Upload trainer_state.json with huggingface_hub
Browse files- trainer_state.json +217 -0
trainer_state.json
ADDED
@@ -0,0 +1,217 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_metric": 0.9069767594337463,
|
3 |
+
"best_model_checkpoint": "./llama3/27-06-24-Weni-ZeroShot-Agents-Llama3-4.0.37-DPO_Experiment with DPO and Llama3 8B, zeroshot 4.0.37-2_max_steps-570_batch_16_2024-06-27_ppid_9/checkpoint-80",
|
4 |
+
"epoch": 0.8366013071895425,
|
5 |
+
"eval_steps": 20,
|
6 |
+
"global_step": 80,
|
7 |
+
"is_hyper_param_search": false,
|
8 |
+
"is_local_process_zero": true,
|
9 |
+
"is_world_process_zero": true,
|
10 |
+
"log_history": [
|
11 |
+
{
|
12 |
+
"epoch": 0.10457516339869281,
|
13 |
+
"grad_norm": 3.655749559402466,
|
14 |
+
"learning_rate": 3.3333333333333337e-06,
|
15 |
+
"logits/chosen": -0.22948360443115234,
|
16 |
+
"logits/rejected": -0.22978875041007996,
|
17 |
+
"logps/chosen": -39.710899353027344,
|
18 |
+
"logps/rejected": -39.52346420288086,
|
19 |
+
"loss": 0.662,
|
20 |
+
"rewards/accuracies": 0.625,
|
21 |
+
"rewards/chosen": 0.11001075804233551,
|
22 |
+
"rewards/margins": 0.06790003925561905,
|
23 |
+
"rewards/rejected": 0.04211071878671646,
|
24 |
+
"step": 10
|
25 |
+
},
|
26 |
+
{
|
27 |
+
"epoch": 0.20915032679738563,
|
28 |
+
"grad_norm": 2.4934024810791016,
|
29 |
+
"learning_rate": 5.978260869565218e-06,
|
30 |
+
"logits/chosen": -0.22906163334846497,
|
31 |
+
"logits/rejected": -0.22876068949699402,
|
32 |
+
"logps/chosen": -30.569751739501953,
|
33 |
+
"logps/rejected": -33.86491012573242,
|
34 |
+
"loss": 0.5549,
|
35 |
+
"rewards/accuracies": 0.7437499761581421,
|
36 |
+
"rewards/chosen": 0.9981435537338257,
|
37 |
+
"rewards/margins": 0.38557058572769165,
|
38 |
+
"rewards/rejected": 0.6125729084014893,
|
39 |
+
"step": 20
|
40 |
+
},
|
41 |
+
{
|
42 |
+
"epoch": 0.20915032679738563,
|
43 |
+
"eval_logits/chosen": -0.19793638586997986,
|
44 |
+
"eval_logits/rejected": -0.197507843375206,
|
45 |
+
"eval_logps/chosen": -26.848573684692383,
|
46 |
+
"eval_logps/rejected": -35.13609313964844,
|
47 |
+
"eval_loss": 0.43225446343421936,
|
48 |
+
"eval_rewards/accuracies": 0.8081395626068115,
|
49 |
+
"eval_rewards/chosen": 1.355695366859436,
|
50 |
+
"eval_rewards/margins": 0.861485481262207,
|
51 |
+
"eval_rewards/rejected": 0.4942099153995514,
|
52 |
+
"eval_runtime": 76.6712,
|
53 |
+
"eval_samples_per_second": 2.23,
|
54 |
+
"eval_steps_per_second": 1.122,
|
55 |
+
"step": 20
|
56 |
+
},
|
57 |
+
{
|
58 |
+
"epoch": 0.3137254901960784,
|
59 |
+
"grad_norm": 1.9067373275756836,
|
60 |
+
"learning_rate": 5.869565217391305e-06,
|
61 |
+
"logits/chosen": -0.23883526027202606,
|
62 |
+
"logits/rejected": -0.2374078780412674,
|
63 |
+
"logps/chosen": -30.109268188476562,
|
64 |
+
"logps/rejected": -43.55142593383789,
|
65 |
+
"loss": 0.3732,
|
66 |
+
"rewards/accuracies": 0.84375,
|
67 |
+
