"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-90224 → checkpoint-90847}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-90224 → checkpoint-90847}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-90224 → checkpoint-90847}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-90224 → checkpoint-90847}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-90224 → checkpoint-90847}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-90224 → checkpoint-90847}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-90224 → checkpoint-90847}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-90224 → checkpoint-90847}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-90224 → checkpoint-90847}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629946430.8098779/events.out.tfevents.1629946430.8e89bd551565.924.31 +3 -0
- model-bin/finetune/base/log/1629946866.7205796/events.out.tfevents.1629946866.8e89bd551565.924.33 +3 -0
- model-bin/finetune/base/log/1629947293.5282943/events.out.tfevents.1629947293.8e89bd551565.924.35 +3 -0
- model-bin/finetune/base/log/1629947731.4501543/events.out.tfevents.1629947731.8e89bd551565.924.37 +3 -0
- model-bin/finetune/base/log/1629948172.226034/events.out.tfevents.1629948172.8e89bd551565.924.39 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629946430.8e89bd551565.924.30 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629946865.8e89bd551565.924.32 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629947293.8e89bd551565.924.34 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629947731.8e89bd551565.924.36 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629948172.8e89bd551565.924.38 +3 -0
model-bin/finetune/base/{checkpoint-90224 → checkpoint-90847}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-90224 → checkpoint-90847}/optimizer.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 722165393
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:059f1becf8b39dcee76e6892a2b081bc70757f47a310129d0b5ea65b31bf37ad
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-90224 → checkpoint-90847}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-90224 → checkpoint-90847}/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 sha256:5fb13c34fb11dd331464d813a241c63c38b52eb802c711b52f064013c38883db
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-90224 → checkpoint-90847}/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 sha256:04abce0d0284c6da65004c6da2c9ba91b9646db5efd6f5375a9b927541f4a0fd
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-90224 → checkpoint-90847}/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 sha256:1a28146e60e7482fd8bf1168040f8ce03267a19d616aee5d2b7fadfd2ef5a1a7
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-90224 → checkpoint-90847}/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 sha256:597f477ce136a0d19546bbeaf1d90265ad546212e55347ca9da84764d96083ef
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-90224 → checkpoint-90847}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17731766728757703,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-88857",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -214155,11 +214155,806 @@
|
|
| 214155 |
"eval_steps_per_second": 0.685,
|
| 214156 |
"eval_wer": 0.19095368633674795,
|
| 214157 |
"step": 90224
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 214158 |
}
|
| 214159 |
],
|
| 214160 |
-
"max_steps":
|
| 214161 |
"num_train_epochs": 5000,
|
| 214162 |
-
"total_flos": 2.
|
| 214163 |
"trial_name": null,
|
| 214164 |
"trial_params": null
|
| 214165 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17731766728757703,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-88857",
|
| 4 |
+
"epoch": 731.995983935743,
|
| 5 |
+
"global_step": 90847,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 214155 |
