"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-33843 → checkpoint-34466}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-33843 → checkpoint-34466}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-33843 → checkpoint-34466}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-33843 → checkpoint-34466}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-33843 → checkpoint-34466}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-33843 → checkpoint-34466}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-33843 → checkpoint-34466}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-33843 → checkpoint-34466}/trainer_state.json +420 -3
- model-bin/finetune/base/{checkpoint-33843 → checkpoint-34466}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629651690.711412/events.out.tfevents.1629651690.91f414e9573a.903.181 +3 -0
- model-bin/finetune/base/log/1629652353.4624758/events.out.tfevents.1629652353.91f414e9573a.903.183 +3 -0
- model-bin/finetune/base/log/1629653002.462609/events.out.tfevents.1629653002.91f414e9573a.903.185 +3 -0
- model-bin/finetune/base/log/1629653643.1680498/events.out.tfevents.1629653643.91f414e9573a.903.187 +3 -0
- model-bin/finetune/base/log/1629654289.2915192/events.out.tfevents.1629654289.91f414e9573a.903.189 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629651690.91f414e9573a.903.180 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629652353.91f414e9573a.903.182 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629653002.91f414e9573a.903.184 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629653643.91f414e9573a.903.186 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629654289.91f414e9573a.903.188 +3 -0
model-bin/finetune/base/{checkpoint-33843 → checkpoint-34466}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-33843 → checkpoint-34466}/optimizer.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 722165009
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:54e5d4b62d076d7706403c43f3d911dda14a24fed0b62e4dfe4a411b67794ae4
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-33843 → checkpoint-34466}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-33843 → checkpoint-34466}/pytorch_model.bin
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 377909911
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2870fa68c9327c358c08db8f684ba8b84bb32e6784a09cbfb21e74d21e2239ee
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-33843 → checkpoint-34466}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 14503
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c3abf278e08e091b125ef1cfed6a86bdb4e4f34ccb3045ada38644ba277dae7b
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-33843 → checkpoint-34466}/scaler.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 559
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c862b9c320e4bf6b87cd21078da8c93e03620b480fd3d470ebcd9d2b60dd8a6c
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-33843 → checkpoint-34466}/scheduler.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 623
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:782694dffae77e741d7c7074465d7dd4963fd0a01bf72db18a10f713edf05a2b
