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_14-07-55_uclaml04.cs.ucla.edu/events.out.tfevents.1705874931.uclaml04.cs.ucla.edu.725178.0 +2 -2
- runs/Jan21_14-07-55_uclaml04.cs.ucla.edu/events.out.tfevents.1705900772.uclaml04.cs.ucla.edu.725178.1 +3 -0
- train_results.json +4 -4
- trainer_state.json +107 -107
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: 0.5656
|
19 |
+
- Rewards/chosen: 0.0673
|
20 |
+
- Rewards/rejected: -0.5344
|
21 |
+
- Rewards/accuracies: 0.7000
|
22 |
+
- Rewards/margins: 0.6018
|
23 |
+
- Logps/rejected: -264.6442
|
24 |
+
- Logps/chosen: -283.5841
|
25 |
+
- Logits/rejected: -2.8292
|
26 |
+
- Logits/chosen: -2.8303
|
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: 5e-06
|
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 |
+
| 0.5689 | 1.0 | 955 | 0.5656 | 0.0673 | -0.5344 | 0.7000 | 0.6018 | -264.6442 | -283.5841 | -2.8292 | -2.8303 | 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 sha256:52d5571e6c85d40b4ab603d26a4d7d7716eaa1a22f8e22bebc933ad16dbbb8ae
|
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": -2.8302695751190186,
|
4 |
+
"eval_logits/rejected": -2.8291618824005127,
|
5 |
+
"eval_logps/chosen": -283.5841369628906,
|
6 |
+
"eval_logps/rejected": -264.64422607421875,
|
7 |
+
"eval_loss": 0.5655555129051208,
|
8 |
"eval_pred_label": 0.0,
|
9 |
+
"eval_rewards/accuracies": 0.699999988079071,
|
10 |
+
"eval_rewards/chosen": 0.06732505559921265,
|
11 |
+
"eval_rewards/margins": 0.6017746329307556,
|
12 |
+
"eval_rewards/rejected": -0.534449577331543,
|
13 |
+
"eval_runtime": 457.1465,
|
14 |
"eval_samples": 2000,
|
15 |
+
"eval_samples_per_second": 4.375,
|
16 |
+
"eval_steps_per_second": 0.273,
|
17 |
"eval_use_label": 0.0,
|
18 |
+
"train_loss": 0.5870625535855118,
|
19 |
+
"train_runtime": 25384.4334,
|
20 |
"train_samples": 61135,
|
21 |
+
"train_samples_per_second": 2.408,
|
22 |
+
"train_steps_per_second": 0.038
|
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": -2.8302695751190186,
|
4 |
+
"eval_logits/rejected": -2.8291618824005127,
|
5 |
+
"eval_logps/chosen": -283.5841369628906,
|
6 |
+
"eval_logps/rejected": -264.64422607421875,
|
7 |
+
"eval_loss": 0.5655555129051208,
|
8 |
"eval_pred_label": 0.0,
|
9 |
+
"eval_rewards/accuracies": 0.699999988079071,
|
10 |
+
"eval_rewards/chosen": 0.06732505559921265,
|
11 |
+
"eval_rewards/margins": 0.6017746329307556,
|
12 |
+
"eval_rewards/rejected": -0.534449577331543,
|
13 |
+
"eval_runtime": 457.1465,
|
14 |
"eval_samples": 2000,
|
15 |
+
"eval_samples_per_second": 4.375,
|
16 |
+
"eval_steps_per_second": 0.273,
|
17 |
"eval_use_label": 0.0
|
18 |
}
|
runs/Jan21_14-07-55_uclaml04.cs.ucla.edu/events.out.tfevents.1705874931.uclaml04.cs.ucla.edu.725178.0
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 sha256:47c97b55e57bc0004ea9c1ec4e8880902f5488120bd4caddf86cb78fedfcc63f
|
3 |
+
size 13039
|
runs/Jan21_14-07-55_uclaml04.cs.ucla.edu/events.out.tfevents.1705900772.uclaml04.cs.ucla.edu.725178.1
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:866b58e1ab9cceff104815835a51153047f74404ef944895a1dbbf5ab4c7b723
|
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": 0.5870625535855118,
|
4 |
+
"train_runtime": 25384.4334,
|
5 |
"train_samples": 61135,
|
6 |
+
"train_samples_per_second": 2.408,
|
7 |
+
"train_steps_per_second": 0.038
|
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": -
|
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": -283.
|
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": 0.