"rewards/chosen": 1.0676757097244263,
|
68 |
+
"rewards/margins": 1.4297826290130615,
|
69 |
+
"rewards/rejected": -0.3621070086956024,
|
70 |
+
"step": 30
|
71 |
+
},
|
72 |
+
{
|
73 |
+
"epoch": 0.41830065359477125,
|
74 |
+
"grad_norm": 3.629997730255127,
|
75 |
+
"learning_rate": 5.760869565217392e-06,
|
76 |
+
"logits/chosen": -0.2035980522632599,
|
77 |
+
"logits/rejected": -0.20044592022895813,
|
78 |
+
"logps/chosen": -28.863727569580078,
|
79 |
+
"logps/rejected": -50.26195526123047,
|
80 |
+
"loss": 0.3276,
|
81 |
+
"rewards/accuracies": 0.84375,
|
82 |
+
"rewards/chosen": 1.200588345527649,
|
83 |
+
"rewards/margins": 2.2083239555358887,
|
84 |
+
"rewards/rejected": -1.0077354907989502,
|
85 |
+
"step": 40
|
86 |
+
},
|
87 |
+
{
|
88 |
+
"epoch": 0.41830065359477125,
|
89 |
+
"eval_logits/chosen": -0.16349415481090546,
|
90 |
+
"eval_logits/rejected": -0.15954892337322235,
|
91 |
+
"eval_logps/chosen": -24.61951446533203,
|
92 |
+
"eval_logps/rejected": -48.91069412231445,
|
93 |
+
"eval_loss": 0.34378084540367126,
|
94 |
+
"eval_rewards/accuracies": 0.8895348906517029,
|
95 |
+
"eval_rewards/chosen": 1.5786010026931763,
|
96 |
+
"eval_rewards/margins": 2.4618515968322754,
|
97 |
+
"eval_rewards/rejected": -0.8832504749298096,
|
98 |
+
"eval_runtime": 76.7395,
|
99 |
+
"eval_samples_per_second": 2.228,
|
100 |
+
"eval_steps_per_second": 1.121,
|
101 |
+
"step": 40
|
102 |
+
},
|
103 |
+
{
|
104 |
+
"epoch": 0.5228758169934641,
|
105 |
+
"grad_norm": 2.5461719036102295,
|
106 |
+
"learning_rate": 5.652173913043479e-06,
|
107 |
+
"logits/chosen": -0.16493651270866394,
|
108 |
+
"logits/rejected": -0.16144290566444397,
|
109 |
+
"logps/chosen": -24.76712989807129,
|
110 |
+
"logps/rejected": -46.993003845214844,
|
111 |
+
"loss": 0.3648,
|
112 |
+
"rewards/accuracies": 0.875,
|
113 |
+
"rewards/chosen": 1.5913469791412354,
|
114 |
+
"rewards/margins": 2.25089955329895,
|
115 |
+
"rewards/rejected": -0.6595526933670044,
|
116 |
+
"step": 50
|
117 |
+
},
|
118 |
+
{
|
119 |
+
"epoch": 0.6274509803921569,
|
120 |
+
"grad_norm": 2.0358729362487793,
|
121 |
+
"learning_rate": 5.543478260869566e-06,
|
122 |
+
"logits/chosen": -0.2202371060848236,
|
123 |
+
"logits/rejected": -0.21734721958637238,
|
124 |
+
"logps/chosen": -26.78680419921875,
|
125 |
+
"logps/rejected": -47.03938674926758,
|
126 |
+
"loss": 0.3448,
|
127 |
+
"rewards/accuracies": 0.831250011920929,
|
128 |
+
"rewards/chosen": 1.3999578952789307,
|
129 |
+
"rewards/margins": 2.0680160522460938,
|
130 |
+
"rewards/rejected": -0.6680583357810974,
|
131 |
+
"step": 60
|
132 |
+
},
|
133 |
+
{
|
134 |
+
"epoch": 0.6274509803921569,
|
135 |
+
"eval_logits/chosen": -0.17944073677062988,
|
136 |
+
"eval_logits/rejected": -0.1763658970594406,
|
137 |
+
"eval_logps/chosen": -27.3469295501709,
|
138 |
+
"eval_logps/rejected": -48.937984466552734,
|
139 |
+
"eval_loss": 0.2700035870075226,
