Model save
Browse files- README.md +62 -0
- adapter_model.safetensors +1 -1
- all_results.json +8 -0
- runs/May15_22-49-58_gpu4-119-5/events.out.tfevents.1715777510.gpu4-119-5.3384151.0 +2 -2
- train_results.json +8 -0
- trainer_state.json +590 -0
README.md
ADDED
@@ -0,0 +1,62 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: mit
|
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-v28-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-v28-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: 3
|
44 |
+
- gradient_accumulation_steps: 4
|
45 |
+
- total_train_batch_size: 48
|
46 |
+
- total_eval_batch_size: 12
|
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:b34fee57762819ccb37993169e959b1ce25a4347e4dc942a5976b84529b69ff7
|
3 |
size 167807296
|
all_results.json
ADDED
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 1.0,
|
3 |
+
"train_loss": 0.19170005051862626,
|
4 |
+
"train_runtime": 5499.6327,
|
5 |
+
"train_samples": 19170,
|
6 |
+
"train_samples_per_second": 3.486,
|
7 |
+
"train_steps_per_second": 0.073
|
8 |
+
}
|
runs/May15_22-49-58_gpu4-119-5/events.out.tfevents.1715777510.gpu4-119-5.3384151.0
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 sha256:69116fe6d6aab3cdb47bcf1a1f406f3607bc38aff9634cf2c35f664ebcbc66a3
|
3 |
+
size 30406
|
train_results.json
ADDED
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 1.0,
|
3 |
+
"train_loss": 0.19170005051862626,
|
4 |
+
"train_runtime": 5499.6327,
|
5 |
+
"train_samples": 19170,
|
6 |
+
"train_samples_per_second": 3.486,
|
7 |
+
"train_steps_per_second": 0.073
|
8 |
+
}
|
trainer_state.json
ADDED
@@ -0,0 +1,590 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_metric": null,
|
3 |
+
"best_model_checkpoint": null,
|
4 |
+
"epoch": 0.9987484355444305,
|
5 |
+
"eval_steps": 500,
|
6 |
+
"global_step": 399,
|
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": 1.2500000000000002e-07,
|
14 |
+
"logits/chosen": 0.3541638255119324,
|
15 |
+
"logits/rejected": 0.2986571788787842,
|
16 |
+
"logps/chosen": -321.2114562988281,
|
17 |
+
"logps/rejected": -353.251708984375,
|
18 |
+
"loss": 0.2236,
|
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": 0.03,
|
27 |
+
"learning_rate": 1.25e-06,
|
28 |
+
"logits/chosen": 0.1456776261329651,
|
29 |
+
"logits/rejected": 0.25098955631256104,
|
30 |
+
"logps/chosen": -287.8302917480469,
|
31 |
+
"logps/rejected": -297.34259033203125,
|
32 |
+
"loss": 0.2182,
|
33 |
+
"rewards/accuracies": 0.3541666567325592,
|
34 |
+
"rewards/chosen": 5.259945464786142e-05,
|
35 |
+
"rewards/margins": -8.035360224312171e-06,
|
36 |
+
"rewards/rejected": 6.063482214813121e-05,
|
37 |
+
"step": 10
|
38 |
+
},
|
39 |
+
{
|
40 |
+
"epoch": 0.05,
|
41 |
+
"learning_rate": 2.5e-06,
|
42 |
+
"logits/chosen": 0.14523428678512573,
|
43 |
+
"logits/rejected": 0.19231222569942474,
|
44 |
+
"logps/chosen": -328.94781494140625,
|
45 |
+
"logps/rejected": -320.3280029296875,
|
46 |
+
"loss": 0.2358,
|
47 |
+
