Model save
Browse files- README.md +11 -11
- adapter_model.safetensors +1 -1
- all_results.json +16 -16
- eval_results.json +12 -12
- runs/Jan21_22-53-06_uclaml04.cs.ucla.edu/events.out.tfevents.1705906443.uclaml04.cs.ucla.edu.843444.0 +2 -2
- runs/Jan21_22-53-06_uclaml04.cs.ucla.edu/events.out.tfevents.1705932667.uclaml04.cs.ucla.edu.843444.1 +3 -0
- train_results.json +4 -4
- trainer_state.json +108 -108
README.md
CHANGED
@@ -15,15 +15,15 @@ should probably proofread and complete it, then remove this comment. -->
|
|
15 |
|
16 |
This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full) on the None dataset.
|
17 |
It achieves the following results on the evaluation set:
|
18 |
-
- Loss: 0.
|
19 |
-
- Rewards/chosen: 0.
|
20 |
-
- Rewards/rejected: -0.
|
21 |
-
- Rewards/accuracies: 0.
|
22 |
-
- Rewards/margins: 0.
|
23 |
-
- Logps/rejected: -
|
24 |
-
- Logps/chosen: -283.
|
25 |
-
- Logits/rejected: -2.
|
26 |
-
- Logits/chosen: -2.
|
27 |
- Use Label: 0.0
|
28 |
- Pred Label: 0.0
|
29 |
|
@@ -44,7 +44,7 @@ More information needed
|
|
44 |
### Training hyperparameters
|
45 |
|
46 |
The following hyperparameters were used during training:
|
47 |
-
- learning_rate:
|
48 |
- train_batch_size: 4
|
49 |
- eval_batch_size: 4
|
50 |
- seed: 42
|
@@ -62,7 +62,7 @@ The following hyperparameters were used during training:
|
|
62 |
|
63 |
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen | Use Label | Pred Label |
|
64 |
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|:---------:|:----------:|
|
65 |
-
| 0.
|
66 |
|
67 |
|
68 |
### Framework versions
|
|
|
15 |
|
16 |
This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full) on the None dataset.
|
17 |
It achieves the following results on the evaluation set:
|
18 |
+
- Loss: 0.5460
|
19 |
+
- Rewards/chosen: 0.0878
|
20 |
+
- Rewards/rejected: -0.6697
|
21 |
+
- Rewards/accuracies: 0.7140
|
22 |
+
- Rewards/margins: 0.7575
|
23 |
+
- Logps/rejected: -265.9970
|
24 |
+
- Logps/chosen: -283.3795
|
25 |
+
- Logits/rejected: -2.8187
|
26 |
+
- Logits/chosen: -2.8274
|
27 |
- Use Label: 0.0
|
28 |
- Pred Label: 0.0
|
29 |
|
|
|
44 |
### Training hyperparameters
|
45 |
|
46 |
The following hyperparameters were used during training:
|
47 |
+
- learning_rate: 1e-05
|
48 |
- train_batch_size: 4
|
49 |
- eval_batch_size: 4
|
50 |
- seed: 42
|
|
|
62 |
|
63 |
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen | Use Label | Pred Label |
|
64 |
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|:---------:|:----------:|
|
65 |
+
| 0.545 | 1.0 | 955 | 0.5460 | 0.0878 | -0.6697 | 0.7140 | 0.7575 | -265.9970 | -283.3795 | -2.8187 | -2.8274 | 0.0 | 0.0 |
|
66 |
|
67 |
|
68 |
### Framework versions
|
adapter_model.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 218138576
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:ee44cce906c7a6f93e762633279c05b8cb98e432f3c9ac83bd6f6213cf55c20a
|
3 |
size 218138576
|
all_results.json
CHANGED
@@ -1,23 +1,23 @@
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
-
"eval_logits/chosen": -2.
|
4 |
-
"eval_logits/rejected": -2.
|
5 |
-
"eval_logps/chosen": -283.