|
140 |
+
"eval_rewards/accuracies": 0.9069767594337463,
|
141 |
+
"eval_rewards/chosen": 1.3058594465255737,
|
142 |
+
"eval_rewards/margins": 2.191838026046753,
|
143 |
+
"eval_rewards/rejected": -0.8859787583351135,
|
144 |
+
"eval_runtime": 76.7029,
|
145 |
+
"eval_samples_per_second": 2.229,
|
146 |
+
"eval_steps_per_second": 1.121,
|
147 |
+
"step": 60
|
148 |
+
},
|
149 |
+
{
|
150 |
+
"epoch": 0.7320261437908496,
|
151 |
+
"grad_norm": 1.7749762535095215,
|
152 |
+
"learning_rate": 5.4347826086956525e-06,
|
153 |
+
"logits/chosen": -0.1932402402162552,
|
154 |
+
"logits/rejected": -0.19045117497444153,
|
155 |
+
"logps/chosen": -26.827350616455078,
|
156 |
+
"logps/rejected": -46.88459014892578,
|
157 |
+
"loss": 0.2909,
|
158 |
+
"rewards/accuracies": 0.8812500238418579,
|
159 |
+
"rewards/chosen": 1.3978151082992554,
|
160 |
+
"rewards/margins": 2.0628795623779297,
|
161 |
+
"rewards/rejected": -0.6650643348693848,
|
162 |
+
"step": 70
|
163 |
+
},
|
164 |
+
{
|
165 |
+
"epoch": 0.8366013071895425,
|
166 |
+
"grad_norm": 2.9730594158172607,
|
167 |
+
"learning_rate": 5.326086956521739e-06,
|
168 |
+
"logits/chosen": -0.15139932930469513,
|
169 |
+
"logits/rejected": -0.14830470085144043,
|
170 |
+
"logps/chosen": -26.229084014892578,
|
171 |
+
"logps/rejected": -47.40288162231445,
|
172 |
+
"loss": 0.2844,
|
173 |
+
"rewards/accuracies": 0.875,
|
174 |
+
"rewards/chosen": 1.493841290473938,
|
175 |
+
"rewards/margins": 2.2263193130493164,
|
176 |
+
"rewards/rejected": -0.7324780225753784,
|
177 |
+
"step": 80
|
178 |
+
},
|
179 |
+
{
|
180 |
+
"epoch": 0.8366013071895425,
|
181 |
+
"eval_logits/chosen": -0.15539461374282837,
|
182 |
+
"eval_logits/rejected": -0.15173271298408508,
|
183 |
+
"eval_logps/chosen": -30.73025894165039,
|
184 |
+
"eval_logps/rejected": -54.970027923583984,
|
185 |
+
"eval_loss": 0.262494832277298,
|
186 |
+
"eval_rewards/accuracies": 0.9069767594337463,
|
187 |
+
"eval_rewards/chosen": 0.967526912689209,
|
188 |
+
"eval_rewards/margins": 2.4567105770111084,
|
189 |
+
"eval_rewards/rejected": -1.4891836643218994,
|
190 |
+
"eval_runtime": 76.69,
|
191 |
+
"eval_samples_per_second": 2.23,
|
192 |
+
"eval_steps_per_second": 1.121,
|
193 |
+
"step": 80
|
194 |
+
}
|
195 |
+
],
|
196 |
+
"logging_steps": 10,
|
197 |
+
"max_steps": 570,
|
198 |
+
"num_input_tokens_seen": 0,
|
199 |
+
"num_train_epochs": 6,
|
200 |
+
"save_steps": 80,
|
201 |
+
"stateful_callbacks": {
|
202 |
+
"TrainerControl": {
|
203 |
+
"args": {
|
204 |
+
"should_epoch_stop": false,
|
205 |
+
"should_evaluate": false,
|
206 |
+
"should_log": false,
|
207 |
+
"should_save": true,
|
208 |
+
"should_training_stop": false
|
209 |
+
},
|
210 |
+
"attributes": {}
|
211 |
+
}
|
212 |
+
},
|
213 |
+
"total_flos": 0.0,
|
214 |
+
"train_batch_size": 2,
|
215 |
+
"trial_name": null,
|
216 |
+
"trial_params": null
|
217 |
+
}
|