Hugging Face's logo

CharlesLi
/
OpenELM-1_1B-DPO-full-max-10-reward

Text Generation
Transformers
TensorBoard
Safetensors
openelm
trl
dpo
alignment-handbook
Generated from Trainer
conversational
custom_code
Model card Files Metrics Community
CharlesLi commited on
Commit
7a8c06f
·
verified ·
1 Parent(s): 6676f5b

Model save

Browse files
Files changed (9) hide show
  1. README.md +38 -37
  2. all_results.json +17 -4
  3. eval_results.json +16 -0
  4. model.safetensors +1 -1
  5. runs/Oct05_06-10-28_xe8545-a100-15/events.out.tfevents.1728114605.xe8545-a100-15.1041997.1 +3 -0
  6. runs/Oct06_14-29-28_xe8545-a100-15/events.out.tfevents.1728218504.xe8545-a100-15.1082351.0 +3 -0
  7. train_results.json +4 -4
  8. trainer_state.json +0 -0
  9. training_args.bin +1 -1
README.md CHANGED
@@ -3,6 +3,7 @@ library_name: transformers
3
  tags:
4
  - trl
5
  - dpo
 
6
  - generated_from_trainer
7
  model-index:
8
  - name: OpenELM-1_1B-DPO-full-max-10-reward
@@ -16,15 +17,15 @@ should probably proofread and complete it, then remove this comment. -->
16
 
17
  This model was trained from scratch on an unknown dataset.
18
  It achieves the following results on the evaluation set:
19
- - Loss: 1.5814
20
- - Rewards/chosen: -10.375
21
- - Rewards/rejected: -12.25
22
- - Rewards/accuracies: 0.5938
23
- - Rewards/margins: 1.8906
24
- - Logps/rejected: -1520.0
25
- - Logps/chosen: -1360.0
26
- - Logits/rejected: -5.6875
27
- - Logits/chosen: -7.5938
28
 
