ahmedabdelwahed
commited on
Commit
•
309e492
1
Parent(s):
1548700
Mojiz-DPO
Browse files- README.md +19 -49
- adapter_config.json +2 -2
- adapter_model.safetensors +1 -1
- training_args.bin +1 -1
README.md
CHANGED
@@ -17,14 +17,14 @@ should probably proofread and complete it, then remove this comment. -->
|
|
17 |
This model is a fine-tuned version of [ahmedabdelwahed/Mojiz-sft](https://huggingface.co/ahmedabdelwahed/Mojiz-sft) on the None dataset.
|
18 |
It achieves the following results on the evaluation set:
|
19 |
- Loss: 0.0000
|
20 |
-
- Rewards/chosen:
|
21 |
-
- Rewards/rejected: -
|
22 |
- Rewards/accuracies: 1.0
|
23 |
-
- Rewards/margins:
|
24 |
-
- Logps/rejected: -
|
25 |
-
- Logps/chosen: -
|
26 |
-
- Logits/rejected: -11.
|
27 |
-
- Logits/chosen: -12.
|
28 |
|
29 |
## Model description
|
30 |
|
@@ -43,59 +43,29 @@ More information needed
|
|
43 |
### Training hyperparameters
|
44 |
|
45 |
The following hyperparameters were used during training:
|
46 |
-
- learning_rate: 0.
|
47 |
- train_batch_size: 4
|
48 |
- eval_batch_size: 8
|
49 |
- seed: 42
|
50 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
51 |
- lr_scheduler_type: linear
|
52 |
- lr_scheduler_warmup_steps: 150
|
53 |
-
- training_steps:
|
54 |
|
55 |
### Training results
|
56 |
|
57 |
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
58 |
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
59 |
-
| 0.
|
60 |
-
| 0.
|
61 |
-
| 0.
|
62 |
-
| 0.0 | 1.63 | 400 | 0.0000 |
|
63 |
-
| 0.0 | 2.04 | 500 | 0.0000 |
|
64 |
-
| 0.0 | 2.45 | 600 | 0.0000 |
|
65 |
-
| 0.0 | 2.86 | 700 | 0.0000 | 16.
|
66 |
-
| 0.0 | 3.27 | 800 | 0.0000 |
|
67 |
-
| 0.0 | 3.67 | 900 | 0.0000 |
|
68 |
-
| 0.0 | 4.08 | 1000 | 0.0000 |
|
69 |
-
| 0.0 | 4.49 | 1100 | 0.0000 | 21.2289 | -8.8084 | 1.0 | 30.0373 | -88.6562 | -283.8553 | -11.7555 | -12.8438 |
|
70 |
-
| 0.0 | 4.9 | 1200 | 0.0000 | 21.2556 | -8.8423 | 1.0 | 30.0979 | -88.7241 | -283.8019 | -11.7538 | -12.8404 |
|
71 |
-
| 0.0 | 5.31 | 1300 | 0.0000 | 21.3255 | -9.0397 | 1.0 | 30.3653 | -89.1189 | -283.6620 | -11.7587 | -12.8441 |
|
72 |
-
| 0.0 | 5.71 | 1400 | 0.0000 | 21.3586 | -9.0546 | 1.0 | 30.4132 | -89.1486 | -283.5958 | -11.7606 | -12.8458 |
|
73 |
-
| 0.0 | 6.12 | 1500 | 0.0000 | 21.3779 | -9.0803 | 1.0 | 30.4582 | -89.2000 | -283.5573 | -11.7600 | -12.8442 |
|
74 |
-
| 0.0001 | 6.53 | 1600 | 0.0000 | 21.4632 | -8.9069 | 1.0 | 30.3700 | -88.8531 | -283.3867 | -11.7795 | -12.8770 |
|
75 |
-
| 0.0 | 6.94 | 1700 | 0.0000 | 21.4992 | -8.8450 | 1.0 | 30.3441 | -88.7293 | -283.3147 | -11.7894 | -12.8933 |
|
76 |
-
| 0.0 | 7.35 | 1800 | 0.0000 | 21.5109 | -8.8690 | 1.0 | 30.3799 | -88.7774 | -283.2912 | -11.7896 | -12.8931 |
|
77 |
-
| 0.0 | 7.76 | 1900 | 0.0000 | 21.5322 | -8.9067 | 1.0 | 30.4390 | -88.8529 | -283.2486 | -11.7905 | -12.8917 |
