Model save
Browse files- README.md +62 -0
- adapter_model.safetensors +1 -1
- all_results.json +8 -0
- runs/May12_11-10-31_gpu4-119-5/events.out.tfevents.1715476347.gpu4-119-5.2325374.0 +2 -2
- train_results.json +8 -0
- trainer_state.json +954 -0
README.md
ADDED
@@ -0,0 +1,62 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: apache-2.0
|
3 |
+
library_name: peft
|
4 |
+
tags:
|
5 |
+
- trl
|
6 |
+
- dpo
|
7 |
+
- generated_from_trainer
|
8 |
+
base_model: DUAL-GPO/phi-2-gpo-new-i0
|
9 |
+
model-index:
|
10 |
+
- name: phi-2-gpo-v3-i1
|
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 |
+
# phi-2-gpo-v3-i1
|
18 |
+
|
19 |
+
This model is a fine-tuned version of [DUAL-GPO/phi-2-gpo-new-i0](https://huggingface.co/DUAL-GPO/phi-2-gpo-new-i0) on the None dataset.
|
20 |
+
|
21 |
+
## Model description
|
22 |
+
|
23 |
+
More information needed
|
24 |
+
|
25 |
+
## Intended uses & limitations
|
26 |
+
|
27 |
+
More information needed
|
28 |
+
|
29 |
+
## Training and evaluation data
|
30 |
+
|
31 |
+
More information needed
|
32 |
+
|
33 |
+
## Training procedure
|
34 |
+
|
35 |
+
### Training hyperparameters
|
36 |
+
|
37 |
+
The following hyperparameters were used during training:
|
38 |
+
- learning_rate: 5e-06
|
39 |
+
- train_batch_size: 4
|
40 |
+
- eval_batch_size: 4
|
41 |
+
- seed: 42
|
42 |
+
- distributed_type: multi-GPU
|
43 |
+
- num_devices: 2
|
44 |
+
- gradient_accumulation_steps: 4
|
45 |
+
- total_train_batch_size: 32
|
46 |
+
- total_eval_batch_size: 8
|
47 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
48 |
+
- lr_scheduler_type: cosine
|
49 |
+
- lr_scheduler_warmup_ratio: 0.1
|
50 |
+
- num_epochs: 1
|
51 |
+
|
52 |
+
### Training results
|
53 |
+
|
54 |
+
|
55 |
+
|
56 |
+
### Framework versions
|
57 |
+
|
58 |
+
- PEFT 0.7.1
|
59 |
+
- Transformers 4.36.2
|
60 |
+
- Pytorch 2.1.2+cu121
|
61 |
+
- Datasets 2.14.6
|
62 |
+
- 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 167807296
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:7b0d48962bbf459eb9481e45741955de11ee52e3e06b132cbddf3d662da7f745
|
3 |
size 167807296
|
all_results.json
ADDED
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 1.0,
|
3 |
+
"train_loss": 0.2497515143236009,
|
4 |
+
"train_runtime": 7980.5775,
|
5 |
+
"train_samples": 21000,
|
6 |
+
"train_samples_per_second": 2.631,
|
7 |
+
"train_steps_per_second": 0.082
|
8 |
+
}
|
runs/May12_11-10-31_gpu4-119-5/events.out.tfevents.1715476347.gpu4-119-5.2325374.0
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:aefc2f6a59bf125b28d6db3cd943c742e7238b0ade2963d4ab3b6065c3ef6cc8
|
3 |
+
size 46878
|
train_results.json
ADDED
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 1.0,
|
3 |
+
"train_loss": 0.2497515143236009,
|
4 |
+
"train_runtime": 7980.5775,
|
5 |
+
"train_samples": 21000,
|
6 |
+
"train_samples_per_second": 2.631,
|
7 |
+
"train_steps_per_second": 0.082
|
8 |
+
}
|
trainer_state.json
ADDED
@@ -0,0 +1,954 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_metric": null,
|
3 |
+
"best_model_checkpoint": null,
|
4 |
+
"epoch": 0.9996190476190476,
|
5 |
+
"eval_steps": 500,
|
6 |
+
"global_step": 656,
|
7 |
+
"is_hyper_param_search": false,
|
8 |
+
"is_local_process_zero": true,
|
9 |
+
"is_world_process_zero": true,
|
10 |
+
"log_history": [
|
11 |
+
{
|
12 |
+
"epoch": 0.0,
|
13 |
+
"learning_rate": 7.575757575757576e-08,
|
14 |
+
"logits/chosen": -0.06230628490447998,
|
15 |
+
"logits/rejected": 0.387611985206604,
|
16 |
+
"logps/chosen": -299.96368408203125,
|
17 |
+
"logps/rejected": -309.7692565917969,
|
18 |
+
"loss": 0.3457,
|
19 |
+
"rewards/accuracies": 0.0,
|
20 |
+
"rewards/chosen": 0.0,
|
21 |
+
"rewards/margins": 0.0,
|
22 |
+
"rewards/rejected": 0.0,
|
23 |
+
"step": 1
|
24 |
+
},
|
25 |
+
{
|
26 |
+
"epoch": 0.02,
|
27 |
+
"learning_rate": 7.575757575757576e-07,
|
28 |
+
"logits/chosen": 0.10603617876768112,
|
29 |
+
"logits/rejected": 0.24432696402072906,
|
30 |
+
"logps/chosen": -349.7105712890625,
|
31 |
+
"logps/rejected": -287.6678466796875,
|
32 |
+
"loss": 0.3317,
|
33 |
+
"rewards/accuracies": 0.3611111044883728,
|
34 |
+
"rewards/chosen": -0.00022038168390281498,
|
35 |
+
"rewards/margins": -0.0001302291639149189,
|
36 |
+
"rewards/rejected": -9.01525272638537e-05,
|
37 |
+
"step": 10
|
38 |
+
},
|
39 |
+
{
|
40 |
+
"epoch": 0.03,
|
41 |
+
"learning_rate": 1.5151515151515152e-06,
|
42 |
+
"logits/chosen": 0.09682648628950119,
|
43 |
+
"logits/rejected": 0.25353121757507324,
|
44 |
+
"logps/chosen": -334.0283203125,
|
45 |
+
"logps/rejected": -270.40179443359375,
|
46 |
+
"loss": 0.3365,
|
47 |
+
"rewards/accuracies": 0.4375,
|
48 |
+
"rewards/chosen": 8.945484069045051e-07,
|
49 |
+
"rewards/margins": 2.460894847899908e-06,
|
50 |
+
"rewards/rejected": -1.5663565591239603e-06,
|
51 |
+
"step": 20
|
52 |
+
},
|
53 |
+
{
|
54 |
+
"epoch": 0.05,
|
55 |
+
"learning_rate": 2.2727272727272728e-06,
|
56 |
+