|
6 |
-
"eval_logps/rejected": -
|
7 |
-
"eval_loss": 0.
|
8 |
"eval_pred_label": 0.0,
|
9 |
-
"eval_rewards/accuracies": 0.
|
10 |
-
"eval_rewards/chosen": 0.
|
11 |
-
"eval_rewards/margins": 0.
|
12 |
-
"eval_rewards/rejected": -0.
|
13 |
-
"eval_runtime":
|
14 |
"eval_samples": 2000,
|
15 |
-
"eval_samples_per_second": 4.
|
16 |
-
"eval_steps_per_second": 0.
|
17 |
"eval_use_label": 0.0,
|
18 |
-
"train_loss": 0.
|
19 |
-
"train_runtime":
|
20 |
"train_samples": 61135,
|
21 |
-
"train_samples_per_second": 2.
|
22 |
-
"train_steps_per_second": 0.
|
23 |
}
|
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
+
"eval_logits/chosen": -2.827404737472534,
|
4 |
+
"eval_logits/rejected": -2.818655014038086,
|
5 |
+
"eval_logps/chosen": -283.37945556640625,
|
6 |
+
"eval_logps/rejected": -265.9969787597656,
|
7 |
+
"eval_loss": 0.5459502935409546,
|
8 |
"eval_pred_label": 0.0,
|
9 |
+
"eval_rewards/accuracies": 0.7139999866485596,
|
10 |
+
"eval_rewards/chosen": 0.08778975158929825,
|
11 |
+
"eval_rewards/margins": 0.7575166821479797,
|
12 |
+
"eval_rewards/rejected": -0.6697269678115845,
|
13 |
+
"eval_runtime": 478.4263,
|
14 |
"eval_samples": 2000,
|
15 |
+
"eval_samples_per_second": 4.18,
|
16 |
+
"eval_steps_per_second": 0.261,
|
17 |
"eval_use_label": 0.0,
|
18 |
+
"train_loss": 0.5628191218950361,
|
19 |
+
"train_runtime": 25746.1298,
|
20 |
"train_samples": 61135,
|
21 |
+
"train_samples_per_second": 2.375,
|
22 |
+
"train_steps_per_second": 0.037
|
23 |
}
|
eval_results.json
CHANGED
@@ -1,18 +1,18 @@
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
-
"eval_logits/chosen": -2.
|
4 |
-
"eval_logits/rejected": -2.
|
5 |
-
"eval_logps/chosen": -283.
|
6 |
-
"eval_logps/rejected": -
|
7 |
-
"eval_loss": 0.
|
8 |
"eval_pred_label": 0.0,
|
9 |
-
"eval_rewards/accuracies": 0.
|
10 |
-
"eval_rewards/chosen": 0.
|
11 |
-
"eval_rewards/margins": 0.
|
12 |
-
"eval_rewards/rejected": -0.
|
13 |
-
"eval_runtime":
|
14 |
"eval_samples": 2000,
|
15 |
-
"eval_samples_per_second": 4.
|
16 |
-
"eval_steps_per_second": 0.