"eval_steps_per_second": 0.685,
|
| 214156 |
"eval_wer": 0.19095368633674795,
|
| 214157 |
"step": 90224
|
| 214158 |
+
},
|
| 214159 |
+
{
|
| 214160 |
+
"epoch": 721.01,
|
| 214161 |
+
"learning_rate": 8.570689102564103e-06,
|
| 214162 |
+
"loss": 0.3959,
|
| 214163 |
+
"step": 90225
|
| 214164 |
+
},
|
| 214165 |
+
{
|
| 214166 |
+
"epoch": 721.05,
|
| 214167 |
+
"learning_rate": 8.570608974358975e-06,
|
| 214168 |
+
"loss": 0.3308,
|
| 214169 |
+
"step": 90230
|
| 214170 |
+
},
|
| 214171 |
+
{
|
| 214172 |
+
"epoch": 721.09,
|
| 214173 |
+
"learning_rate": 8.570528846153848e-06,
|
| 214174 |
+
"loss": 0.314,
|
| 214175 |
+
"step": 90235
|
| 214176 |
+
},
|
| 214177 |
+
{
|
| 214178 |
+
"epoch": 721.13,
|
| 214179 |
+
"learning_rate": 8.570448717948719e-06,
|
| 214180 |
+
"loss": 0.318,
|
| 214181 |
+
"step": 90240
|
| 214182 |
+
},
|
| 214183 |
+
{
|
| 214184 |
+
"epoch": 721.17,
|
| 214185 |
+
"learning_rate": 8.57036858974359e-06,
|
| 214186 |
+
"loss": 0.5316,
|
| 214187 |
+
"step": 90245
|
| 214188 |
+
},
|
| 214189 |
+
{
|
| 214190 |
+
"epoch": 721.21,
|
| 214191 |
+
"learning_rate": 8.570288461538462e-06,
|
| 214192 |
+
"loss": 1.0654,
|
| 214193 |
+
"step": 90250
|
| 214194 |
+
},
|
| 214195 |
+
{
|
| 214196 |
+
"epoch": 721.25,
|
| 214197 |
+
"learning_rate": 8.570208333333335e-06,
|
| 214198 |
+
"loss": 0.3506,
|
| 214199 |
+
"step": 90255
|
| 214200 |
+
},
|
| 214201 |
+
{
|
| 214202 |
+
"epoch": 721.29,
|
| 214203 |
+
"learning_rate": 8.570128205128206e-06,
|
| 214204 |
+
"loss": 0.3143,
|
| 214205 |
+
"step": 90260
|
| 214206 |
+
},
|
| 214207 |
+
{
|
| 214208 |
+
"epoch": 721.33,
|
| 214209 |
+
"learning_rate": 8.570048076923077e-06,
|
| 214210 |
+
"loss": 0.2882,
|
| 214211 |
+
"step": 90265
|
| 214212 |
+
},
|
| 214213 |
+
{
|
| 214214 |
+
"epoch": 721.37,
|
| 214215 |
+
"learning_rate": 8.56996794871795e-06,
|
| 214216 |
+
"loss": 0.5043,
|
| 214217 |
+
"step": 90270
|
| 214218 |
+
},
|
| 214219 |
+
{
|
| 214220 |
+
"epoch": 721.41,
|
| 214221 |
+
"learning_rate": 8.569887820512822e-06,
|
| 214222 |
+
"loss": 1.2249,
|
| 214223 |
+
"step": 90275
|
| 214224 |
+
},
|
| 214225 |
+
{
|
| 214226 |
+
"epoch": 721.45,
|
| 214227 |
+
"learning_rate": 8.569807692307693e-06,
|
| 214228 |
+
"loss": 0.2814,
|
| 214229 |
+
"step": 90280
|
| 214230 |
+
},
|
| 214231 |
+
{
|
| 214232 |
+
"epoch": 721.49,
|
| 214233 |
+
"learning_rate": 8.569727564102565e-06,
|
| 214234 |
+
"loss": 0.3371,
|
| 214235 |
+
"step": 90285
|
| 214236 |
+
},
|
| 214237 |
+
{
|
| 214238 |
+
"epoch": 721.53,
|
| 214239 |
+
"learning_rate": 8.569647435897438e-06,
|
| 214240 |
+
"loss": 0.323,
|
| 214241 |
+
"step": 90290
|
| 214242 |
+
},
|
| 214243 |
+
{
|
| 214244 |
+
"epoch": 721.57,
|
| 214245 |
+
"learning_rate": 8.569567307692307e-06,
|
| 214246 |
+
"loss": 0.5027,
|
| 214247 |
+
"step": 90295
|
| 214248 |
+
},
|
| 214249 |
+
{
|
| 214250 |
+
"epoch": 721.61,
|
| 214251 |
+
"learning_rate": 8.56948717948718e-06,
|
| 214252 |
+
"loss": 0.9879,
|
| 214253 |
+
"step": 90300
|
| 214254 |
+
},
|
| 214255 |
+
{
|
| 214256 |
+
"epoch": 721.65,
|
| 214257 |
+
"learning_rate": 8.569407051282052e-06,
|
| 214258 |
+
"loss": 0.2975,
|
| 214259 |
+
"step": 90305
|
| 214260 |
+
},
|
| 214261 |
+
{
|
| 214262 |
+
"epoch": 721.69,
|
| 214263 |
+
"learning_rate": 8.569326923076923e-06,
|
| 214264 |
+
"loss": 0.2783,
|
| 214265 |
+
"step": 90310
|
| 214266 |
+
},
|
| 214267 |
+
{
|
| 214268 |
+
"epoch": 721.73,
|
| 214269 |
+
"learning_rate": 8.569246794871796e-06,
|
| 214270 |
+
"loss": 0.3157,
|
| 214271 |
+
"step": 90315
|
| 214272 |
+
},
|
| 214273 |
+
{
|
| 214274 |
+
"epoch": 721.77,
|
| 214275 |
+
"learning_rate": 8.569166666666667e-06,
|
| 214276 |
+
"loss": 0.5318,
|
| 214277 |
+
"step": 90320
|
| 214278 |
+
},
|
| 214279 |
+
{
|
| 214280 |
+
"epoch": 721.81,
|
| 214281 |
+
"learning_rate": 8.569086538461539e-06,