"logits/chosen": 0.12519071996212006,
|
57 |
+
"logits/rejected": 0.20205554366111755,
|
58 |
+
"logps/chosen": -287.5645446777344,
|
59 |
+
"logps/rejected": -266.9909362792969,
|
60 |
+
"loss": 0.3417,
|
61 |
+
"rewards/accuracies": 0.518750011920929,
|
62 |
+
"rewards/chosen": 0.0003230299334973097,
|
63 |
+
"rewards/margins": 0.0002776235924102366,
|
64 |
+
"rewards/rejected": 4.540634108707309e-05,
|
65 |
+
"step": 30
|
66 |
+
},
|
67 |
+
{
|
68 |
+
"epoch": 0.06,
|
69 |
+
"learning_rate": 3.0303030303030305e-06,
|
70 |
+
"logits/chosen": 0.04107561707496643,
|
71 |
+
"logits/rejected": 0.20281007885932922,
|
72 |
+
"logps/chosen": -319.4472351074219,
|
73 |
+
"logps/rejected": -262.85418701171875,
|
74 |
+
"loss": 0.3615,
|
75 |
+
"rewards/accuracies": 0.59375,
|
76 |
+
"rewards/chosen": 0.0008775835740379989,
|
77 |
+
"rewards/margins": 0.0006280745146796107,
|
78 |
+
"rewards/rejected": 0.0002495090593583882,
|
79 |
+
"step": 40
|
80 |
+
},
|
81 |
+
{
|
82 |
+
"epoch": 0.08,
|
83 |
+
"learning_rate": 3.7878787878787882e-06,
|
84 |
+
"logits/chosen": 0.12531228363513947,
|
85 |
+
"logits/rejected": 0.27179789543151855,
|
86 |
+
"logps/chosen": -359.4334411621094,
|
87 |
+
"logps/rejected": -278.2669982910156,
|
88 |
+
"loss": 0.3412,
|
89 |
+
"rewards/accuracies": 0.6937500238418579,
|
90 |
+
"rewards/chosen": 0.0017220426816493273,
|
91 |
+
"rewards/margins": 0.001711520948447287,
|
92 |
+
"rewards/rejected": 1.0521779586269986e-05,
|
93 |
+
"step": 50
|
94 |
+
},
|
95 |
+
{
|
96 |
+
"epoch": 0.09,
|
97 |
+
"learning_rate": 4.5454545454545455e-06,
|
98 |
+
"logits/chosen": 0.07350507378578186,
|
99 |
+
"logits/rejected": 0.22942480444908142,
|
100 |
+
"logps/chosen": -346.56927490234375,
|
101 |
+
"logps/rejected": -306.72686767578125,
|
102 |
+
"loss": 0.3553,
|
103 |
+
"rewards/accuracies": 0.78125,
|
104 |
+
"rewards/chosen": 0.0025817144196480513,
|
105 |
+
"rewards/margins": 0.003724290756508708,
|
106 |
+
"rewards/rejected": -0.001142576104030013,
|
107 |
+
"step": 60
|
108 |
+
},
|
109 |
+
{
|
110 |
+
"epoch": 0.11,
|
111 |
+
"learning_rate": 4.999432965739786e-06,
|
112 |
+
"logits/chosen": 0.09025775641202927,
|
113 |
+
"logits/rejected": 0.31351155042648315,
|
114 |
+
"logps/chosen": -381.5528869628906,
|
115 |
+
"logps/rejected": -286.2391662597656,
|
116 |
+
"loss": 0.3633,
|
117 |
+
"rewards/accuracies": 0.625,
|
118 |
+
"rewards/chosen": 0.0029081101529300213,
|
119 |
+
"rewards/margins": 0.0062509761191904545,
|
120 |
+
"rewards/rejected": -0.003342865500599146,
|
121 |
+
"step": 70
|
122 |
+
},
|
123 |
+
{
|
124 |
+
"epoch": 0.12,
|
125 |
+
"learning_rate": 4.9930567839810125e-06,
|
126 |
+
"logits/chosen": 0.0820033922791481,
|
127 |
+
"logits/rejected": 0.2647712826728821,
|
128 |
+
"logps/chosen": -350.4310607910156,
|
129 |
+
"logps/rejected": -280.1022644042969,
|
130 |
+
"loss": 0.3175,
|
131 |
+
"rewards/accuracies": 0.6812499761581421,
|
132 |
+
"rewards/chosen": 0.0049164495430886745,
|
133 |
+
"rewards/margins": 0.010990149341523647,
|
134 |
+
"rewards/rejected": -0.006073700729757547,
|
135 |
+
"step": 80
|
136 |
+
},
|
137 |
+
{
|
138 |
+
"epoch": 0.14,
|
139 |
+
"learning_rate": 4.979613761906212e-06,
|
140 |
+
"logits/chosen": 0.11785046756267548,
|
141 |
+
"logits/rejected": 0.3276062607765198,
|
142 |
+
"logps/chosen": -370.00860595703125,
|
143 |
+
"logps/rejected": -322.6431579589844,
|
144 |
+
"loss": 0.3244,
|
145 |
+
"rewards/accuracies": 0.762499988079071,
|
146 |
+
"rewards/chosen": 0.004425252787768841,
|
147 |
+
"rewards/margins": 0.019340159371495247,
|
148 |
+
"rewards/rejected": -0.014914905652403831,
|
149 |
+
"step": 90
|
150 |
+
},
|
151 |
+
{
|
152 |
+
"epoch": 0.15,
|
153 |
+
"learning_rate": 4.959142005221991e-06,
|
154 |
+
"logits/chosen": 0.19294002652168274,
|
155 |
+
"logits/rejected": 0.3200022578239441,
|
156 |
+
"logps/chosen": -369.6689147949219,
|
157 |
+
"logps/rejected": -329.09893798828125,
|
158 |
+
"loss": 0.2978,
|
159 |
+
"rewards/accuracies": 0.75,
|
160 |
+
"rewards/chosen": -0.018663963302969933,
|
161 |
+
"rewards/margins": 0.04274021461606026,
|
162 |
+
"rewards/rejected": -0.06140417978167534,
|
163 |
+
"step": 100
|
164 |
+
},
|
165 |
+
{
|
166 |
+
"epoch": 0.17,
|
167 |
+
"learning_rate": 4.931699543346854e-06,
|
168 |
+
"logits/chosen": 0.23536458611488342,
|
169 |
+
"logits/rejected": 0.35565608739852905,
|
170 |
+
"logps/chosen": -432.9308166503906,
|
171 |
+
"logps/rejected": -432.2447204589844,
|
172 |
+
"loss": 0.2948,
|
173 |
+
"rewards/accuracies": 0.699999988079071,
|
174 |
+
"rewards/chosen": -0.07575322687625885,
|
175 |
+
"rewards/margins": 0.06633279472589493,
|
176 |
+
"rewards/rejected": -0.14208602905273438,
|
177 |
+
"step": 110
|
178 |
+
},
|
179 |
+
{
|
180 |
+
"epoch": 0.18,
|
181 |
+
"learning_rate": 4.897364164920515e-06,
|
182 |
+
"logits/chosen": 0.24238090217113495,
|
183 |
+
"logits/rejected": 0.2669451832771301,
|
