Model save
Browse files- README.md +11 -11
- adapter_model.safetensors +1 -1
- all_results.json +14 -14
- eval_results.json +12 -12
- runs/Jan21_01-22-38_uclaml04.cs.ucla.edu/events.out.tfevents.1705829022.uclaml04.cs.ucla.edu.475602.0 +2 -2
- runs/Jan21_01-22-38_uclaml04.cs.ucla.edu/events.out.tfevents.1705855119.uclaml04.cs.ucla.edu.475602.1 +3 -0
- train_results.json +2 -2
- trainer_state.json +105 -105
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: -
|
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: 0.6433
|
19 |
+
- Rewards/chosen: 0.0460
|
20 |
+
- Rewards/rejected: -0.0949
|
21 |
+
- Rewards/accuracies: 0.6740
|
22 |
+
- Rewards/margins: 0.1409
|
23 |
+
- Logps/rejected: -260.2492
|
24 |
+
- Logps/chosen: -283.7975
|
25 |
+
- Logits/rejected: -2.8439
|
26 |
+
- Logits/chosen: -2.8406
|
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-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 |
+
| 0.6441 | 1.0 | 955 | 0.6433 | 0.0460 | -0.0949 | 0.6740 | 0.1409 | -260.2492 | -283.7975 | -2.8439 | -2.8406 | 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 sha256:5cb8cb0361b2338d889c4aeb84f54608df0a08612cd81d53bfe35d6340c2e7b0
|
3 |
size 218138576
|
all_results.json
CHANGED
@@ -1,22 +1,22 @@
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
-
"eval_logits/chosen": -2.
|
4 |
-
"eval_logits/rejected": -2.
|
5 |
-
"eval_logps/chosen": -
|
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": 2.385,
|
22 |
"train_steps_per_second": 0.037
|
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
+
"eval_logits/chosen": -2.8405630588531494,
|
4 |
+
"eval_logits/rejected": -2.8439478874206543,
|
5 |
+
"eval_logps/chosen": -283.7974548339844,
|
6 |
+
"eval_logps/rejected": -260.2491760253906,
|
7 |
+
"eval_loss": 0.6432895660400391,
|
8 |
"eval_pred_label": 0.0,
|
9 |
+
"eval_rewards/accuracies": 0.6740000247955322,
|
10 |
+
"eval_rewards/chosen": 0.04599040374159813,
|
11 |
+
"eval_rewards/margins": 0.14093737304210663,
|
12 |
+
"eval_rewards/rejected": -0.0949469730257988,
|
13 |
+
"eval_runtime": 468.6797,
|
14 |
"eval_samples": 2000,
|
15 |
+
"eval_samples_per_second": 4.267,
|
16 |
+
"eval_steps_per_second": 0.267,
|
17 |
"eval_use_label": 0.0,
|
18 |
+
"train_loss": 0.660746982334796,
|
19 |
+
"train_runtime": 25629.2924,
|
20 |
"train_samples": 61135,
|
21 |
"train_samples_per_second": 2.385,
|
22 |
"train_steps_per_second": 0.037
|
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": -
|
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": -2.8405630588531494,
|
4 |
+
"eval_logits/rejected": -2.8439478874206543,
|
5 |
+
"eval_logps/chosen": -283.7974548339844,
|
6 |
+
"eval_logps/rejected": -260.2491760253906,
|
7 |
+
"eval_loss": 0.6432895660400391,
|
8 |
"eval_pred_label": 0.0,
|
9 |
+
"eval_rewards/accuracies": 0.6740000247955322,
|
10 |
+
"eval_rewards/chosen": 0.04599040374159813,
|
11 |
+
"eval_rewards/margins": 0.14093737304210663,
|
12 |
+
"eval_rewards/rejected": -0.0949469730257988,
|
13 |
+
"eval_runtime": 468.6797,
|
14 |
"eval_samples": 2000,
|
15 |
+
"eval_samples_per_second": 4.267,
|
16 |
+
"eval_steps_per_second": 0.267,
|
17 |
"eval_use_label": 0.0
|
18 |
}
|
runs/Jan21_01-22-38_uclaml04.cs.ucla.edu/events.out.tfevents.1705829022.uclaml04.cs.ucla.edu.475602.0
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:e676c35e95fb49da088feb27d029428aa16e45cec2b9e90b44763b8a1695219a
|
3 |
+
size 13039
|
runs/Jan21_01-22-38_uclaml04.cs.ucla.edu/events.out.tfevents.1705855119.uclaml04.cs.ucla.edu.475602.1
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:7960fe6cc975d99bac6bab44dc4b36c37ee587fc654ea62de4d68de7e46d302a
|
3 |
+
size 935
|
train_results.json
CHANGED
@@ -1,7 +1,7 @@
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
-
"train_loss": 0.
|
4 |
-
"train_runtime":
|
5 |
"train_samples": 61135,
|
6 |
"train_samples_per_second": 2.385,
|
7 |
"train_steps_per_second": 0.037
|
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
+
"train_loss": 0.660746982334796,
|
4 |
+
"train_runtime": 25629.2924,
|
5 |
"train_samples": 61135,
|
6 |
"train_samples_per_second": 2.385,
|
7 |
"train_steps_per_second": 0.037
|
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": -270.
