": -0.002478615380823612,
|
47 |
+
"rewards/margins_std": 0.002669532783329487,
|
48 |
+
"rewards/rejected": 0.0011379148345440626,
|
49 |
+
"step": 10
|
50 |
+
},
|
51 |
+
{
|
52 |
+
"dpo_losses": 0.6901537775993347,
|
53 |
+
"epoch": 0.06,
|
54 |
+
"grad_norm": 1.829780676576113,
|
55 |
+
"learning_rate": 2.7777777777777783e-06,
|
56 |
+
"logits/chosen": -2.7248008251190186,
|
57 |
+
"logits/rejected": -2.7065372467041016,
|
58 |
+
"logps/chosen": -291.9751892089844,
|
59 |
+
"logps/rejected": -214.52914428710938,
|
60 |
+
"loss": 0.69,
|
61 |
+
"positive_losses": 0.00235748291015625,
|
62 |
+
"rewards/accuracies": 0.7875000238418579,
|
63 |
+
"rewards/chosen": 0.01850745640695095,
|
64 |
+
"rewards/margins": 0.006009287666529417,
|
65 |
+
"rewards/margins_max": 0.013369890861213207,
|
66 |
+
"rewards/margins_min": -0.0006899007130414248,
|
67 |
+
"rewards/margins_std": 0.006301888730376959,
|
68 |
+
"rewards/rejected": 0.01249817106872797,
|
69 |
+
"step": 20
|
70 |
+
},
|
71 |
+
{
|
72 |
+
"dpo_losses": 0.6790497303009033,
|
73 |
+
"epoch": 0.08,
|
74 |
+
"grad_norm": 2.096661038575657,
|
75 |
+
"learning_rate": 4.166666666666667e-06,
|
76 |
+
"logits/chosen": -2.8153939247131348,
|
77 |
+
"logits/rejected": -2.7460672855377197,
|
78 |
+
"logps/chosen": -298.10052490234375,
|
79 |
+
"logps/rejected": -229.7678680419922,
|
80 |
+
"loss": 0.677,
|
81 |
+
"positive_losses": 0.0,
|
82 |
+
"rewards/accuracies": 0.887499988079071,
|
83 |
+
"rewards/chosen": 0.05605363845825195,
|
84 |
+
"rewards/margins": 0.02858993411064148,
|
85 |
+
"rewards/margins_max": 0.058357615023851395,
|
86 |
+
"rewards/margins_min": 0.004640273749828339,
|
87 |
+
"rewards/margins_std": 0.02467900700867176,
|
88 |
+
"rewards/rejected": 0.027463700622320175,
|
89 |
+
"step": 30
|
90 |
+
},
|
91 |
+
{
|
92 |
+
"dpo_losses": 0.6675597429275513,
|
93 |
+
"epoch": 0.11,
|
94 |
+
"grad_norm": 1.7320035926217752,
|
95 |
+
"learning_rate": 4.998060489154965e-06,
|
96 |
+
"logits/chosen": -2.8310070037841797,
|
97 |
+
"logits/rejected": -2.751425266265869,
|
98 |
+
"logps/chosen": -268.48809814453125,
|
99 |
+
"logps/rejected": -222.01107788085938,
|
100 |
+
"loss": 0.6662,
|
101 |
+
"positive_losses": 0.054492950439453125,
|
102 |
+
"rewards/accuracies": 0.862500011920929,
|
103 |
+
"rewards/chosen": 0.08996561169624329,
|
104 |
+
"rewards/margins": 0.05272960662841797,
|
105 |
+
"rewards/margins_max": 0.1101265698671341,
|
106 |
+
"rewards/margins_min": 0.003616312053054571,
|
107 |
+
"rewards/margins_std": 0.048521898686885834,
|
108 |
+
"rewards/rejected": 0.03723599761724472,
|
109 |
+
"step": 40
|
110 |
+
},
|
111 |
+
{
|
112 |
+
"dpo_losses": 0.6397972106933594,
|
113 |
+
"epoch": 0.14,
|
114 |
+
"grad_norm": 9.583890638870626,
|
115 |
+
"learning_rate": 4.976275538042932e-06,
|
116 |
+
"logits/chosen": -2.7891061305999756,
|
117 |
+
"logits/rejected": -2.7175135612487793,
|
118 |
+
"logps/chosen": -262.20794677734375,
|
119 |
+
"logps/rejected": -231.79653930664062,
|
120 |
+
"loss": 0.6446,
|
121 |
+
"positive_losses": 0.0,
|
122 |
+
"rewards/accuracies": 0.9375,
|
123 |
+
"rewards/chosen": 0.13362163305282593,
|
124 |
+
"rewards/margins": 0.11281381547451019,
|
125 |
+
"rewards/margins_max": 0.23626498878002167,
|
126 |
+
"rewards/margins_min": 0.022470083087682724,
|
127 |
+
"rewards/margins_std": 0.0988926962018013,
|
128 |
+
"rewards/rejected": 0.02080780453979969,
|
129 |
+
"step": 50
|
130 |
+
},
|
131 |
+
{
|
132 |
+
"dpo_losses": 0.6110584139823914,
|
133 |
+
"epoch": 0.17,
|
134 |
+
"grad_norm": 2.0747443213986694,
|
135 |
+
"learning_rate": 4.93049306999712e-06,
|
136 |
+
"logits/chosen": -2.7118520736694336,
|
137 |
+
"logits/rejected": -2.6753315925598145,
|
138 |
+
"logps/chosen": -296.9767150878906,
|
139 |
+
"logps/rejected": -263.8233947753906,
|
140 |
+
"loss": 0.628,
|
141 |
+
"positive_losses": 0.011554336175322533,
|
142 |
+
"rewards/accuracies": 0.987500011920929,
|
143 |
+
"rewards/chosen": 0.16662926971912384,
|
144 |