184 |
+
"logps/chosen": -502.5521545410156,
|
185 |
+
"logps/rejected": -526.5723876953125,
|
186 |
+
"loss": 0.286,
|
187 |
+
"rewards/accuracies": 0.65625,
|
188 |
+
"rewards/chosen": -0.16139724850654602,
|
189 |
+
"rewards/margins": 0.09947613626718521,
|
190 |
+
"rewards/rejected": -0.26087337732315063,
|
191 |
+
"step": 120
|
192 |
+
},
|
193 |
+
{
|
194 |
+
"epoch": 0.2,
|
195 |
+
"learning_rate": 4.8562331973035396e-06,
|
196 |
+
"logits/chosen": 0.24233976006507874,
|
197 |
+
"logits/rejected": 0.37175804376602173,
|
198 |
+
"logps/chosen": -561.4464721679688,
|
199 |
+
"logps/rejected": -616.5276489257812,
|
200 |
+
"loss": 0.2701,
|
201 |
+
"rewards/accuracies": 0.7250000238418579,
|
202 |
+
"rewards/chosen": -0.1863406002521515,
|
203 |
+
"rewards/margins": 0.11555895954370499,
|
204 |
+
"rewards/rejected": -0.30189958214759827,
|
205 |
+
"step": 130
|
206 |
+
},
|
207 |
+
{
|
208 |
+
"epoch": 0.21,
|
209 |
+
"learning_rate": 4.808423230692374e-06,
|
210 |
+
"logits/chosen": 0.2752463221549988,
|
211 |
+
"logits/rejected": 0.41067615151405334,
|
212 |
+
"logps/chosen": -544.420654296875,
|
213 |
+
"logps/rejected": -609.9964599609375,
|
214 |
+
"loss": 0.2437,
|
215 |
+
"rewards/accuracies": 0.762499988079071,
|
216 |
+
"rewards/chosen": -0.16954469680786133,
|
217 |
+
"rewards/margins": 0.15670140087604523,
|
218 |
+
"rewards/rejected": -0.32624611258506775,
|
219 |
+
"step": 140
|
220 |
+
},
|
221 |
+
{
|
222 |
+
"epoch": 0.23,
|
223 |
+
"learning_rate": 4.754069787631761e-06,
|
224 |
+
"logits/chosen": 0.22061054408550262,
|
225 |
+
"logits/rejected": 0.38979893922805786,
|
226 |
+
"logps/chosen": -499.3285217285156,
|
227 |
+
"logps/rejected": -558.7503051757812,
|
228 |
+
"loss": 0.2709,
|
229 |
+
"rewards/accuracies": 0.637499988079071,
|
230 |
+
"rewards/chosen": -0.1981675624847412,
|
231 |
+
"rewards/margins": 0.09831173717975616,
|
232 |
+
"rewards/rejected": -0.2964792847633362,
|
233 |
+
"step": 150
|
234 |
+
},
|
235 |
+
{
|
236 |
+
"epoch": 0.24,
|
237 |
+
"learning_rate": 4.693326938861367e-06,
|
238 |
+
"logits/chosen": 0.2467188537120819,
|
239 |
+
"logits/rejected": 0.35028940439224243,
|
240 |
+
"logps/chosen": -526.26513671875,
|
241 |
+
"logps/rejected": -599.4227294921875,
|
242 |
+
"loss": 0.2605,
|
243 |
+
"rewards/accuracies": 0.643750011920929,
|
244 |
+
"rewards/chosen": -0.1936112642288208,
|
245 |
+
"rewards/margins": 0.1178494542837143,
|
246 |
+
"rewards/rejected": -0.3114607036113739,
|
247 |
+
"step": 160
|
248 |
+
},
|
249 |
+
{
|
250 |
+
"epoch": 0.26,
|
251 |
+
"learning_rate": 4.626366866585528e-06,
|
252 |
+
"logits/chosen": 0.22420647740364075,
|
253 |
+
"logits/rejected": 0.4213325083255768,
|
254 |
+
"logps/chosen": -622.0628662109375,
|
255 |
+
"logps/rejected": -708.9386596679688,
|
256 |
+
"loss": 0.2402,
|
257 |
+
"rewards/accuracies": 0.7562500238418579,
|
258 |
+
"rewards/chosen": -0.2440744936466217,
|
259 |
+
"rewards/margins": 0.15784001350402832,
|
260 |
+
"rewards/rejected": -0.40191444754600525,
|
261 |
+
"step": 170
|
262 |
+
},
|
263 |
+
{
|
264 |
+
"epoch": 0.27,
|
265 |
+
"learning_rate": 4.553379376404085e-06,
|
266 |
+
"logits/chosen": 0.27953073382377625,
|
267 |
+
"logits/rejected": 0.3603507876396179,
|
268 |
+
"logps/chosen": -601.0642700195312,
|
269 |
+
"logps/rejected": -663.9993896484375,
|
270 |
+
"loss": 0.2735,
|
271 |
+
"rewards/accuracies": 0.675000011920929,
|
272 |
+
"rewards/chosen": -0.2422056943178177,
|
273 |
+
"rewards/margins": 0.1294884979724884,
|
274 |
+
"rewards/rejected": -0.3716942071914673,
|
275 |
+
"step": 180
|
276 |
+
},
|
277 |
+
{
|
278 |
+
"epoch": 0.29,
|
279 |
+
"learning_rate": 4.474571359287791e-06,
|
280 |
+
"logits/chosen": 0.23259811103343964,
|
281 |
+
"logits/rejected": 0.4210383892059326,
|
282 |
+
"logps/chosen": -572.4783935546875,
|
283 |
+
"logps/rejected": -687.89306640625,
|
284 |
+
"loss": 0.2247,
|
285 |
+
"rewards/accuracies": 0.75,
|
286 |
+
"rewards/chosen": -0.20176899433135986,
|
287 |
+
"rewards/margins": 0.18673528730869293,
|
288 |
+
"rewards/rejected": -0.3885042667388916,
|
289 |
+
"step": 190
|
290 |
+
},
|
291 |
+
{
|
292 |
+
"epoch": 0.3,
|
293 |
+
"learning_rate": 4.3901662051233755e-06,
|
294 |
+
"logits/chosen": 0.2610084116458893,
|
295 |
+
"logits/rejected": 0.3713424801826477,
|
296 |
+
"logps/chosen": -539.9381103515625,
|
297 |
+
"logps/rejected": -680.5265502929688,
|
298 |
+
"loss": 0.2243,
|
299 |
+
"rewards/accuracies": 0.7562500238418579,
|
300 |
+
"rewards/chosen": -0.20527365803718567,
|
301 |
+
"rewards/margins": 0.1885356605052948,
|
302 |
+
"rewards/rejected": -0.3938092887401581,
|
303 |
+
"step": 200
|
304 |
+
},
|
305 |
+
{
|
306 |
+
"epoch": 0.32,
|
307 |
+
"learning_rate": 4.30040316949064e-06,
|
308 |
+
"logits/chosen": 0.2715567946434021,
|
309 |
+
"logits/rejected": 0.3666667640209198,
|
310 |
+
"logps/chosen": -536.0437622070312,
|
311 |
+
