"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-118592 → checkpoint-119214}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-118592 → checkpoint-119214}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-118592 → checkpoint-119214}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-118592 → checkpoint-119214}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-118592 → checkpoint-119214}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-118592 → checkpoint-119214}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-118592 → checkpoint-119214}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-118592 → checkpoint-119214}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-118592 → checkpoint-119214}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630124544.4129226/events.out.tfevents.1630124544.86bb0ddabf9b.4092.11 +3 -0
- model-bin/finetune/base/log/1630124931.7639797/events.out.tfevents.1630124931.86bb0ddabf9b.4092.13 +3 -0
- model-bin/finetune/base/log/1630125329.5501342/events.out.tfevents.1630125329.86bb0ddabf9b.4092.15 +3 -0
- model-bin/finetune/base/log/1630125717.7793531/events.out.tfevents.1630125717.86bb0ddabf9b.4092.17 +3 -0
- model-bin/finetune/base/log/1630126111.5479174/events.out.tfevents.1630126111.86bb0ddabf9b.4092.19 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630124544.86bb0ddabf9b.4092.10 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630124931.86bb0ddabf9b.4092.12 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630125329.86bb0ddabf9b.4092.14 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630125717.86bb0ddabf9b.4092.16 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630126111.86bb0ddabf9b.4092.18 +3 -0
model-bin/finetune/base/{checkpoint-118592 → checkpoint-119214}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-118592 → checkpoint-119214}/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 sha256:55aacdd914cb2655a54021a1bbb63ec8acd27d7f7dc9f63a06ac84429949b5bd
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-118592 → checkpoint-119214}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-118592 → checkpoint-119214}/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 sha256:711b3e9f623cd6253c1fed8482d7044c6b00fae1e5394e154a5d6cfeadeef50b
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-118592 → checkpoint-119214}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 14567
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:95b46a99ec26524229d1e43a1ea32dbc0368ea83b1202d48e849e2c2adcbc24f
|
| 3 |
size 14567
|
model-bin/finetune/base/{checkpoint-118592 → checkpoint-119214}/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 sha256:4241cd19c98b2b0c4446fb3c2c03a1cbb1fef2851254b06939606ced7f358c8c