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-33843 → checkpoint-34466}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.20553532410779315,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-31352",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -145038,11 +145038,428 @@
|
|
| 145038 |
"eval_steps_per_second": 0.655,
|
| 145039 |
"eval_wer": 0.21317914030111282,
|
| 145040 |
"step": 33843
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 145041 |
}
|
| 145042 |
],
|
| 145043 |
"max_steps": 625000,
|
| 145044 |
"num_train_epochs": 5000,
|
| 145045 |
-
"total_flos": 9.
|
| 145046 |
"trial_name": null,
|
| 145047 |
"trial_params": null
|
| 145048 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.20553532410779315,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-31352",
|
| 4 |
+
"epoch": 275.0,
|
| 5 |
+
"global_step": 34466,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 145038 |
"eval_steps_per_second": 0.655,
|
| 145039 |
"eval_wer": 0.21317914030111282,
|
| 145040 |
"step": 33843
|
| 145041 |
+
},
|
| 145042 |
+
{
|
| 145043 |
+
"epoch": 272.06,
|
| 145044 |
+
"learning_rate": 9.46951534733441e-06,
|
| 145045 |
+
"loss": 0.4285,
|
| 145046 |
+
"step": 33850
|
| 145047 |
+
},
|
| 145048 |
+
{
|
| 145049 |
+
"epoch": 272.14,
|
| 145050 |
+
"learning_rate": 9.46935379644588e-06,
|
| 145051 |
+
"loss": 0.4423,
|
| 145052 |
+
"step": 33860
|
| 145053 |
+
},
|
| 145054 |
+
{
|
| 145055 |
+
"epoch": 272.22,
|
| 145056 |
+
"learning_rate": 9.46919224555735e-06,
|
| 145057 |
+
"loss": 0.8837,
|
| 145058 |
+
"step": 33870
|
| 145059 |
+
},
|
| 145060 |
+
{
|
| 145061 |
+
"epoch": 272.3,
|
| 145062 |
+
"learning_rate": 9.469030694668822e-06,
|
| 145063 |
+
"loss": 0.3521,
|
| 145064 |
+
"step": 33880
|
| 145065 |
+
},
|
| 145066 |
+
{
|
| 145067 |
+
"epoch": 272.38,
|
| 145068 |
+
"learning_rate": 9.468869143780292e-06,
|
| 145069 |
+
"loss": 0.6718,
|
| 145070 |
+
"step": 33890
|
| 145071 |
+
},
|
| 145072 |
+
{
|
| 145073 |
+
"epoch": 272.46,
|
| 145074 |
+
"learning_rate": 9.468707592891762e-06,
|
| 145075 |
+
"loss": 0.7548,
|
| 145076 |
+
"step": 33900
|
| 145077 |
+
},
|
| 145078 |
+
{
|
| 145079 |
+
"epoch": 272.54,
|
| 145080 |
+
"learning_rate": 9.468546042003232e-06,
|
| 145081 |
+
"loss": 0.4065,
|
| 145082 |
+
"step": 33910
|
| 145083 |
+
},
|
| 145084 |
+
{
|
| 145085 |
+
"epoch": 272.62,
|
| 145086 |
+
"learning_rate": 9.468384491114702e-06,
|
| 145087 |
+
"loss": 0.9252,
|
| 145088 |
+
"step": 33920
|
| 145089 |
+
},
|
| 145090 |
+
{
|
| 145091 |
+
"epoch": 272.7,
|
| 145092 |
+
"learning_rate": 9.468222940226172e-06,
|
| 145093 |
+
"loss": 0.3799,
|
| 145094 |
+
"step": 33930
|
| 145095 |
+
},
|
| 145096 |
+
{
|
| 145097 |
+
"epoch": 272.78,
|
| 145098 |
+
"learning_rate": 9.468061389337642e-06,
|
| 145099 |
+
"loss": 0.6615,
|
| 145100 |
+
"step": 33940
|
| 145101 |
+
},
|
| 145102 |
+
{
|
| 145103 |
+
"epoch": 272.86,
|
| 145104 |
+
"learning_rate": 9.467899838449112e-06,
|
| 145105 |
+
"loss": 0.7824,
|
| 145106 |
+
"step": 33950
|
| 145107 |
+
},
|
| 145108 |
+
{
|
| 145109 |
+
"epoch": 272.94,
|
| 145110 |
+
"learning_rate": 9.467738287560582e-06,
|
| 145111 |
+
"loss": 0.3931,
|
| 145112 |
+
"step": 33960
|
| 145113 |
+
},
|
| 145114 |
+
{
|
| 145115 |
+
"epoch": 273.0,
|
| 145116 |
+
"eval_loss": 0.4480145573616028,
|
| 145117 |
+
"eval_runtime": 41.5679,
|
| 145118 |
+