"logps/rejected": -629.3165283203125,
|
312 |
+
"loss": 0.248,
|
313 |
+
"rewards/accuracies": 0.762499988079071,
|
314 |
+
"rewards/chosen": -0.18539923429489136,
|
315 |
+
"rewards/margins": 0.16864952445030212,
|
316 |
+
"rewards/rejected": -0.35404878854751587,
|
317 |
+
"step": 210
|
318 |
+
},
|
319 |
+
{
|
320 |
+
"epoch": 0.34,
|
321 |
+
"learning_rate": 4.205536695466524e-06,
|
322 |
+
"logits/chosen": 0.23633995652198792,
|
323 |
+
"logits/rejected": 0.3769295811653137,
|
324 |
+
"logps/chosen": -584.9187622070312,
|
325 |
+
"logps/rejected": -705.7189331054688,
|
326 |
+
"loss": 0.2185,
|
327 |
+
"rewards/accuracies": 0.7562500238418579,
|
328 |
+
"rewards/chosen": -0.2217777669429779,
|
329 |
+
"rewards/margins": 0.20478694140911102,
|
330 |
+
"rewards/rejected": -0.4265647530555725,
|
331 |
+
"step": 220
|
332 |
+
},
|
333 |
+
{
|
334 |
+
"epoch": 0.35,
|
335 |
+
"learning_rate": 4.105835692378557e-06,
|
336 |
+
"logits/chosen": 0.29004430770874023,
|
337 |
+
"logits/rejected": 0.4667590260505676,
|
338 |
+
"logps/chosen": -701.4776611328125,
|
339 |
+
"logps/rejected": -798.4501953125,
|
340 |
+
"loss": 0.2323,
|
341 |
+
"rewards/accuracies": 0.7250000238418579,
|
342 |
+
"rewards/chosen": -0.316620796918869,
|
343 |
+
"rewards/margins": 0.18385566771030426,
|
344 |
+
"rewards/rejected": -0.5004764795303345,
|
345 |
+
"step": 230
|
346 |
+
},
|
347 |
+
{
|
348 |
+
"epoch": 0.37,
|
349 |
+
"learning_rate": 4.001582773552153e-06,
|
350 |
+
"logits/chosen": 0.27633368968963623,
|
351 |
+
"logits/rejected": 0.42621952295303345,
|
352 |
+
"logps/chosen": -621.6236572265625,
|
353 |
+
"logps/rejected": -752.4913940429688,
|
354 |
+
"loss": 0.2466,
|
355 |
+
"rewards/accuracies": 0.731249988079071,
|
356 |
+
"rewards/chosen": -0.2713666260242462,
|
357 |
+
"rewards/margins": 0.17953212559223175,
|
358 |
+
"rewards/rejected": -0.4508987367153168,
|
359 |
+
"step": 240
|
360 |
+
},
|
361 |
+
{
|
362 |
+
"epoch": 0.38,
|
363 |
+
"learning_rate": 3.893073455212438e-06,
|
364 |
+
"logits/chosen": 0.2567233443260193,
|
365 |
+
"logits/rejected": 0.3767459988594055,
|
366 |
+
"logps/chosen": -553.5460205078125,
|
367 |
+
"logps/rejected": -691.7283935546875,
|
368 |
+
"loss": 0.2438,
|
369 |
+
"rewards/accuracies": 0.75,
|
370 |
+
"rewards/chosen": -0.23452623188495636,
|
371 |
+
"rewards/margins": 0.17362844944000244,
|
372 |
+
"rewards/rejected": -0.4081546664237976,
|
373 |
+
"step": 250
|
374 |
+
},
|
375 |
+
{
|
376 |
+
"epoch": 0.4,
|
377 |
+
"learning_rate": 3.7806153188114027e-06,
|
378 |
+
"logits/chosen": 0.24300554394721985,
|
379 |
+
"logits/rejected": 0.30316272377967834,
|
380 |
+
"logps/chosen": -578.8638916015625,
|
381 |
+
"logps/rejected": -715.2518310546875,
|
382 |
+
"loss": 0.2468,
|
383 |
+
"rewards/accuracies": 0.731249988079071,
|
384 |
+
"rewards/chosen": -0.2516656219959259,
|
385 |
+
"rewards/margins": 0.16586793959140778,
|
386 |
+
"rewards/rejected": -0.4175335466861725,
|
387 |
+
"step": 260
|
388 |
+
},
|
389 |
+
{
|
390 |
+
"epoch": 0.41,
|
391 |
+
"learning_rate": 3.6645271391548542e-06,
|
392 |
+
"logits/chosen": 0.23739242553710938,
|
393 |
+
"logits/rejected": 0.426413357257843,
|
394 |
+
"logps/chosen": -634.893798828125,
|
395 |
+
"logps/rejected": -719.962890625,
|
396 |
+
"loss": 0.2531,
|
397 |
+
"rewards/accuracies": 0.7437499761581421,
|
398 |
+
"rewards/chosen": -0.25505417585372925,
|
399 |
+
"rewards/margins": 0.18401430547237396,
|
400 |
+
"rewards/rejected": -0.4390684962272644,
|
401 |
+
"step": 270
|
402 |
+
},
|
403 |
+
{
|
404 |
+
"epoch": 0.43,
|
405 |
+
"learning_rate": 3.5451379808006014e-06,
|
406 |
+
"logits/chosen": 0.2224760502576828,
|
407 |
+
"logits/rejected": 0.3537066578865051,
|
408 |
+
"logps/chosen": -589.7808227539062,
|
409 |
+
"logps/rejected": -747.9609375,
|
410 |
+
"loss": 0.2283,
|
411 |
+
"rewards/accuracies": 0.762499988079071,
|
412 |
+
"rewards/chosen": -0.2638700306415558,
|
413 |
+
"rewards/margins": 0.1954762488603592,
|
414 |
+
"rewards/rejected": -0.45934629440307617,
|
415 |
+
"step": 280
|
416 |
+
},
|
417 |
+
{
|
418 |
+
"epoch": 0.44,
|
419 |
+
"learning_rate": 3.4227862652892106e-06,
|
420 |
+
"logits/chosen": 0.23644611239433289,
|
421 |
+
"logits/rejected": 0.4177249073982239,
|
422 |
+
"logps/chosen": -645.470703125,
|
423 |
+
"logps/rejected": -779.50927734375,
|
424 |
+
"loss": 0.2437,
|
425 |
+
"rewards/accuracies": 0.731249988079071,
|
426 |
+
"rewards/chosen": -0.27952614426612854,
|
427 |
+
"rewards/margins": 0.19378896057605743,
|
428 |
+
"rewards/rejected": -0.4733150601387024,
|
429 |
+
"step": 290
|
430 |
+
},
|
431 |
+
{
|
432 |
+
"epoch": 0.46,
|
433 |
+
"learning_rate": 3.2978188118513814e-06,
|
434 |
+
"logits/chosen": 0.2835056483745575,
|
435 |
+
"logits/rejected": 0.3992118239402771,
|
436 |
+
"logps/chosen": -620.6287841796875,
|
437 |
+
"logps/rejected": -733.6190185546875,
|
438 |
+
"loss": 0.2377,
|
439 |
+