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-118592 → checkpoint-119214}/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 sha256:3f23628dd3f5669933750ad348a61e506a0eb60fcfd41c156d4f6fea9051c095
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-118592 → checkpoint-119214}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -250251,11 +250251,800 @@
|
|
| 250251 |
"eval_steps_per_second": 0.744,
|
| 250252 |
"eval_wer": 0.19291367167646845,
|
| 250253 |
"step": 118592
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 250254 |
}
|
| 250255 |
],
|
| 250256 |
"max_steps": 620000,
|
| 250257 |
"num_train_epochs": 5000,
|
| 250258 |
-
"total_flos": 3.
|
| 250259 |
"trial_name": null,
|
| 250260 |
"trial_params": null
|
| 250261 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 960.995983935743,
|
| 5 |
+
"global_step": 119214,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 250251 |
"eval_steps_per_second": 0.744,
|
| 250252 |
"eval_wer": 0.19291367167646845,
|
| 250253 |
"step": 118592
|
| 250254 |
+
},
|
| 250255 |
+
{
|
| 250256 |
+
"epoch": 948.02,
|
| 250257 |
+
"learning_rate": 8.11621794871795e-06,
|
| 250258 |
+
"loss": 0.3007,
|
| 250259 |
+
"step": 118595
|
| 250260 |
+
},
|
| 250261 |
+
{
|
| 250262 |
+
"epoch": 948.06,
|
| 250263 |
+
"learning_rate": 8.116137820512821e-06,
|
| 250264 |
+
"loss": 0.3091,
|
| 250265 |
+
"step": 118600
|
| 250266 |
+
},
|
| 250267 |
+
{
|
| 250268 |
+
"epoch": 948.1,
|
| 250269 |
+
"learning_rate": 8.116057692307693e-06,
|
| 250270 |
+
"loss": 0.273,
|
| 250271 |
+
"step": 118605
|
| 250272 |
+
},
|
| 250273 |
+
{
|
| 250274 |
+
"epoch": 948.14,
|
| 250275 |
+
"learning_rate": 8.115977564102566e-06,
|
| 250276 |
+
"loss": 0.3906,
|
| 250277 |
+
"step": 118610
|
| 250278 |
+
},
|
| 250279 |
+
{
|
| 250280 |
+
"epoch": 948.18,
|
| 250281 |
+
"learning_rate": 8.115897435897437e-06,
|
| 250282 |
+
"loss": 0.6425,
|
| 250283 |
+
"step": 118615
|
| 250284 |
+
},
|
| 250285 |
+
{
|
| 250286 |
+
"epoch": 948.22,
|
| 250287 |
+
"learning_rate": 8.115817307692308e-06,
|
| 250288 |
+
"loss": 0.7462,
|
| 250289 |
+
"step": 118620
|
| 250290 |
+
},
|
| 250291 |
+
{
|
| 250292 |
+
"epoch": 948.26,
|
| 250293 |
+
"learning_rate": 8.11573717948718e-06,
|
| 250294 |
+
"loss": 0.2641,
|
| 250295 |
+
"step": 118625
|
| 250296 |
+
},
|
| 250297 |
+
{
|
| 250298 |
+
"epoch": 948.3,
|
| 250299 |
+
"learning_rate": 8.115657051282053e-06,
|
| 250300 |
+
"loss": 0.2792,
|
| 250301 |
+
"step": 118630
|
| 250302 |
+
},
|
| 250303 |
+
{
|
| 250304 |
+
"epoch": 948.34,
|
| 250305 |
+
"learning_rate": 8.115576923076923e-06,
|
| 250306 |
+
"loss": 0.439,
|
| 250307 |
+
"step": 118635
|
| 250308 |
+
},
|
| 250309 |
+
{
|
| 250310 |
+
"epoch": 948.38,
|
| 250311 |
+
"learning_rate": 8.115496794871796e-06,
|
| 250312 |
+
"loss": 0.7356,
|
| 250313 |
+
"step": 118640
|
| 250314 |
+
},
|
| 250315 |
+
{
|
| 250316 |