|
17 |
"eval_use_label": 0.0
|
18 |
}
|
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
+
"eval_logits/chosen": -2.827404737472534,
|
4 |
+
"eval_logits/rejected": -2.818655014038086,
|
5 |
+
"eval_logps/chosen": -283.37945556640625,
|
6 |
+
"eval_logps/rejected": -265.9969787597656,
|
7 |
+
"eval_loss": 0.5459502935409546,
|
8 |
"eval_pred_label": 0.0,
|
9 |
+
"eval_rewards/accuracies": 0.7139999866485596,
|
10 |
+
"eval_rewards/chosen": 0.08778975158929825,
|
11 |
+
"eval_rewards/margins": 0.7575166821479797,
|
12 |
+
"eval_rewards/rejected": -0.6697269678115845,
|
13 |
+
"eval_runtime": 478.4263,
|
14 |
"eval_samples": 2000,
|
15 |
+
"eval_samples_per_second": 4.18,
|
16 |
+
"eval_steps_per_second": 0.261,
|
17 |
"eval_use_label": 0.0
|
18 |
}
|
runs/Jan21_22-53-06_uclaml04.cs.ucla.edu/events.out.tfevents.1705906443.uclaml04.cs.ucla.edu.843444.0
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 sha256:70cef13045c99f0b2e0174d81d676e33577c08d8132d6a7dd5fe8a82051a671a
|
3 |
+
size 13038
|
runs/Jan21_22-53-06_uclaml04.cs.ucla.edu/events.out.tfevents.1705932667.uclaml04.cs.ucla.edu.843444.1
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:7e4a8aa0de3f53d33d6f9f03d407834c788dfffa1cfc429c328ae7af5f7003e3
|
3 |
+
size 935
|
train_results.json
CHANGED
@@ -1,8 +1,8 @@
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
-
"train_loss": 0.
|
4 |
-
"train_runtime":
|
5 |
"train_samples": 61135,
|
6 |
-
"train_samples_per_second": 2.
|
7 |
-
"train_steps_per_second": 0.
|
8 |
}
|
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
+
"train_loss": 0.5628191218950361,
|
4 |
+
"train_runtime": 25746.1298,
|
5 |
"train_samples": 61135,
|
6 |
+
"train_samples_per_second": 2.375,
|
7 |
+
"train_steps_per_second": 0.037
|
8 |
}
|
trainer_state.json
CHANGED
@@ -10,7 +10,7 @@
|
|
10 |
"log_history": [
|
11 |
{
|
12 |
"epoch": 0.0,
|
13 |
-
"learning_rate":
|
14 |
"logits/chosen": -2.980285167694092,
|
15 |
"logits/rejected": -2.87275767326355,
|
16 |
"logps/chosen": -313.4390563964844,
|
@@ -26,163 +26,163 @@
|
|
26 |
},
|
27 |
{
|
28 |
"epoch": 0.1,
|
29 |
-
"learning_rate":
|
30 |
-
"logits/chosen": -2.
|
31 |
-
"logits/rejected": -2.
|
32 |
-
"logps/chosen": -285.
|
33 |
-
"logps/rejected": -271.
|
34 |
-
"loss": 0.
|
35 |
"pred_label": 0.0,
|
36 |
-
"rewards/accuracies": 0.
|
37 |
-
"rewards/chosen": 0.
|
38 |
-
"rewards/margins": 0.
|
39 |
-
"rewards/rejected": -0.
|
40 |
"step": 100,
|
41 |
"use_label": 0.0
|
42 |
},
|
43 |
{
|
44 |
"epoch": 0.21,
|
45 |
-
"learning_rate":
|
46 |
-
"logits/chosen": -2.
|
47 |
-
"logits/rejected": -2.
|
48 |
-
"logps/chosen": -278.
|
49 |
-
"logps/rejected": -
|
50 |
-
"loss": 0.
|
51 |
"pred_label": 0.0,
|
52 |
-
"rewards/accuracies": 0.
|
53 |
-
"rewards/chosen": 0.
|
54 |
-
"rewards/margins": 0.
|
55 |
-
"rewards/rejected": -0.
|
56 |
"step": 200,
|
57 |
"use_label": 0.0
|
58 |
},
|
59 |
{
|
60 |
"epoch": 0.31,
|
61 |
-
"learning_rate":
|
62 |
-
"logits/chosen": -2.
|
63 |
-
"logits/rejected": -2.
|
64 |
-
"logps/chosen": -
|
65 |
-
"logps/rejected": -
|
66 |
-
"loss": 0.
|
67 |
"pred_label": 0.0,
|
68 |
-
"rewards/accuracies": 0.
|
69 |
-
"rewards/chosen": 0.
|
70 |
-
"rewards/margins": 0.
|
71 |
-
"rewards/rejected": -0.