"rewards/accuracies": 0.4625000059604645,
|
48 |
+
"rewards/chosen": -2.648494228196796e-05,
|
49 |
+
"rewards/margins": 9.429633064428344e-05,
|
50 |
+
"rewards/rejected": -0.0001207812674692832,
|
51 |
+
"step": 20
|
52 |
+
},
|
53 |
+
{
|
54 |
+
"epoch": 0.08,
|
55 |
+
"learning_rate": 3.7500000000000005e-06,
|
56 |
+
"logits/chosen": 0.1521739363670349,
|
57 |
+
"logits/rejected": 0.23981845378875732,
|
58 |
+
"logps/chosen": -326.76171875,
|
59 |
+
"logps/rejected": -321.0068054199219,
|
60 |
+
"loss": 0.2182,
|
61 |
+
"rewards/accuracies": 0.574999988079071,
|
62 |
+
"rewards/chosen": -3.38268764608074e-05,
|
63 |
+
"rewards/margins": 0.0005001984536647797,
|
64 |
+
"rewards/rejected": -0.0005340253701433539,
|
65 |
+
"step": 30
|
66 |
+
},
|
67 |
+
{
|
68 |
+
"epoch": 0.1,
|
69 |
+
"learning_rate": 5e-06,
|
70 |
+
"logits/chosen": 0.12187125533819199,
|
71 |
+
"logits/rejected": 0.22642993927001953,
|
72 |
+
"logps/chosen": -315.1197204589844,
|
73 |
+
"logps/rejected": -298.721435546875,
|
74 |
+
"loss": 0.2174,
|
75 |
+
"rewards/accuracies": 0.5562499761581421,
|
76 |
+
"rewards/chosen": -0.0005543121951632202,
|
77 |
+
"rewards/margins": 0.0006722904508933425,
|
78 |
+
"rewards/rejected": -0.0012266027042642236,
|
79 |
+
"step": 40
|
80 |
+
},
|
81 |
+
{
|
82 |
+
"epoch": 0.13,
|
83 |
+
"learning_rate": 4.990433706654258e-06,
|
84 |
+
"logits/chosen": 0.17174628376960754,
|
85 |
+
"logits/rejected": 0.21399536728858948,
|
86 |
+
"logps/chosen": -306.0259094238281,
|
87 |
+
"logps/rejected": -306.9112243652344,
|
88 |
+
"loss": 0.1905,
|
89 |
+
"rewards/accuracies": 0.543749988079071,
|
90 |
+
"rewards/chosen": -0.0015839699190109968,
|
91 |
+
"rewards/margins": 0.002141424687579274,
|
92 |
+
"rewards/rejected": -0.0037253950722515583,
|
93 |
+
"step": 50
|
94 |
+
},
|
95 |
+
{
|
96 |
+
"epoch": 0.15,
|
97 |
+
"learning_rate": 4.961808037791733e-06,
|
98 |
+
"logits/chosen": 0.18885941803455353,
|
99 |
+
"logits/rejected": 0.18989692628383636,
|
100 |
+
"logps/chosen": -333.9552307128906,
|
101 |
+
"logps/rejected": -316.93634033203125,
|
102 |
+
"loss": 0.2192,
|
103 |
+
"rewards/accuracies": 0.543749988079071,
|
104 |
+
"rewards/chosen": -0.00444876030087471,
|
105 |
+
"rewards/margins": 0.0024011698551476,
|
106 |
+
"rewards/rejected": -0.0068499306216835976,
|
107 |
+
"step": 60
|
108 |
+
},
|
109 |
+
{
|
110 |
+
"epoch": 0.18,
|
111 |
+
"learning_rate": 4.914342066648869e-06,
|
112 |
+
"logits/chosen": 0.12300668656826019,
|
113 |
+
"logits/rejected": 0.20358912646770477,
|
114 |
+
"logps/chosen": -338.68841552734375,
|
115 |
+
"logps/rejected": -346.804443359375,
|
116 |
+
"loss": 0.2047,
|
117 |
+
"rewards/accuracies": 0.612500011920929,
|
118 |
+
"rewards/chosen": -0.005950641818344593,
|
119 |
+
"rewards/margins": 0.0063765160739421844,
|
120 |
+
"rewards/rejected": -0.012327157892286777,
|
121 |
+
"step": 70
|
122 |
+
},
|
123 |
+
{
|
124 |
+
"epoch": 0.2,
|
125 |
+
"learning_rate": 4.848399051948784e-06,
