End of training
Browse files- README.md +15 -2
- all_results.json +22 -0
- eval_results.json +17 -0
- train_results.json +8 -0
- trainer_state.json +302 -0
- training_loss.png +0 -0
- training_rewards_accuracies.png +0 -0
- training_sft_loss.png +0 -0
README.md
CHANGED
@@ -2,9 +2,10 @@
|
|
2 |
license: apache-2.0
|
3 |
library_name: peft
|
4 |
tags:
|
|
|
|
|
5 |
- trl
|
6 |
- dpo
|
7 |
-
- llama-factory
|
8 |
- generated_from_trainer
|
9 |
base_model: mistralai/Mistral-7B-Instruct-v0.2
|
10 |
model-index:
|
@@ -17,7 +18,19 @@ should probably proofread and complete it, then remove this comment. -->
|
|
17 |
|
18 |
# Mistral-7B-Instruct-v0.2-ORPO
|
19 |
|
20 |
-
This model is a fine-tuned version of [mistralai/Mistral-7B-Instruct-v0.2](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2) on
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
|
22 |
## Model description
|
23 |
|
|
|
2 |
license: apache-2.0
|
3 |
library_name: peft
|
4 |
tags:
|
5 |
+
- llama-factory
|
6 |
+
- lora
|
7 |
- trl
|
8 |
- dpo
|
|
|
9 |
- generated_from_trainer
|
10 |
base_model: mistralai/Mistral-7B-Instruct-v0.2
|
11 |
model-index:
|
|
|
18 |
|
19 |
# Mistral-7B-Instruct-v0.2-ORPO
|
20 |
|
21 |
+
This model is a fine-tuned version of [mistralai/Mistral-7B-Instruct-v0.2](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2) on the dpo_mix_en dataset.
|
22 |
+
It achieves the following results on the evaluation set:
|
23 |
+
- Loss: 1.1249
|
24 |
+
- Rewards/chosen: -0.1062
|
25 |
+
- Rewards/rejected: -0.1197
|
26 |
+
- Rewards/accuracies: 0.4400
|
27 |
+
- Rewards/margins: 0.0135
|
28 |
+
- Logps/rejected: -1.1975
|
29 |
+
- Logps/chosen: -1.0620
|
30 |
+
- Logits/rejected: -2.6819
|
31 |
+
- Logits/chosen: -2.6777
|
32 |
+
- Sft Loss: 1.0620
|
33 |
+
- Odds Ratio Loss: 0.6295
|
34 |
|
35 |
## Model description
|
36 |
|
all_results.json
ADDED
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 2.986666666666667,
|
3 |
+
"eval_logits/chosen": -2.677677869796753,
|
4 |
+
"eval_logits/rejected": -2.6818528175354004,
|
5 |
+
"eval_logps/chosen": -1.0619781017303467,
|
6 |
+
"eval_logps/rejected": -1.1974726915359497,
|
7 |
+
"eval_loss": 1.1249287128448486,
|
8 |
+
"eval_odds_ratio_loss": 0.62950599193573,
|
9 |
+
"eval_rewards/accuracies": 0.4399999976158142,
|
10 |
+
"eval_rewards/chosen": -0.10619781166315079,
|
11 |
+
"eval_rewards/margins": 0.013549459166824818,
|
12 |
+
"eval_rewards/rejected": -0.11974727362394333,
|
13 |
+
"eval_runtime": 57.437,
|
14 |
+
"eval_samples_per_second": 1.741,
|
15 |
+
"eval_sft_loss": 1.0619781017303467,
|
16 |
+
"eval_steps_per_second": 0.435,
|
17 |
+
"total_flos": 2.303993975490478e+17,
|
18 |