"rewards/accuracies": 0.7562500238418579,
|
440 |
+
"rewards/chosen": -0.2810095250606537,
|
441 |
+
"rewards/margins": 0.17397195100784302,
|
442 |
+
"rewards/rejected": -0.4549815058708191,
|
443 |
+
"step": 300
|
444 |
+
},
|
445 |
+
{
|
446 |
+
"epoch": 0.47,
|
447 |
+
"learning_rate": 3.1705898543111576e-06,
|
448 |
+
"logits/chosen": 0.2648586630821228,
|
449 |
+
"logits/rejected": 0.36566048860549927,
|
450 |
+
"logps/chosen": -600.49853515625,
|
451 |
+
"logps/rejected": -694.8892822265625,
|
452 |
+
"loss": 0.2408,
|
453 |
+
"rewards/accuracies": 0.675000011920929,
|
454 |
+
"rewards/chosen": -0.2665795385837555,
|
455 |
+
"rewards/margins": 0.14690735936164856,
|
456 |
+
"rewards/rejected": -0.41348689794540405,
|
457 |
+
"step": 310
|
458 |
+
},
|
459 |
+
{
|
460 |
+
"epoch": 0.49,
|
461 |
+
"learning_rate": 3.041460036971664e-06,
|
462 |
+
"logits/chosen": 0.2214949131011963,
|
463 |
+
"logits/rejected": 0.38505813479423523,
|
464 |
+
"logps/chosen": -669.3818969726562,
|
465 |
+
"logps/rejected": -743.3441162109375,
|
466 |
+
"loss": 0.2197,
|
467 |
+
"rewards/accuracies": 0.7749999761581421,
|
468 |
+
"rewards/chosen": -0.27733737230300903,
|
469 |
+
"rewards/margins": 0.16739198565483093,
|
470 |
+
"rewards/rejected": -0.4447293281555176,
|
471 |
+
"step": 320
|
472 |
+
},
|
473 |
+
{
|
474 |
+
"epoch": 0.5,
|
475 |
+
"learning_rate": 2.910795392329649e-06,
|
476 |
+
"logits/chosen": 0.19933928549289703,
|
477 |
+
"logits/rejected": 0.3905678391456604,
|
478 |
+
"logps/chosen": -631.4853515625,
|
479 |
+
"logps/rejected": -736.2476806640625,
|
480 |
+
"loss": 0.2262,
|
481 |
+
"rewards/accuracies": 0.7124999761581421,
|
482 |
+
"rewards/chosen": -0.2883307933807373,
|
483 |
+
"rewards/margins": 0.17361871898174286,
|
484 |
+
"rewards/rejected": -0.46194949746131897,
|
485 |
+
"step": 330
|
486 |
+
},
|
487 |
+
{
|
488 |
+
"epoch": 0.52,
|
489 |
+
"learning_rate": 2.7789663035166035e-06,
|
490 |
+
"logits/chosen": 0.19208988547325134,
|
491 |
+
"logits/rejected": 0.3280327022075653,
|
492 |
+
"logps/chosen": -637.7025146484375,
|
493 |
+
"logps/rejected": -742.5934448242188,
|
494 |
+
"loss": 0.229,
|
495 |
+
"rewards/accuracies": 0.7124999761581421,
|
496 |
+
"rewards/chosen": -0.2803892493247986,
|
497 |
+
"rewards/margins": 0.18691286444664001,
|
498 |
+
"rewards/rejected": -0.46730202436447144,
|
499 |
+
"step": 340
|
500 |
+
},
|
501 |
+
{
|
502 |
+
"epoch": 0.53,
|
503 |
+
"learning_rate": 2.6463464544075344e-06,
|
504 |
+
"logits/chosen": 0.2192709892988205,
|
505 |
+
"logits/rejected": 0.354716956615448,
|
506 |
+
"logps/chosen": -666.2679443359375,
|
507 |
+
"logps/rejected": -781.5311889648438,
|
508 |
+
"loss": 0.2365,
|
509 |
+
"rewards/accuracies": 0.731249988079071,
|
510 |
+
"rewards/chosen": -0.313618928194046,
|
511 |
+
"rewards/margins": 0.2000730335712433,
|
512 |
+
"rewards/rejected": -0.5136920213699341,
|
513 |
+
"step": 350
|
514 |
+
},
|
515 |
+
{
|
516 |
+
"epoch": 0.55,
|
517 |
+
"learning_rate": 2.513311770373421e-06,
|
518 |
+
"logits/chosen": 0.24396447837352753,
|
519 |
+
"logits/rejected": 0.3740311563014984,
|
520 |
+
"logps/chosen": -632.1403198242188,
|
521 |
+
"logps/rejected": -780.8329467773438,
|
522 |
+
"loss": 0.2137,
|
523 |
+
"rewards/accuracies": 0.800000011920929,
|
524 |
+
"rewards/chosen": -0.29758018255233765,
|
525 |
+
"rewards/margins": 0.20380613207817078,
|
526 |
+
"rewards/rejected": -0.501386284828186,
|
527 |
+
"step": 360
|
528 |
+
},
|
529 |
+
{
|
530 |
+
"epoch": 0.56,
|
531 |
+
"learning_rate": 2.380239352679908e-06,
|
532 |
+
"logits/chosen": 0.2605392634868622,
|
533 |
+
"logits/rejected": 0.3702816367149353,
|
534 |
+
"logps/chosen": -659.3692626953125,
|
535 |
+
"logps/rejected": -771.1994018554688,
|
536 |
+
"loss": 0.2134,
|
537 |
+
"rewards/accuracies": 0.75,
|
538 |
+
"rewards/chosen": -0.2961762547492981,
|
539 |
+
"rewards/margins": 0.19712337851524353,
|
540 |
+
"rewards/rejected": -0.49329957365989685,
|
541 |
+
"step": 370
|
542 |
+
},
|
543 |
+
{
|
544 |
+
"epoch": 0.58,
|
545 |
+
"learning_rate": 2.247506409552795e-06,
|
546 |
+
"logits/chosen": 0.2626807987689972,
|
547 |
+
"logits/rejected": 0.4523330330848694,
|
548 |
+
"logps/chosen": -723.3246459960938,
|
549 |
+
"logps/rejected": -843.2516479492188,
|
550 |
+
"loss": 0.2222,
|
551 |
+
"rewards/accuracies": 0.7875000238418579,
|
552 |
+
"rewards/chosen": -0.33206573128700256,
|
553 |
+
"rewards/margins": 0.2155541181564331,
|
554 |
+
"rewards/rejected": -0.5476198792457581,
|
555 |
+
"step": 380
|
556 |
+
},
|
557 |
+
{
|
558 |
+
"epoch": 0.59,
|
559 |
+
"learning_rate": 2.1154891869403436e-06,
|
560 |
+
"logits/chosen": 0.2673259675502777,
|
561 |
+
"logits/rejected": 0.3809369206428528,
|
562 |
+
"logps/chosen": -627.7708740234375,
|
563 |
+
"logps/rejected": -776.5219116210938,
|
564 |
+
"loss": 0.2345,
|
565 |
+
"rewards/accuracies": 0.737500011920929,
|
566 |
+