|
34 |
-
"loss": 0.
|
35 |
"pred_label": 0.0,
|
36 |
-
"rewards/accuracies": 0.
|
37 |
-
"rewards/chosen":
|
38 |
-
"rewards/margins":
|
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": -260.
|
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": -269.
|
97 |
-
"logps/rejected": -252.
|
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": -
|
113 |
-
"logps/rejected": -265.
|
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": -281.
|
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": -288.
|
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": -277.
|
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": -
|
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 |
},
|
@@ -190,8 +190,8 @@
|
|
190 |
"epoch": 1.0,
|
191 |
"step": 955,
|
192 |
"total_flos": 0.0,
|
193 |
-
"train_loss": 0.
|
194 |
-
"train_runtime":
|
195 |
"train_samples_per_second": 2.385,
|
196 |
"train_steps_per_second": 0.037
|
197 |
}
|
|
|
10 |
"log_history": [
|
11 |
{
|
12 |
"epoch": 0.0,
|
13 |
+
"learning_rate": 1.0416666666666666e-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": 9.953434225844004e-07,
|
30 |
+
"logits/chosen": -2.8195509910583496,
|
31 |
+
"logits/rejected": -2.8288235664367676,
|
32 |
+
"logps/chosen": -285.27215576171875,
|
33 |
+
"logps/rejected": -270.96209716796875,
|
34 |
+
"loss": 0.6923,
|
35 |
"pred_label": 0.0,
|
36 |
+
"rewards/accuracies": 0.5183081030845642,
|
37 |
+
"rewards/chosen": 0.0008437540382146835,
|
38 |
+
"rewards/margins": 0.0020915681961923838,
|
39 |
+
"rewards/rejected": -0.0012478141579777002,
|
40 |
"step": 100,
|
41 |
"use_label": 0.0
|
42 |
},
|
43 |
{
|
44 |
"epoch": 0.21,
|
45 |
+
"learning_rate": 8.789289871944121e-07,
|
46 |
+
"logits/chosen": -2.82773494720459,
|
47 |
+
"logits/rejected": -2.816188335418701,
|
48 |
+
"logps/chosen": -278.6949768066406,
|
49 |
+
"logps/rejected": -260.6236572265625,
|
50 |
+
"loss": 0.6855,
|
51 |
"pred_label": 0.0,
|
52 |
+
"rewards/accuracies": 0.5956249833106995,
|
53 |
+
"rewards/chosen": 0.00869889184832573,
|
54 |
+
"rewards/margins": 0.017191100865602493,
|
55 |
+
"rewards/rejected": -0.008492209017276764,
|
56 |
"step": 200,
|
57 |
"use_label": 0.0
|
58 |
},
|
59 |
{
|
60 |
"epoch": 0.31,
|
61 |
+
"learning_rate": 7.625145518044237e-07,
|
62 |
+
"logits/chosen": -2.813408613204956,
|
63 |
+
"logits/rejected": -2.8151955604553223,
|
64 |
+
"logps/chosen": -283.8744812011719,
|
65 |
+
"logps/rejected": -254.1025390625,
|
66 |
+
"loss": 0.675,
|
67 |
"pred_label": 0.0,
|
68 |
+
"rewards/accuracies": 0.653124988079071,
|
69 |
+
"rewards/chosen": 0.01931346394121647,
|
70 |
+
"rewards/margins": 0.042313817888498306,
|
71 |
+
"rewards/rejected": -0.023000353947281837,
|
72 |
"step": 300,
|
73 |
"use_label": 0.0
|
74 |
},
|
75 |
{
|
76 |
"epoch": 0.42,
|
77 |
+
"learning_rate": 6.461001164144354e-07,
|
78 |
+
"logits/chosen": -2.8259572982788086,
|
79 |
+
"logits/rejected": -2.808323383331299,
|
80 |
+
"logps/chosen": -284.1655578613281,
|
81 |
+
"logps/rejected": -263.1576843261719,
|
82 |
+
"loss": 0.6649,
|
83 |
"pred_label": 0.0,
|
84 |
+
"rewards/accuracies": 0.6837499737739563,
|
85 |
+
"rewards/chosen": 0.026492329314351082,
|
86 |
+
"rewards/margins": 0.06956954300403595,
|
87 |
+
"rewards/rejected": -0.04307721555233002,
|
88 |
"step": 400,
|
89 |
"use_label": 0.0
|
90 |
},
|
91 |
{
|
92 |
"epoch": 0.52,
|
93 |
+
"learning_rate": 5.29685681024447e-07,
|
94 |
+
"logits/chosen": -2.806447744369507,
|
95 |
+
"logits/rejected": -2.794842004776001,
|
96 |
+
"logps/chosen": -269.8316345214844,
|
97 |
+