+
"epoch": 948.42,
|
| 250317 |
+
"learning_rate": 8.115416666666669e-06,
|
| 250318 |
+
"loss": 0.7915,
|
| 250319 |
+
"step": 118645
|
| 250320 |
+
},
|
| 250321 |
+
{
|
| 250322 |
+
"epoch": 948.46,
|
| 250323 |
+
"learning_rate": 8.115336538461538e-06,
|
| 250324 |
+
"loss": 0.2884,
|
| 250325 |
+
"step": 118650
|
| 250326 |
+
},
|
| 250327 |
+
{
|
| 250328 |
+
"epoch": 948.5,
|
| 250329 |
+
"learning_rate": 8.115256410256411e-06,
|
| 250330 |
+
"loss": 0.3225,
|
| 250331 |
+
"step": 118655
|
| 250332 |
+
},
|
| 250333 |
+
{
|
| 250334 |
+
"epoch": 948.54,
|
| 250335 |
+
"learning_rate": 8.115176282051283e-06,
|
| 250336 |
+
"loss": 0.4471,
|
| 250337 |
+
"step": 118660
|
| 250338 |
+
},
|
| 250339 |
+
{
|
| 250340 |
+
"epoch": 948.58,
|
| 250341 |
+
"learning_rate": 8.115096153846154e-06,
|
| 250342 |
+
"loss": 0.8053,
|
| 250343 |
+
"step": 118665
|
| 250344 |
+
},
|
| 250345 |
+
{
|
| 250346 |
+
"epoch": 948.62,
|
| 250347 |
+
"learning_rate": 8.115016025641025e-06,
|
| 250348 |
+
"loss": 0.7982,
|
| 250349 |
+
"step": 118670
|
| 250350 |
+
},
|
| 250351 |
+
{
|
| 250352 |
+
"epoch": 948.66,
|
| 250353 |
+
"learning_rate": 8.114935897435899e-06,
|
| 250354 |
+
"loss": 0.2521,
|
| 250355 |
+
"step": 118675
|
| 250356 |
+
},
|
| 250357 |
+
{
|
| 250358 |
+
"epoch": 948.7,
|
| 250359 |
+
"learning_rate": 8.11485576923077e-06,
|
| 250360 |
+
"loss": 0.2354,
|
| 250361 |
+
"step": 118680
|
| 250362 |
+
},
|
| 250363 |
+
{
|
| 250364 |
+
"epoch": 948.74,
|
| 250365 |
+
"learning_rate": 8.114775641025641e-06,
|
| 250366 |
+
"loss": 0.389,
|
| 250367 |
+
"step": 118685
|
| 250368 |
+
},
|
| 250369 |
+
{
|
| 250370 |
+
"epoch": 948.78,
|
| 250371 |
+
"learning_rate": 8.114695512820513e-06,
|
| 250372 |
+
"loss": 0.7431,
|
| 250373 |
+
"step": 118690
|
| 250374 |
+
},
|
| 250375 |
+
{
|
| 250376 |
+
"epoch": 948.82,
|
| 250377 |
+
"learning_rate": 8.114615384615386e-06,
|
| 250378 |
+
"loss": 0.8635,
|
| 250379 |
+
"step": 118695
|
| 250380 |
+
},
|
| 250381 |
+
{
|
| 250382 |
+
"epoch": 948.86,
|
| 250383 |
+
"learning_rate": 8.114535256410257e-06,
|
| 250384 |
+
"loss": 0.2869,
|
| 250385 |
+
"step": 118700
|
| 250386 |
+
},
|
| 250387 |
+
{
|
| 250388 |
+
"epoch": 948.9,
|
| 250389 |
+
"learning_rate": 8.114455128205128e-06,
|
| 250390 |
+
"loss": 0.2877,
|
| 250391 |
+
"step": 118705
|
| 250392 |
+
},
|
| 250393 |
+
{
|
| 250394 |
+
"epoch": 948.94,
|
| 250395 |
+
"learning_rate": 8.114375000000001e-06,
|
| 250396 |
+
"loss": 0.4119,
|
| 250397 |
+
"step": 118710
|
| 250398 |
+
},
|
| 250399 |
+
{
|
| 250400 |
+
"epoch": 948.98,
|
| 250401 |
+
"learning_rate": 8.114294871794873e-06,
|
| 250402 |
+
"loss": 0.7172,
|
| 250403 |
+
"step": 118715
|
| 250404 |
+
},
|
| 250405 |
+
{
|
| 250406 |
+
"epoch": 949.0,
|
| 250407 |
+
"eval_loss": 0.36025649309158325,
|