"rewards/chosen": -0.2913658320903778,
|
567 |
+
"rewards/margins": 0.19852493703365326,
|
568 |
+
"rewards/rejected": -0.48989081382751465,
|
569 |
+
"step": 390
|
570 |
+
},
|
571 |
+
{
|
572 |
+
"epoch": 0.61,
|
573 |
+
"learning_rate": 1.9845619020032552e-06,
|
574 |
+
"logits/chosen": 0.22264230251312256,
|
575 |
+
"logits/rejected": 0.36639919877052307,
|
576 |
+
"logps/chosen": -656.4054565429688,
|
577 |
+
"logps/rejected": -789.1956787109375,
|
578 |
+
"loss": 0.2189,
|
579 |
+
"rewards/accuracies": 0.737500011920929,
|
580 |
+
"rewards/chosen": -0.30770599842071533,
|
581 |
+
"rewards/margins": 0.20198026299476624,
|
582 |
+
"rewards/rejected": -0.5096862316131592,
|
583 |
+
"step": 400
|
584 |
+
},
|
585 |
+
{
|
586 |
+
"epoch": 0.62,
|
587 |
+
"learning_rate": 1.8550956823554708e-06,
|
588 |
+
"logits/chosen": 0.25498437881469727,
|
589 |
+
"logits/rejected": 0.39235639572143555,
|
590 |
+
"logps/chosen": -597.4380493164062,
|
591 |
+
"logps/rejected": -743.4146728515625,
|
592 |
+
"loss": 0.2504,
|
593 |
+
"rewards/accuracies": 0.75,
|
594 |
+
"rewards/chosen": -0.2715555429458618,
|
595 |
+
"rewards/margins": 0.1882806420326233,
|
596 |
+
"rewards/rejected": -0.4598361849784851,
|
597 |
+
"step": 410
|
598 |
+
},
|
599 |
+
{
|
600 |
+
"epoch": 0.64,
|
601 |
+
"learning_rate": 1.7274575140626318e-06,
|
602 |
+
"logits/chosen": 0.2268383502960205,
|
603 |
+
"logits/rejected": 0.3757990598678589,
|
604 |
+
"logps/chosen": -628.30517578125,
|
605 |
+
"logps/rejected": -744.66455078125,
|
606 |
+
"loss": 0.2356,
|
607 |
+
"rewards/accuracies": 0.75,
|
608 |
+
"rewards/chosen": -0.2700210213661194,
|
609 |
+
"rewards/margins": 0.19132563471794128,
|
610 |
+
"rewards/rejected": -0.4613465666770935,
|
611 |
+
"step": 420
|
612 |
+
},
|
613 |
+
{
|
614 |
+
"epoch": 0.66,
|
615 |
+
"learning_rate": 1.6020092013802002e-06,
|
616 |
+
"logits/chosen": 0.26629549264907837,
|
617 |
+
"logits/rejected": 0.38806790113449097,
|
618 |
+
"logps/chosen": -643.1737060546875,
|
619 |
+
"logps/rejected": -752.5285034179688,
|
620 |
+
"loss": 0.2192,
|
621 |
+
"rewards/accuracies": 0.7437499761581421,
|
622 |
+
"rewards/chosen": -0.28404101729393005,
|
623 |
+
"rewards/margins": 0.18914134800434113,
|
624 |
+
"rewards/rejected": -0.4731822907924652,
|
625 |
+
"step": 430
|
626 |
+
},
|
627 |
+
{
|
628 |
+
"epoch": 0.67,
|
629 |
+
"learning_rate": 1.4791063411799938e-06,
|
630 |
+
"logits/chosen": 0.17316266894340515,
|
631 |
+
"logits/rejected": 0.3880611062049866,
|
632 |
+
"logps/chosen": -670.3546142578125,
|
633 |
+
"logps/rejected": -764.1770629882812,
|
634 |
+
"loss": 0.2366,
|
635 |
+
"rewards/accuracies": 0.7124999761581421,
|
636 |
+
"rewards/chosen": -0.3103855550289154,
|
637 |
+
"rewards/margins": 0.17021675407886505,
|
638 |
+
"rewards/rejected": -0.48060232400894165,
|
639 |
+
"step": 440
|
640 |
+
},
|
641 |
+
{
|
642 |
+
"epoch": 0.69,
|
643 |
+
"learning_rate": 1.3590973149722103e-06,
|
644 |
+
"logits/chosen": 0.24187886714935303,
|
645 |
+
"logits/rejected": 0.37198665738105774,
|
646 |
+
"logps/chosen": -628.3953857421875,
|
647 |
+
"logps/rejected": -773.7864990234375,
|
648 |
+
"loss": 0.2157,
|
649 |
+
"rewards/accuracies": 0.7437499761581421,
|
650 |
+
"rewards/chosen": -0.3083949089050293,
|
651 |
+
"rewards/margins": 0.18738897144794464,
|
652 |
+
"rewards/rejected": -0.49578380584716797,
|
653 |
+
"step": 450
|
654 |
+
},
|
655 |
+
{
|
656 |
+
"epoch": 0.7,
|
657 |
+
"learning_rate": 1.2423223013801946e-06,
|
658 |
+
"logits/chosen": 0.19144193828105927,
|
659 |
+
"logits/rejected": 0.3342723846435547,
|
660 |
+
"logps/chosen": -679.7613525390625,
|
661 |
+
"logps/rejected": -820.7508544921875,
|
662 |
+
"loss": 0.2274,
|
663 |
+
"rewards/accuracies": 0.7250000238418579,
|
664 |
+
"rewards/chosen": -0.3411591649055481,
|
665 |
+
"rewards/margins": 0.18641141057014465,
|
666 |
+
"rewards/rejected": -0.5275705456733704,
|
667 |
+
"step": 460
|
668 |
+
},
|
669 |
+
{
|
670 |
+
"epoch": 0.72,
|
671 |
+
"learning_rate": 1.1291123118671665e-06,
|
672 |
+
"logits/chosen": 0.21762657165527344,
|
673 |
+
"logits/rejected": 0.307900607585907,
|
674 |
+
"logps/chosen": -651.0516967773438,
|
675 |
+
"logps/rejected": -805.5900268554688,
|
676 |
+
"loss": 0.2194,
|
677 |
+
"rewards/accuracies": 0.762499988079071,
|
678 |
+
"rewards/chosen": -0.309937983751297,
|
679 |
+
"rewards/margins": 0.20782515406608582,
|
680 |
+
"rewards/rejected": -0.5177631378173828,
|
681 |
+
"step": 470
|
682 |
+
},
|
683 |
+
{
|
684 |
+
"epoch": 0.73,
|
685 |
+
"learning_rate": 1.019788252448267e-06,
|
686 |
+
"logits/chosen": 0.184129536151886,
|
687 |
+
"logits/rejected": 0.3611859083175659,
|
688 |
+
"logps/chosen": -619.1923828125,
|
689 |
+
"logps/rejected": -762.8614501953125,
|
690 |
+
"loss": 0.2191,
|
691 |
+
"rewards/accuracies": 0.7250000238418579,
|
692 |
+
"rewards/chosen": -0.30183008313179016,
|
693 |
+
"rewards/margins": 0.18515999615192413,