|
126 |
+
"logits/chosen": 0.13564321398735046,
|
127 |
+
"logits/rejected": 0.17065586149692535,
|
128 |
+
"logps/chosen": -361.12530517578125,
|
129 |
+
"logps/rejected": -399.69268798828125,
|
130 |
+
"loss": 0.2007,
|
131 |
+
"rewards/accuracies": 0.6812499761581421,
|
132 |
+
"rewards/chosen": -0.014969304203987122,
|
133 |
+
"rewards/margins": 0.016210615634918213,
|
134 |
+
"rewards/rejected": -0.031179919838905334,
|
135 |
+
"step": 80
|
136 |
+
},
|
137 |
+
{
|
138 |
+
"epoch": 0.23,
|
139 |
+
"learning_rate": 4.764483657869654e-06,
|
140 |
+
"logits/chosen": 0.08848075568675995,
|
141 |
+
"logits/rejected": 0.10415778309106827,
|
142 |
+
"logps/chosen": -377.1496887207031,
|
143 |
+
"logps/rejected": -380.4229431152344,
|
144 |
+
"loss": 0.2091,
|
145 |
+
"rewards/accuracies": 0.543749988079071,
|
146 |
+
"rewards/chosen": -0.03349294513463974,
|
147 |
+
"rewards/margins": 0.019886447116732597,
|
148 |
+
"rewards/rejected": -0.05337939411401749,
|
149 |
+
"step": 90
|
150 |
+
},
|
151 |
+
{
|
152 |
+
"epoch": 0.25,
|
153 |
+
"learning_rate": 4.663238091832267e-06,
|
154 |
+
"logits/chosen": 0.18175415694713593,
|
155 |
+
"logits/rejected": 0.18852263689041138,
|
156 |
+
"logps/chosen": -354.5711364746094,
|
157 |
+
"logps/rejected": -434.7327575683594,
|
158 |
+
"loss": 0.1934,
|
159 |
+
"rewards/accuracies": 0.643750011920929,
|
160 |
+
"rewards/chosen": -0.049878861755132675,
|
161 |
+
"rewards/margins": 0.05035804584622383,
|
162 |
+
"rewards/rejected": -0.1002369150519371,
|
163 |
+
"step": 100
|
164 |
+
},
|
165 |
+
{
|
166 |
+
"epoch": 0.28,
|
167 |
+
"learning_rate": 4.54543718966436e-06,
|
168 |
+
"logits/chosen": 0.09073193371295929,
|
169 |
+
"logits/rejected": 0.16552071273326874,
|
170 |
+
"logps/chosen": -385.48699951171875,
|
171 |
+
"logps/rejected": -428.86865234375,
|
172 |
+
"loss": 0.1949,
|
173 |
+
"rewards/accuracies": 0.550000011920929,
|
174 |
+
"rewards/chosen": -0.0685494989156723,
|
175 |
+
"rewards/margins": 0.03330964595079422,
|
176 |
+
"rewards/rejected": -0.10185913741588593,
|
177 |
+
"step": 110
|
178 |
+
},
|
179 |
+
{
|
180 |
+
"epoch": 0.3,
|
181 |
+
"learning_rate": 4.411982485755156e-06,
|
182 |
+
"logits/chosen": 0.1523444950580597,
|
183 |
+
"logits/rejected": 0.21539661288261414,
|
184 |
+
"logps/chosen": -384.61822509765625,
|
185 |
+
"logps/rejected": -413.7029724121094,
|
186 |
+
"loss": 0.1987,
|
187 |
+
"rewards/accuracies": 0.65625,
|
188 |
+
"rewards/chosen": -0.05241132900118828,
|
189 |
+
"rewards/margins": 0.04536561295390129,
|
190 |
+
"rewards/rejected": -0.09777694940567017,
|
191 |
+
"step": 120
|
192 |
+
},
|
193 |
+
{
|
194 |
+
"epoch": 0.33,
|
195 |
+
"learning_rate": 4.263895313581429e-06,
|
196 |
+
"logits/chosen": 0.16519860923290253,
|
197 |
+
"logits/rejected": 0.22018559277057648,
|
198 |
+
"logps/chosen": -457.999267578125,
|
199 |
+
"logps/rejected": -465.42724609375,
|
200 |
+
"loss": 0.1924,
|
201 |
+
"rewards/accuracies": 0.606249988079071,