| 250408 |
+
"eval_runtime": 36.5931,
|
| 250409 |
+
"eval_samples_per_second": 22.928,
|
| 250410 |
+
"eval_steps_per_second": 0.738,
|
| 250411 |
+
"eval_wer": 0.18660429197297096,
|
| 250412 |
+
"step": 118717
|
| 250413 |
+
},
|
| 250414 |
+
{
|
| 250415 |
+
"epoch": 957.02,
|
| 250416 |
+
"learning_rate": 8.114214743589744e-06,
|
| 250417 |
+
"loss": 0.3572,
|
| 250418 |
+
"step": 118720
|
| 250419 |
+
},
|
| 250420 |
+
{
|
| 250421 |
+
"epoch": 957.06,
|
| 250422 |
+
"learning_rate": 8.114134615384615e-06,
|
| 250423 |
+
"loss": 0.2555,
|
| 250424 |
+
"step": 118725
|
| 250425 |
+
},
|
| 250426 |
+
{
|
| 250427 |
+
"epoch": 957.1,
|
| 250428 |
+
"learning_rate": 8.114054487179489e-06,
|
| 250429 |
+
"loss": 0.2978,
|
| 250430 |
+
"step": 118730
|
| 250431 |
+
},
|
| 250432 |
+
{
|
| 250433 |
+
"epoch": 957.14,
|
| 250434 |
+
"learning_rate": 8.11397435897436e-06,
|
| 250435 |
+
"loss": 0.3445,
|
| 250436 |
+
"step": 118735
|
| 250437 |
+
},
|
| 250438 |
+
{
|
| 250439 |
+
"epoch": 957.18,
|
| 250440 |
+
"learning_rate": 8.113894230769231e-06,
|
| 250441 |
+
"loss": 0.7311,
|
| 250442 |
+
"step": 118740
|
| 250443 |
+
},
|
| 250444 |
+
{
|
| 250445 |
+
"epoch": 957.22,
|
| 250446 |
+
"learning_rate": 8.113814102564104e-06,
|
| 250447 |
+
"loss": 0.7177,
|
| 250448 |
+
"step": 118745
|
| 250449 |
+
},
|
| 250450 |
+
{
|
| 250451 |
+
"epoch": 957.27,
|
| 250452 |
+
"learning_rate": 8.113733974358976e-06,
|
| 250453 |
+
"loss": 0.32,
|
| 250454 |
+
"step": 118750
|
| 250455 |
+
},
|
| 250456 |
+
{
|
| 250457 |
+
"epoch": 957.31,
|
| 250458 |
+
"learning_rate": 8.113653846153847e-06,
|
| 250459 |
+
"loss": 0.3072,
|
| 250460 |
+
"step": 118755
|
| 250461 |
+
},
|
| 250462 |
+
{
|
| 250463 |
+
"epoch": 957.35,
|
| 250464 |
+
"learning_rate": 8.113573717948718e-06,
|
| 250465 |
+
"loss": 0.3509,
|
| 250466 |
+
"step": 118760
|
| 250467 |
+
},
|
| 250468 |
+
{
|
| 250469 |
+
"epoch": 957.39,
|
| 250470 |
+
"learning_rate": 8.113493589743591e-06,
|
| 250471 |
+
"loss": 0.7122,
|
| 250472 |
+
"step": 118765
|
| 250473 |
+
},
|
| 250474 |
+
{
|
| 250475 |
+
"epoch": 957.43,
|
| 250476 |
+
"learning_rate": 8.113413461538461e-06,
|
| 250477 |
+
"loss": 0.9206,
|
| 250478 |
+
"step": 118770
|
| 250479 |
+
},
|
| 250480 |
+
{
|
| 250481 |
+
"epoch": 957.47,
|
| 250482 |
+
"learning_rate": 8.113333333333334e-06,
|
| 250483 |
+
"loss": 0.3058,
|
| 250484 |
+
"step": 118775
|
| 250485 |
+
},
|
| 250486 |
+
{
|
| 250487 |
+
"epoch": 957.51,
|
| 250488 |
+
"learning_rate": 8.113253205128206e-06,
|
| 250489 |
+
"loss": 0.2759,
|
| 250490 |
+
"step": 118780
|
| 250491 |
+
},
|
| 250492 |
+
{
|
| 250493 |
+
"epoch": 957.55,
|
| 250494 |
+
"learning_rate": 8.113173076923077e-06,
|
| 250495 |
+
"loss": 0.3676,
|
| 250496 |
+
"step": 118785
|
| 250497 |
+
},
|
| 250498 |
+
{
|
| 250499 |