+
"rewards/margins": 0.17714819312095642,
|
145 |
+
"rewards/margins_max": 0.30765318870544434,
|
146 |
+
"rewards/margins_min": 0.05318903177976608,
|
147 |
+
"rewards/margins_std": 0.11578011512756348,
|
148 |
+
"rewards/rejected": -0.010518952272832394,
|
149 |
+
"step": 60
|
150 |
+
},
|
151 |
+
{
|
152 |
+
"dpo_losses": 0.6022371053695679,
|
153 |
+
"epoch": 0.2,
|
154 |
+
"grad_norm": 1.5871888283763238,
|
155 |
+
"learning_rate": 4.861156761634014e-06,
|
156 |
+
"logits/chosen": -2.7271430492401123,
|
157 |
+
"logits/rejected": -2.6688759326934814,
|
158 |
+
"logps/chosen": -303.47613525390625,
|
159 |
+
"logps/rejected": -236.2406463623047,
|
160 |
+
"loss": 0.6175,
|
161 |
+
"positive_losses": 0.19450588524341583,
|
162 |
+
"rewards/accuracies": 0.949999988079071,
|
163 |
+
"rewards/chosen": 0.19385087490081787,
|
164 |
+
"rewards/margins": 0.19984133541584015,
|
165 |
+
"rewards/margins_max": 0.4134605824947357,
|
166 |
+
"rewards/margins_min": 0.04761160537600517,
|
167 |
+
"rewards/margins_std": 0.16880682110786438,
|
168 |
+
"rewards/rejected": -0.00599044980481267,
|
169 |
+
"step": 70
|
170 |
+
},
|
171 |
+
{
|
172 |
+
"dpo_losses": 0.5768495798110962,
|
173 |
+
"epoch": 0.23,
|
174 |
+
"grad_norm": 1.804849988880195,
|
175 |
+
"learning_rate": 4.7689385491773934e-06,
|
176 |
+
"logits/chosen": -2.738285779953003,
|
177 |
+
"logits/rejected": -2.684203863143921,
|
178 |
+
"logps/chosen": -300.8853454589844,
|
179 |
+
"logps/rejected": -292.05633544921875,
|
180 |
+
"loss": 0.6017,
|
181 |
+
"positive_losses": 0.328561395406723,
|
182 |
+
"rewards/accuracies": 0.987500011920929,
|
183 |
+
"rewards/chosen": 0.20385125279426575,
|
184 |
+
"rewards/margins": 0.26062771677970886,
|
185 |
+
"rewards/margins_max": 0.4970013201236725,
|
186 |
+
"rewards/margins_min": 0.05170217156410217,
|
187 |
+
"rewards/margins_std": 0.2058703452348709,
|
188 |
+
"rewards/rejected": -0.056776486337184906,
|
189 |
+
"step": 80
|
190 |
+
},
|
191 |
+
{
|
192 |
+
"dpo_losses": 0.5672236084938049,
|
193 |
+
"epoch": 0.25,
|
194 |
+
"grad_norm": 2.184742961229221,
|
195 |
+
"learning_rate": 4.654732116743193e-06,
|
196 |
+
"logits/chosen": -2.6370556354522705,
|
197 |
+
"logits/rejected": -2.601066827774048,
|
198 |
+
"logps/chosen": -252.70535278320312,
|
199 |
+
"logps/rejected": -203.89418029785156,
|
200 |
+
"loss": 0.5769,
|
201 |
+
"positive_losses": 0.07196970283985138,
|
202 |
+
"rewards/accuracies": 0.9750000238418579,
|
203 |
+
"rewards/chosen": 0.2328944206237793,
|
204 |
+
"rewards/margins": 0.2819642424583435,
|
205 |
+
"rewards/margins_max": 0.514846682548523,
|
206 |
+
"rewards/margins_min": 0.09985215216875076,
|
207 |
+
"rewards/margins_std": 0.19404996931552887,
|
208 |
+
"rewards/rejected": -0.049069829285144806,
|
209 |
+
"step": 90
|
210 |
+
},
|
211 |
+
{
|
212 |
+
"dpo_losses": 0.5702880620956421,
|
213 |
+
"epoch": 0.28,
|
214 |
+
"grad_norm": 2.550586173059517,
|
215 |
+
"learning_rate": 4.5196442356717526e-06,
|
216 |
+
"logits/chosen": -2.6703598499298096,
|
217 |
+
"logits/rejected": -2.6374478340148926,
|
218 |
+
"logps/chosen": -264.9583740234375,
|
219 |
+
"logps/rejected": -273.49615478515625,
|
220 |
+
"loss": 0.6232,
|
221 |
+
"positive_losses": 1.2302151918411255,
|
222 |
+
"rewards/accuracies": 0.949999988079071,
|
223 |
+
"rewards/chosen": 0.16453364491462708,
|
224 |
+
"rewards/margins": 0.27762115001678467,
|
225 |
+
"rewards/margins_max": 0.5491287708282471,
|
226 |
+
"rewards/margins_min": 0.05581303685903549,
|
227 |
+
"rewards/margins_std": 0.22483690083026886,
|
228 |
+
"rewards/rejected": -0.113087497651577,
|
229 |
+
"step": 100
|
230 |
+
},
|
231 |
+
{
|
232 |
+
"epoch": 0.28,
|
233 |
+
"eval_dpo_losses": 0.6656126976013184,
|
234 |
+
"eval_logits/chosen": -2.67258620262146,
|
235 |
+
"eval_logits/rejected": -2.6360833644866943,
|
236 |
+
"eval_logps/chosen": -280.30804443359375,
|
237 |
+
"eval_logps/rejected": -261.0971984863281,
|
238 |
+
"eval_loss": 1.1412982940673828,
|