+
"train_loss": 1.1586682626179285,
|
19 |
+
"train_runtime": 5352.1206,
|
20 |
+
"train_samples_per_second": 0.504,
|
21 |
+
"train_steps_per_second": 0.031
|
22 |
+
}
|
eval_results.json
ADDED
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 2.986666666666667,
|
3 |
+
"eval_logits/chosen": -2.677677869796753,
|
4 |
+
"eval_logits/rejected": -2.6818528175354004,
|
5 |
+
"eval_logps/chosen": -1.0619781017303467,
|
6 |
+
"eval_logps/rejected": -1.1974726915359497,
|
7 |
+
"eval_loss": 1.1249287128448486,
|
8 |
+
"eval_odds_ratio_loss": 0.62950599193573,
|
9 |
+
"eval_rewards/accuracies": 0.4399999976158142,
|
10 |
+
"eval_rewards/chosen": -0.10619781166315079,
|
11 |
+
"eval_rewards/margins": 0.013549459166824818,
|
12 |
+
"eval_rewards/rejected": -0.11974727362394333,
|
13 |
+
"eval_runtime": 57.437,
|
14 |
+
"eval_samples_per_second": 1.741,
|
15 |
+
"eval_sft_loss": 1.0619781017303467,
|
16 |
+
"eval_steps_per_second": 0.435
|
17 |
+
}
|
train_results.json
ADDED
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 2.986666666666667,
|
3 |
+
"total_flos": 2.303993975490478e+17,
|
4 |
+
"train_loss": 1.1586682626179285,
|
5 |
+
"train_runtime": 5352.1206,
|
6 |
+
"train_samples_per_second": 0.504,
|
7 |
+
"train_steps_per_second": 0.031
|
8 |
+
}
|
trainer_state.json
ADDED
@@ -0,0 +1,302 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_metric": null,
|
3 |
+
"best_model_checkpoint": null,
|
4 |
+
"epoch": 2.986666666666667,
|
5 |
+
"eval_steps": 500,
|
6 |
+
"global_step": 168,
|
7 |
+
"is_hyper_param_search": false,
|
8 |
+
"is_local_process_zero": true,
|
9 |
+
"is_world_process_zero": true,
|
10 |
+
"log_history": [
|
11 |
+
{
|
12 |
+
"epoch": 0.17777777777777778,
|
13 |
+
"grad_norm": 1.5870693922042847,
|
14 |
+
"learning_rate": 4.957230266673969e-06,
|
15 |
+
"logits/chosen": -2.5848517417907715,
|
16 |
+
"logits/rejected": -2.592822313308716,
|
17 |
+
"logps/chosen": -1.2300995588302612,
|
18 |
+
"logps/rejected": -1.2750391960144043,
|
19 |
+
"loss": 1.3004,
|
20 |
+
"odds_ratio_loss": 0.7031511068344116,
|
21 |
+
"rewards/accuracies": 0.45625001192092896,
|
22 |
+
"rewards/chosen": -0.12300996482372284,
|
23 |
+
"rewards/margins": 0.004493957385420799,
|
24 |
+
"rewards/rejected": -0.1275039166212082,
|
25 |
+
"sft_loss": 1.2300995588302612,
|
26 |
+
"step": 10
|
27 |
+
},
|
28 |
+
{
|
29 |
+
"epoch": 0.35555555555555557,
|
30 |
+
"grad_norm": 2.5478012561798096,
|
31 |
+
"learning_rate": 4.84529614391025e-06,
|
32 |
+
"logits/chosen": -2.568402051925659,
|
33 |
+
"logits/rejected": -2.5743260383605957,
|
34 |
+
"logps/chosen": -1.1148786544799805,
|
35 |
+
"logps/rejected": -1.2647678852081299,