|
| 214282 |
+
"loss": 1.0942,
|
| 214283 |
+
"step": 90325
|
| 214284 |
+
},
|
| 214285 |
+
{
|
| 214286 |
+
"epoch": 721.85,
|
| 214287 |
+
"learning_rate": 8.56900641025641e-06,
|
| 214288 |
+
"loss": 0.3259,
|
| 214289 |
+
"step": 90330
|
| 214290 |
+
},
|
| 214291 |
+
{
|
| 214292 |
+
"epoch": 721.89,
|
| 214293 |
+
"learning_rate": 8.568926282051283e-06,
|
| 214294 |
+
"loss": 0.2652,
|
| 214295 |
+
"step": 90335
|
| 214296 |
+
},
|
| 214297 |
+
{
|
| 214298 |
+
"epoch": 721.93,
|
| 214299 |
+
"learning_rate": 8.568846153846155e-06,
|
| 214300 |
+
"loss": 0.3069,
|
| 214301 |
+
"step": 90340
|
| 214302 |
+
},
|
| 214303 |
+
{
|
| 214304 |
+
"epoch": 721.97,
|
| 214305 |
+
"learning_rate": 8.568766025641026e-06,
|
| 214306 |
+
"loss": 0.5629,
|
| 214307 |
+
"step": 90345
|
| 214308 |
+
},
|
| 214309 |
+
{
|
| 214310 |
+
"epoch": 722.0,
|
| 214311 |
+
"eval_loss": 0.39299121499061584,
|
| 214312 |
+
"eval_runtime": 39.1237,
|
| 214313 |
+
"eval_samples_per_second": 21.496,
|
| 214314 |
+
"eval_steps_per_second": 0.69,
|
| 214315 |
+
"eval_wer": 0.18836970851032664,
|
| 214316 |
+
"step": 90349
|
| 214317 |
+
},
|
| 214318 |
+
{
|
| 214319 |
+
"epoch": 722.01,
|
| 214320 |
+
"learning_rate": 8.568685897435897e-06,
|
| 214321 |
+
"loss": 0.5475,
|
| 214322 |
+
"step": 90350
|
| 214323 |
+
},
|
| 214324 |
+
{
|
| 214325 |
+
"epoch": 722.05,
|
| 214326 |
+
"learning_rate": 8.56860576923077e-06,
|
| 214327 |
+
"loss": 0.2935,
|
| 214328 |
+
"step": 90355
|
| 214329 |
+
},
|
| 214330 |
+
{
|
| 214331 |
+
"epoch": 722.09,
|
| 214332 |
+
"learning_rate": 8.568525641025642e-06,
|
| 214333 |
+
"loss": 0.306,
|
| 214334 |
+
"step": 90360
|
| 214335 |
+
},
|
| 214336 |
+
{
|
| 214337 |
+
"epoch": 722.13,
|
| 214338 |
+
"learning_rate": 8.568445512820513e-06,
|
| 214339 |
+
"loss": 0.3145,
|
| 214340 |
+
"step": 90365
|
| 214341 |
+
},
|
| 214342 |
+
{
|
| 214343 |
+
"epoch": 722.17,
|
| 214344 |
+
"learning_rate": 8.568365384615386e-06,
|
| 214345 |
+
"loss": 0.5436,
|
| 214346 |
+
"step": 90370
|
| 214347 |
+
},
|
| 214348 |
+
{
|
| 214349 |
+
"epoch": 722.21,
|
| 214350 |
+
"learning_rate": 8.568285256410258e-06,
|
| 214351 |
+
"loss": 1.26,
|
| 214352 |
+
"step": 90375
|
| 214353 |
+
},
|
| 214354 |
+
{
|
| 214355 |
+
"epoch": 722.25,
|
| 214356 |
+
"learning_rate": 8.568205128205129e-06,
|
| 214357 |
+
"loss": 0.295,
|
| 214358 |
+
"step": 90380
|
| 214359 |
+
},
|
| 214360 |
+
{
|
| 214361 |
+
"epoch": 722.29,
|
| 214362 |
+
"learning_rate": 8.568125e-06,
|
| 214363 |
+
"loss": 0.2853,
|
| 214364 |
+
"step": 90385
|
| 214365 |
+
},
|
| 214366 |
+
{
|
| 214367 |
+
"epoch": 722.33,
|
| 214368 |
+
"learning_rate": 8.568044871794873e-06,
|
| 214369 |
+
"loss": 0.3319,
|
| 214370 |
+
"step": 90390
|
| 214371 |
+
},
|
| 214372 |
+
{
|
| 214373 |
+
"epoch": 722.37,
|
| 214374 |
+
"learning_rate": 8.567964743589745e-06,
|
| 214375 |
+
"loss": 0.5341,
|
| 214376 |
+
"step": 90395
|
| 214377 |
+
},
|
| 214378 |
+
{
|
| 214379 |
+
"epoch": 722.41,
|
| 214380 |
+
"learning_rate": 8.567884615384616e-06,
|
| 214381 |
+
"loss": 1.1765,
|
| 214382 |
+
"step": 90400
|
| 214383 |
+
},
|
| 214384 |
+
{
|
| 214385 |
+
"epoch": 722.45,
|
| 214386 |
+
"learning_rate": 8.567804487179487e-06,
|
| 214387 |
+
"loss": 0.336,
|
| 214388 |
+
"step": 90405
|
| 214389 |
+
},
|
| 214390 |
+
{
|
| 214391 |
+
"epoch": 722.49,
|
| 214392 |
+
"learning_rate": 8.56772435897436e-06,
|
| 214393 |
+
"loss": 0.2876,
|
| 214394 |
+
"step": 90410
|
| 214395 |
+
},
|
| 214396 |
+
{
|
| 214397 |
+
"epoch": 722.53,
|
| 214398 |
+
"learning_rate": 8.567644230769232e-06,
|
| 214399 |
+
"loss": 0.3503,
|
| 214400 |
+
"step": 90415
|
| 214401 |
+
},
|
| 214402 |
+
{
|
| 214403 |
+
"epoch": 722.57,
|
| 214404 |
+
"learning_rate": 8.567564102564103e-06,
|
| 214405 |
+
"loss": 0.5086,
|
| 214406 |
+
"step": 90420
|
| 214407 |
+
},