|
202 |
+
"rewards/chosen": -0.08990312367677689,
|
203 |
+
"rewards/margins": 0.03736043721437454,
|
204 |
+
"rewards/rejected": -0.12726356089115143,
|
205 |
+
"step": 130
|
206 |
+
},
|
207 |
+
{
|
208 |
+
"epoch": 0.35,
|
209 |
+
"learning_rate": 4.1023089894069815e-06,
|
210 |
+
"logits/chosen": 0.16239216923713684,
|
211 |
+
"logits/rejected": 0.1908317506313324,
|
212 |
+
"logps/chosen": -464.09332275390625,
|
213 |
+
"logps/rejected": -468.3551330566406,
|
214 |
+
"loss": 0.185,
|
215 |
+
"rewards/accuracies": 0.643750011920929,
|
216 |
+
"rewards/chosen": -0.1087506040930748,
|
217 |
+
"rewards/margins": 0.043623246252536774,
|
218 |
+
"rewards/rejected": -0.15237386524677277,
|
219 |
+
"step": 140
|
220 |
+
},
|
221 |
+
{
|
222 |
+
"epoch": 0.38,
|
223 |
+
"learning_rate": 3.928460138973984e-06,
|
224 |
+
"logits/chosen": 0.1521633267402649,
|
225 |
+
"logits/rejected": 0.21451088786125183,
|
226 |
+
"logps/chosen": -425.41912841796875,
|
227 |
+
"logps/rejected": -481.1216735839844,
|
228 |
+
"loss": 0.1898,
|
229 |
+
"rewards/accuracies": 0.518750011920929,
|
230 |
+
"rewards/chosen": -0.10458217561244965,
|
231 |
+
"rewards/margins": 0.03971753641963005,
|
232 |
+
"rewards/rejected": -0.14429971575737,
|
233 |
+
"step": 150
|
234 |
+
},
|
235 |
+
{
|
236 |
+
"epoch": 0.4,
|
237 |
+
"learning_rate": 3.7436792335632864e-06,
|
238 |
+
"logits/chosen": 0.24725870788097382,
|
239 |
+
"logits/rejected": 0.1957666426897049,
|
240 |
+
"logps/chosen": -455.04736328125,
|
241 |
+
"logps/rejected": -517.961181640625,
|
242 |
+
"loss": 0.1817,
|
243 |
+
"rewards/accuracies": 0.668749988079071,
|
244 |
+
"rewards/chosen": -0.10722579061985016,
|
245 |
+
"rewards/margins": 0.048337507992982864,
|
246 |
+
"rewards/rejected": -0.15556327998638153,
|
247 |
+
"step": 160
|
248 |
+
},
|
249 |
+
{
|
250 |
+
"epoch": 0.43,
|
251 |
+
"learning_rate": 3.5493804078515697e-06,
|
252 |
+
"logits/chosen": 0.2167464792728424,
|
253 |
+
"logits/rejected": 0.2552216649055481,
|
254 |
+
"logps/chosen": -427.9091796875,
|
255 |
+
"logps/rejected": -485.00732421875,
|
256 |
+
"loss": 0.178,
|
257 |
+
"rewards/accuracies": 0.59375,
|
258 |
+
"rewards/chosen": -0.09426573663949966,
|
259 |
+
"rewards/margins": 0.05705801397562027,
|
260 |
+
"rewards/rejected": -0.15132373571395874,
|
261 |
+
"step": 170
|
262 |
+
},
|
263 |
+
{
|
264 |
+
"epoch": 0.45,
|
265 |
+
"learning_rate": 3.3470506374896266e-06,
|
266 |
+
"logits/chosen": 0.20716889202594757,
|
267 |
+
"logits/rejected": 0.22972209751605988,
|
268 |
+
"logps/chosen": -466.29718017578125,
|
269 |
+
"logps/rejected": -509.15911865234375,
|
270 |
+
"loss": 0.1718,
|
271 |
+
"rewards/accuracies": 0.6499999761581421,
|
272 |
+
"rewards/chosen": -0.1065358892083168,
|
273 |
+
"rewards/margins": 0.058396100997924805,
|
274 |
+
"rewards/rejected": -0.164931982755661,
|
275 |
+
"step": 180
|
276 |
+
},
|
277 |
+
{
|
278 |
+
"epoch": 0.48,
|
279 |
+
"learning_rate": 3.1382383592261445e-06,