239 |
+
"eval_positive_losses": 4.261031627655029,
|
240 |
+
"eval_rewards/accuracies": 0.6230000257492065,
|
241 |
+
"eval_rewards/chosen": 0.04285382851958275,
|
242 |
+
"eval_rewards/margins": 0.06803657114505768,
|
243 |
+
"eval_rewards/margins_max": 0.40864306688308716,
|
244 |
+
"eval_rewards/margins_min": -0.22808942198753357,
|
245 |
+
"eval_rewards/margins_std": 0.2094314992427826,
|
246 |
+
"eval_rewards/rejected": -0.02518274076282978,
|
247 |
+
"eval_runtime": 429.2755,
|
248 |
+
"eval_samples_per_second": 4.659,
|
249 |
+
"eval_steps_per_second": 0.291,
|
250 |
+
"step": 100
|
251 |
+
},
|
252 |
+
{
|
253 |
+
"dpo_losses": 0.5097740888595581,
|
254 |
+
"epoch": 0.31,
|
255 |
+
"grad_norm": 6.336382416368574,
|
256 |
+
"learning_rate": 4.364984038837727e-06,
|
257 |
+
"logits/chosen": -2.742903709411621,
|
258 |
+
"logits/rejected": -2.654869318008423,
|
259 |
+
"logps/chosen": -349.24517822265625,
|
260 |
+
"logps/rejected": -304.54730224609375,
|
261 |
+
"loss": 0.543,
|
262 |
+
"positive_losses": 0.44344156980514526,
|
263 |
+
"rewards/accuracies": 1.0,
|
264 |
+
"rewards/chosen": 0.25360527634620667,
|
265 |
+
"rewards/margins": 0.43474069237709045,
|
266 |
+
"rewards/margins_max": 0.7704421281814575,
|
267 |
+
"rewards/margins_min": 0.1366521120071411,
|
268 |
+
"rewards/margins_std": 0.2834155559539795,
|
269 |
+
"rewards/rejected": -0.1811354160308838,
|
270 |
+
"step": 110
|
271 |
+
},
|
272 |
+
{
|
273 |
+
"dpo_losses": 0.518837571144104,
|
274 |
+
"epoch": 0.34,
|
275 |
+
"grad_norm": 2.194144050007341,
|
276 |
+
"learning_rate": 4.192250333880045e-06,
|
277 |
+
"logits/chosen": -2.7281386852264404,
|
278 |
+
"logits/rejected": -2.670868396759033,
|
279 |
+
"logps/chosen": -321.75982666015625,
|
280 |
+
"logps/rejected": -280.87091064453125,
|
281 |
+
"loss": 0.5524,
|
282 |
+
"positive_losses": 0.46012669801712036,
|
283 |
+
"rewards/accuracies": 0.9750000238418579,
|
284 |
+
"rewards/chosen": 0.26626402139663696,
|
285 |
+
"rewards/margins": 0.4130307137966156,
|
286 |
+
"rewards/margins_max": 0.7945607900619507,
|
287 |
+
"rewards/margins_min": 0.14706461131572723,
|
288 |
+
"rewards/margins_std": 0.2963123917579651,
|
289 |
+
"rewards/rejected": -0.14676669239997864,
|
290 |
+
"step": 120
|
291 |
+
},
|
292 |
+
{
|
293 |
+
"dpo_losses": 0.4917011260986328,
|
294 |
+
"epoch": 0.37,
|
295 |
+
"grad_norm": 1.7534787479023215,
|
296 |
+
"learning_rate": 4.0031170782990214e-06,
|
297 |
+
"logits/chosen": -2.711912155151367,
|
298 |
+
"logits/rejected": -2.634033441543579,
|
299 |
+
"logps/chosen": -353.554443359375,
|
300 |
+
"logps/rejected": -320.6388244628906,
|
301 |
+
"loss": 0.5518,
|
302 |
+
"positive_losses": 0.8977662920951843,
|
303 |
+
"rewards/accuracies": 0.987500011920929,
|
304 |
+
"rewards/chosen": 0.2880980372428894,
|
305 |
+
"rewards/margins": 0.4901772439479828,
|
306 |
+
"rewards/margins_max": 0.8924927711486816,
|
307 |
+
"rewards/margins_min": 0.1499636471271515,
|
308 |
+
"rewards/margins_std": 0.3346417546272278,
|
309 |
+
"rewards/rejected": -0.20207922160625458,
|
310 |
+
"step": 130
|
311 |
+
},
|
312 |
+
{
|
313 |
+
"dpo_losses": 0.4866393208503723,
|
314 |
+
"epoch": 0.39,
|
315 |
+
"grad_norm": 21.27134583914694,
|
316 |
+
"learning_rate": 3.7994171571810756e-06,
|
317 |
+
"logits/chosen": -2.6895060539245605,
|
318 |
+
"logits/rejected": -2.6512811183929443,
|
319 |
+
"logps/chosen": -291.05548095703125,
|
320 |
+
"logps/rejected": -294.4687805175781,
|
321 |
+
"loss": 0.5718,
|
322 |
+
"positive_losses": 0.2207096517086029,
|
323 |
+
"rewards/accuracies": 0.987500011920929,
|
324 |
+
"rewards/chosen": 0.2735855281352997,
|
325 |
+
"rewards/margins": 0.5197780132293701,
|
326 |
+
"rewards/margins_max": 1.003483772277832,
|
327 |
+
"rewards/margins_min": 0.1269286870956421,
|
328 |
+
"rewards/margins_std": 0.3979441523551941,
|
329 |
+
"rewards/rejected": -0.24619252979755402,
|
330 |
+
"step": 140
|
331 |
+
},
|
332 |
+
{
|
333 |
+
"dpo_losses": 0.5046078562736511,