|
| 214408 |
+
{
|
| 214409 |
+
"epoch": 722.61,
|
| 214410 |
+
"learning_rate": 8.567483974358976e-06,
|
| 214411 |
+
"loss": 1.1796,
|
| 214412 |
+
"step": 90425
|
| 214413 |
+
},
|
| 214414 |
+
{
|
| 214415 |
+
"epoch": 722.65,
|
| 214416 |
+
"learning_rate": 8.567403846153846e-06,
|
| 214417 |
+
"loss": 0.3368,
|
| 214418 |
+
"step": 90430
|
| 214419 |
+
},
|
| 214420 |
+
{
|
| 214421 |
+
"epoch": 722.69,
|
| 214422 |
+
"learning_rate": 8.567323717948719e-06,
|
| 214423 |
+
"loss": 0.2575,
|
| 214424 |
+
"step": 90435
|
| 214425 |
+
},
|
| 214426 |
+
{
|
| 214427 |
+
"epoch": 722.73,
|
| 214428 |
+
"learning_rate": 8.56724358974359e-06,
|
| 214429 |
+
"loss": 0.2965,
|
| 214430 |
+
"step": 90440
|
| 214431 |
+
},
|
| 214432 |
+
{
|
| 214433 |
+
"epoch": 722.77,
|
| 214434 |
+
"learning_rate": 8.567163461538462e-06,
|
| 214435 |
+
"loss": 0.52,
|
| 214436 |
+
"step": 90445
|
| 214437 |
+
},
|
| 214438 |
+
{
|
| 214439 |
+
"epoch": 722.81,
|
| 214440 |
+
"learning_rate": 8.567083333333333e-06,
|
| 214441 |
+
"loss": 1.0366,
|
| 214442 |
+
"step": 90450
|
| 214443 |
+
},
|
| 214444 |
+
{
|
| 214445 |
+
"epoch": 722.85,
|
| 214446 |
+
"learning_rate": 8.567003205128206e-06,
|
| 214447 |
+
"loss": 0.3497,
|
| 214448 |
+
"step": 90455
|
| 214449 |
+
},
|
| 214450 |
+
{
|
| 214451 |
+
"epoch": 722.89,
|
| 214452 |
+
"learning_rate": 8.566923076923077e-06,
|
| 214453 |
+
"loss": 0.3024,
|
| 214454 |
+
"step": 90460
|
| 214455 |
+
},
|
| 214456 |
+
{
|
| 214457 |
+
"epoch": 722.93,
|
| 214458 |
+
"learning_rate": 8.566842948717949e-06,
|
| 214459 |
+
"loss": 0.3712,
|
| 214460 |
+
"step": 90465
|
| 214461 |
+
},
|
| 214462 |
+
{
|
| 214463 |
+
"epoch": 722.97,
|
| 214464 |
+
"learning_rate": 8.566762820512822e-06,
|
| 214465 |
+
"loss": 0.5973,
|
| 214466 |
+
"step": 90470
|
| 214467 |
+
},
|
| 214468 |
+
{
|
| 214469 |
+
"epoch": 723.0,
|
| 214470 |
+
"eval_loss": 0.40732017159461975,
|
| 214471 |
+
"eval_runtime": 37.8665,
|
| 214472 |
+
"eval_samples_per_second": 22.21,
|
| 214473 |
+
"eval_steps_per_second": 0.713,
|
| 214474 |
+
"eval_wer": 0.18586561967553153,
|
| 214475 |
+
"step": 90474
|
| 214476 |
+
},
|
| 214477 |
+
{
|
| 214478 |
+
"epoch": 729.01,
|
| 214479 |
+
"learning_rate": 8.566682692307693e-06,
|
| 214480 |
+
"loss": 0.3401,
|
| 214481 |
+
"step": 90475
|
| 214482 |
+
},
|
| 214483 |
+
{
|
| 214484 |
+
"epoch": 729.05,
|
| 214485 |
+
"learning_rate": 8.566602564102565e-06,
|
| 214486 |
+
"loss": 0.2985,
|
| 214487 |
+
"step": 90480
|
| 214488 |
+
},
|
| 214489 |
+
{
|
| 214490 |
+
"epoch": 729.09,
|
| 214491 |
+
"learning_rate": 8.566522435897436e-06,
|
| 214492 |
+
"loss": 0.3321,
|
| 214493 |
+
"step": 90485
|
| 214494 |
+
},
|
| 214495 |
+
{
|
| 214496 |
+
"epoch": 729.13,
|
| 214497 |
+
"learning_rate": 8.566442307692309e-06,
|
| 214498 |
+
"loss": 0.3825,
|
| 214499 |
+
"step": 90490
|
| 214500 |
+
},
|
| 214501 |
+
{
|
| 214502 |
+
"epoch": 729.17,
|
| 214503 |
+
"learning_rate": 8.56636217948718e-06,
|
| 214504 |
+
"loss": 0.5136,
|
| 214505 |
+
"step": 90495
|
| 214506 |
+
},
|
| 214507 |
+
{
|
| 214508 |
+
"epoch": 729.21,
|
| 214509 |
+
"learning_rate": 8.566282051282052e-06,
|
| 214510 |
+
"loss": 1.1282,
|
| 214511 |
+
"step": 90500
|
| 214512 |
+
},
|
| 214513 |
+
{
|
| 214514 |
+
"epoch": 729.25,
|
| 214515 |
+
"learning_rate": 8.566201923076923e-06,
|
| 214516 |
+
"loss": 0.3183,
|
| 214517 |
+
"step": 90505
|
| 214518 |
+
},
|
| 214519 |
+
{
|
| 214520 |
+
"epoch": 729.29,
|
| 214521 |
+
"learning_rate": 8.566121794871796e-06,
|
| 214522 |
+
"loss": 0.2711,
|
| 214523 |
+
"step": 90510
|
| 214524 |
+
},
|
| 214525 |
+
{
|
| 214526 |
+
"epoch": 729.33,
|
| 214527 |
+
"learning_rate": 8.566041666666667e-06,
|
| 214528 |
+
"loss": 0.3284,
|
| 214529 |
+
"step": 90515
|
| 214530 |
+
},
|
| 214531 |
+
{
|
| 214532 |
+
"epoch": 729.37,
|
| 214533 |
+