|
36 |
+
"loss": 1.1815,
|
37 |
+
"odds_ratio_loss": 0.6664617657661438,
|
38 |
+
"rewards/accuracies": 0.48124998807907104,
|
39 |
+
"rewards/chosen": -0.11148785054683685,
|
40 |
+
"rewards/margins": 0.014988938346505165,
|
41 |
+
"rewards/rejected": -0.12647680938243866,
|
42 |
+
"sft_loss": 1.1148786544799805,
|
43 |
+
"step": 20
|
44 |
+
},
|
45 |
+
{
|
46 |
+
"epoch": 0.5333333333333333,
|
47 |
+
"grad_norm": 1.738682746887207,
|
48 |
+
"learning_rate": 4.667009949002349e-06,
|
49 |
+
"logits/chosen": -2.5901498794555664,
|
50 |
+
"logits/rejected": -2.6051435470581055,
|
51 |
+
"logps/chosen": -1.2092516422271729,
|
52 |
+
"logps/rejected": -1.3586838245391846,
|
53 |
+
"loss": 1.2768,
|
54 |
+
"odds_ratio_loss": 0.6759374737739563,
|
55 |
+
"rewards/accuracies": 0.4375,
|
56 |
+
"rewards/chosen": -0.120925173163414,
|
57 |
+
"rewards/margins": 0.01494320947676897,
|
58 |
+
"rewards/rejected": -0.1358683854341507,
|
59 |
+
"sft_loss": 1.2092516422271729,
|
60 |
+
"step": 30
|
61 |
+
},
|
62 |
+
{
|
63 |
+
"epoch": 0.7111111111111111,
|
64 |
+
"grad_norm": 10.248797416687012,
|
65 |
+
"learning_rate": 4.397288409237892e-06,
|
66 |
+
"logits/chosen": -2.4943628311157227,
|
67 |
+
"logits/rejected": -2.5201773643493652,
|
68 |
+
"logps/chosen": -1.1567463874816895,
|
69 |
+
"logps/rejected": -1.3316763639450073,
|
70 |
+
"loss": 1.2213,
|
71 |
+
"odds_ratio_loss": 0.6450860500335693,
|
72 |
+
"rewards/accuracies": 0.48124998807907104,
|
73 |
+
"rewards/chosen": -0.11567463725805283,
|
74 |
+
"rewards/margins": 0.017492998391389847,
|
75 |
+
"rewards/rejected": -0.13316763937473297,
|
76 |
+
"sft_loss": 1.1567463874816895,
|
77 |
+
"step": 40
|
78 |
+
},
|
79 |
+
{
|
80 |
+
"epoch": 0.8888888888888888,
|
81 |
+
"grad_norm": 6.022939682006836,
|
82 |
+
"learning_rate": 4.097593272948218e-06,
|
83 |
+
"logits/chosen": -2.6041746139526367,
|
84 |
+
"logits/rejected": -2.6204562187194824,
|
85 |
+
"logps/chosen": -1.2715274095535278,
|
86 |
+
"logps/rejected": -1.6485040187835693,
|
87 |
+
"loss": 1.3385,
|
88 |
+
"odds_ratio_loss": 0.6699766516685486,
|
89 |
+
"rewards/accuracies": 0.48124998807907104,
|
90 |
+
"rewards/chosen": -0.12715274095535278,
|
91 |
+
"rewards/margins": 0.03769766539335251,
|
92 |
+
"rewards/rejected": -0.1648503988981247,
|
93 |
+
"sft_loss": 1.2715274095535278,
|
94 |
+
"step": 50
|
95 |
+
},
|
96 |
+
{
|
97 |
+
"epoch": 1.0666666666666667,
|
98 |
+
"grad_norm": 1.5987056493759155,
|
99 |
+
"learning_rate": 3.712001321256744e-06,
|
100 |
+
"logits/chosen": -2.6027374267578125,
|
101 |
+
"logits/rejected": -2.627061128616333,
|
102 |
+
"logps/chosen": -1.0649330615997314,
|
103 |
+