"eval_samples_per_second": 20.184,
|
| 145119 |
+
"eval_steps_per_second": 0.65,
|
| 145120 |
+
"eval_wer": 0.2201854962389542,
|
| 145121 |
+
"step": 33967
|
| 145122 |
+
},
|
| 145123 |
+
{
|
| 145124 |
+
"epoch": 271.02,
|
| 145125 |
+
"learning_rate": 9.467576736672052e-06,
|
| 145126 |
+
"loss": 0.4824,
|
| 145127 |
+
"step": 33970
|
| 145128 |
+
},
|
| 145129 |
+
{
|
| 145130 |
+
"epoch": 271.1,
|
| 145131 |
+
"learning_rate": 9.467415185783522e-06,
|
| 145132 |
+
"loss": 0.3883,
|
| 145133 |
+
"step": 33980
|
| 145134 |
+
},
|
| 145135 |
+
{
|
| 145136 |
+
"epoch": 271.18,
|
| 145137 |
+
"learning_rate": 9.467253634894992e-06,
|
| 145138 |
+
"loss": 0.6886,
|
| 145139 |
+
"step": 33990
|
| 145140 |
+
},
|
| 145141 |
+
{
|
| 145142 |
+
"epoch": 271.26,
|
| 145143 |
+
"learning_rate": 9.467092084006462e-06,
|
| 145144 |
+
"loss": 0.7794,
|
| 145145 |
+
"step": 34000
|
| 145146 |
+
},
|
| 145147 |
+
{
|
| 145148 |
+
"epoch": 271.34,
|
| 145149 |
+
"learning_rate": 9.466930533117932e-06,
|
| 145150 |
+
"loss": 0.4534,
|
| 145151 |
+
"step": 34010
|
| 145152 |
+
},
|
| 145153 |
+
{
|
| 145154 |
+
"epoch": 271.42,
|
| 145155 |
+
"learning_rate": 9.466768982229404e-06,
|
| 145156 |
+
"loss": 0.9204,
|
| 145157 |
+
"step": 34020
|
| 145158 |
+
},
|
| 145159 |
+
{
|
| 145160 |
+
"epoch": 271.5,
|
| 145161 |
+
"learning_rate": 9.466607431340874e-06,
|
| 145162 |
+
"loss": 0.4433,
|
| 145163 |
+
"step": 34030
|
| 145164 |
+
},
|
| 145165 |
+
{
|
| 145166 |
+
"epoch": 271.58,
|
| 145167 |
+
"learning_rate": 9.466445880452344e-06,
|
| 145168 |
+
"loss": 0.6585,
|
| 145169 |
+
"step": 34040
|
| 145170 |
+
},
|
| 145171 |
+
{
|
| 145172 |
+
"epoch": 271.66,
|
| 145173 |
+
"learning_rate": 9.466284329563814e-06,
|
| 145174 |
+
"loss": 0.7418,
|
| 145175 |
+
"step": 34050
|
| 145176 |
+
},
|
| 145177 |
+
{
|
| 145178 |
+
"epoch": 271.74,
|
| 145179 |
+
"learning_rate": 9.466122778675284e-06,
|
| 145180 |
+
"loss": 0.4903,
|
| 145181 |
+
"step": 34060
|
| 145182 |
+
},
|
| 145183 |
+
{
|
| 145184 |
+
"epoch": 271.82,
|
| 145185 |
+
"learning_rate": 9.465961227786754e-06,
|
| 145186 |
+
"loss": 0.9372,
|
| 145187 |
+
"step": 34070
|
| 145188 |
+
},
|
| 145189 |
+
{
|
| 145190 |
+
"epoch": 271.9,
|
| 145191 |
+
"learning_rate": 9.465799676898224e-06,
|
| 145192 |
+
"loss": 0.3894,
|
| 145193 |
+
"step": 34080
|
| 145194 |
+
},
|
| 145195 |
+
{
|
| 145196 |
+
"epoch": 271.98,
|
| 145197 |
+
"learning_rate": 9.465638126009694e-06,
|
| 145198 |
+
"loss": 0.8142,
|
| 145199 |
+
"step": 34090
|
| 145200 |
+
},
|
| 145201 |
+
{
|
| 145202 |
+
"epoch": 272.0,
|
| 145203 |
+
"eval_loss": 0.4223657548427582,
|
| 145204 |
+
"eval_runtime": 40.6428,
|
| 145205 |
+
"eval_samples_per_second": 20.668,
|
| 145206 |
+
"eval_steps_per_second": 0.664,
|
| 145207 |
+
"eval_wer": 0.2171622401847575,
|
| 145208 |
+
"step": 34092
|
| 145209 |
+
},
|
| 145210 |
+
{
|
| 145211 |
+
"epoch": 272.06,
|
| 145212 |
+
"learning_rate": 9.465476575121163e-06,
|
| 145213 |
+
"loss": 0.4175,
|
| 145214 |
+
"step": 34100
|
| 145215 |
+
},
|
| 145216 |
+
{
|
| 145217 |
+
"epoch": 272.14,
|
| 145218 |
+
"learning_rate": 9.465315024232633e-06,
|
| 145219 |
+
"loss": 0.8738,
|
| 145220 |
+
"step": 34110
|
| 145221 |
+
},
|
| 145222 |
+
{
|
| 145223 |
+
"epoch": 272.22,
|
| 145224 |
+