|
694 |
+
"rewards/rejected": -0.4869900643825531,
|
695 |
+
"step": 480
|
696 |
+
},
|
697 |
+
{
|
698 |
+
"epoch": 0.75,
|
699 |
+
"learning_rate": 9.146600140475945e-07,
|
700 |
+
"logits/chosen": 0.20372629165649414,
|
701 |
+
"logits/rejected": 0.33465132117271423,
|
702 |
+
"logps/chosen": -636.4930419921875,
|
703 |
+
"logps/rejected": -756.65576171875,
|
704 |
+
"loss": 0.2029,
|
705 |
+
"rewards/accuracies": 0.762499988079071,
|
706 |
+
"rewards/chosen": -0.2839277386665344,
|
707 |
+
"rewards/margins": 0.197633758187294,
|
708 |
+
"rewards/rejected": -0.4815615117549896,
|
709 |
+
"step": 490
|
710 |
+
},
|
711 |
+
{
|
712 |
+
"epoch": 0.76,
|
713 |
+
"learning_rate": 8.140255940787059e-07,
|
714 |
+
"logits/chosen": 0.18791750073432922,
|
715 |
+
"logits/rejected": 0.4121164381504059,
|
716 |
+
"logps/chosen": -636.9640502929688,
|
717 |
+
"logps/rejected": -798.8194580078125,
|
718 |
+
"loss": 0.2107,
|
719 |
+
"rewards/accuracies": 0.737500011920929,
|
720 |
+
"rewards/chosen": -0.27898019552230835,
|
721 |
+
"rewards/margins": 0.21688583493232727,
|
722 |
+
"rewards/rejected": -0.49586600065231323,
|
723 |
+
"step": 500
|
724 |
+
},
|
725 |
+
{
|
726 |
+
"epoch": 0.78,
|
727 |
+
"learning_rate": 7.181702517385789e-07,
|
728 |
+
"logits/chosen": 0.19490660727024078,
|
729 |
+
"logits/rejected": 0.3919000029563904,
|
730 |
+
"logps/chosen": -693.1792602539062,
|
731 |
+
"logps/rejected": -796.6826171875,
|
732 |
+
"loss": 0.2212,
|
733 |
+
"rewards/accuracies": 0.7749999761581421,
|
734 |
+
"rewards/chosen": -0.30116355419158936,
|
735 |
+
"rewards/margins": 0.22377757728099823,
|
736 |
+
"rewards/rejected": -0.524941086769104,
|
737 |
+
"step": 510
|
738 |
+
},
|
739 |
+
{
|
740 |
+
"epoch": 0.79,
|
741 |
+
"learning_rate": 6.273656994094232e-07,
|
742 |
+
"logits/chosen": 0.21901503205299377,
|
743 |
+
"logits/rejected": 0.33780670166015625,
|
744 |
+
"logps/chosen": -620.2666015625,
|
745 |
+
"logps/rejected": -768.0850830078125,
|
746 |
+
"loss": 0.2174,
|
747 |
+
"rewards/accuracies": 0.7250000238418579,
|
748 |
+
"rewards/chosen": -0.29874101281166077,
|
749 |
+
"rewards/margins": 0.19197914004325867,
|
750 |
+
"rewards/rejected": -0.49072012305259705,
|
751 |
+
"step": 520
|
752 |
+
},
|
753 |
+
{
|
754 |
+
"epoch": 0.81,
|
755 |
+
"learning_rate": 5.418693324604082e-07,
|
756 |
+
"logits/chosen": 0.16905078291893005,
|
757 |
+
"logits/rejected": 0.3485548198223114,
|
758 |
+
"logps/chosen": -657.9064331054688,
|
759 |
+
"logps/rejected": -817.0252075195312,
|
760 |
+
"loss": 0.22,
|
761 |
+
"rewards/accuracies": 0.8125,
|
762 |
+
"rewards/chosen": -0.3066551983356476,
|
763 |
+
"rewards/margins": 0.22827258706092834,
|
764 |
+
"rewards/rejected": -0.5349277853965759,
|
765 |
+
"step": 530
|
766 |
+
},
|
767 |
+
{
|
768 |
+
"epoch": 0.82,
|
769 |
+
"learning_rate": 4.619234996325314e-07,
|
770 |
+
"logits/chosen": 0.2440187931060791,
|
771 |
+
"logits/rejected": 0.34625181555747986,
|
772 |
+
"logps/chosen": -640.2761840820312,
|
773 |
+
"logps/rejected": -762.24609375,
|
774 |
+
"loss": 0.2044,
|
775 |
+
"rewards/accuracies": 0.7250000238418579,
|
776 |
+
"rewards/chosen": -0.2988249659538269,
|
777 |
+
"rewards/margins": 0.19742272794246674,
|
778 |
+
"rewards/rejected": -0.49624767899513245,
|
779 |
+
"step": 540
|
780 |
+
},
|
781 |
+
{
|
782 |
+
"epoch": 0.84,
|
783 |
+
"learning_rate": 3.877548160747768e-07,
|
784 |
+
"logits/chosen": 0.20869365334510803,
|
785 |
+
"logits/rejected": 0.31174546480178833,
|
786 |
+
"logps/chosen": -660.8912353515625,
|
787 |
+
"logps/rejected": -801.5670166015625,
|
788 |
+
"loss": 0.2247,
|
789 |
+
"rewards/accuracies": 0.737500011920929,
|
790 |
+
"rewards/chosen": -0.3307650685310364,
|
791 |
+
"rewards/margins": 0.18925593793392181,
|
792 |
+
"rewards/rejected": -0.5200210213661194,
|
793 |
+
"step": 550
|
794 |
+
},
|
795 |
+
{
|
796 |
+
"epoch": 0.85,
|
797 |
+
"learning_rate": 3.195735209788528e-07,
|
798 |
+
"logits/chosen": 0.20122747123241425,
|
799 |
+
"logits/rejected": 0.35053348541259766,
|
800 |
+
"logps/chosen": -679.0135498046875,
|
801 |
+
"logps/rejected": -827.4410400390625,
|
802 |
+
"loss": 0.2419,
|
803 |
+
"rewards/accuracies": 0.7437499761581421,
|
804 |
+
"rewards/chosen": -0.3423236012458801,
|
805 |
+
"rewards/margins": 0.19195982813835144,
|
806 |
+
"rewards/rejected": -0.534283459186554,
|
807 |
+
"step": 560
|
808 |
+
},
|
809 |
+
{
|
810 |
+
"epoch": 0.87,
|
811 |
+
"learning_rate": 2.5757288163336806e-07,
|
812 |
+
"logits/chosen": 0.26682180166244507,
|
813 |
+
"logits/rejected": 0.3439808785915375,
|
814 |
+
"logps/chosen": -672.2916259765625,
|
815 |
+
"logps/rejected": -777.9793701171875,
|
816 |
+
"loss": 0.2386,
|
817 |
+
"rewards/accuracies": 0.706250011920929,
|
818 |
+
"rewards/chosen": -0.312061071395874,
|
819 |
+
"rewards/margins": 0.18959736824035645,
|
820 |
+
"rewards/rejected": -0.5016584396362305,