|
334 |
+
"epoch": 0.42,
|
335 |
+
"grad_norm": 3.3011186957688583,
|
336 |
+
"learning_rate": 3.5831246207606597e-06,
|
337 |
+
"logits/chosen": -2.6959190368652344,
|
338 |
+
"logits/rejected": -2.658679962158203,
|
339 |
+
"logps/chosen": -264.2646179199219,
|
340 |
+
"logps/rejected": -234.5491180419922,
|
341 |
+
"loss": 0.5366,
|
342 |
+
"positive_losses": 0.490040123462677,
|
343 |
+
"rewards/accuracies": 0.949999988079071,
|
344 |
+
"rewards/chosen": 0.24420371651649475,
|
345 |
+
"rewards/margins": 0.45886701345443726,
|
346 |
+
"rewards/margins_max": 0.8680801391601562,
|
347 |
+
"rewards/margins_min": 0.1154303103685379,
|
348 |
+
"rewards/margins_std": 0.34930768609046936,
|
349 |
+
"rewards/rejected": -0.2146632969379425,
|
350 |
+
"step": 150
|
351 |
+
},
|
352 |
+
{
|
353 |
+
"dpo_losses": 0.48088502883911133,
|
354 |
+
"epoch": 0.45,
|
355 |
+
"grad_norm": 2.135658014816511,
|
356 |
+
"learning_rate": 3.3563355539546795e-06,
|
357 |
+
"logits/chosen": -2.665548801422119,
|
358 |
+
"logits/rejected": -2.6138901710510254,
|
359 |
+
"logps/chosen": -274.263427734375,
|
360 |
+
"logps/rejected": -260.50518798828125,
|
361 |
+
"loss": 0.5724,
|
362 |
+
"positive_losses": 0.9731669425964355,
|
363 |
+
"rewards/accuracies": 0.987500011920929,
|
364 |
+
"rewards/chosen": 0.262834370136261,
|
365 |
+
"rewards/margins": 0.5239533185958862,
|
366 |
+
"rewards/margins_max": 0.9884392023086548,
|
367 |
+
"rewards/margins_min": 0.15575796365737915,
|
368 |
+
"rewards/margins_std": 0.3754872977733612,
|
369 |
+
"rewards/rejected": -0.26111894845962524,
|
370 |
+
"step": 160
|
371 |
+
},
|
372 |
+
{
|
373 |
+
"dpo_losses": 0.4504636824131012,
|
374 |
+
"epoch": 0.48,
|
375 |
+
"grad_norm": 3.940043763048366,
|
376 |
+
"learning_rate": 3.121247763262235e-06,
|
377 |
+
"logits/chosen": -2.708754777908325,
|
378 |
+
"logits/rejected": -2.657917022705078,
|
379 |
+
"logps/chosen": -297.7489013671875,
|
380 |
+
"logps/rejected": -327.0563049316406,
|
381 |
+
"loss": 0.4813,
|
382 |
+
"positive_losses": 0.03098602220416069,
|
383 |
+
"rewards/accuracies": 0.9624999761581421,
|
384 |
+
"rewards/chosen": 0.3187271058559418,
|
385 |
+
"rewards/margins": 0.6266334652900696,
|
386 |
+
"rewards/margins_max": 1.0517089366912842,
|
387 |
+
"rewards/margins_min": 0.17500966787338257,
|
388 |
+
"rewards/margins_std": 0.3909396231174469,
|
389 |
+
"rewards/rejected": -0.3079063296318054,
|
390 |
+
"step": 170
|
391 |
+
},
|
392 |
+
{
|
393 |
+
"dpo_losses": 0.4588772654533386,
|
394 |
+
"epoch": 0.51,
|
395 |
+
"grad_norm": 8.823245159881209,
|
396 |
+
"learning_rate": 2.8801394778833475e-06,
|
397 |
+
"logits/chosen": -2.6968963146209717,
|
398 |
+
"logits/rejected": -2.6140356063842773,
|
399 |
+
"logps/chosen": -305.4325866699219,
|
400 |
+
"logps/rejected": -326.99798583984375,
|
401 |
+
"loss": 0.5468,
|
402 |
+
"positive_losses": 0.8232825994491577,
|
403 |
+
"rewards/accuracies": 0.987500011920929,
|
404 |
+
"rewards/chosen": 0.2646820843219757,
|
405 |
+
"rewards/margins": 0.5928131937980652,
|
406 |
+
"rewards/margins_max": 1.0361554622650146,
|
407 |
+
"rewards/margins_min": 0.28750157356262207,
|
408 |
+
"rewards/margins_std": 0.33570224046707153,
|
409 |
+
"rewards/rejected": -0.32813113927841187,
|
410 |
+
"step": 180
|
411 |
+
},
|
412 |
+
{
|
413 |
+
"dpo_losses": 0.45539379119873047,
|
414 |
+
"epoch": 0.54,
|
415 |
+
"grad_norm": 3.517893013000186,
|
416 |
+
"learning_rate": 2.6353472714635443e-06,
|
417 |
+
"logits/chosen": -2.6537580490112305,
|
418 |
+
"logits/rejected": -2.5634191036224365,
|
419 |
+
"logps/chosen": -287.6109619140625,
|
420 |
+
"logps/rejected": -265.6959228515625,
|
421 |
+
"loss": 0.5435,
|
422 |
+
"positive_losses": 0.9886103868484497,
|
423 |
+
"rewards/accuracies": 0.949999988079071,
|
424 |
+
"rewards/chosen": 0.34349915385246277,
|
425 |
+
"rewards/margins": 0.6255816221237183,
|
426 |
+
"rewards/margins_max": 1.1905597448349,
|
427 |
+
"rewards/margins_min": 0.168921560049057,
|
428 |