+
"epoch": 957.59,
|
| 250500 |
+
"learning_rate": 8.113092948717948e-06,
|
| 250501 |
+
"loss": 0.7807,
|
| 250502 |
+
"step": 118790
|
| 250503 |
+
},
|
| 250504 |
+
{
|
| 250505 |
+
"epoch": 957.63,
|
| 250506 |
+
"learning_rate": 8.113012820512821e-06,
|
| 250507 |
+
"loss": 0.8521,
|
| 250508 |
+
"step": 118795
|
| 250509 |
+
},
|
| 250510 |
+
{
|
| 250511 |
+
"epoch": 957.67,
|
| 250512 |
+
"learning_rate": 8.112932692307693e-06,
|
| 250513 |
+
"loss": 0.3331,
|
| 250514 |
+
"step": 118800
|
| 250515 |
+
},
|
| 250516 |
+
{
|
| 250517 |
+
"epoch": 957.71,
|
| 250518 |
+
"learning_rate": 8.112852564102564e-06,
|
| 250519 |
+
"loss": 0.2894,
|
| 250520 |
+
"step": 118805
|
| 250521 |
+
},
|
| 250522 |
+
{
|
| 250523 |
+
"epoch": 957.75,
|
| 250524 |
+
"learning_rate": 8.112772435897437e-06,
|
| 250525 |
+
"loss": 0.3214,
|
| 250526 |
+
"step": 118810
|
| 250527 |
+
},
|
| 250528 |
+
{
|
| 250529 |
+
"epoch": 957.79,
|
| 250530 |
+
"learning_rate": 8.112692307692308e-06,
|
| 250531 |
+
"loss": 0.7028,
|
| 250532 |
+
"step": 118815
|
| 250533 |
+
},
|
| 250534 |
+
{
|
| 250535 |
+
"epoch": 957.83,
|
| 250536 |
+
"learning_rate": 8.11261217948718e-06,
|
| 250537 |
+
"loss": 0.8221,
|
| 250538 |
+
"step": 118820
|
| 250539 |
+
},
|
| 250540 |
+
{
|
| 250541 |
+
"epoch": 957.87,
|
| 250542 |
+
"learning_rate": 8.112532051282051e-06,
|
| 250543 |
+
"loss": 0.3689,
|
| 250544 |
+
"step": 118825
|
| 250545 |
+
},
|
| 250546 |
+
{
|
| 250547 |
+
"epoch": 957.91,
|
| 250548 |
+
"learning_rate": 8.112451923076924e-06,
|
| 250549 |
+
"loss": 0.3002,
|
| 250550 |
+
"step": 118830
|
| 250551 |
+
},
|
| 250552 |
+
{
|
| 250553 |
+
"epoch": 957.95,
|
| 250554 |
+
"learning_rate": 8.112371794871796e-06,
|
| 250555 |
+
"loss": 0.3331,
|
| 250556 |
+
"step": 118835
|
| 250557 |
+
},
|
| 250558 |
+
{
|
| 250559 |
+
"epoch": 957.99,
|
| 250560 |
+
"learning_rate": 8.112291666666667e-06,
|
| 250561 |
+
"loss": 0.905,
|
| 250562 |
+
"step": 118840
|
| 250563 |
+
},
|
| 250564 |
+
{
|
| 250565 |
+
"epoch": 958.0,
|
| 250566 |
+
"eval_loss": 0.32400813698768616,
|
| 250567 |
+
"eval_runtime": 36.9467,
|
| 250568 |
+
"eval_samples_per_second": 22.708,
|
| 250569 |
+
"eval_steps_per_second": 0.731,
|
| 250570 |
+
"eval_wer": 0.1764069264069264,
|
| 250571 |
+
"step": 118841
|
| 250572 |
+
},
|
| 250573 |
+
{
|
| 250574 |
+
"epoch": 958.03,
|
| 250575 |
+
"learning_rate": 8.11221153846154e-06,
|
| 250576 |
+
"loss": 0.3427,
|
| 250577 |
+
"step": 118845
|
| 250578 |
+
},
|
| 250579 |
+
{
|
| 250580 |
+
"epoch": 958.07,
|
| 250581 |
+
"learning_rate": 8.112131410256411e-06,
|
| 250582 |
+
"loss": 0.3329,
|
| 250583 |
+
"step": 118850
|
| 250584 |
+
},
|
| 250585 |
+
{
|
| 250586 |
+
"epoch": 958.11,
|
| 250587 |
+
"learning_rate": 8.112051282051283e-06,
|
| 250588 |
+
"loss": 0.2775,
|