|
78 |
-
| 0.0 | 8.16 | 2000 | 0.0000 | 21.5638 | -8.9260 | 1.0 | 30.4898 | -88.8915 | -283.1855 | -11.7910 | -12.8912 |
|
79 |
-
| 0.0 | 8.57 | 2100 | 0.0000 | 21.5688 | -8.9323 | 1.0 | 30.5011 | -88.9041 | -283.1755 | -11.7910 | -12.8909 |
|
80 |
-
| 0.0 | 8.98 | 2200 | 0.0000 | 21.5834 | -8.9761 | 1.0 | 30.5596 | -88.9917 | -283.1462 | -11.7917 | -12.8894 |
|
81 |
-
| 0.0 | 9.39 | 2300 | 0.0000 | 21.7552 | -9.1179 | 1.0 | 30.8731 | -89.2752 | -282.8026 | -11.7852 | -12.8848 |
|
82 |
-
| 0.0 | 9.8 | 2400 | 0.0000 | 21.7651 | -9.1413 | 1.0 | 30.9064 | -89.3221 | -282.7829 | -11.7850 | -12.8839 |
|
83 |
-
| 0.0 | 10.2 | 2500 | 0.0000 | 21.7780 | -9.1572 | 1.0 | 30.9351 | -89.3538 | -282.7572 | -11.7861 | -12.8846 |
|
84 |
-
| 0.0 | 10.61 | 2600 | 0.0000 | 21.8023 | -9.1708 | 1.0 | 30.9731 | -89.3810 | -282.7084 | -11.7872 | -12.8865 |
|
85 |
-
| 0.0 | 11.02 | 2700 | 0.0000 | 21.8155 | -9.1793 | 1.0 | 30.9948 | -89.3980 | -282.6820 | -11.7879 | -12.8875 |
|
86 |
-
| 0.0 | 11.43 | 2800 | 0.0000 | 21.9589 | -9.1294 | 1.0 | 31.0883 | -89.2983 | -282.3953 | -11.8192 | -12.9290 |
|
87 |
-
| 0.0 | 11.84 | 2900 | 0.0000 | 21.9736 | -9.1440 | 1.0 | 31.1177 | -89.3275 | -282.3658 | -11.8210 | -12.9310 |
|
88 |
-
| 0.0 | 12.24 | 3000 | 0.0000 | 21.9922 | -9.1667 | 1.0 | 31.1589 | -89.3729 | -282.3287 | -11.8189 | -12.9279 |
|
89 |
-
| 0.0 | 12.65 | 3100 | 0.0000 | 21.9939 | -9.1697 | 1.0 | 31.1636 | -89.3788 | -282.3252 | -11.8189 | -12.9278 |
|
90 |
-
| 0.0 | 13.06 | 3200 | 0.0000 | 22.0187 | -9.2172 | 1.0 | 31.2360 | -89.4739 | -282.2755 | -11.8188 | -12.9264 |
|
91 |
-
| 0.0 | 13.47 | 3300 | 0.0000 | 22.0256 | -9.2325 | 1.0 | 31.2581 | -89.5044 | -282.2618 | -11.8194 | -12.9272 |
|
92 |
-
| 0.0 | 13.88 | 3400 | 0.0000 | 22.0341 | -9.2495 | 1.0 | 31.2836 | -89.5385 | -282.2448 | -11.8192 | -12.9267 |
|
93 |
-
| 0.0 | 14.29 | 3500 | 0.0000 | 22.0343 | -9.2506 | 1.0 | 31.2849 | -89.5407 | -282.2444 | -11.8192 | -12.9267 |
|
94 |
-
| 0.0 | 14.69 | 3600 | 0.0000 | 22.2061 | -9.2685 | 1.0 | 31.4746 | -89.5764 | -281.9009 | -11.8502 | -12.9633 |
|
95 |
-
| 0.0 | 15.1 | 3700 | 0.0000 | 22.2073 | -9.2691 | 1.0 | 31.4763 | -89.5776 | -281.8986 | -11.8503 | -12.9634 |
|
96 |
-
| 0.0 | 15.51 | 3800 | 0.0000 | 22.2107 | -9.2755 | 1.0 | 31.4862 | -89.5905 | -281.8916 | -11.8501 | -12.9628 |
|
97 |
-
| 0.0 | 15.92 | 3900 | 0.0000 | 22.2207 | -9.2795 | 1.0 | 31.5003 | -89.5985 | -281.8716 | -11.8493 | -12.9618 |
|
98 |
-
| 0.0 | 16.33 | 4000 | 0.0000 | 22.2209 | -9.2802 | 1.0 | 31.5011 | -89.5998 | -281.8713 | -11.8496 | -12.9621 |
|
99 |
|
100 |
|
101 |
### Framework versions
|
|
|
17 |
This model is a fine-tuned version of [ahmedabdelwahed/Mojiz-sft](https://huggingface.co/ahmedabdelwahed/Mojiz-sft) on the None dataset.