+
"rewards/margins_std": 0.453277051448822,
|
429 |
+
"rewards/rejected": -0.2820824980735779,
|
430 |
+
"step": 190
|
431 |
+
},
|
432 |
+
{
|
433 |
+
"dpo_losses": 0.44315657019615173,
|
434 |
+
"epoch": 0.56,
|
435 |
+
"grad_norm": 27.976402148032502,
|
436 |
+
"learning_rate": 2.3892434184240536e-06,
|
437 |
+
"logits/chosen": -2.7400636672973633,
|
438 |
+
"logits/rejected": -2.662397623062134,
|
439 |
+
"logps/chosen": -309.39691162109375,
|
440 |
+
"logps/rejected": -299.7530212402344,
|
441 |
+
"loss": 0.5625,
|
442 |
+
"positive_losses": 0.9616166353225708,
|
443 |
+
"rewards/accuracies": 0.9624999761581421,
|
444 |
+
"rewards/chosen": 0.30135902762413025,
|
445 |
+
"rewards/margins": 0.6429153084754944,
|
446 |
+
"rewards/margins_max": 1.131412148475647,
|
447 |
+
"rewards/margins_min": 0.17879006266593933,
|
448 |
+
"rewards/margins_std": 0.4260264039039612,
|
449 |
+
"rewards/rejected": -0.34155628085136414,
|
450 |
+
"step": 200
|
451 |
+
},
|
452 |
+
{
|
453 |
+
"epoch": 0.56,
|
454 |
+
"eval_dpo_losses": 0.6469283699989319,
|
455 |
+
"eval_logits/chosen": -2.678022623062134,
|
456 |
+
"eval_logits/rejected": -2.6380200386047363,
|
457 |
+
"eval_logps/chosen": -286.4236145019531,
|
458 |
+
"eval_logps/rejected": -272.83990478515625,
|
459 |
+
"eval_loss": 1.7185667753219604,
|
460 |
+
"eval_positive_losses": 9.667731285095215,
|
461 |
+
"eval_rewards/accuracies": 0.6420000195503235,
|
462 |
+
"eval_rewards/chosen": -0.018302178010344505,
|
463 |
+
"eval_rewards/margins": 0.1243075579404831,
|
464 |
+
"eval_rewards/margins_max": 0.6361650228500366,
|
465 |
+
"eval_rewards/margins_min": -0.3433184325695038,
|
466 |
+
"eval_rewards/margins_std": 0.32774004340171814,
|
467 |
+
"eval_rewards/rejected": -0.14260973036289215,
|
468 |
+
"eval_runtime": 428.2243,
|
469 |
+
"eval_samples_per_second": 4.67,
|
470 |
+
"eval_steps_per_second": 0.292,
|
471 |
+
"step": 200
|
472 |
+
},
|
473 |
+
{
|
474 |
+
"dpo_losses": 0.4354400634765625,
|
475 |
+
"epoch": 0.59,
|
476 |
+
"grad_norm": 23.522369776083625,
|
477 |
+
"learning_rate": 2.1442129043167877e-06,
|
478 |
+
"logits/chosen": -2.6434009075164795,
|
479 |
+
"logits/rejected": -2.6138339042663574,
|
480 |
+
"logps/chosen": -286.7272033691406,
|
481 |
+
"logps/rejected": -291.8896789550781,
|
482 |
+
"loss": 0.513,
|
483 |
+
"positive_losses": 0.665066123008728,
|
484 |
+
"rewards/accuracies": 0.9624999761581421,
|
485 |
+
"rewards/chosen": 0.33457106351852417,
|
486 |
+
"rewards/margins": 0.6892200708389282,
|
487 |
+
"rewards/margins_max": 1.2269551753997803,
|
488 |
+
"rewards/margins_min": 0.18099449574947357,
|
489 |
+
"rewards/margins_std": 0.46556130051612854,
|
490 |
+
"rewards/rejected": -0.35464900732040405,
|
491 |
+
"step": 210
|
492 |
+
},
|
493 |
+
{
|
494 |
+
"dpo_losses": 0.4387238025665283,
|
495 |
+
"epoch": 0.62,
|
496 |
+
"grad_norm": 11.92404423048434,
|
497 |
+
"learning_rate": 1.9026303129961049e-06,
|
498 |
+
"logits/chosen": -2.7612462043762207,
|
499 |
+
"logits/rejected": -2.664234161376953,
|
500 |
+
"logps/chosen": -319.7461853027344,
|
501 |
+
"logps/rejected": -306.0053405761719,
|
502 |
+
"loss": 0.5894,
|
503 |
+
"positive_losses": 1.1452913284301758,
|
504 |
+
"rewards/accuracies": 0.9750000238418579,
|
505 |
+
"rewards/chosen": 0.33710065484046936,
|
506 |
+
"rewards/margins": 0.6538791656494141,
|
507 |
+
"rewards/margins_max": 1.1509373188018799,
|
508 |
+
"rewards/margins_min": 0.19225715100765228,
|
509 |
+
"rewards/margins_std": 0.4403897225856781,
|
510 |
+
"rewards/rejected": -0.3167785704135895,
|
511 |
+
"step": 220
|
512 |
+
},
|
513 |
+
{
|
514 |
+
"dpo_losses": 0.44511428475379944,
|
515 |
+
"epoch": 0.65,
|
516 |
+
"grad_norm": 2.419282473127918,
|
517 |
+
"learning_rate": 1.66683681459314e-06,
|
518 |
+
"logits/chosen": -2.773876428604126,
|
519 |
+
"logits/rejected": -2.67607045173645,
|
520 |
+
"logps/chosen": -339.04718017578125,
|
521 |
+
"logps/rejected": -293.1225891113281,
|
522 |
+
"loss": 0.4763,
|