"logps/rejected": -252.43292236328125,
|
98 |
+
"loss": 0.6596,
|
99 |
"pred_label": 0.0,
|
100 |
+
"rewards/accuracies": 0.628125011920929,
|
101 |
+
"rewards/chosen": 0.022135533392429352,
|
102 |
+
"rewards/margins": 0.06696329265832901,
|
103 |
+
"rewards/rejected": -0.04482775181531906,
|
104 |
"step": 500,
|
105 |
"use_label": 0.0
|
106 |
},
|
107 |
{
|
108 |
"epoch": 0.63,
|
109 |
+
"learning_rate": 4.1327124563445866e-07,
|
110 |
+
"logits/chosen": -2.8327181339263916,
|
111 |
+
"logits/rejected": -2.819202184677124,
|
112 |
+
"logps/chosen": -284.8338623046875,
|
113 |
+
"logps/rejected": -265.977294921875,
|
114 |
+
"loss": 0.6493,
|
115 |
"pred_label": 0.0,
|
116 |
+
"rewards/accuracies": 0.6675000190734863,
|
117 |
+
"rewards/chosen": 0.03970767930150032,
|
118 |
+
"rewards/margins": 0.11121242493391037,
|
119 |
+
"rewards/rejected": -0.07150473445653915,
|
120 |
"step": 600,
|
121 |
"use_label": 0.0
|
122 |
},
|
123 |
{
|
124 |
"epoch": 0.73,
|
125 |
+
"learning_rate": 2.968568102444703e-07,
|
126 |
+
"logits/chosen": -2.8258111476898193,
|
127 |
+
"logits/rejected": -2.8097949028015137,
|
128 |
+
"logps/chosen": -281.1001892089844,
|
129 |
+
"logps/rejected": -249.4017791748047,
|
130 |
+
"loss": 0.6463,
|
131 |
"pred_label": 0.0,
|
132 |
+
"rewards/accuracies": 0.6806250214576721,
|
133 |
+
"rewards/chosen": 0.04594804719090462,
|
134 |
+
"rewards/margins": 0.12001516669988632,
|
135 |
+
"rewards/rejected": -0.0740671157836914,
|
136 |
"step": 700,
|
137 |
"use_label": 0.0
|
138 |
},
|
139 |
{
|
140 |
"epoch": 0.84,
|
141 |
+
"learning_rate": 1.8044237485448194e-07,
|
142 |
+
"logits/chosen": -2.82832407951355,
|
143 |
+
"logits/rejected": -2.833012580871582,
|
144 |
+
"logps/chosen": -288.21307373046875,
|
145 |
+
"logps/rejected": -262.298095703125,
|
146 |
+
"loss": 0.64,
|
147 |
"pred_label": 0.0,
|
148 |
+
"rewards/accuracies": 0.6831250190734863,
|
149 |
+
"rewards/chosen": 0.05439276248216629,
|
150 |
+
"rewards/margins": 0.14299620687961578,
|
151 |
+
"rewards/rejected": -0.08860345184803009,
|
152 |
"step": 800,
|
153 |
"use_label": 0.0
|
154 |
},
|
155 |
{
|
156 |
"epoch": 0.94,
|
157 |
+
"learning_rate": 6.40279394644936e-08,
|
158 |
+
"logits/chosen": -2.8194103240966797,
|
159 |
+
"logits/rejected": -2.8261044025421143,
|
160 |
+
"logps/chosen": -277.05517578125,
|
161 |
+
"logps/rejected": -257.2059020996094,
|
162 |
+
"loss": 0.6441,
|
163 |
"pred_label": 0.0,
|
164 |
+
"rewards/accuracies": 0.6762499809265137,
|
165 |
+
"rewards/chosen": 0.04530614987015724,
|
166 |
+
"rewards/margins": 0.13231344521045685,
|
167 |
+
"rewards/rejected": -0.0870073139667511,
|
168 |
"step": 900,
|
169 |
"use_label": 0.0
|
170 |
},
|
171 |
{
|
172 |
"epoch": 1.0,
|
173 |
+
"eval_logits/chosen": -2.8405630588531494,
|
174 |
+
"eval_logits/rejected": -2.8439478874206543,
|
175 |
+
"eval_logps/chosen": -283.7974548339844,
|
176 |
+
"eval_logps/rejected": -260.2491760253906,
|
177 |
+
"eval_loss": 0.6432895660400391,
|
178 |
"eval_pred_label": 0.0,
|
179 |
+
"eval_rewards/accuracies": 0.6740000247955322,
|
180 |
+
"eval_rewards/chosen": 0.04599040374159813,
|
181 |
+
"eval_rewards/margins": 0.14093737304210663,
|
182 |
+
"eval_rewards/rejected": -0.0949469730257988,
|
183 |
+
"eval_runtime": 468.103,
|
184 |
+
"eval_samples_per_second": 4.273,
|
185 |
+
"eval_steps_per_second": 0.267,
|
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": 0.660746982334796,
|
194 |
+
"train_runtime": 25629.2924,
|
195 |
"train_samples_per_second": 2.385,
|
196 |
"train_steps_per_second": 0.037
|
197 |
}
|