|
72 |
"step": 300,
|
73 |
"use_label": 0.0
|
74 |
},
|
75 |
{
|
76 |
"epoch": 0.42,
|
77 |
-
"learning_rate":
|
78 |
-
"logits/chosen": -2.
|
79 |
-
"logits/rejected": -2.
|
80 |
-
"logps/chosen": -284.
|
81 |
-
"logps/rejected": -
|
82 |
-
"loss": 0.
|
83 |
"pred_label": 0.0,
|
84 |
-
"rewards/accuracies": 0.
|
85 |
-
"rewards/chosen": 0.
|
86 |
-
"rewards/margins": 0.
|
87 |
-
"rewards/rejected": -0.
|
88 |
"step": 400,
|
89 |
"use_label": 0.0
|
90 |
},
|
91 |
{
|
92 |
"epoch": 0.52,
|
93 |
-
"learning_rate":
|
94 |
-
"logits/chosen": -2.
|
95 |
-
"logits/rejected": -2.
|
96 |
-
"logps/chosen": -
|
97 |
-
"logps/rejected": -
|
98 |
-
"loss": 0.
|
99 |
"pred_label": 0.0,
|
100 |
-
"rewards/accuracies": 0.
|
101 |
-
"rewards/chosen":
|
102 |
-
"rewards/margins": 0.
|
103 |
-
"rewards/rejected": -0.
|
104 |
"step": 500,
|
105 |
"use_label": 0.0
|
106 |
},
|
107 |
{
|
108 |
"epoch": 0.63,
|
109 |
-
"learning_rate":
|
110 |
-
"logits/chosen": -2.
|
111 |
-
"logits/rejected": -2.
|
112 |
-
"logps/chosen": -284.
|
113 |
-
"logps/rejected": -
|
114 |
-
"loss": 0.
|
115 |
"pred_label": 0.0,
|
116 |
-
"rewards/accuracies": 0.
|
117 |
-
"rewards/chosen": 0.
|
118 |
-
"rewards/margins": 0.
|
119 |
-
"rewards/rejected": -0.
|
120 |
"step": 600,
|
121 |
"use_label": 0.0
|
122 |
},
|
123 |
{
|
124 |
"epoch": 0.73,
|
125 |
-
"learning_rate":
|
126 |
-
"logits/chosen": -2.
|
127 |
-
"logits/rejected": -2.
|
128 |
-
"logps/chosen": -280.
|
129 |
-
"logps/rejected": -
|
130 |
-
"loss": 0.
|
131 |
"pred_label": 0.0,
|
132 |
-
"rewards/accuracies": 0.
|
133 |
-
"rewards/chosen": 0.
|
134 |
-
"rewards/margins": 0.
|
135 |
-
"rewards/rejected": -0.
|
136 |
"step": 700,
|
137 |
"use_label": 0.0
|
138 |
},
|
139 |
{
|
140 |
"epoch": 0.84,
|
141 |
-
"learning_rate":
|
142 |
-
"logits/chosen": -2.
|
143 |
-
"logits/rejected": -2.
|
144 |
-
"logps/chosen": -287.
|
145 |
-
"logps/rejected": -
|
146 |
-
"loss": 0.
|
147 |
"pred_label": 0.0,
|
148 |
-
"rewards/accuracies": 0.
|
149 |
-
"rewards/chosen": 0.
|
150 |
-
"rewards/margins": 0.
|
151 |
-
"rewards/rejected": -0.
|
152 |
"step": 800,
|
153 |
"use_label": 0.0
|
154 |
},
|
155 |
{
|
156 |
"epoch": 0.94,
|
157 |
-
"learning_rate":
|
158 |
-
"logits/chosen": -2.
|
159 |
-
"logits/rejected": -2.
|
160 |
-
"logps/chosen": -276.
|
161 |
-
"logps/rejected": -
|
162 |
-
"loss": 0.
|
163 |
"pred_label": 0.0,
|
164 |
-
"rewards/accuracies": 0.