"learning_rate": 8.565961538461539e-06,
|
| 214534 |
+
"loss": 0.5451,
|
| 214535 |
+
"step": 90520
|
| 214536 |
+
},
|
| 214537 |
+
{
|
| 214538 |
+
"epoch": 729.41,
|
| 214539 |
+
"learning_rate": 8.565881410256412e-06,
|
| 214540 |
+
"loss": 1.1487,
|
| 214541 |
+
"step": 90525
|
| 214542 |
+
},
|
| 214543 |
+
{
|
| 214544 |
+
"epoch": 729.45,
|
| 214545 |
+
"learning_rate": 8.565801282051283e-06,
|
| 214546 |
+
"loss": 0.3511,
|
| 214547 |
+
"step": 90530
|
| 214548 |
+
},
|
| 214549 |
+
{
|
| 214550 |
+
"epoch": 729.49,
|
| 214551 |
+
"learning_rate": 8.565721153846155e-06,
|
| 214552 |
+
"loss": 0.3186,
|
| 214553 |
+
"step": 90535
|
| 214554 |
+
},
|
| 214555 |
+
{
|
| 214556 |
+
"epoch": 729.53,
|
| 214557 |
+
"learning_rate": 8.565641025641026e-06,
|
| 214558 |
+
"loss": 0.3456,
|
| 214559 |
+
"step": 90540
|
| 214560 |
+
},
|
| 214561 |
+
{
|
| 214562 |
+
"epoch": 729.57,
|
| 214563 |
+
"learning_rate": 8.565560897435899e-06,
|
| 214564 |
+
"loss": 0.5414,
|
| 214565 |
+
"step": 90545
|
| 214566 |
+
},
|
| 214567 |
+
{
|
| 214568 |
+
"epoch": 729.61,
|
| 214569 |
+
"learning_rate": 8.56548076923077e-06,
|
| 214570 |
+
"loss": 1.1278,
|
| 214571 |
+
"step": 90550
|
| 214572 |
+
},
|
| 214573 |
+
{
|
| 214574 |
+
"epoch": 729.65,
|
| 214575 |
+
"learning_rate": 8.565400641025642e-06,
|
| 214576 |
+
"loss": 0.3432,
|
| 214577 |
+
"step": 90555
|
| 214578 |
+
},
|
| 214579 |
+
{
|
| 214580 |
+
"epoch": 729.69,
|
| 214581 |
+
"learning_rate": 8.565320512820513e-06,
|
| 214582 |
+
"loss": 0.2526,
|
| 214583 |
+
"step": 90560
|
| 214584 |
+
},
|
| 214585 |
+
{
|
| 214586 |
+
"epoch": 729.73,
|
| 214587 |
+
"learning_rate": 8.565240384615386e-06,
|
| 214588 |
+
"loss": 0.4858,
|
| 214589 |
+
"step": 90565
|
| 214590 |
+
},
|
| 214591 |
+
{
|
| 214592 |
+
"epoch": 729.77,
|
| 214593 |
+
"learning_rate": 8.565160256410257e-06,
|
| 214594 |
+
"loss": 0.5345,
|
| 214595 |
+
"step": 90570
|
| 214596 |
+
},
|
| 214597 |
+
{
|
| 214598 |
+
"epoch": 729.81,
|
| 214599 |
+
"learning_rate": 8.565080128205129e-06,
|
| 214600 |
+
"loss": 1.1488,
|
| 214601 |
+
"step": 90575
|
| 214602 |
+
},
|
| 214603 |
+
{
|
| 214604 |
+
"epoch": 729.85,
|
| 214605 |
+
"learning_rate": 8.565000000000002e-06,
|
| 214606 |
+
"loss": 0.249,
|
| 214607 |
+
"step": 90580
|
| 214608 |
+
},
|
| 214609 |
+
{
|
| 214610 |
+
"epoch": 729.89,
|
| 214611 |
+
"learning_rate": 8.564919871794872e-06,
|
| 214612 |
+
"loss": 0.3071,
|
| 214613 |
+
"step": 90585
|
| 214614 |
+
},
|
| 214615 |
+
{
|
| 214616 |
+
"epoch": 729.93,
|
| 214617 |
+
"learning_rate": 8.564839743589745e-06,
|
| 214618 |
+
"loss": 0.3299,
|
| 214619 |
+
"step": 90590
|
| 214620 |
+
},
|
| 214621 |
+
{
|
| 214622 |
+
"epoch": 729.97,
|
| 214623 |
+
"learning_rate": 8.564759615384616e-06,
|
| 214624 |
+
"loss": 0.5956,
|
| 214625 |
+
"step": 90595
|
| 214626 |
+
},
|
| 214627 |
+
{
|
| 214628 |
+
"epoch": 730.0,
|
| 214629 |
+
"eval_loss": 0.4340682923793793,
|
| 214630 |
+
"eval_runtime": 39.1955,
|
| 214631 |
+
"eval_samples_per_second": 21.482,
|
| 214632 |
+
"eval_steps_per_second": 0.689,
|
| 214633 |
+
"eval_wer": 0.18922451350566366,
|
| 214634 |
+
"step": 90598
|
| 214635 |
+
},
|
| 214636 |
+
{
|
| 214637 |
+
"epoch": 724.02,
|
| 214638 |
+
"learning_rate": 8.564679487179487e-06,
|
| 214639 |
+
"loss": 0.3322,
|
| 214640 |
+
"step": 90600
|
| 214641 |
+
},
|
| 214642 |
+
{
|
| 214643 |
+
"epoch": 724.06,
|
| 214644 |
+
"learning_rate": 8.564599358974359e-06,
|
| 214645 |
+
"loss": 0.305,
|
| 214646 |
+
"step": 90605
|
| 214647 |
+
},
|
| 214648 |
+
{
|
| 214649 |
+
"epoch": 724.1,
|
| 214650 |
+
"learning_rate": 8.564519230769232e-06,
|
| 214651 |
+
"loss": 0.3531,
|
| 214652 |
+
"step": 90610
|
| 214653 |
+
},
|
| 214654 |
+
{
|
| 214655 |
+
"epoch": 724.14,
|
| 214656 |
+
"learning_rate": 8.564439102564103e-06,
|
| 214657 |
+
"loss": 0.3253,
|