"logps/rejected": -1.3192559480667114,
|
104 |
+
"loss": 1.1266,
|
105 |
+
"odds_ratio_loss": 0.616315484046936,
|
106 |
+
"rewards/accuracies": 0.550000011920929,
|
107 |
+
"rewards/chosen": -0.10649331659078598,
|
108 |
+
"rewards/margins": 0.025432268157601357,
|
109 |
+
"rewards/rejected": -0.13192559778690338,
|
110 |
+
"sft_loss": 1.0649330615997314,
|
111 |
+
"step": 60
|
112 |
+
},
|
113 |
+
{
|
114 |
+
"epoch": 1.2444444444444445,
|
115 |
+
"grad_norm": 9.42609977722168,
|
116 |
+
"learning_rate": 3.284100248421622e-06,
|
117 |
+
"logits/chosen": -2.592658758163452,
|
118 |
+
"logits/rejected": -2.612031936645508,
|
119 |
+
"logps/chosen": -1.0701837539672852,
|
120 |
+
"logps/rejected": -1.2456352710723877,
|
121 |
+
"loss": 1.1365,
|
122 |
+
"odds_ratio_loss": 0.6634659767150879,
|
123 |
+
"rewards/accuracies": 0.512499988079071,
|
124 |
+
"rewards/chosen": -0.10701838880777359,
|
125 |
+
"rewards/margins": 0.017545154318213463,
|
126 |
+
"rewards/rejected": -0.1245635375380516,
|
127 |
+
"sft_loss": 1.0701837539672852,
|
128 |
+
"step": 70
|
129 |
+
},
|
130 |
+
{
|
131 |
+
"epoch": 1.4222222222222223,
|
132 |
+
"grad_norm": 1.1247073411941528,
|
133 |
+
"learning_rate": 2.8288274295187397e-06,
|
134 |
+
"logits/chosen": -2.6690917015075684,
|
135 |
+
"logits/rejected": -2.6884384155273438,
|
136 |
+
"logps/chosen": -1.0463426113128662,
|
137 |
+
"logps/rejected": -1.2134628295898438,
|
138 |
+
"loss": 1.1088,
|
139 |
+
"odds_ratio_loss": 0.6244794130325317,
|
140 |
+
"rewards/accuracies": 0.4749999940395355,
|
141 |
+
"rewards/chosen": -0.10463427007198334,
|
142 |
+
"rewards/margins": 0.016712015494704247,
|
143 |
+
"rewards/rejected": -0.12134629487991333,
|
144 |
+
"sft_loss": 1.0463426113128662,
|
145 |
+
"step": 80
|
146 |
+
},
|
147 |
+
{
|
148 |
+
"epoch": 1.6,
|
149 |
+
"grad_norm": 8.079381942749023,
|
150 |
+
"learning_rate": 2.3620757456122797e-06,
|
151 |
+
"logits/chosen": -2.638796329498291,
|
152 |
+
"logits/rejected": -2.6489083766937256,
|
153 |
+
"logps/chosen": -1.0774247646331787,
|
154 |
+
"logps/rejected": -1.465980887413025,
|
155 |
+
"loss": 1.1411,
|
156 |
+
"odds_ratio_loss": 0.6369243860244751,
|
157 |
+
"rewards/accuracies": 0.512499988079071,
|
158 |
+
"rewards/chosen": -0.10774248838424683,
|
159 |
+
"rewards/margins": 0.03885561600327492,
|
160 |
+
"rewards/rejected": -0.14659810066223145,
|
161 |
+
"sft_loss": 1.0774247646331787,
|
162 |
+
"step": 90
|
163 |
+
},
|
164 |
+
{
|
165 |
+
"epoch": 1.7777777777777777,
|
166 |
+
"grad_norm": 3.1488540172576904,
|
167 |
+
"learning_rate": 1.9001387874698865e-06,
|
168 |
+
"logits/chosen": -2.659306049346924,
|
169 |
+
"logits/rejected": -2.67085599899292,