"learning_rate": 9.465153473344103e-06,
|
| 145225 |
+
"loss": 0.9068,
|
| 145226 |
+
"step": 34120
|
| 145227 |
+
},
|
| 145228 |
+
{
|
| 145229 |
+
"epoch": 272.3,
|
| 145230 |
+
"learning_rate": 9.464991922455573e-06,
|
| 145231 |
+
"loss": 0.3807,
|
| 145232 |
+
"step": 34130
|
| 145233 |
+
},
|
| 145234 |
+
{
|
| 145235 |
+
"epoch": 272.38,
|
| 145236 |
+
"learning_rate": 9.464830371567043e-06,
|
| 145237 |
+
"loss": 0.7017,
|
| 145238 |
+
"step": 34140
|
| 145239 |
+
},
|
| 145240 |
+
{
|
| 145241 |
+
"epoch": 272.46,
|
| 145242 |
+
"learning_rate": 9.464668820678515e-06,
|
| 145243 |
+
"loss": 0.6221,
|
| 145244 |
+
"step": 34150
|
| 145245 |
+
},
|
| 145246 |
+
{
|
| 145247 |
+
"epoch": 272.54,
|
| 145248 |
+
"learning_rate": 9.464507269789985e-06,
|
| 145249 |
+
"loss": 0.8577,
|
| 145250 |
+
"step": 34160
|
| 145251 |
+
},
|
| 145252 |
+
{
|
| 145253 |
+
"epoch": 272.62,
|
| 145254 |
+
"learning_rate": 9.464345718901455e-06,
|
| 145255 |
+
"loss": 1.0584,
|
| 145256 |
+
"step": 34170
|
| 145257 |
+
},
|
| 145258 |
+
{
|
| 145259 |
+
"epoch": 272.7,
|
| 145260 |
+
"learning_rate": 9.464184168012925e-06,
|
| 145261 |
+
"loss": 0.3823,
|
| 145262 |
+
"step": 34180
|
| 145263 |
+
},
|
| 145264 |
+
{
|
| 145265 |
+
"epoch": 272.78,
|
| 145266 |
+
"learning_rate": 9.464022617124395e-06,
|
| 145267 |
+
"loss": 0.6333,
|
| 145268 |
+
"step": 34190
|
| 145269 |
+
},
|
| 145270 |
+
{
|
| 145271 |
+
"epoch": 272.86,
|
| 145272 |
+
"learning_rate": 9.463861066235865e-06,
|
| 145273 |
+
"loss": 0.6998,
|
| 145274 |
+
"step": 34200
|
| 145275 |
+
},
|
| 145276 |
+
{
|
| 145277 |
+
"epoch": 272.94,
|
| 145278 |
+
"learning_rate": 9.463699515347335e-06,
|
| 145279 |
+
"loss": 0.4155,
|
| 145280 |
+
"step": 34210
|
| 145281 |
+
},
|
| 145282 |
+
{
|
| 145283 |
+
"epoch": 273.0,
|
| 145284 |
+
"eval_loss": 0.4414938688278198,
|
| 145285 |
+
"eval_runtime": 40.0649,
|
| 145286 |
+
"eval_samples_per_second": 20.966,
|
| 145287 |
+
"eval_steps_per_second": 0.674,
|
| 145288 |
+
"eval_wer": 0.2120436086924709,
|
| 145289 |
+
"step": 34217
|
| 145290 |
+
},
|
| 145291 |
+
{
|
| 145292 |
+
"epoch": 275.02,
|
| 145293 |
+
"learning_rate": 9.463537964458805e-06,
|
| 145294 |
+
"loss": 0.4167,
|
| 145295 |
+
"step": 34220
|
| 145296 |
+
},
|
| 145297 |
+
{
|
| 145298 |
+
"epoch": 275.1,
|
| 145299 |
+
"learning_rate": 9.463376413570275e-06,
|
| 145300 |
+
"loss": 0.3276,
|
| 145301 |
+
"step": 34230
|
| 145302 |
+
},
|
| 145303 |
+
{
|
| 145304 |
+
"epoch": 275.18,
|
| 145305 |
+
"learning_rate": 9.463214862681745e-06,
|
| 145306 |
+
"loss": 0.6905,
|
| 145307 |
+
"step": 34240
|
| 145308 |
+
},
|
| 145309 |
+
{
|
| 145310 |
+
"epoch": 275.27,
|
| 145311 |
+
"learning_rate": 9.463053311793215e-06,
|
| 145312 |
+
"loss": 0.723,
|
| 145313 |
+
"step": 34250
|
| 145314 |
+
},
|
| 145315 |
+
{
|
| 145316 |
+
"epoch": 275.35,
|
| 145317 |
+
"learning_rate": 9.462891760904685e-06,
|
| 145318 |
+
"loss": 0.4073,
|
| 145319 |
+
"step": 34260
|
| 145320 |
+
},
|
| 145321 |
+
{
|
| 145322 |
+
"epoch": 275.43,
|
| 145323 |
+
"learning_rate": 9.462730210016155e-06,
|
| 145324 |
+
"loss": 0.9286,
|
| 145325 |
+
"step": 34270
|
| 145326 |
+
},
|
| 145327 |
+
{
|
| 145328 |
+
"epoch": 275.51,
|
| 145329 |
+
"learning_rate": 9.462568659127625e-06,
|
| 145330 |
+