| 214658 |
+
"step": 90615
|
| 214659 |
+
},
|
| 214660 |
+
{
|
| 214661 |
+
"epoch": 724.18,
|
| 214662 |
+
"learning_rate": 8.564358974358974e-06,
|
| 214663 |
+
"loss": 0.7224,
|
| 214664 |
+
"step": 90620
|
| 214665 |
+
},
|
| 214666 |
+
{
|
| 214667 |
+
"epoch": 724.22,
|
| 214668 |
+
"learning_rate": 8.564278846153847e-06,
|
| 214669 |
+
"loss": 1.1113,
|
| 214670 |
+
"step": 90625
|
| 214671 |
+
},
|
| 214672 |
+
{
|
| 214673 |
+
"epoch": 724.26,
|
| 214674 |
+
"learning_rate": 8.564198717948719e-06,
|
| 214675 |
+
"loss": 0.2906,
|
| 214676 |
+
"step": 90630
|
| 214677 |
+
},
|
| 214678 |
+
{
|
| 214679 |
+
"epoch": 724.3,
|
| 214680 |
+
"learning_rate": 8.56411858974359e-06,
|
| 214681 |
+
"loss": 0.3082,
|
| 214682 |
+
"step": 90635
|
| 214683 |
+
},
|
| 214684 |
+
{
|
| 214685 |
+
"epoch": 724.34,
|
| 214686 |
+
"learning_rate": 8.564038461538462e-06,
|
| 214687 |
+
"loss": 0.3138,
|
| 214688 |
+
"step": 90640
|
| 214689 |
+
},
|
| 214690 |
+
{
|
| 214691 |
+
"epoch": 724.38,
|
| 214692 |
+
"learning_rate": 8.563958333333335e-06,
|
| 214693 |
+
"loss": 0.6324,
|
| 214694 |
+
"step": 90645
|
| 214695 |
+
},
|
| 214696 |
+
{
|
| 214697 |
+
"epoch": 724.42,
|
| 214698 |
+
"learning_rate": 8.563878205128206e-06,
|
| 214699 |
+
"loss": 0.8806,
|
| 214700 |
+
"step": 90650
|
| 214701 |
+
},
|
| 214702 |
+
{
|
| 214703 |
+
"epoch": 724.46,
|
| 214704 |
+
"learning_rate": 8.563798076923077e-06,
|
| 214705 |
+
"loss": 0.2908,
|
| 214706 |
+
"step": 90655
|
| 214707 |
+
},
|
| 214708 |
+
{
|
| 214709 |
+
"epoch": 724.5,
|
| 214710 |
+
"learning_rate": 8.563717948717949e-06,
|
| 214711 |
+
"loss": 0.2909,
|
| 214712 |
+
"step": 90660
|
| 214713 |
+
},
|
| 214714 |
+
{
|
| 214715 |
+
"epoch": 724.54,
|
| 214716 |
+
"learning_rate": 8.563637820512822e-06,
|
| 214717 |
+
"loss": 0.3203,
|
| 214718 |
+
"step": 90665
|
| 214719 |
+
},
|
| 214720 |
+
{
|
| 214721 |
+
"epoch": 724.58,
|
| 214722 |
+
"learning_rate": 8.563557692307693e-06,
|
| 214723 |
+
"loss": 0.6635,
|
| 214724 |
+
"step": 90670
|
| 214725 |
+
},
|
| 214726 |
+
{
|
| 214727 |
+
"epoch": 724.62,
|
| 214728 |
+
"learning_rate": 8.563477564102564e-06,
|
| 214729 |
+
"loss": 1.0963,
|
| 214730 |
+
"step": 90675
|
| 214731 |
+
},
|
| 214732 |
+
{
|
| 214733 |
+
"epoch": 724.66,
|
| 214734 |
+
"learning_rate": 8.563397435897437e-06,
|
| 214735 |
+
"loss": 0.3117,
|
| 214736 |
+
"step": 90680
|
| 214737 |
+
},
|
| 214738 |
+
{
|
| 214739 |
+
"epoch": 724.7,
|
| 214740 |
+
"learning_rate": 8.563317307692309e-06,
|
| 214741 |
+
"loss": 0.2845,
|
| 214742 |
+
"step": 90685
|
| 214743 |
+
},
|
| 214744 |
+
{
|
| 214745 |
+
"epoch": 724.74,
|
| 214746 |
+
"learning_rate": 8.56323717948718e-06,
|
| 214747 |
+
"loss": 0.4016,
|
| 214748 |
+
"step": 90690
|
| 214749 |
+
},
|
| 214750 |
+
{
|
| 214751 |
+
"epoch": 724.78,
|
| 214752 |
+
"learning_rate": 8.563157051282052e-06,
|
| 214753 |
+
"loss": 0.6272,
|
| 214754 |
+
"step": 90695
|
| 214755 |
+
},
|
| 214756 |
+
{
|
| 214757 |
+
"epoch": 724.82,
|
| 214758 |
+
"learning_rate": 8.563076923076925e-06,
|
| 214759 |
+
"loss": 1.0456,
|
| 214760 |
+
"step": 90700
|
| 214761 |
+
},
|
| 214762 |
+
{
|
| 214763 |
+
"epoch": 724.86,
|
| 214764 |
+
"learning_rate": 8.562996794871796e-06,
|
| 214765 |
+
"loss": 0.3503,
|
| 214766 |
+
"step": 90705
|
| 214767 |
+
},
|
| 214768 |
+
{
|
| 214769 |
+
"epoch": 724.9,
|
| 214770 |
+
"learning_rate": 8.562916666666667e-06,
|
| 214771 |
+
"loss": 0.4135,
|
| 214772 |
+
"step": 90710
|
| 214773 |
+
},
|
| 214774 |
+
{
|
| 214775 |
+
"epoch": 724.94,
|
| 214776 |
+
"learning_rate": 8.562836538461539e-06,
|
| 214777 |
+
"loss": 0.4302,
|
| 214778 |
+
"step": 90715
|
| 214779 |
+
},
|
| 214780 |
+
{
|
| 214781 |
+
"epoch": 724.98,
|
| 214782 |
+
"learning_rate": 8.562756410256412e-06,
|
| 214783 |
+
"loss": 0.6428,
|
| 214784 |
+
"step": 90720
|
| 214785 |