|
165 |
-
"rewards/chosen": 0.
|
166 |
-
"rewards/margins": 0.
|
167 |
-
"rewards/rejected": -0.
|
168 |
"step": 900,
|
169 |
"use_label": 0.0
|
170 |
},
|
171 |
{
|
172 |
"epoch": 1.0,
|
173 |
-
"eval_logits/chosen": -2.
|
174 |
-
"eval_logits/rejected": -2.
|
175 |
-
"eval_logps/chosen": -283.
|
176 |
-
"eval_logps/rejected": -
|
177 |
-
"eval_loss": 0.
|
178 |
"eval_pred_label": 0.0,
|
179 |
-
"eval_rewards/accuracies": 0.
|
180 |
-
"eval_rewards/chosen": 0.
|
181 |
-
"eval_rewards/margins": 0.
|
182 |
-
"eval_rewards/rejected": -0.
|
183 |
-
"eval_runtime":
|
184 |
-
"eval_samples_per_second": 4.
|
185 |
-
"eval_steps_per_second": 0.
|
186 |
"eval_use_label": 0.0,
|
187 |
"step": 955
|
188 |
},
|
@@ -190,16 +190,16 @@
|
|
190 |
"epoch": 1.0,
|
191 |
"step": 955,
|
192 |
"total_flos": 0.0,
|
193 |
-
"train_loss": 0.
|
194 |
-
"train_runtime":
|
195 |
-
"train_samples_per_second": 2.
|
196 |
-
"train_steps_per_second": 0.
|
197 |
}
|
198 |
],
|
199 |
"logging_steps": 100,
|
200 |
"max_steps": 955,
|
201 |
"num_train_epochs": 1,
|
202 |
-
"save_steps":
|
203 |
"total_flos": 0.0,
|
204 |
"trial_name": null,
|
205 |
"trial_params": null
|
|
|
10 |
"log_history": [
|
11 |
{
|
12 |
"epoch": 0.0,
|
13 |
+
"learning_rate": 1.0416666666666667e-07,
|
14 |
"logits/chosen": -2.980285167694092,
|
15 |
"logits/rejected": -2.87275767326355,
|
16 |
"logps/chosen": -313.4390563964844,
|
|
|
26 |
},
|
27 |
{
|
28 |
"epoch": 0.1,
|
29 |
+
"learning_rate": 9.953434225844005e-06,
|
30 |
+
"logits/chosen": -2.8180909156799316,
|
31 |
+
"logits/rejected": -2.8273613452911377,
|
32 |
+
"logps/chosen": -285.13623046875,
|
33 |
+
"logps/rejected": -271.66839599609375,
|
34 |
+
"loss": 0.6624,
|
35 |
"pred_label": 0.0,
|
36 |
+
"rewards/accuracies": 0.6041666865348816,
|
37 |
+
"rewards/chosen": 0.014434419572353363,
|
38 |
+
"rewards/margins": 0.08631344139575958,
|
39 |
+
"rewards/rejected": -0.07187902927398682,
|
40 |
"step": 100,
|
41 |
"use_label": 0.0
|
42 |
},
|
43 |
{
|
44 |
"epoch": 0.21,
|
45 |
+
"learning_rate": 8.789289871944122e-06,
|
46 |
+
"logits/chosen": -2.8230364322662354,
|
47 |
+
"logits/rejected": -2.8086395263671875,
|
48 |
+
"logps/chosen": -278.2524108886719,
|
49 |
+
"logps/rejected": -263.9921569824219,
|
50 |
+
"loss": 0.5868,
|
51 |
"pred_label": 0.0,
|
52 |
+
"rewards/accuracies": 0.6949999928474426,
|
53 |
+
"rewards/chosen": 0.05295524746179581,
|
54 |
+
"rewards/margins": 0.39829620718955994,
|
55 |
+
"rewards/rejected": -0.3453409671783447,
|
56 |
"step": 200,
|
57 |
"use_label": 0.0
|
58 |
},
|
59 |
{
|
60 |
"epoch": 0.31,
|
61 |
+