523 |
+
"positive_losses": 0.6133368611335754,
|
524 |
+
"rewards/accuracies": 0.987500011920929,
|
525 |
+
"rewards/chosen": 0.32628515362739563,
|
526 |
+
"rewards/margins": 0.6365767121315002,
|
527 |
+
"rewards/margins_max": 1.125410795211792,
|
528 |
+
"rewards/margins_min": 0.21782192587852478,
|
529 |
+
"rewards/margins_std": 0.4051855504512787,
|
530 |
+
"rewards/rejected": -0.3102915287017822,
|
531 |
+
"step": 230
|
532 |
+
},
|
533 |
+
{
|
534 |
+
"dpo_losses": 0.4544529318809509,
|
535 |
+
"epoch": 0.68,
|
536 |
+
"grad_norm": 13.447116267552904,
|
537 |
+
"learning_rate": 1.4391174773015836e-06,
|
538 |
+
"logits/chosen": -2.7197587490081787,
|
539 |
+
"logits/rejected": -2.649749279022217,
|
540 |
+
"logps/chosen": -302.6105041503906,
|
541 |
+
"logps/rejected": -321.8402404785156,
|
542 |
+
"loss": 0.692,
|
543 |
+
"positive_losses": 2.48455810546875,
|
544 |
+
"rewards/accuracies": 0.9375,
|
545 |
+
"rewards/chosen": 0.22186538577079773,
|
546 |
+
"rewards/margins": 0.6085190773010254,
|
547 |
+
"rewards/margins_max": 1.1415433883666992,
|
548 |
+
"rewards/margins_min": 0.23370866477489471,
|
549 |
+
"rewards/margins_std": 0.41311854124069214,
|
550 |
+
"rewards/rejected": -0.38665369153022766,
|
551 |
+
"step": 240
|
552 |
+
},
|
553 |
+
{
|
554 |
+
"dpo_losses": 0.45861634612083435,
|
555 |
+
"epoch": 0.7,
|
556 |
+
"grad_norm": 5.111403689556549,
|
557 |
+
"learning_rate": 1.2216791228457778e-06,
|
558 |
+
"logits/chosen": -2.716823101043701,
|
559 |
+
"logits/rejected": -2.640800952911377,
|
560 |
+
"logps/chosen": -280.11114501953125,
|
561 |
+
"logps/rejected": -281.67138671875,
|
562 |
+
"loss": 0.4992,
|
563 |
+
"positive_losses": 0.6084854006767273,
|
564 |
+
"rewards/accuracies": 0.987500011920929,
|
565 |
+
"rewards/chosen": 0.31169968843460083,
|
566 |
+
"rewards/margins": 0.6179708242416382,
|
567 |
+
"rewards/margins_max": 1.2185614109039307,
|
568 |
+
"rewards/margins_min": 0.1615341305732727,
|
569 |
+
"rewards/margins_std": 0.4740964472293854,
|
570 |
+
"rewards/rejected": -0.30627113580703735,
|
571 |
+
"step": 250
|
572 |
+
},
|
573 |
+
{
|
574 |
+
"dpo_losses": 0.4628082811832428,
|
575 |
+
"epoch": 0.73,
|
576 |
+
"grad_norm": 2.699692592075128,
|
577 |
+
"learning_rate": 1.0166289402331391e-06,
|
578 |
+
"logits/chosen": -2.7728962898254395,
|
579 |
+
"logits/rejected": -2.684753894805908,
|
580 |
+
"logps/chosen": -263.36126708984375,
|
581 |
+
"logps/rejected": -289.21661376953125,
|
582 |
+
"loss": 0.5624,
|
583 |
+
"positive_losses": 0.9304378628730774,
|
584 |
+
"rewards/accuracies": 0.9624999761581421,
|
585 |
+
"rewards/chosen": 0.28732261061668396,
|
586 |
+
"rewards/margins": 0.5901791453361511,
|
587 |
+
"rewards/margins_max": 1.102694034576416,
|
588 |
+
"rewards/margins_min": 0.17682021856307983,
|
589 |
+
"rewards/margins_std": 0.4229150712490082,
|
590 |
+
"rewards/rejected": -0.30285659432411194,
|
591 |
+
"step": 260
|
592 |
+
},
|
593 |
+
{
|
594 |
+
"dpo_losses": 0.4588424265384674,
|
595 |
+
"epoch": 0.76,
|
596 |
+
"grad_norm": 2.4735784513371377,
|
597 |
+
"learning_rate": 8.259540650444736e-07,
|
598 |
+
"logits/chosen": -2.717153787612915,
|
599 |
+
"logits/rejected": -2.662932872772217,
|
600 |
+
"logps/chosen": -278.75482177734375,
|
601 |
+
"logps/rejected": -291.56866455078125,
|
602 |
+
"loss": 0.5853,
|
603 |
+
"positive_losses": 0.9098857641220093,
|
604 |
+
"rewards/accuracies": 0.949999988079071,
|
605 |
+
"rewards/chosen": 0.30360764265060425,
|
606 |
+
"rewards/margins": 0.5942984223365784,
|
607 |
+
"rewards/margins_max": 1.0322821140289307,
|
608 |
+
"rewards/margins_min": 0.21275146305561066,
|
609 |
+
"rewards/margins_std": 0.36198341846466064,
|
610 |
+
"rewards/rejected": -0.2906908392906189,
|
611 |
+
"step": 270
|
612 |
+
},
|
613 |
+
{
|
614 |
+
"dpo_losses": 0.4629085958003998,
|
615 |
+
"epoch": 0.79,
|
616 |
+
"grad_norm": 13.451546074592132,
|
617 |
+
"learning_rate": 6.515023221586722e-07,
|
618 |
+
"logits/chosen": -2.6962451934814453,