|
170 |
+
"logps/chosen": -1.0682704448699951,
|
171 |
+
"logps/rejected": -1.2112770080566406,
|
172 |
+
"loss": 1.1326,
|
173 |
+
"odds_ratio_loss": 0.6432015299797058,
|
174 |
+
"rewards/accuracies": 0.5625,
|
175 |
+
"rewards/chosen": -0.10682705789804459,
|
176 |
+
"rewards/margins": 0.01430064719170332,
|
177 |
+
"rewards/rejected": -0.12112770229578018,
|
178 |
+
"sft_loss": 1.0682704448699951,
|
179 |
+
"step": 100
|
180 |
+
},
|
181 |
+
{
|
182 |
+
"epoch": 1.9555555555555557,
|
183 |
+
"grad_norm": 2.191756010055542,
|
184 |
+
"learning_rate": 1.4591420711120425e-06,
|
185 |
+
"logits/chosen": -2.685157299041748,
|
186 |
+
"logits/rejected": -2.695525646209717,
|
187 |
+
"logps/chosen": -1.0902485847473145,
|
188 |
+
"logps/rejected": -1.2659653425216675,
|
189 |
+
"loss": 1.1523,
|
190 |
+
"odds_ratio_loss": 0.6206409335136414,
|
191 |
+
"rewards/accuracies": 0.4937500059604645,
|
192 |
+
"rewards/chosen": -0.10902485996484756,
|
193 |
+
"rewards/margins": 0.017571674659848213,
|
194 |
+
"rewards/rejected": -0.12659654021263123,
|
195 |
+
"sft_loss": 1.0902485847473145,
|
196 |
+
"step": 110
|
197 |
+
},
|
198 |
+
{
|
199 |
+
"epoch": 2.1333333333333333,
|
200 |
+
"grad_norm": 1.3168927431106567,
|
201 |
+
"learning_rate": 1.0544801205950327e-06,
|
202 |
+
"logits/chosen": -2.6532204151153564,
|
203 |
+
"logits/rejected": -2.6720328330993652,
|
204 |
+
"logps/chosen": -1.0528920888900757,
|
205 |
+
"logps/rejected": -1.2378777265548706,
|
206 |
+
"loss": 1.1173,
|
207 |
+
"odds_ratio_loss": 0.6442069411277771,
|
208 |
+
"rewards/accuracies": 0.48750001192092896,
|
209 |
+
"rewards/chosen": -0.10528920590877533,
|
210 |
+
"rewards/margins": 0.018498573452234268,
|
211 |
+
"rewards/rejected": -0.1237877756357193,
|
212 |
+
"sft_loss": 1.0528920888900757,
|
213 |
+
"step": 120
|
214 |
+
},
|
215 |
+
{
|
216 |
+
"epoch": 2.311111111111111,
|
217 |
+
"grad_norm": 0.9034324288368225,
|
218 |
+
"learning_rate": 7.002790686106056e-07,
|
219 |
+
"logits/chosen": -2.6588597297668457,
|
220 |
+
"logits/rejected": -2.6591033935546875,
|
221 |
+
"logps/chosen": -0.9490475654602051,
|
222 |
+
"logps/rejected": -1.0968741178512573,
|
223 |
+
"loss": 1.0107,
|
224 |
+
"odds_ratio_loss": 0.616373598575592,
|
225 |
+
"rewards/accuracies": 0.48750001192092896,
|
226 |
+
"rewards/chosen": -0.09490474313497543,
|
227 |
+
"rewards/margins": 0.01478266529738903,
|
228 |
+
"rewards/rejected": -0.1096874251961708,
|
229 |
+
"sft_loss": 0.9490475654602051,
|
230 |
+
"step": 130
|
231 |
+
},
|
232 |
+
{
|
233 |
+
"epoch": 2.488888888888889,
|
234 |
+
"grad_norm": 1.7593573331832886,
|
235 |
+
"learning_rate": 4.0890353469735717e-07,