"learning_rate": 7.625145518044238e-06,
|
62 |
+
"logits/chosen": -2.803905725479126,
|
63 |
+
"logits/rejected": -2.802032232284546,
|
64 |
+
"logps/chosen": -284.01385498046875,
|
65 |
+
"logps/rejected": -259.5546569824219,
|
66 |
+
"loss": 0.562,
|
67 |
"pred_label": 0.0,
|
68 |
+
"rewards/accuracies": 0.7056249976158142,
|
69 |
+
"rewards/chosen": 0.005377008114010096,
|
70 |
+
"rewards/margins": 0.5735920667648315,
|
71 |
+
"rewards/rejected": -0.5682151317596436,
|
72 |
"step": 300,
|
73 |
"use_label": 0.0
|
74 |
},
|
75 |
{
|
76 |
"epoch": 0.42,
|
77 |
+
"learning_rate": 6.461001164144355e-06,
|
78 |
+
"logits/chosen": -2.8141045570373535,
|
79 |
+
"logits/rejected": -2.7911813259124756,
|
80 |
+
"logps/chosen": -284.3139953613281,
|
81 |
+
"logps/rejected": -269.4837951660156,
|
82 |
+
"loss": 0.5527,
|
83 |
"pred_label": 0.0,
|
84 |
+
"rewards/accuracies": 0.7250000238418579,
|
85 |
+
"rewards/chosen": 0.011648621410131454,
|
86 |
+
"rewards/margins": 0.6873368620872498,
|
87 |
+
"rewards/rejected": -0.6756882667541504,
|
88 |
"step": 400,
|
89 |
"use_label": 0.0
|
90 |
},
|
91 |
{
|
92 |
"epoch": 0.52,
|
93 |
+
"learning_rate": 5.2968568102444705e-06,
|
94 |
+
"logits/chosen": -2.7915823459625244,
|
95 |
+
"logits/rejected": -2.776299476623535,
|
96 |
+
"logps/chosen": -269.73016357421875,
|
97 |
+
"logps/rejected": -257.2498474121094,
|
98 |
+
"loss": 0.5556,
|
99 |
"pred_label": 0.0,
|
100 |
+
"rewards/accuracies": 0.6862499713897705,
|
101 |
+
"rewards/chosen": 0.03227977454662323,
|
102 |
+
"rewards/margins": 0.5588020086288452,
|
103 |
+
"rewards/rejected": -0.5265222191810608,
|
104 |
"step": 500,
|
105 |
"use_label": 0.0
|
106 |
},
|
107 |
{
|
108 |
"epoch": 0.63,
|
109 |
+
"learning_rate": 4.132712456344587e-06,
|
110 |
+
"logits/chosen": -2.8199498653411865,
|
111 |
+
"logits/rejected": -2.8022332191467285,
|
112 |
+
"logps/chosen": -284.0947265625,
|
113 |
+
"logps/rejected": -271.3774108886719,
|
114 |
+
"loss": 0.5422,
|
115 |
"pred_label": 0.0,
|
116 |
+
"rewards/accuracies": 0.7143750190734863,
|
117 |
+
"rewards/chosen": 0.11361943930387497,
|
118 |
+
"rewards/margins": 0.7251341938972473,
|
119 |
+
"rewards/rejected": -0.6115147471427917,
|
120 |
"step": 600,
|
121 |
"use_label": 0.0
|
122 |
},
|
123 |
{
|
124 |
"epoch": 0.73,
|
125 |
+
"learning_rate": 2.9685681024447033e-06,
|
126 |
+
"logits/chosen": -2.8110527992248535,
|
127 |
+
"logits/rejected": -2.788975477218628,
|
128 |
+
"logps/chosen": -280.3959045410156,
|
129 |
+
"logps/rejected": -254.49673461914062,
|
130 |
+
"loss": 0.5404,
|
131 |
"pred_label": 0.0,
|
132 |
+