|
619 |
+
"logits/rejected": -2.6575076580047607,
|
620 |
+
"logps/chosen": -274.9664001464844,
|
621 |
+
"logps/rejected": -304.9722595214844,
|
622 |
+
"loss": 0.5625,
|
623 |
+
"positive_losses": 1.4465850591659546,
|
624 |
+
"rewards/accuracies": 0.9750000238418579,
|
625 |
+
"rewards/chosen": 0.2849060893058777,
|
626 |
+
"rewards/margins": 0.60865718126297,
|
627 |
+
"rewards/margins_max": 1.1329301595687866,
|
628 |
+
"rewards/margins_min": 0.1755952537059784,
|
629 |
+
"rewards/margins_std": 0.4414794445037842,
|
630 |
+
"rewards/rejected": -0.3237510919570923,
|
631 |
+
"step": 280
|
632 |
+
},
|
633 |
+
{
|
634 |
+
"dpo_losses": 0.47258663177490234,
|
635 |
+
"epoch": 0.82,
|
636 |
+
"grad_norm": 2.654477953260434,
|
637 |
+
"learning_rate": 4.949643185335288e-07,
|
638 |
+
"logits/chosen": -2.707307815551758,
|
639 |
+
"logits/rejected": -2.652792453765869,
|
640 |
+
"logps/chosen": -259.1030578613281,
|
641 |
+
"logps/rejected": -292.6324462890625,
|
642 |
+
"loss": 0.6149,
|
643 |
+
"positive_losses": 1.7202523946762085,
|
644 |
+
"rewards/accuracies": 0.949999988079071,
|
645 |
+
"rewards/chosen": 0.27813172340393066,
|
646 |
+
"rewards/margins": 0.5642735958099365,
|
647 |
+
"rewards/margins_max": 1.0385398864746094,
|
648 |
+
"rewards/margins_min": 0.12702254951000214,
|
649 |
+
"rewards/margins_std": 0.4158683717250824,
|
650 |
+
"rewards/rejected": -0.28614187240600586,
|
651 |
+
"step": 290
|
652 |
+
},
|
653 |
+
{
|
654 |
+
"dpo_losses": 0.4324049949645996,
|
655 |
+
"epoch": 0.85,
|
656 |
+
"grad_norm": 11.591501845708454,
|
657 |
+
"learning_rate": 3.578570595810274e-07,
|
658 |
+
"logits/chosen": -2.7821717262268066,
|
659 |
+
"logits/rejected": -2.6995315551757812,
|
660 |
+
"logps/chosen": -309.7518310546875,
|
661 |
+
"logps/rejected": -320.70916748046875,
|
662 |
+
"loss": 0.4748,
|
663 |
+
"positive_losses": 0.8444260358810425,
|
664 |
+
"rewards/accuracies": 1.0,
|
665 |
+
"rewards/chosen": 0.3676120638847351,
|
666 |
+
"rewards/margins": 0.6803697943687439,
|
667 |
+
"rewards/margins_max": 1.199285864830017,
|
668 |
+
"rewards/margins_min": 0.21351738274097443,
|
669 |
+
"rewards/margins_std": 0.4415613114833832,
|
670 |
+
"rewards/rejected": -0.312757670879364,
|
671 |
+
"step": 300
|
672 |
+
},
|
673 |
+
{
|
674 |
+
"epoch": 0.85,
|
675 |
+
"eval_dpo_losses": 0.6448404788970947,
|
676 |
+
"eval_logits/chosen": -2.715327739715576,
|
677 |
+
"eval_logits/rejected": -2.6732916831970215,
|
678 |
+
"eval_logps/chosen": -284.4527587890625,
|
679 |
+
"eval_logps/rejected": -271.32244873046875,
|
680 |
+
"eval_loss": 1.6048117876052856,
|
681 |
+
"eval_positive_losses": 8.706162452697754,
|
682 |
+
"eval_rewards/accuracies": 0.6470000147819519,
|
683 |
+
"eval_rewards/chosen": 0.0014067561132833362,
|
684 |
+
"eval_rewards/margins": 0.12884218990802765,
|
685 |
+
"eval_rewards/margins_max": 0.6374967098236084,
|
686 |
+
"eval_rewards/margins_min": -0.34605804085731506,
|
687 |
+
"eval_rewards/margins_std": 0.3295030891895294,
|
688 |
+
"eval_rewards/rejected": -0.12743544578552246,
|
689 |
+
"eval_runtime": 428.2498,
|
690 |
+
"eval_samples_per_second": 4.67,
|
691 |
+
"eval_steps_per_second": 0.292,
|
692 |
+
"step": 300
|
693 |
+
},
|
694 |
+
{
|
695 |
+
"dpo_losses": 0.45941466093063354,
|
696 |
+
"epoch": 0.87,
|
697 |
+
"grad_norm": 2.6085680781835205,
|
698 |
+
"learning_rate": 2.4150924791035037e-07,
|
699 |
+
"logits/chosen": -2.774445056915283,
|
700 |
+
"logits/rejected": -2.673360824584961,
|
701 |
+
"logps/chosen": -267.74237060546875,
|
702 |
+
"logps/rejected": -243.88473510742188,
|
703 |
+
"loss": 0.5697,
|
704 |
+
"positive_losses": 1.3653801679611206,
|
705 |
+
"rewards/accuracies": 0.987500011920929,
|
706 |
+
"rewards/chosen": 0.30073457956314087,
|
707 |
+
"rewards/margins": 0.5973426699638367,
|
708 |
+
"rewards/margins_max": 1.1060882806777954,
|
709 |
+
"rewards/margins_min": 0.18351522088050842,
|
710 |
+
"rewards/margins_std": 0.4086340069770813,
|
711 |
+