29
  ## Model description
30
 
@@ -61,34 +62,34 @@ The following hyperparameters were used during training:
61
 
62
  | Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
63
  |:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
64
- | 0.3737 | 0.1047 | 100 | 0.6771 | -2.0469 | -2.3906 | 0.5918 | 0.3301 | -528.0 | -524.0 | -13.625 | -13.875 |
65
- | 0.2761 | 0.2094 | 200 | 0.7596 | -3.1719 | -3.7344 | 0.6035 | 0.5703 | -664.0 | -636.0 | -8.5625 | -9.3125 |
66
- | 0.2447 | 0.3141 | 300 | 0.8172 | -3.8125 | -4.4688 | 0.6074 | 0.6484 | -736.0 | -700.0 | -13.0625 | -13.75 |
67
- | 0.2872 | 0.4188 | 400 | 0.7366 | -3.5469 | -4.0312 | 0.5977 | 0.4707 | -692.0 | -672.0 | -10.5625 | -11.4375 |
68
- | 0.255 | 0.5236 | 500 | 0.8523 | -4.875 | -5.75 | 0.6035 | 0.8672 | -864.0 | -808.0 | -11.4375 | -12.6875 |
69
- | 0.2543 | 0.6283 | 600 | 0.8723 | -4.6562 | -5.2812 | 0.5723 | 0.6328 | -820.0 | -784.0 | -13.5 | -14.1875 |
70
- | 0.2317 | 0.7330 | 700 | 0.8776 | -4.875 | -5.8125 | 0.6035 | 0.9141 | -868.0 | -808.0 | -9.125 | -10.8125 |
71
- | 0.2231 | 0.8377 | 800 | 1.0687 | -7.7188 | -8.8125 | 0.5723 | 1.0859 | -1168.0 | -1088.0 | -10.3125 | -11.5625 |
72
- | 0.2171 | 0.9424 | 900 | 0.9579 | -7.75 | -8.5625 | 0.5762 | 0.8398 | -1144.0 | -1096.0 | -10.125 | -11.0 |
73
- | 0.0657 | 1.0471 | 1000 | 1.0863 | -7.2812 | -8.375 | 0.5840 | 1.1016 | -1128.0 | -1048.0 | -8.3125 | -9.625 |
74
- | 0.0641 | 1.1518 | 1100 | 1.1397 | -8.0 | -9.1875 | 0.5977 | 1.1719 | -1208.0 | -1120.0 | -8.9375 | -10.125 |
75
- | 0.0464 | 1.2565 | 1200 | 1.2448 | -8.3125 | -9.5625 | 0.6016 | 1.2734 | -1248.0 | -1152.0 | -6.3438 | -7.9688 |
76
- | 0.0455 | 1.3613 | 1300 | 1.2368 | -8.1875 | -9.4375 | 0.5859 | 1.25 | -1232.0 | -1136.0 | -8.5 | -10.0625 |
77
- | 0.0614 | 1.4660 | 1400 | 1.2305 | -8.0 | -9.3125 | 0.5996 | 1.2578 | -1216.0 | -1120.0 | -7.875 | -9.375 |
78
- | 0.0573 | 1.5707 | 1500 | 1.2170 | -7.8125 | -9.0 | 0.5723 | 1.1641 | -1184.0 | -1104.0 | -11.0625 | -12.375 |
79
- | 0.0561 | 1.6754 | 1600 | 1.1795 | -9.1875 | -10.5 | 0.6094 | 1.2656 | -1336.0 | -1240.0 | -6.9062 | -8.75 |
80
- | 0.0735 | 1.7801 | 1700 | 1.3456 | -9.0625 | -10.625 | 0.6055 | 1.5859 | -1352.0 | -1224.0 | -7.0938 | -8.8125 |
81
- | 0.0646 | 1.8848 | 1800 | 1.1507 | -8.25 | -9.625 | 0.6055 | 1.3438 | -1248.0 | -1144.0 | -6.3438 | -7.875 |
82
- | 0.0256 | 1.9895 | 1900 | 1.2247 | -8.625 | -10.125 | 0.5977 | 1.4922 | -1304.0 | -1184.0 | -6.9062 | -8.5 |
83
- | 0.002 | 2.0942 | 2000 | 1.4274 | -9.9375 | -11.6875 | 0.5957 | 1.75 | -1456.0 | -1312.0 | -5.4688 | -7.3125 |
84
- | 0.0035 | 2.1990 | 2100 | 1.5121 | -10.125 | -12.0 | 0.6016 | 1.8438 | -1488.0 | -1328.0 | -5.625 | -7.5625 |
85
- | 0.0073 | 2.3037 | 2200 | 1.4838 | -10.25 | -12.0625 | 0.5996 | 1.8125 | -1496.0 | -1344.0 | -5.4062 | -7.25 |
86
- | 0.0081 | 2.4084 | 2300 | 1.5810 | -10.875 | -12.75 | 0.5957 | 1.8984 | -1568.0 | -1408.0 | -4.8125 | -6.6562 |
87
- | 0.0074 | 2.5131 | 2400 | 1.6088 | -10.8125 | -12.6875 | 0.5918 | 1.9141 | -1560.0 | -1400.0 | -5.3438 | -7.2188 |
88
- | 0.0063 | 2.6178 | 2500 | 1.5735 | -10.5 | -12.375 | 0.5938 | 1.8828 | -1528.0 | -1368.0 | -5.5625 | -7.4375 |
89
- | 0.0058 | 2.7225 | 2600 | 1.5902 | -10.5 | -12.4375 | 0.5918 | 1.8906 | -1528.0 | -1368.0 | -5.6562 | -7.5312 |
90
- | 0.0071 | 2.8272 | 2700 | 1.5780 | -10.375 | -12.25 | 0.5898 | 1.8828 | -1512.0 | -1360.0 | -5.7188 | -7.625 |
91
- | 0.0063 | 2.9319 | 2800 | 1.5814 | -10.375 | -12.25 | 0.5938 | 1.8906 | -1520.0 | -1360.0 | -5.6875 | -7.5938 |
92
 
93
 
94
  ### Framework versions
 
3
  tags:
4
  - trl
5
  - dpo
6
+ - alignment-handbook
7
  - generated_from_trainer
8
  model-index:
9
  - name: OpenELM-1_1B-DPO-full-max-10-reward
 
17
 
18
  This model was trained from scratch on an unknown dataset.
19
  It achieves the following results on the evaluation set:
20
+ - Loss: 1.5774
21
+ - Rewards/chosen: -11.5
22
+ - Rewards/rejected: -13.5625
23
+ - Rewards/accuracies: 0.5879
24
+ - Rewards/margins: 2.0938
25
+ - Logps/rejected: -1648.0
26
+ - Logps/chosen: -1464.0
27
+ - Logits/rejected: -0.6484
28
+ - Logits/chosen: -2.5938
29
 