+
},
|
| 214786 |
+
{
|
| 214787 |
+
"epoch": 725.0,
|
| 214788 |
+
"eval_loss": 0.39161810278892517,
|
| 214789 |
+
"eval_runtime": 39.4302,
|
| 214790 |
+
"eval_samples_per_second": 21.354,
|
| 214791 |
+
"eval_steps_per_second": 0.685,
|
| 214792 |
+
"eval_wer": 0.18582597364992037,
|
| 214793 |
+
"step": 90723
|
| 214794 |
+
},
|
| 214795 |
+
{
|
| 214796 |
+
"epoch": 731.02,
|
| 214797 |
+
"learning_rate": 8.562676282051283e-06,
|
| 214798 |
+
"loss": 0.4568,
|
| 214799 |
+
"step": 90725
|
| 214800 |
+
},
|
| 214801 |
+
{
|
| 214802 |
+
"epoch": 731.06,
|
| 214803 |
+
"learning_rate": 8.562596153846154e-06,
|
| 214804 |
+
"loss": 0.2456,
|
| 214805 |
+
"step": 90730
|
| 214806 |
+
},
|
| 214807 |
+
{
|
| 214808 |
+
"epoch": 731.1,
|
| 214809 |
+
"learning_rate": 8.562516025641028e-06,
|
| 214810 |
+
"loss": 0.3276,
|
| 214811 |
+
"step": 90735
|
| 214812 |
+
},
|
| 214813 |
+
{
|
| 214814 |
+
"epoch": 731.14,
|
| 214815 |
+
"learning_rate": 8.562435897435897e-06,
|
| 214816 |
+
"loss": 0.3748,
|
| 214817 |
+
"step": 90740
|
| 214818 |
+
},
|
| 214819 |
+
{
|
| 214820 |
+
"epoch": 731.18,
|
| 214821 |
+
"learning_rate": 8.56235576923077e-06,
|
| 214822 |
+
"loss": 0.675,
|
| 214823 |
+
"step": 90745
|
| 214824 |
+
},
|
| 214825 |
+
{
|
| 214826 |
+
"epoch": 731.22,
|
| 214827 |
+
"learning_rate": 8.562275641025642e-06,
|
| 214828 |
+
"loss": 1.0248,
|
| 214829 |
+
"step": 90750
|
| 214830 |
+
},
|
| 214831 |
+
{
|
| 214832 |
+
"epoch": 731.26,
|
| 214833 |
+
"learning_rate": 8.562195512820513e-06,
|
| 214834 |
+
"loss": 0.3,
|
| 214835 |
+
"step": 90755
|
| 214836 |
+
},
|
| 214837 |
+
{
|
| 214838 |
+
"epoch": 731.3,
|
| 214839 |
+
"learning_rate": 8.562115384615384e-06,
|
| 214840 |
+
"loss": 0.285,
|
| 214841 |
+
"step": 90760
|
| 214842 |
+
},
|
| 214843 |
+
{
|
| 214844 |
+
"epoch": 731.34,
|
| 214845 |
+
"learning_rate": 8.562035256410257e-06,
|
| 214846 |
+
"loss": 0.3437,
|
| 214847 |
+
"step": 90765
|
| 214848 |
+
},
|
| 214849 |
+
{
|
| 214850 |
+
"epoch": 731.38,
|
| 214851 |
+
"learning_rate": 8.561955128205129e-06,
|
| 214852 |
+
"loss": 0.6141,
|
| 214853 |
+
"step": 90770
|
| 214854 |
+
},
|
| 214855 |
+
{
|
| 214856 |
+
"epoch": 731.42,
|
| 214857 |
+
"learning_rate": 8.561875e-06,
|
| 214858 |
+
"loss": 1.1171,
|
| 214859 |
+
"step": 90775
|
| 214860 |
+
},
|
| 214861 |
+
{
|
| 214862 |
+
"epoch": 731.46,
|
| 214863 |
+
"learning_rate": 8.561794871794873e-06,
|
| 214864 |
+
"loss": 0.2953,
|
| 214865 |
+
"step": 90780
|
| 214866 |
+
},
|
| 214867 |
+
{
|
| 214868 |
+
"epoch": 731.5,
|
| 214869 |
+
"learning_rate": 8.561714743589744e-06,
|
| 214870 |
+
"loss": 0.3753,
|
| 214871 |
+
"step": 90785
|
| 214872 |
+
},
|
| 214873 |
+
{
|
| 214874 |
+
"epoch": 731.54,
|
| 214875 |
+
"learning_rate": 8.561634615384616e-06,
|
| 214876 |
+
"loss": 0.3691,
|
| 214877 |
+
"step": 90790
|
| 214878 |
+
},
|
| 214879 |
+
{
|
| 214880 |
+
"epoch": 731.58,
|
| 214881 |
+
"learning_rate": 8.561554487179487e-06,
|
| 214882 |
+
"loss": 0.6694,
|
| 214883 |
+
"step": 90795
|
| 214884 |
+
},
|
| 214885 |
+
{
|
| 214886 |
+
"epoch": 731.62,
|
| 214887 |
+
"learning_rate": 8.56147435897436e-06,
|
| 214888 |
+
"loss": 1.198,
|
| 214889 |
+
"step": 90800
|
| 214890 |
+
},
|
| 214891 |
+
{
|
| 214892 |
+
"epoch": 731.66,
|
| 214893 |
+
"learning_rate": 8.561394230769232e-06,
|
| 214894 |
+
"loss": 0.2969,
|
| 214895 |
+
"step": 90805
|
| 214896 |
+
},
|
| 214897 |
+
{
|
| 214898 |
+
"epoch": 731.7,
|
| 214899 |
+
"learning_rate": 8.561314102564103e-06,
|
| 214900 |
+
"loss": 0.2967,
|
| 214901 |
+
"step": 90810
|
| 214902 |
+
},
|
| 214903 |
+
{
|
| 214904 |
+
"epoch": 731.74,
|
| 214905 |
+
"learning_rate": 8.561233974358974e-06,
|
| 214906 |
+
"loss": 0.4696,
|
| 214907 |
+
"step": 90815
|
| 214908 |
+
},
|
| 214909 |
+
{
|
| 214910 |
+
"epoch": 731.78,
|
| 214911 |
+