|
236 |
+
"logits/chosen": -2.686373472213745,
|
237 |
+
"logits/rejected": -2.7087228298187256,
|
238 |
+
"logps/chosen": -1.0427912473678589,
|
239 |
+
"logps/rejected": -1.486315131187439,
|
240 |
+
"loss": 1.1034,
|
241 |
+
"odds_ratio_loss": 0.6059055328369141,
|
242 |
+
"rewards/accuracies": 0.6000000238418579,
|
243 |
+
"rewards/chosen": -0.10427911579608917,
|
244 |
+
"rewards/margins": 0.04435240477323532,
|
245 |
+
"rewards/rejected": -0.1486315280199051,
|
246 |
+
"sft_loss": 1.0427912473678589,
|
247 |
+
"step": 140
|
248 |
+
},
|
249 |
+
{
|
250 |
+
"epoch": 2.6666666666666665,
|
251 |
+
"grad_norm": 5.990176200866699,
|
252 |
+
"learning_rate": 1.905249951948071e-07,
|
253 |
+
"logits/chosen": -2.715557813644409,
|
254 |
+
"logits/rejected": -2.731482982635498,
|
255 |
+
"logps/chosen": -1.144046425819397,
|
256 |
+
"logps/rejected": -1.2883504629135132,
|
257 |
+
"loss": 1.2074,
|
258 |
+
"odds_ratio_loss": 0.6331667304039001,
|
259 |
+
"rewards/accuracies": 0.48124998807907104,
|
260 |
+
"rewards/chosen": -0.11440464109182358,
|
261 |
+
"rewards/margins": 0.014430420473217964,
|
262 |
+
"rewards/rejected": -0.1288350522518158,
|
263 |
+
"sft_loss": 1.144046425819397,
|
264 |
+
"step": 150
|
265 |
+
},
|
266 |
+
{
|
267 |
+
"epoch": 2.8444444444444446,
|
268 |
+
"grad_norm": 2.793104887008667,
|
269 |
+
"learning_rate": 5.2766712488021566e-08,
|
270 |
+
"logits/chosen": -2.695949077606201,
|
271 |
+
"logits/rejected": -2.7109360694885254,
|
272 |
+
"logps/chosen": -0.9868586659431458,
|
273 |
+
"logps/rejected": -1.2109596729278564,
|
274 |
+
"loss": 1.0456,
|
275 |
+
"odds_ratio_loss": 0.5878169536590576,
|
276 |
+
"rewards/accuracies": 0.581250011920929,
|
277 |
+
"rewards/chosen": -0.0986858680844307,
|
278 |
+
"rewards/margins": 0.022410111501812935,
|
279 |
+
"rewards/rejected": -0.12109597772359848,
|
280 |
+
"sft_loss": 0.9868586659431458,
|
281 |
+
"step": 160
|
282 |
+
},
|
283 |
+
{
|
284 |
+
"epoch": 2.986666666666667,
|
285 |
+
"step": 168,
|
286 |
+
"total_flos": 2.303993975490478e+17,
|
287 |
+
"train_loss": 1.1586682626179285,
|
288 |
+
"train_runtime": 5352.1206,
|
289 |
+
"train_samples_per_second": 0.504,
|
290 |
+
"train_steps_per_second": 0.031
|
291 |
+
}
|
292 |
+
],
|
293 |
+
"logging_steps": 10,
|
294 |
+
"max_steps": 168,
|
295 |
+
"num_input_tokens_seen": 0,
|
296 |
+
"num_train_epochs": 3,
|
297 |
+
"save_steps": 500,
|
298 |
+
"total_flos": 2.303993975490478e+17,
|
299 |
+
"train_batch_size": 2,
|
300 |
+
"trial_name": null,
|
301 |
+
"trial_params": null
|
302 |
+
}
|
training_loss.png
ADDED
training_rewards_accuracies.png
ADDED
training_sft_loss.png
ADDED