"rewards/rejected": -0.2966081500053406,
|
712 |
+
"step": 310
|
713 |
+
},
|
714 |
+
{
|
715 |
+
"dpo_losses": 0.45310109853744507,
|
716 |
+
"epoch": 0.9,
|
717 |
+
"grad_norm": 10.060071948421735,
|
718 |
+
"learning_rate": 1.4704840690808658e-07,
|
719 |
+
"logits/chosen": -2.738978385925293,
|
720 |
+
"logits/rejected": -2.680860757827759,
|
721 |
+
"logps/chosen": -279.5138854980469,
|
722 |
+
"logps/rejected": -293.9893493652344,
|
723 |
+
"loss": 0.5692,
|
724 |
+
"positive_losses": 1.6892318725585938,
|
725 |
+
"rewards/accuracies": 0.9624999761581421,
|
726 |
+
"rewards/chosen": 0.2875928282737732,
|
727 |
+
"rewards/margins": 0.6207860708236694,
|
728 |
+
"rewards/margins_max": 1.124011754989624,
|
729 |
+
"rewards/margins_min": 0.14557920396327972,
|
730 |
+
"rewards/margins_std": 0.44626301527023315,
|
731 |
+
"rewards/rejected": -0.33319321274757385,
|
732 |
+
"step": 320
|
733 |
+
},
|
734 |
+
{
|
735 |
+
"dpo_losses": 0.42673492431640625,
|
736 |
+
"epoch": 0.93,
|
737 |
+
"grad_norm": 9.476085880429812,
|
738 |
+
"learning_rate": 7.538995394063996e-08,
|
739 |
+
"logits/chosen": -2.8187005519866943,
|
740 |
+
"logits/rejected": -2.7311813831329346,
|
741 |
+
"logps/chosen": -318.88360595703125,
|
742 |
+
"logps/rejected": -302.66058349609375,
|
743 |
+
"loss": 0.5314,
|
744 |
+
"positive_losses": 0.5069873929023743,
|
745 |
+
"rewards/accuracies": 0.987500011920929,
|
746 |
+
"rewards/chosen": 0.35436224937438965,
|
747 |
+
"rewards/margins": 0.7115713953971863,
|
748 |
+
"rewards/margins_max": 1.23550546169281,
|
749 |
+
"rewards/margins_min": 0.2139424830675125,
|
750 |
+
"rewards/margins_std": 0.4558965563774109,
|
751 |
+
"rewards/rejected": -0.35720914602279663,
|
752 |
+
"step": 330
|
753 |
+
},
|
754 |
+
{
|
755 |
+
"dpo_losses": 0.4437997341156006,
|
756 |
+
"epoch": 0.96,
|
757 |
+
"grad_norm": 2.682118994824555,
|
758 |
+
"learning_rate": 2.722832907015971e-08,
|
759 |
+
"logits/chosen": -2.6981847286224365,
|
760 |
+
"logits/rejected": -2.6440398693084717,
|
761 |
+
"logps/chosen": -266.6497802734375,
|
762 |
+
"logps/rejected": -282.98199462890625,
|
763 |
+
"loss": 0.5024,
|
764 |
+
"positive_losses": 0.9627658724784851,
|
765 |
+
"rewards/accuracies": 0.9750000238418579,
|
766 |
+
"rewards/chosen": 0.3319571018218994,
|
767 |
+
"rewards/margins": 0.6500804424285889,
|
768 |
+
"rewards/margins_max": 1.2494922876358032,
|
769 |
+
"rewards/margins_min": 0.25120097398757935,
|
770 |
+
"rewards/margins_std": 0.4507668614387512,
|
771 |
+
"rewards/rejected": -0.31812337040901184,
|
772 |
+
"step": 340
|
773 |
+
},
|
774 |
+
{
|
775 |
+
"dpo_losses": 0.4518283009529114,
|
776 |
+
"epoch": 0.99,
|
777 |
+
"grad_norm": 5.762126574549782,
|
778 |
+
"learning_rate": 3.030265255329623e-09,
|
779 |
+
"logits/chosen": -2.6820361614227295,
|
780 |
+
"logits/rejected": -2.6376953125,
|
781 |
+
"logps/chosen": -285.1527404785156,
|
782 |
+
"logps/rejected": -317.6675720214844,
|
783 |
+
"loss": 0.5059,
|
784 |
+
"positive_losses": 0.9290813207626343,
|
785 |
+
"rewards/accuracies": 0.987500011920929,
|
786 |
+
"rewards/chosen": 0.2980636656284332,
|
787 |
+
"rewards/margins": 0.6102195978164673,
|
788 |
+
"rewards/margins_max": 1.0686355829238892,
|
789 |
+
"rewards/margins_min": 0.20541608333587646,
|
790 |
+
"rewards/margins_std": 0.38572338223457336,
|
791 |
+
"rewards/rejected": -0.31215590238571167,
|
792 |
+
"step": 350
|
793 |
+
},
|
794 |
+
{
|
795 |
+
"epoch": 1.0,
|
796 |
+
"step": 355,
|
797 |
+
"total_flos": 0.0,
|
798 |
+
"train_loss": 0.5743894765074824,
|
799 |
+
"train_runtime": 4311.1014,
|
800 |
+
"train_samples_per_second": 1.317,
|
801 |
+
"train_steps_per_second": 0.082
|
802 |
+
}
|
803 |
+
],
|
804 |
+
"logging_steps": 10,
|
805 |
+
"max_steps": 355,
|
806 |
+
"num_input_tokens_seen": 0,
|
807 |
+
"num_train_epochs": 1,
|
808 |
+
"save_steps": 100,
|
809 |
+
"total_flos": 0.0,
|
810 |
+
"train_batch_size": 4,
|
811 |
+
"trial_name": null,
|
812 |
+
"trial_params": null
|
813 |
+
}
|