|
18 |
It achieves the following results on the evaluation set:
|
19 |
- Loss: 0.0000
|
20 |
+
- Rewards/chosen: 16.5213
|
21 |
+
- Rewards/rejected: -7.9440
|
22 |
- Rewards/accuracies: 1.0
|
23 |
+
- Rewards/margins: 24.4653
|
24 |
+
- Logps/rejected: -86.9274
|
25 |
+
- Logps/chosen: -293.2704
|
26 |
+
- Logits/rejected: -11.2915
|
27 |
+
- Logits/chosen: -12.2516
|
28 |
|
29 |
## Model description
|
30 |
|
|
|
43 |
### Training hyperparameters
|
44 |
|
45 |
The following hyperparameters were used during training:
|
46 |
+
- learning_rate: 0.0001
|
47 |
- train_batch_size: 4
|
48 |
- eval_batch_size: 8
|
49 |
- seed: 42
|
50 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
51 |
- lr_scheduler_type: linear
|
52 |
- lr_scheduler_warmup_steps: 150
|
53 |
+
- training_steps: 1000
|
54 |
|
55 |
### Training results
|
56 |
|
57 |
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
58 |
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
59 |
+
| 0.0017 | 0.41 | 100 | 0.0000 | 9.9359 | -3.7597 | 1.0 | 13.6956 | -78.5589 | -306.4413 | -11.4127 | -12.4541 |
|
60 |
+
| 0.0002 | 0.82 | 200 | 0.0000 | 14.1969 | -5.8004 | 1.0 | 19.9973 | -82.6403 | -297.9192 | -11.3000 | -12.2682 |
|
61 |
+
| 0.0037 | 1.22 | 300 | 0.0000 | 14.8615 | -6.7633 | 1.0 | 21.6248 | -84.5661 | -296.5901 | -11.2673 | -12.2269 |
|
62 |
+
| 0.0 | 1.63 | 400 | 0.0000 | 15.4935 | -7.6471 | 1.0 | 23.1406 | -86.3337 | -295.3261 | -11.2271 | -12.1591 |
|
63 |
+
| 0.0 | 2.04 | 500 | 0.0000 | 15.8634 | -7.8871 | 1.0 | 23.7505 | -86.8136 | -294.5863 | -11.2316 | -12.1672 |
|
64 |
+
| 0.0 | 2.45 | 600 | 0.0000 | 16.1624 | -7.8756 | 1.0 | 24.0380 | -86.7906 | -293.9882 | -11.2578 | -12.2052 |
|
65 |
+
| 0.0 | 2.86 | 700 | 0.0000 | 16.1247 | -8.2229 | 1.0 | 24.3476 | -87.4853 | -294.0637 | -11.2414 | -12.1705 |
|
66 |
+
| 0.0 | 3.27 | 800 | 0.0000 | 16.4219 | -7.9771 | 1.0 | 24.3989 | -86.9936 | -293.4693 | -11.2814 | -12.2344 |
|
67 |
+
| 0.0 | 3.67 | 900 | 0.0000 | 16.4248 | -7.9873 | 1.0 | 24.4122 | -87.0141 | -293.4634 | -11.2812 | -12.2342 |
|
68 |
+
| 0.0 | 4.08 | 1000 | 0.0000 | 16.5213 | -7.9440 | 1.0 | 24.4653 | -86.9274 | -293.2704 | -11.2915 | -12.2516 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
69 |
|
70 |
|
71 |
### Framework versions
|
adapter_config.json
CHANGED
@@ -19,8 +19,8 @@
|
|
19 |
"rank_pattern": {},
|
20 |
"revision": null,
|
21 |
"target_modules": [
|
22 |
-
"
|
23 |
-
"
|
24 |
],
|
25 |
"task_type": "SEQ_2_SEQ_LM"
|
26 |
}
|
|
|
19 |
"rank_pattern": {},
|
20 |
"revision": null,
|
21 |
"target_modules": [
|
22 |
+
"v",
|
23 |
+
"q"
|
24 |
],
|
25 |
"task_type": "SEQ_2_SEQ_LM"
|
26 |
}
|
adapter_model.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 7098016
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:79535a0e2a032c784c61d90fcbb50b82a79fd7b9df89d4272ab43984c2421865
|
3 |
size 7098016
|
training_args.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 4219
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:304e26e2abb03ba7eaa0fadfbe1f573864d1c879025abe293c6ffe715100b200
|
3 |
size 4219
|