|
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": -
|
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": -
|
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": -
|
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 |
},
|
@@ -190,10 +190,10 @@
|
|
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,
|
|
|
10 |
"log_history": [
|
11 |
{
|
12 |
"epoch": 0.0,
|
13 |
+
"learning_rate": 5.208333333333333e-08,
|
14 |
"logits/chosen": -2.980285167694092,
|
15 |
"logits/rejected": -2.87275767326355,
|
16 |
"logps/chosen": -313.4390563964844,
|
|
|
26 |
},
|
27 |
{
|
28 |
"epoch": 0.1,
|
29 |
+
"learning_rate": 4.9767171129220025e-06,
|
30 |
+
"logits/chosen": -2.8189077377319336,
|
31 |
+
"logits/rejected": -2.828260660171509,
|
32 |
+
"logps/chosen": -285.1784362792969,
|
33 |
+
"logps/rejected": -271.1445617675781,
|
34 |
+
"loss": 0.6803,
|
35 |
"pred_label": 0.0,
|
36 |
+
"rewards/accuracies": 0.5738636255264282,
|
37 |
+
"rewards/chosen": 0.01021653600037098,
|
38 |
+
"rewards/margins": 0.029710056260228157,
|
39 |
+
"rewards/rejected": -0.019493522122502327,
|
40 |
"step": 100,
|
41 |
"use_label": 0.0
|
42 |
},
|
43 |
{
|
44 |
"epoch": 0.21,
|
45 |
+
"learning_rate": 4.394644935972061e-06,
|
46 |
+
"logits/chosen": -2.8255398273468018,
|
47 |
+
"logits/rejected": -2.813203811645508,
|
48 |
+
"logps/chosen": -278.2582092285156,
|
49 |
+
"logps/rejected": -262.1144714355469,
|
50 |
+
"loss": 0.6236,
|
51 |
"pred_label": 0.0,
|
52 |
+
"rewards/accuracies": 0.6868749856948853,
|
53 |
+
"rewards/chosen": 0.05237884446978569,
|
54 |
+
"rewards/margins": 0.20995216071605682,
|
55 |
+
"rewards/rejected": -0.15757331252098083,
|
56 |
"step": 200,
|
57 |
"use_label": 0.0
|
58 |
},
|
59 |
{
|
60 |
"epoch": 0.31,
|
61 |
+
"learning_rate": 3.812572759022119e-06,
|
62 |
+
"logits/chosen": -2.809633255004883,
|
63 |
+
"logits/rejected": -2.80940580368042,
|
64 |
+
"logps/chosen": -283.6641540527344,
|
65 |
+
"logps/rejected": -257.23779296875,
|
66 |
+
"loss": 0.5901,
|
67 |
"pred_label": 0.0,
|
68 |
+
"rewards/accuracies": 0.6956250071525574,
|
69 |
+
"rewards/chosen": 0.04034877195954323,
|
70 |
+
"rewards/margins": 0.3768764138221741,
|
71 |
+
"rewards/rejected": -0.33652764558792114,
|
72 |
"step": 300,
|
73 |
"use_label": 0.0
|
74 |
},
|
75 |
{
|
76 |
"epoch": 0.42,
|
77 |
+
"learning_rate": 3.2305005820721774e-06,
|
78 |
+
"logits/chosen": -2.821267604827881,
|
79 |
+
"logits/rejected": -2.8011868000030518,
|
80 |
+
"logps/chosen": -284.2022399902344,
|
81 |
+
"logps/rejected": -267.48358154296875,
|
82 |
+
"loss": 0.5757,
|
83 |
"pred_label": 0.0,
|
84 |
+
"rewards/accuracies": 0.7081249952316284,
|
85 |
+
"rewards/chosen": 0.022827474400401115,
|
86 |
+
"rewards/margins": 0.4984941780567169,
|
87 |
+
"rewards/rejected": -0.47566673159599304,
|
88 |
"step": 400,
|
89 |
"use_label": 0.0
|
90 |
},
|
91 |
{
|
92 |
"epoch": 0.52,
|
93 |
+
"learning_rate": 2.6484284051222353e-06,
|
94 |
+
"logits/chosen": -2.7993886470794678,
|
95 |
+
"logits/rejected": -2.7845959663391113,
|
96 |
+
"logps/chosen": -270.09637451171875,
|
97 |
+
"logps/rejected": -256.13458251953125,
|
98 |
+
"loss": 0.5798,
|
99 |
"pred_label": 0.0,
|
100 |
+
"rewards/accuracies": 0.6612499952316284,
|
101 |
+
"rewards/chosen": -0.004339172504842281,
|
102 |
+
"rewards/margins": 0.4106566905975342,
|
103 |
+