"loss": 0.3704,
|
| 145331 |
+
"step": 34280
|
| 145332 |
+
},
|
| 145333 |
+
{
|
| 145334 |
+
"epoch": 275.59,
|
| 145335 |
+
"learning_rate": 9.462407108239096e-06,
|
| 145336 |
+
"loss": 0.6891,
|
| 145337 |
+
"step": 34290
|
| 145338 |
+
},
|
| 145339 |
+
{
|
| 145340 |
+
"epoch": 275.67,
|
| 145341 |
+
"learning_rate": 9.462245557350566e-06,
|
| 145342 |
+
"loss": 0.7574,
|
| 145343 |
+
"step": 34300
|
| 145344 |
+
},
|
| 145345 |
+
{
|
| 145346 |
+
"epoch": 275.75,
|
| 145347 |
+
"learning_rate": 9.462084006462036e-06,
|
| 145348 |
+
"loss": 0.4023,
|
| 145349 |
+
"step": 34310
|
| 145350 |
+
},
|
| 145351 |
+
{
|
| 145352 |
+
"epoch": 275.83,
|
| 145353 |
+
"learning_rate": 9.461922455573506e-06,
|
| 145354 |
+
"loss": 1.016,
|
| 145355 |
+
"step": 34320
|
| 145356 |
+
},
|
| 145357 |
+
{
|
| 145358 |
+
"epoch": 275.91,
|
| 145359 |
+
"learning_rate": 9.461760904684976e-06,
|
| 145360 |
+
"loss": 0.4436,
|
| 145361 |
+
"step": 34330
|
| 145362 |
+
},
|
| 145363 |
+
{
|
| 145364 |
+
"epoch": 275.99,
|
| 145365 |
+
"learning_rate": 9.461599353796446e-06,
|
| 145366 |
+
"loss": 0.7421,
|
| 145367 |
+
"step": 34340
|
| 145368 |
+
},
|
| 145369 |
+
{
|
| 145370 |
+
"epoch": 276.0,
|
| 145371 |
+
"eval_loss": 0.5068224668502808,
|
| 145372 |
+
"eval_runtime": 39.7733,
|
| 145373 |
+
"eval_samples_per_second": 21.12,
|
| 145374 |
+
"eval_steps_per_second": 0.679,
|
| 145375 |
+
"eval_wer": 0.2221556437720021,
|
| 145376 |
+
"step": 34341
|
| 145377 |
+
},
|
| 145378 |
+
{
|
| 145379 |
+
"epoch": 274.07,
|
| 145380 |
+
"learning_rate": 9.461437802907916e-06,
|
| 145381 |
+
"loss": 0.4031,
|
| 145382 |
+
"step": 34350
|
| 145383 |
+
},
|
| 145384 |
+
{
|
| 145385 |
+
"epoch": 274.15,
|
| 145386 |
+
"learning_rate": 9.461276252019386e-06,
|
| 145387 |
+
"loss": 0.4577,
|
| 145388 |
+
"step": 34360
|
| 145389 |
+
},
|
| 145390 |
+
{
|
| 145391 |
+
"epoch": 274.23,
|
| 145392 |
+
"learning_rate": 9.461114701130856e-06,
|
| 145393 |
+
"loss": 1.0469,
|
| 145394 |
+
"step": 34370
|
| 145395 |
+
},
|
| 145396 |
+
{
|
| 145397 |
+
"epoch": 274.31,
|
| 145398 |
+
"learning_rate": 9.460953150242326e-06,
|
| 145399 |
+
"loss": 0.4011,
|
| 145400 |
+
"step": 34380
|
| 145401 |
+
},
|
| 145402 |
+
{
|
| 145403 |
+
"epoch": 274.39,
|
| 145404 |
+
"learning_rate": 9.460791599353796e-06,
|
| 145405 |
+
"loss": 0.8314,
|
| 145406 |
+
"step": 34390
|
| 145407 |
+
},
|
| 145408 |
+
{
|
| 145409 |
+
"epoch": 274.47,
|
| 145410 |
+
"learning_rate": 9.460630048465266e-06,
|
| 145411 |
+
"loss": 0.5669,
|
| 145412 |
+
"step": 34400
|
| 145413 |
+
},
|
| 145414 |
+
{
|
| 145415 |
+
"epoch": 274.55,
|
| 145416 |
+
"learning_rate": 9.460468497576736e-06,
|
| 145417 |
+
"loss": 0.4753,
|
| 145418 |
+
"step": 34410
|
| 145419 |
+
},
|
| 145420 |
+
{
|
| 145421 |
+
"epoch": 274.63,
|
| 145422 |
+
"learning_rate": 9.460306946688208e-06,
|
| 145423 |
+
"loss": 1.0193,
|
| 145424 |
+
"step": 34420
|
| 145425 |
+
},
|
| 145426 |
+
{
|
| 145427 |
+
"epoch": 274.71,
|
| 145428 |
+
"learning_rate": 9.460145395799678e-06,
|
| 145429 |
+
"loss": 0.3737,
|
| 145430 |
+
"step": 34430
|
| 145431 |
+
},
|
| 145432 |
+
{
|
| 145433 |
+
"epoch": 274.79,
|
| 145434 |
+
"learning_rate": 9.459983844911148e-06,
|
| 145435 |
+
"loss": 0.8249,
|
| 145436 |
+
"step": 34440
|
| 145437 |