"rewards/accuracies": 0.7212499976158142,
|
133 |
+
"rewards/chosen": 0.11637673527002335,
|
134 |
+
"rewards/margins": 0.6999369263648987,
|
135 |
+
"rewards/rejected": -0.5835601687431335,
|
136 |
"step": 700,
|
137 |
"use_label": 0.0
|
138 |
},
|
139 |
{
|
140 |
"epoch": 0.84,
|
141 |
+
"learning_rate": 1.8044237485448196e-06,
|
142 |
+
"logits/chosen": -2.8146722316741943,
|
143 |
+
"logits/rejected": -2.812129020690918,
|
144 |
+
"logps/chosen": -287.4331359863281,
|
145 |
+
"logps/rejected": -267.59161376953125,
|
146 |
+
"loss": 0.5343,
|
147 |
"pred_label": 0.0,
|
148 |
+
"rewards/accuracies": 0.721875011920929,
|
149 |
+
"rewards/chosen": 0.13239255547523499,
|
150 |
+
"rewards/margins": 0.7503484487533569,
|
151 |
+
"rewards/rejected": -0.6179558634757996,
|
152 |
"step": 800,
|
153 |
"use_label": 0.0
|
154 |
},
|
155 |
{
|
156 |
"epoch": 0.94,
|
157 |
+
"learning_rate": 6.402793946449361e-07,
|
158 |
+
"logits/chosen": -2.8043179512023926,
|
159 |
+
"logits/rejected": -2.8079681396484375,
|
160 |
+
"logps/chosen": -276.8100891113281,
|
161 |
+
"logps/rejected": -262.690673828125,
|
162 |
+
"loss": 0.545,
|
163 |
"pred_label": 0.0,
|
164 |
+
"rewards/accuracies": 0.7193750143051147,
|
165 |
+
"rewards/chosen": 0.06981150805950165,
|
166 |
+
"rewards/margins": 0.7052963972091675,
|
167 |
+
"rewards/rejected": -0.6354848742485046,
|
168 |
"step": 900,
|
169 |
"use_label": 0.0
|
170 |
},
|
171 |
{
|
172 |
"epoch": 1.0,
|
173 |
+
"eval_logits/chosen": -2.827404737472534,
|
174 |
+
"eval_logits/rejected": -2.818655014038086,
|
175 |
+
"eval_logps/chosen": -283.37945556640625,
|
176 |
+
"eval_logps/rejected": -265.9969787597656,
|
177 |
+
"eval_loss": 0.5459502935409546,
|
178 |
"eval_pred_label": 0.0,
|
179 |
+
"eval_rewards/accuracies": 0.7139999866485596,
|
180 |
+
"eval_rewards/chosen": 0.08778975158929825,
|
181 |
+
"eval_rewards/margins": 0.7575166821479797,
|
182 |
+
"eval_rewards/rejected": -0.6697269678115845,
|
183 |
+
"eval_runtime": 479.5351,
|
184 |
+
"eval_samples_per_second": 4.171,
|
185 |
+
"eval_steps_per_second": 0.261,
|
186 |
"eval_use_label": 0.0,
|
187 |
"step": 955
|
188 |
},
|
|
|
190 |
"epoch": 1.0,
|
191 |
"step": 955,
|
192 |
"total_flos": 0.0,
|
193 |
+
"train_loss": 0.5628191218950361,
|
194 |
+
"train_runtime": 25746.1298,
|
195 |
+
"train_samples_per_second": 2.375,
|
196 |
+
"train_steps_per_second": 0.037
|
197 |
}
|
198 |
],
|
199 |
"logging_steps": 100,
|
200 |
"max_steps": 955,
|
201 |
"num_train_epochs": 1,
|
202 |
+
"save_steps": 50,
|
203 |
"total_flos": 0.0,
|
204 |
"trial_name": null,
|
205 |
"trial_params": null
|