|
821 |
+
"step": 570
|
822 |
+
},
|
823 |
+
{
|
824 |
+
"epoch": 0.88,
|
825 |
+
"learning_rate": 2.019286455866981e-07,
|
826 |
+
"logits/chosen": 0.23167696595191956,
|
827 |
+
"logits/rejected": 0.3346760869026184,
|
828 |
+
"logps/chosen": -632.02587890625,
|
829 |
+
"logps/rejected": -782.4437255859375,
|
830 |
+
"loss": 0.224,
|
831 |
+
"rewards/accuracies": 0.762499988079071,
|
832 |
+
"rewards/chosen": -0.29868921637535095,
|
833 |
+
"rewards/margins": 0.1968630850315094,
|
834 |
+
"rewards/rejected": -0.49555230140686035,
|
835 |
+
"step": 580
|
836 |
+
},
|
837 |
+
{
|
838 |
+
"epoch": 0.9,
|
839 |
+
"learning_rate": 1.5279854247146703e-07,
|
840 |
+
"logits/chosen": 0.19181525707244873,
|
841 |
+
"logits/rejected": 0.3878282904624939,
|
842 |
+
"logps/chosen": -682.7297973632812,
|
843 |
+
"logps/rejected": -799.3133544921875,
|
844 |
+
"loss": 0.221,
|
845 |
+
"rewards/accuracies": 0.762499988079071,
|
846 |
+
"rewards/chosen": -0.3136585056781769,
|
847 |
+
"rewards/margins": 0.19802789390087128,
|
848 |
+
"rewards/rejected": -0.5116864442825317,
|
849 |
+
"step": 590
|
850 |
+
},
|
851 |
+
{
|
852 |
+
"epoch": 0.91,
|
853 |
+
"learning_rate": 1.1032183690276754e-07,
|
854 |
+
"logits/chosen": 0.20669761300086975,
|
855 |
+
"logits/rejected": 0.3533198833465576,
|
856 |
+
"logps/chosen": -694.8823852539062,
|
857 |
+
"logps/rejected": -831.9597778320312,
|
858 |
+
"loss": 0.1987,
|
859 |
+
"rewards/accuracies": 0.7749999761581421,
|
860 |
+
"rewards/chosen": -0.3144986033439636,
|
861 |
+
"rewards/margins": 0.2219509333372116,
|
862 |
+
"rewards/rejected": -0.5364495515823364,
|
863 |
+
"step": 600
|
864 |
+
},
|
865 |
+
{
|
866 |
+
"epoch": 0.93,
|
867 |
+
"learning_rate": 7.46189337174788e-08,
|
868 |
+
"logits/chosen": 0.2405429631471634,
|
869 |
+
"logits/rejected": 0.3472765386104584,
|
870 |
+
"logps/chosen": -657.0903930664062,
|
871 |
+
"logps/rejected": -797.1058959960938,
|
872 |
+
"loss": 0.2369,
|
873 |
+
"rewards/accuracies": 0.737500011920929,
|
874 |
+
"rewards/chosen": -0.310764878988266,
|
875 |
+
"rewards/margins": 0.20058993995189667,
|
876 |
+
"rewards/rejected": -0.5113548040390015,
|
877 |
+
"step": 610
|
878 |
+
},
|
879 |
+
{
|
880 |
+
"epoch": 0.94,
|
881 |
+
"learning_rate": 4.579103667367385e-08,
|
882 |
+
"logits/chosen": 0.2070445716381073,
|
883 |
+
"logits/rejected": 0.4006822109222412,
|
884 |
+
"logps/chosen": -655.3494873046875,
|
885 |
+
"logps/rejected": -780.321533203125,
|
886 |
+
"loss": 0.222,
|
887 |
+
"rewards/accuracies": 0.6937500238418579,
|
888 |
+
"rewards/chosen": -0.32885631918907166,
|
889 |
+
"rewards/margins": 0.17823253571987152,
|
890 |
+
"rewards/rejected": -0.507088840007782,
|
891 |
+
"step": 620
|
892 |
+
},
|
893 |
+
{
|
894 |
+
"epoch": 0.96,
|
895 |
+
"learning_rate": 2.3919861577572924e-08,
|
896 |
+
"logits/chosen": 0.23205919563770294,
|
897 |
+
"logits/rejected": 0.3889433741569519,
|
898 |
+
"logps/chosen": -668.5777587890625,
|
899 |
+
"logps/rejected": -799.8650512695312,
|
900 |
+
"loss": 0.235,
|
901 |
+
"rewards/accuracies": 0.75,
|
902 |
+
"rewards/chosen": -0.3039870858192444,
|
903 |
+
"rewards/margins": 0.20831787586212158,
|
904 |
+
"rewards/rejected": -0.5123049020767212,
|
905 |
+
"step": 630
|
906 |
+
},
|
907 |
+
{
|
908 |
+
"epoch": 0.98,
|
909 |
+
"learning_rate": 9.067404651211808e-09,
|
910 |
+
"logits/chosen": 0.1636372208595276,
|
911 |
+
"logits/rejected": 0.3309451639652252,
|
912 |
+
"logps/chosen": -615.0830688476562,
|
913 |
+
"logps/rejected": -778.7999267578125,
|
914 |
+
"loss": 0.2582,
|
915 |
+
"rewards/accuracies": 0.7250000238418579,
|
916 |
+
"rewards/chosen": -0.3035675883293152,
|
917 |
+
"rewards/margins": 0.19927778840065002,
|
918 |
+
"rewards/rejected": -0.5028454065322876,
|
919 |
+
"step": 640
|
920 |
+
},
|
921 |
+
{
|
922 |
+
"epoch": 0.99,
|
923 |
+
"learning_rate": 1.2757667974155896e-09,
|
924 |
+
"logits/chosen": 0.23857417702674866,
|
925 |
+
"logits/rejected": 0.35767877101898193,
|
926 |
+
"logps/chosen": -685.1931762695312,
|
927 |
+
"logps/rejected": -803.7652587890625,
|
928 |
+
"loss": 0.2486,
|
929 |
+
"rewards/accuracies": 0.75,
|
930 |
+
"rewards/chosen": -0.3033232092857361,
|
931 |
+
"rewards/margins": 0.2285349816083908,
|
932 |
+
"rewards/rejected": -0.5318582057952881,
|
933 |
+
"step": 650
|
934 |
+
},
|
935 |
+
{
|
936 |
+
"epoch": 1.0,
|
937 |
+
"step": 656,
|
938 |
+
"total_flos": 0.0,
|
939 |
+
"train_loss": 0.2497515143236009,
|
940 |
+
"train_runtime": 7980.5775,
|
941 |
+
"train_samples_per_second": 2.631,
|
942 |
+
"train_steps_per_second": 0.082
|
943 |
+
}
|
944 |
+
],
|
945 |
+
"logging_steps": 10,
|
946 |
+
"max_steps": 656,
|
947 |
+
"num_input_tokens_seen": 0,
|
948 |
+
"num_train_epochs": 1,
|
949 |
+
"save_steps": 100,
|
950 |
+
"total_flos": 0.0,
|
951 |
+
"train_batch_size": 4,
|
952 |
+
"trial_name": null,
|
953 |
+
"trial_params": null
|
954 |
+
}
|