|
280 |
+
"logits/chosen": 0.22018647193908691,
|
281 |
+
"logits/rejected": 0.2470862865447998,
|
282 |
+
"logps/chosen": -411.8152770996094,
|
283 |
+
"logps/rejected": -447.99456787109375,
|
284 |
+
"loss": 0.1844,
|
285 |
+
"rewards/accuracies": 0.581250011920929,
|
286 |
+
"rewards/chosen": -0.10458564758300781,
|
287 |
+
"rewards/margins": 0.04747054725885391,
|
288 |
+
"rewards/rejected": -0.15205618739128113,
|
289 |
+
"step": 190
|
290 |
+
},
|
291 |
+
{
|
292 |
+
"epoch": 0.5,
|
293 |
+
"learning_rate": 2.9245416206675732e-06,
|
294 |
+
"logits/chosen": 0.20030078291893005,
|
295 |
+
"logits/rejected": 0.2441815435886383,
|
296 |
+
"logps/chosen": -471.25543212890625,
|
297 |
+
"logps/rejected": -512.8248901367188,
|
298 |
+
"loss": 0.1892,
|
299 |
+
"rewards/accuracies": 0.606249988079071,
|
300 |
+
"rewards/chosen": -0.13170954585075378,
|
301 |
+
"rewards/margins": 0.053105730563402176,
|
302 |
+
"rewards/rejected": -0.18481527268886566,
|
303 |
+
"step": 200
|
304 |
+
},
|
305 |
+
{
|
306 |
+
"epoch": 0.53,
|
307 |
+
"learning_rate": 2.7075958503643746e-06,
|
308 |
+
"logits/chosen": 0.18546412885189056,
|
309 |
+
"logits/rejected": 0.22659745812416077,
|
310 |
+
"logps/chosen": -435.7657775878906,
|
311 |
+
"logps/rejected": -492.03955078125,
|
312 |
+
"loss": 0.1854,
|
313 |
+
"rewards/accuracies": 0.6625000238418579,
|
314 |
+
"rewards/chosen": -0.09934359043836594,
|
315 |
+
"rewards/margins": 0.0697723850607872,
|
316 |
+
"rewards/rejected": -0.16911597549915314,
|
317 |
+
"step": 210
|
318 |
+
},
|
319 |
+
{
|
320 |
+
"epoch": 0.55,
|
321 |
+
"learning_rate": 2.4890613418196207e-06,
|
322 |
+
"logits/chosen": 0.2379184067249298,
|
323 |
+
"logits/rejected": 0.2359950989484787,
|
324 |
+
"logps/chosen": -438.42034912109375,
|
325 |
+
"logps/rejected": -483.0135192871094,
|
326 |
+
"loss": 0.1638,
|
327 |
+
"rewards/accuracies": 0.581250011920929,
|
328 |
+
"rewards/chosen": -0.10627691447734833,
|
329 |
+
"rewards/margins": 0.04897074028849602,
|
330 |
+
"rewards/rejected": -0.15524765849113464,
|
331 |
+
"step": 220
|
332 |
+
},
|
333 |
+
{
|
334 |
+
"epoch": 0.58,
|
335 |
+
"learning_rate": 2.2706105472052367e-06,
|
336 |
+
"logits/chosen": 0.1908051073551178,
|
337 |
+
"logits/rejected": 0.22610357403755188,
|
338 |
+
"logps/chosen": -463.33984375,
|
339 |
+
"logps/rejected": -497.69830322265625,
|
340 |
+
"loss": 0.1863,
|
341 |
+
"rewards/accuracies": 0.6625000238418579,
|
342 |
+
"rewards/chosen": -0.11745200306177139,
|
343 |
+
"rewards/margins": 0.06438393890857697,
|
344 |
+
"rewards/rejected": -0.18183593451976776,
|
345 |
+
"step": 230
|
346 |
+
},
|
347 |
+
{
|
348 |
+
"epoch": 0.6,
|
349 |
+
"learning_rate": 2.053915278027536e-06,
|
350 |
+
"logits/chosen": 0.12600186467170715,
|
351 |
+
"logits/rejected": 0.23243729770183563,
|
352 |
+
"logps/chosen": -501.31561279296875,
|
353 |
+
"logps/rejected": -539.3880615234375,
|
354 |
+
"loss": 0.182,
|
355 |
+