"rewards/rejected": -0.41499578952789307,
|
104 |
"step": 500,
|
105 |
"use_label": 0.0
|
106 |
},
|
107 |
{
|
108 |
"epoch": 0.63,
|
109 |
+
"learning_rate": 2.0663562281722936e-06,
|
110 |
+
"logits/chosen": -2.824444532394409,
|
111 |
+
"logits/rejected": -2.808506965637207,
|
112 |
+
"logps/chosen": -284.5413513183594,
|
113 |
+
"logps/rejected": -270.1356201171875,
|
114 |
+
"loss": 0.5643,
|
115 |
"pred_label": 0.0,
|
116 |
+
"rewards/accuracies": 0.706250011920929,
|
117 |
+
"rewards/chosen": 0.06895674020051956,
|
118 |
+
"rewards/margins": 0.5562920570373535,
|
119 |
+
"rewards/rejected": -0.4873352348804474,
|
120 |
"step": 600,
|
121 |
"use_label": 0.0
|
122 |
},
|
123 |
{
|
124 |
"epoch": 0.73,
|
125 |
+
"learning_rate": 1.4842840512223516e-06,
|
126 |
+
"logits/chosen": -2.8160746097564697,
|
127 |
+
"logits/rejected": -2.795208692550659,
|
128 |
+
"logps/chosen": -280.79766845703125,
|
129 |
+
"logps/rejected": -253.41058349609375,
|
130 |
+
"loss": 0.5631,
|
131 |
"pred_label": 0.0,
|
132 |
+
"rewards/accuracies": 0.7231249809265137,
|
133 |
+
"rewards/chosen": 0.07620371133089066,
|
134 |
+
"rewards/margins": 0.5511507391929626,
|
135 |
+
"rewards/rejected": -0.474947065114975,
|
136 |
"step": 700,
|
137 |
"use_label": 0.0
|
138 |
},
|
139 |
{
|
140 |
"epoch": 0.84,
|
141 |
+
"learning_rate": 9.022118742724098e-07,
|
142 |
+
"logits/chosen": -2.8168938159942627,
|
143 |
+
"logits/rejected": -2.8180816173553467,
|
144 |
+
"logps/chosen": -287.77679443359375,
|
145 |
+
"logps/rejected": -266.3511657714844,
|
146 |
+
"loss": 0.5548,
|
147 |
"pred_label": 0.0,
|
148 |
+
"rewards/accuracies": 0.7112500071525574,
|
149 |
+
"rewards/chosen": 0.09802371263504028,
|
150 |
+
"rewards/margins": 0.5919383764266968,
|
151 |
+
"rewards/rejected": -0.49391472339630127,
|
152 |
"step": 800,
|
153 |
"use_label": 0.0
|
154 |
},
|
155 |
{
|
156 |
"epoch": 0.94,
|
157 |
+
"learning_rate": 3.2013969732246806e-07,
|
158 |
+
"logits/chosen": -2.808311700820923,
|
159 |
+
"logits/rejected": -2.812298536300659,
|
160 |
+
"logps/chosen": -276.976806640625,
|
161 |
+
"logps/rejected": -261.3187561035156,
|
162 |
+
"loss": 0.5689,
|
163 |
"pred_label": 0.0,
|
164 |
+
"rewards/accuracies": 0.7093750238418579,
|
165 |
+
"rewards/chosen": 0.05313897505402565,
|
166 |
+
"rewards/margins": 0.551435649394989,
|
167 |
+
"rewards/rejected": -0.4982966482639313,
|
168 |
"step": 900,
|
169 |
"use_label": 0.0
|
170 |
},
|
171 |
{
|
172 |
"epoch": 1.0,
|
173 |
+
"eval_logits/chosen": -2.8302695751190186,
|
174 |
+
"eval_logits/rejected": -2.8291618824005127,
|
175 |
+
"eval_logps/chosen": -283.5841369628906,
|
176 |
+
"eval_logps/rejected": -264.64422607421875,
|
177 |
+
"eval_loss": 0.5655555129051208,
|
178 |
"eval_pred_label": 0.0,
|
179 |
+
"eval_rewards/accuracies": 0.699999988079071,
|
180 |
+
"eval_rewards/chosen": 0.06732505559921265,
|
181 |
+
"eval_rewards/margins": 0.6017746329307556,
|
182 |
+
"eval_rewards/rejected": -0.534449577331543,
|
183 |
+
"eval_runtime": 457.335,
|
184 |
+
"eval_samples_per_second": 4.373,
|
185 |
+
"eval_steps_per_second": 0.273,
|
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": 0.5870625535855118,
|
194 |
+
"train_runtime": 25384.4334,
|
195 |
+
"train_samples_per_second": 2.408,
|
196 |
+
"train_steps_per_second": 0.038
|
197 |
}
|
198 |
],
|
199 |
"logging_steps": 100,
|