30
  ## Model description
31
 
 
62
 
63
  | Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
64
  |:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
65
+ | 0.3697 | 0.1047 | 100 | 0.6861 | -2.2656 | -2.5938 | 0.5957 | 0.3242 | -548.0 | -544.0 | -12.875 | -13.1875 |
66
+ | 0.2859 | 0.2094 | 200 | 0.7473 | -2.8281 | -3.3281 | 0.5957 | 0.5039 | -620.0 | -600.0 | -7.125 | -7.9062 |
67
+ | 0.2457 | 0.3141 | 300 | 0.8945 | -4.25 | -4.8438 | 0.5762 | 0.5781 | -772.0 | -744.0 | -4.875 | -5.875 |
68
+ | 0.2779 | 0.4188 | 400 | 0.7848 | -4.6875 | -5.2812 | 0.5996 | 0.5742 | -816.0 | -788.0 | -5.6562 | -6.8125 |
69
+ | 0.2378 | 0.5236 | 500 | 0.8630 | -5.4062 | -6.4062 | 0.6016 | 1.0156 | -928.0 | -860.0 | -4.4375 | -5.7188 |
70
+ | 0.2478 | 0.6283 | 600 | 0.8302 | -5.4062 | -6.1562 | 0.6270 | 0.7383 | -904.0 | -860.0 | -3.5625 | -4.7812 |
71
+ | 0.2346 | 0.7330 | 700 | 0.8088 | -5.1562 | -5.9688 | 0.6211 | 0.8164 | -884.0 | -836.0 | -7.9062 | -9.0 |
72
+ | 0.2222 | 0.8377 | 800 | 0.9452 | -6.1562 | -6.9375 | 0.5645 | 0.7734 | -984.0 | -936.0 | -5.6875 | -7.0625 |
73
+ | 0.2077 | 0.9424 | 900 | 0.9077 | -6.9688 | -7.7812 | 0.5918 | 0.8125 | -1064.0 | -1016.0 | -6.6562 | -7.75 |
74
+ | 0.0616 | 1.0471 | 1000 | 1.1159 | -8.0 | -9.1875 | 0.5879 | 1.1953 | -1208.0 | -1120.0 | -3.4375 | -5.2188 |
75
+ | 0.0692 | 1.1518 | 1100 | 1.0924 | -7.5938 | -8.625 | 0.5859 | 1.0469 | -1152.0 | -1080.0 | -4.1875 | -5.6562 |
76
+ | 0.0488 | 1.2565 | 1200 | 1.3851 | -11.375 | -12.625 | 0.5938 | 1.2344 | -1552.0 | -1456.0 | -0.3047 | -1.7969 |
77
+ | 0.0483 | 1.3613 | 1300 | 1.1343 | -7.875 | -9.0 | 0.6094 | 1.1641 | -1192.0 | -1104.0 | -5.0312 | -6.75 |
78
+ | 0.072 | 1.4660 | 1400 | 1.1480 | -8.375 | -9.6875 | 0.6230 | 1.3438 | -1256.0 | -1152.0 | -2.0469 | -3.9844 |
79
+ | 0.0465 | 1.5707 | 1500 | 1.3534 | -9.625 | -11.25 | 0.5938 | 1.6016 | -1416.0 | -1280.0 | -2.4375 | -4.375 |
80
+ | 0.0541 | 1.6754 | 1600 | 1.1639 | -8.5 | -9.75 | 0.5938 | 1.2266 | -1264.0 | -1168.0 | -3.1719 | -4.9375 |
81
+ | 0.0826 | 1.7801 | 1700 | 1.2758 | -9.5 | -11.0 | 0.5977 | 1.4922 | -1392.0 | -1272.0 | -1.8203 | -3.6406 |
82
+ | 0.0429 | 1.8848 | 1800 | 1.1922 | -8.5625 | -10.0 | 0.5938 | 1.4375 | -1288.0 | -1176.0 | -1.3672 | -3.3281 |
83
+ | 0.0316 | 1.9895 | 1900 | 1.2850 | -9.25 | -10.9375 | 0.5938 | 1.6875 | -1384.0 | -1240.0 | -2.2969 | -4.2188 |
84
+ | 0.0026 | 2.0942 | 2000 | 1.4026 | -10.5625 | -12.375 | 0.5957 | 1.7891 | -1528.0 | -1376.0 | -1.0938 | -3.0156 |
85
+ | 0.0032 | 2.1990 | 2100 | 1.4889 | -11.0625 | -13.0 | 0.5918 | 1.9297 | -1592.0 | -1424.0 | -0.875 | -2.8281 |
86
+ | 0.0059 | 2.3037 | 2200 | 1.5006 | -11.25 | -13.1875 | 0.5996 | 1.9375 | -1608.0 | -1440.0 | -0.8281 | -2.75 |
87
+ | 0.008 | 2.4084 | 2300 | 1.5364 | -11.5 | -13.5 | 0.6016 | 1.9688 | -1640.0 | -1472.0 | -0.4961 | -2.4062 |
88
+ | 0.0059 | 2.5131 | 2400 | 1.5711 | -11.5 | -13.5 | 0.5898 | 2.0469 | -1640.0 | -1464.0 | -0.6562 | -2.6094 |
89
+ | 0.0063 | 2.6178 | 2500 | 1.5534 | -11.375 | -13.4375 | 0.5879 | 2.0469 | -1632.0 | -1456.0 | -0.6367 | -2.5781 |
90
+ | 0.0075 | 2.7225 | 2600 | 1.5770 | -11.5625 | -13.625 | 0.5859 | 2.0781 | -1648.0 | -1472.0 | -0.5859 | -2.5312 |
91
+ | 0.005 | 2.8272 | 2700 | 1.5729 | -11.4375 | -13.5 | 0.5840 | 2.0938 | -1640.0 | -1464.0 | -0.6562 | -2.6094 |
92
+ | 0.0065 | 2.9319 | 2800 | 1.5774 | -11.5 | -13.5625 | 0.5879 | 2.0938 | -1648.0 | -1464.0 | -0.6484 | -2.5938 |
93
 