"rewards/accuracies": 0.6187499761581421,
|
356 |
+
"rewards/chosen": -0.14145687222480774,
|
357 |
+
"rewards/margins": 0.07019603252410889,
|
358 |
+
"rewards/rejected": -0.21165291965007782,
|
359 |
+
"step": 240
|
360 |
+
},
|
361 |
+
{
|
362 |
+
"epoch": 0.63,
|
363 |
+
"learning_rate": 1.8406339106957888e-06,
|
364 |
+
"logits/chosen": 0.18264178931713104,
|
365 |
+
"logits/rejected": 0.26273074746131897,
|
366 |
+
"logps/chosen": -487.56524658203125,
|
367 |
+
"logps/rejected": -538.5673828125,
|
368 |
+
"loss": 0.19,
|
369 |
+
"rewards/accuracies": 0.675000011920929,
|
370 |
+
"rewards/chosen": -0.13434770703315735,
|
371 |
+
"rewards/margins": 0.07964815199375153,
|
372 |
+
"rewards/rejected": -0.21399584412574768,
|
373 |
+
"step": 250
|
374 |
+
},
|
375 |
+
{
|
376 |
+
"epoch": 0.65,
|
377 |
+
"learning_rate": 1.6323986949100579e-06,
|
378 |
+
"logits/chosen": 0.211247056722641,
|
379 |
+
"logits/rejected": 0.24369597434997559,
|
380 |
+
"logps/chosen": -467.6783142089844,
|
381 |
+
"logps/rejected": -553.1571044921875,
|
382 |
+
"loss": 0.1756,
|
383 |
+
"rewards/accuracies": 0.7124999761581421,
|
384 |
+
"rewards/chosen": -0.13178598880767822,
|
385 |
+
"rewards/margins": 0.08503976464271545,
|
386 |
+
"rewards/rejected": -0.21682576835155487,
|
387 |
+
"step": 260
|
388 |
+
},
|
389 |
+
{
|
390 |
+
"epoch": 0.68,
|
391 |
+
"learning_rate": 1.4308032619976382e-06,
|
392 |
+
"logits/chosen": 0.2250673770904541,
|
393 |
+
"logits/rejected": 0.25937551259994507,
|
394 |
+
"logps/chosen": -474.22991943359375,
|
395 |
+
"logps/rejected": -576.9962768554688,
|
396 |
+
"loss": 0.183,
|
397 |
+
"rewards/accuracies": 0.6499999761581421,
|
398 |
+
"rewards/chosen": -0.12007347494363785,
|
399 |
+
"rewards/margins": 0.10777894407510757,
|
400 |
+
"rewards/rejected": -0.22785241901874542,
|
401 |
+
"step": 270
|
402 |
+
},
|
403 |
+
{
|
404 |
+
"epoch": 0.7,
|
405 |
+
"learning_rate": 1.2373904287972503e-06,
|
406 |
+
"logits/chosen": 0.19488680362701416,
|
407 |
+
"logits/rejected": 0.2598015367984772,
|
408 |
+
"logps/chosen": -463.74713134765625,
|
409 |
+
"logps/rejected": -519.3098754882812,
|
410 |
+
"loss": 0.1688,
|
411 |
+
"rewards/accuracies": 0.637499988079071,
|
412 |
+
"rewards/chosen": -0.11257203668355942,
|
413 |
+
"rewards/margins": 0.07239626348018646,
|
414 |
+
"rewards/rejected": -0.18496832251548767,
|
415 |
+
"step": 280
|
416 |
+
},
|
417 |
+
{
|
418 |
+
"epoch": 0.73,
|
419 |
+
"learning_rate": 1.0536403904282774e-06,
|
420 |
+
"logits/chosen": 0.19138000905513763,
|
421 |
+
"logits/rejected": 0.25364065170288086,
|
422 |
+
"logps/chosen": -437.4383239746094,
|
423 |
+
"logps/rejected": -513.4635009765625,
|
424 |
+
"loss": 0.1822,
|
425 |
+
"rewards/accuracies": 0.6312500238418579,
|
426 |
+
"rewards/chosen": -0.11546895653009415,
|
427 |
+
"rewards/margins": 0.08019855618476868,
|
428 |
+
"rewards/rejected": -0.19566750526428223,
|
429 |
+
"step": 290
|
430 |
+
},
|
431 |
+
{
|
432 |
+
"epoch": 0.75,
|