"learning_rate": 8.561153846153847e-06,
|
| 214912 |
+
"loss": 0.6796,
|
| 214913 |
+
"step": 90820
|
| 214914 |
+
},
|
| 214915 |
+
{
|
| 214916 |
+
"epoch": 731.82,
|
| 214917 |
+
"learning_rate": 8.561073717948719e-06,
|
| 214918 |
+
"loss": 1.075,
|
| 214919 |
+
"step": 90825
|
| 214920 |
+
},
|
| 214921 |
+
{
|
| 214922 |
+
"epoch": 731.86,
|
| 214923 |
+
"learning_rate": 8.56099358974359e-06,
|
| 214924 |
+
"loss": 0.3209,
|
| 214925 |
+
"step": 90830
|
| 214926 |
+
},
|
| 214927 |
+
{
|
| 214928 |
+
"epoch": 731.9,
|
| 214929 |
+
"learning_rate": 8.560913461538463e-06,
|
| 214930 |
+
"loss": 0.3053,
|
| 214931 |
+
"step": 90835
|
| 214932 |
+
},
|
| 214933 |
+
{
|
| 214934 |
+
"epoch": 731.94,
|
| 214935 |
+
"learning_rate": 8.560833333333335e-06,
|
| 214936 |
+
"loss": 0.3938,
|
| 214937 |
+
"step": 90840
|
| 214938 |
+
},
|
| 214939 |
+
{
|
| 214940 |
+
"epoch": 731.98,
|
| 214941 |
+
"learning_rate": 8.560753205128206e-06,
|
| 214942 |
+
"loss": 0.7524,
|
| 214943 |
+
"step": 90845
|
| 214944 |
+
},
|
| 214945 |
+
{
|
| 214946 |
+
"epoch": 732.0,
|
| 214947 |
+
"eval_loss": 0.3970924913883209,
|
| 214948 |
+
"eval_runtime": 39.9515,
|
| 214949 |
+
"eval_samples_per_second": 21.076,
|
| 214950 |
+
"eval_steps_per_second": 0.676,
|
| 214951 |
+
"eval_wer": 0.17860725032239577,
|
| 214952 |
+
"step": 90847
|
| 214953 |
}
|
| 214954 |
],
|
| 214955 |
+
"max_steps": 620000,
|
| 214956 |
"num_train_epochs": 5000,
|
| 214957 |
+
"total_flos": 2.556513208848127e+20,
|
| 214958 |
"trial_name": null,
|
| 214959 |
"trial_params": null
|
| 214960 |
}
|
model-bin/finetune/base/{checkpoint-90224 → checkpoint-90847}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629946430.8098779/events.out.tfevents.1629946430.8e89bd551565.924.31
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0618a1807167f1dcba33af2fffd998886f3965b98f6dbcfde0bb73d3aa883d4c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629946866.7205796/events.out.tfevents.1629946866.8e89bd551565.924.33
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d8461f1cda7357540abedb29d6df5a00a94f312a1d0cf63d78c1bb313dcf6640
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629947293.5282943/events.out.tfevents.1629947293.8e89bd551565.924.35
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f9bb91f64684415110a78bcf66515040a66195052e35a55962e99a1800e6e2e3
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629947731.4501543/events.out.tfevents.1629947731.8e89bd551565.924.37
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f85ba0707a05a4eac1d3b6c23404c5140f2a7ac188ddbf1d36288650efcbb4fb
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629948172.226034/events.out.tfevents.1629948172.8e89bd551565.924.39
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c4befea1585251bc7c99b383b5bfd7684f3a4860350acf26f3a68bff29eab240
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629946430.8e89bd551565.924.30
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3e2718ce0307e47ebd148df27a8de268e07787c55f85ff7f88559b55142085cc
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629946865.8e89bd551565.924.32
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8926a6ebf4ccc025156d8e7d5f8295105dabc670214fae5e7bb09fedc7e5cc51
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629947293.8e89bd551565.924.34
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b4c2821a24b944ef3287bd7b53f04a6866aad0d533603765ff98281bc9705a55
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629947731.8e89bd551565.924.36
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:89a8ec8a55692aadd934e9cf43ff933855378f67141e4dff560486c1043bbdab
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629948172.8e89bd551565.924.38
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3ab273cf29802d043bb6104ea525d62f2b868b25aa762204b1073c9a93428b56
|
| 3 |
+
size 8622
|