+
},
|
| 145438 |
+
{
|
| 145439 |
+
"epoch": 274.87,
|
| 145440 |
+
"learning_rate": 9.459822294022618e-06,
|
| 145441 |
+
"loss": 0.6964,
|
| 145442 |
+
"step": 34450
|
| 145443 |
+
},
|
| 145444 |
+
{
|
| 145445 |
+
"epoch": 274.95,
|
| 145446 |
+
"learning_rate": 9.459660743134088e-06,
|
| 145447 |
+
"loss": 0.5363,
|
| 145448 |
+
"step": 34460
|
| 145449 |
+
},
|
| 145450 |
+
{
|
| 145451 |
+
"epoch": 275.0,
|
| 145452 |
+
"eval_loss": 0.3939749002456665,
|
| 145453 |
+
"eval_runtime": 40.4344,
|
| 145454 |
+
"eval_samples_per_second": 20.774,
|
| 145455 |
+
"eval_steps_per_second": 0.668,
|
| 145456 |
+
"eval_wer": 0.21125762087055153,
|
| 145457 |
+
"step": 34466
|
| 145458 |
}
|
| 145459 |
],
|
| 145460 |
"max_steps": 625000,
|
| 145461 |
"num_train_epochs": 5000,
|
| 145462 |
+
"total_flos": 9.69828256252031e+19,
|
| 145463 |
"trial_name": null,
|
| 145464 |
"trial_params": null
|
| 145465 |
}
|
model-bin/finetune/base/{checkpoint-33843 → checkpoint-34466}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629651690.711412/events.out.tfevents.1629651690.91f414e9573a.903.181
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7e40cbccf4e2bd784825a3333ded87496abf87dcb5538a2d6a3096891b0e6c08
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629652353.4624758/events.out.tfevents.1629652353.91f414e9573a.903.183
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:eaecbaccbb6548734fdd8aba1693bd22fd2aab7b7f1e3aa835e7ed802cfe8fb6
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629653002.462609/events.out.tfevents.1629653002.91f414e9573a.903.185
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d3910f99e70ab0a2617690abcda86d33d87b647cb8a84a76a94e945121ef169e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629653643.1680498/events.out.tfevents.1629653643.91f414e9573a.903.187
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5adfa1ff250f446e5551005731915becc132340b9f84dcb3c760a0bbcdd2f41a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629654289.2915192/events.out.tfevents.1629654289.91f414e9573a.903.189
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2ee3bb62d4cefac710c267b4c52894c24c85aa25acd2c1c833a00e214cca1f2a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629651690.91f414e9573a.903.180
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:681d572cb6a4d30b276153732c6ad802af1dce2b8db27543469fd197c9959ff6
|
| 3 |
+
size 6543
|
model-bin/finetune/base/log/events.out.tfevents.1629652353.91f414e9573a.903.182
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:620cdac77e4d5a20b6aabcc8d78ce5d294418449aa933ab77b7b5ba934460e43
|
| 3 |
+
size 6703
|
model-bin/finetune/base/log/events.out.tfevents.1629653002.91f414e9573a.903.184
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9cf3135632d41691fce9633edaa32a1dba8690a2436acc4d2df339138a4b2205
|
| 3 |
+
size 6543
|
model-bin/finetune/base/log/events.out.tfevents.1629653643.91f414e9573a.903.186
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5b85c3047016e15472ccddb7dafe76de089b1071f5a96430fdac9f20aa989778
|
| 3 |
+
size 6703
|
model-bin/finetune/base/log/events.out.tfevents.1629654289.91f414e9573a.903.188
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:770765a58801538456f04ba2d50b7c969379e8d354990f85390526e50545308a
|
| 3 |
+
size 6543
|