433 |
+
"learning_rate": 8.809593923061812e-07,
|
434 |
+
"logits/chosen": 0.18998712301254272,
|
435 |
+
"logits/rejected": 0.2664821743965149,
|
436 |
+
"logps/chosen": -441.168701171875,
|
437 |
+
"logps/rejected": -503.0856018066406,
|
438 |
+
"loss": 0.2118,
|
439 |
+
"rewards/accuracies": 0.550000011920929,
|
440 |
+
"rewards/chosen": -0.12977081537246704,
|
441 |
+
"rewards/margins": 0.05418657138943672,
|
442 |
+
"rewards/rejected": -0.18395736813545227,
|
443 |
+
"step": 300
|
444 |
+
},
|
445 |
+
{
|
446 |
+
"epoch": 0.78,
|
447 |
+
"learning_rate": 7.206689680975798e-07,
|
448 |
+
"logits/chosen": 0.2218310385942459,
|
449 |
+
"logits/rejected": 0.23224084079265594,
|
450 |
+
"logps/chosen": -409.9444885253906,
|
451 |
+
"logps/rejected": -473.93536376953125,
|
452 |
+
"loss": 0.1855,
|
453 |
+
"rewards/accuracies": 0.65625,
|
454 |
+
"rewards/chosen": -0.09453083574771881,
|
455 |
+
"rewards/margins": 0.0672791600227356,
|
456 |
+
"rewards/rejected": -0.1618099957704544,
|
457 |
+
"step": 310
|
458 |
+
},
|
459 |
+
{
|
460 |
+
"epoch": 0.8,
|
461 |
+
"learning_rate": 5.739958259772663e-07,
|
462 |
+
"logits/chosen": 0.14368923008441925,
|
463 |
+
"logits/rejected": 0.2114243060350418,
|
464 |
+
"logps/chosen": -422.76171875,
|
465 |
+
"logps/rejected": -484.56103515625,
|
466 |
+
"loss": 0.1757,
|
467 |
+
"rewards/accuracies": 0.65625,
|
468 |
+
"rewards/chosen": -0.10932054370641708,
|
469 |
+
"rewards/margins": 0.07388332486152649,
|
470 |
+
"rewards/rejected": -0.18320389091968536,
|
471 |
+
"step": 320
|
472 |
+
},
|
473 |
+
{
|
474 |
+
"epoch": 0.83,
|
475 |
+
"learning_rate": 4.420624605880136e-07,
|
476 |
+
"logits/chosen": 0.18750019371509552,
|
477 |
+
"logits/rejected": 0.23132064938545227,
|
478 |
+
"logps/chosen": -434.45233154296875,
|
479 |
+
"logps/rejected": -525.4594116210938,
|
480 |
+
"loss": 0.2015,
|
481 |
+
"rewards/accuracies": 0.6187499761581421,
|
482 |
+
"rewards/chosen": -0.125051349401474,
|
483 |
+
"rewards/margins": 0.051269471645355225,
|
484 |
+
"rewards/rejected": -0.17632082104682922,
|
485 |
+
"step": 330
|
486 |
+
},
|
487 |
+
{
|
488 |
+
"epoch": 0.85,
|
489 |
+
"learning_rate": 3.258785625501448e-07,
|
490 |
+
"logits/chosen": 0.1886998862028122,
|
491 |
+
"logits/rejected": 0.22365060448646545,
|
492 |
+
"logps/chosen": -438.25457763671875,
|
493 |
+
"logps/rejected": -477.1897888183594,
|
494 |
+
"loss": 0.1806,
|
495 |
+
"rewards/accuracies": 0.643750011920929,
|
496 |
+
"rewards/chosen": -0.11255072057247162,
|
497 |
+
"rewards/margins": 0.06545019894838333,
|
498 |
+
"rewards/rejected": -0.17800091207027435,
|
499 |
+
"step": 340
|
500 |
+
},
|
501 |
+
{
|
502 |
+
"epoch": 0.88,
|
503 |
+
"learning_rate": 2.263332912642066e-07,
|
504 |
+
"logits/chosen": 0.20381930470466614,
|
505 |
+
"logits/rejected": 0.24624279141426086,
|
506 |
+
"logps/chosen": -473.3438415527344,
|
507 |
+
"logps/rejected": -501.24713134765625,
|
508 |
+
"loss": 0.1847,
|
509 |
+