94
 
95
  ### Framework versions
all_results.json CHANGED
@@ -1,9 +1,22 @@
1
  {
2
  "epoch": 3.0,
 
 
 
 
 
 
 
 
 
 
 
 
 
3
  "total_flos": 0.0,
4
- "train_loss": 0.116292925227765,
5
- "train_runtime": 12240.7967,
6
  "train_samples": 61119,
7
- "train_samples_per_second": 14.979,
8
- "train_steps_per_second": 0.234
9
  }
 
1
  {
2
  "epoch": 3.0,
3
+ "eval_logits/chosen": -7.59375,
4
+ "eval_logits/rejected": -5.6875,
5
+ "eval_logps/chosen": -1360.0,
6
+ "eval_logps/rejected": -1520.0,
7
+ "eval_loss": 1.5815156698226929,
8
+ "eval_rewards/accuracies": 0.59375,
9
+ "eval_rewards/chosen": -10.375,
10
+ "eval_rewards/margins": 1.890625,
11
+ "eval_rewards/rejected": -12.3125,
12
+ "eval_runtime": 46.5154,
13
+ "eval_samples": 2000,
14
+ "eval_samples_per_second": 42.996,
15
+ "eval_steps_per_second": 0.688,
16
  "total_flos": 0.0,
17
+ "train_loss": 0.11596642152112088,
18
+ "train_runtime": 12300.8596,
19
  "train_samples": 61119,
20
+ "train_samples_per_second": 14.906,
21
+ "train_steps_per_second": 0.233
22
  }
eval_results.json ADDED
@@ -0,0 +1,16 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 3.0,
3
+ "eval_logits/chosen": -7.59375,
4
+ "eval_logits/rejected": -5.6875,
5
+ "eval_logps/chosen": -1360.0,
6
+ "eval_logps/rejected": -1520.0,
7
+ "eval_loss": 1.5815156698226929,
8
+ "eval_rewards/accuracies": 0.59375,
9
+ "eval_rewards/chosen": -10.375,
10
+ "eval_rewards/margins": 1.890625,
11
+ "eval_rewards/rejected": -12.3125,
12
+ "eval_runtime": 46.5154,
13
+ "eval_samples": 2000,
14
+ "eval_samples_per_second": 42.996,
15
+ "eval_steps_per_second": 0.688
16
+ }
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:e9f75f97f619ab2737b7de16b33bad8dad3884712fd87e87b39df03a974f8892
3
  size 2159808696
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e410ca880e3028083ccd8f31ed3499dd8d016fa5427dfc154e4308fc3691ac3b
3
  size 2159808696
runs/Oct05_06-10-28_xe8545-a100-15/events.out.tfevents.1728114605.xe8545-a100-15.1041997.1 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:741563394edd849ac6aff3c891cb6ce6125f1b09f39c67e7e666769cec3d6fae
3
+ size 828
runs/Oct06_14-29-28_xe8545-a100-15/events.out.tfevents.1728218504.xe8545-a100-15.1082351.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c4312c20a55bc70aa1e9fd0d1b7473be844438633c594fe4186332d912cc25fc
3
+ size 225711
train_results.json CHANGED
@@ -1,9 +1,9 @@
1
  {
2
  "epoch": 3.0,
3
  "total_flos": 0.0,
4
- "train_loss": 0.116292925227765,
5
- "train_runtime": 12240.7967,
6
  "train_samples": 61119,
7
- "train_samples_per_second": 14.979,
8
- "train_steps_per_second": 0.234
9
  }
 
1
  {
2
  "epoch": 3.0,
3
  "total_flos": 0.0,
4
+ "train_loss": 0.11596642152112088,
5
+ "train_runtime": 12300.8596,
6
  "train_samples": 61119,
7
+ "train_samples_per_second": 14.906,
8
+ "train_steps_per_second": 0.233
9
  }
trainer_state.json CHANGED
The diff for this file is too large to render. See raw diff
 
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:a503dca67b11c4161459c676910cb819b0c83cdbb947f51abf87c2e8c4293ebf
3
  size 7672
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:85f6fdae88218d4c18b9a42663ab5b7360c3a620a2c79368cadba98069493a9a
3
  size 7672