"rewards/accuracies": 0.6000000238418579,
|
510 |
+
"rewards/chosen": -0.12005212157964706,
|
511 |
+
"rewards/margins": 0.06751255691051483,
|
512 |
+
"rewards/rejected": -0.1875646859407425,
|
513 |
+
"step": 350
|
514 |
+
},
|
515 |
+
{
|
516 |
+
"epoch": 0.9,
|
517 |
+
"learning_rate": 1.4418847014323944e-07,
|
518 |
+
"logits/chosen": 0.17499907314777374,
|
519 |
+
"logits/rejected": 0.2898316979408264,
|
520 |
+
"logps/chosen": -470.5660095214844,
|
521 |
+
"logps/rejected": -507.25201416015625,
|
522 |
+
"loss": 0.1811,
|
523 |
+
"rewards/accuracies": 0.6312500238418579,
|
524 |
+
"rewards/chosen": -0.11223053932189941,
|
525 |
+
"rewards/margins": 0.06604616343975067,
|
526 |
+
"rewards/rejected": -0.1782766878604889,
|
527 |
+
"step": 360
|
528 |
+
},
|
529 |
+
{
|
530 |
+
"epoch": 0.93,
|
531 |
+
"learning_rate": 8.007275635178585e-08,
|
532 |
+
"logits/chosen": 0.23265579342842102,
|
533 |
+
"logits/rejected": 0.24021784961223602,
|
534 |
+
"logps/chosen": -468.1248474121094,
|
535 |
+
"logps/rejected": -506.1273498535156,
|
536 |
+
"loss": 0.1988,
|
537 |
+
"rewards/accuracies": 0.5562499761581421,
|
538 |
+
"rewards/chosen": -0.12280559539794922,
|
539 |
+
"rewards/margins": 0.051576316356658936,
|
540 |
+
"rewards/rejected": -0.17438191175460815,
|
541 |
+
"step": 370
|
542 |
+
},
|
543 |
+
{
|
544 |
+
"epoch": 0.95,
|
545 |
+
"learning_rate": 3.447682967080618e-08,
|
546 |
+
"logits/chosen": 0.18692392110824585,
|
547 |
+
"logits/rejected": 0.2151685208082199,
|
548 |
+
"logps/chosen": -461.8233947753906,
|
549 |
+
"logps/rejected": -561.45947265625,
|
550 |
+
"loss": 0.1897,
|
551 |
+
"rewards/accuracies": 0.675000011920929,
|
552 |
+
"rewards/chosen": -0.11950306594371796,
|
553 |
+
"rewards/margins": 0.07405074685811996,
|
554 |
+
"rewards/rejected": -0.1935538351535797,
|
555 |
+
"step": 380
|
556 |
+
},
|
557 |
+
{
|
558 |
+
"epoch": 0.98,
|
559 |
+
"learning_rate": 7.749637308301361e-09,
|
560 |
+
"logits/chosen": 0.24688251316547394,
|
561 |
+
"logits/rejected": 0.2674398422241211,
|
562 |
+
"logps/chosen": -467.228515625,
|
563 |
+
"logps/rejected": -535.1661376953125,
|
564 |
+
"loss": 0.1783,
|
565 |
+
"rewards/accuracies": 0.6000000238418579,
|
566 |
+
"rewards/chosen": -0.11184708774089813,
|
567 |
+
"rewards/margins": 0.07584673166275024,
|
568 |
+
"rewards/rejected": -0.18769383430480957,
|
569 |
+
"step": 390
|
570 |
+
},
|
571 |
+
{
|
572 |
+
"epoch": 1.0,
|
573 |
+
"step": 399,
|
574 |
+
"total_flos": 0.0,
|
575 |
+
"train_loss": 0.19170005051862626,
|
576 |
+
"train_runtime": 5499.6327,
|
577 |
+
"train_samples_per_second": 3.486,
|
578 |
+
"train_steps_per_second": 0.073
|
579 |
+
}
|
580 |
+
],
|
581 |
+
"logging_steps": 10,
|
582 |
+
"max_steps": 399,
|
583 |
+
"num_input_tokens_seen": 0,
|
584 |
+
"num_train_epochs": 1,
|
585 |
+
"save_steps": 100,
|
586 |
+
"total_flos": 0.0,
|
587 |
+
"train_batch_size": 4,
|
588 |
+
"trial_name": null,
|
589 |
+
"trial_params": null
|
590 |
+
}
|