| 250589 |
+
"step": 118855
|
| 250590 |
+
},
|
| 250591 |
+
{
|
| 250592 |
+
"epoch": 958.15,
|
| 250593 |
+
"learning_rate": 8.111971153846154e-06,
|
| 250594 |
+
"loss": 0.3691,
|
| 250595 |
+
"step": 118860
|
| 250596 |
+
},
|
| 250597 |
+
{
|
| 250598 |
+
"epoch": 958.19,
|
| 250599 |
+
"learning_rate": 8.111891025641027e-06,
|
| 250600 |
+
"loss": 0.7862,
|
| 250601 |
+
"step": 118865
|
| 250602 |
+
},
|
| 250603 |
+
{
|
| 250604 |
+
"epoch": 958.23,
|
| 250605 |
+
"learning_rate": 8.111810897435898e-06,
|
| 250606 |
+
"loss": 0.6417,
|
| 250607 |
+
"step": 118870
|
| 250608 |
+
},
|
| 250609 |
+
{
|
| 250610 |
+
"epoch": 958.27,
|
| 250611 |
+
"learning_rate": 8.11173076923077e-06,
|
| 250612 |
+
"loss": 0.2834,
|
| 250613 |
+
"step": 118875
|
| 250614 |
+
},
|
| 250615 |
+
{
|
| 250616 |
+
"epoch": 958.31,
|
| 250617 |
+
"learning_rate": 8.111650641025641e-06,
|
| 250618 |
+
"loss": 0.3272,
|
| 250619 |
+
"step": 118880
|
| 250620 |
+
},
|
| 250621 |
+
{
|
| 250622 |
+
"epoch": 958.35,
|
| 250623 |
+
"learning_rate": 8.111570512820514e-06,
|
| 250624 |
+
"loss": 0.3643,
|
| 250625 |
+
"step": 118885
|
| 250626 |
+
},
|
| 250627 |
+
{
|
| 250628 |
+
"epoch": 958.39,
|
| 250629 |
+
"learning_rate": 8.111490384615386e-06,
|
| 250630 |
+
"loss": 0.9751,
|
| 250631 |
+
"step": 118890
|
| 250632 |
+
},
|
| 250633 |
+
{
|
| 250634 |
+
"epoch": 958.43,
|
| 250635 |
+
"learning_rate": 8.111410256410257e-06,
|
| 250636 |
+
"loss": 0.6574,
|
| 250637 |
+
"step": 118895
|
| 250638 |
+
},
|
| 250639 |
+
{
|
| 250640 |
+
"epoch": 958.47,
|
| 250641 |
+
"learning_rate": 8.11133012820513e-06,
|
| 250642 |
+
"loss": 0.2749,
|
| 250643 |
+
"step": 118900
|
| 250644 |
+
},
|
| 250645 |
+
{
|
| 250646 |
+
"epoch": 958.51,
|
| 250647 |
+
"learning_rate": 8.111250000000001e-06,
|
| 250648 |
+
"loss": 0.2626,
|
| 250649 |
+
"step": 118905
|
| 250650 |
+
},
|
| 250651 |
+
{
|
| 250652 |
+
"epoch": 958.55,
|
| 250653 |
+
"learning_rate": 8.111169871794873e-06,
|
| 250654 |
+
"loss": 0.3795,
|
| 250655 |
+
"step": 118910
|
| 250656 |
+
},
|
| 250657 |
+
{
|
| 250658 |
+
"epoch": 958.59,
|
| 250659 |
+
"learning_rate": 8.111089743589744e-06,
|
| 250660 |
+
"loss": 0.7953,
|
| 250661 |
+
"step": 118915
|
| 250662 |
+
},
|
| 250663 |
+
{
|
| 250664 |
+
"epoch": 958.63,
|
| 250665 |
+
"learning_rate": 8.111009615384617e-06,
|
| 250666 |
+
"loss": 0.6629,
|
| 250667 |
+
"step": 118920
|
| 250668 |
+
},
|
| 250669 |
+
{
|
| 250670 |
+
"epoch": 958.67,
|
| 250671 |
+
"learning_rate": 8.110929487179487e-06,
|
| 250672 |
+
"loss": 0.2853,
|
| 250673 |
+
"step": 118925
|
| 250674 |
+
},
|
| 250675 |
+
{
|
| 250676 |
+
"epoch": 958.71,
|
| 250677 |
+
"learning_rate": 8.11084935897436e-06,
|
| 250678 |
+
"loss": 0.3376,
|
| 250679 |
+
"step": 118930
|
| 250680 |
+
},
|
| 250681 |
+
{
|
| 250682 |
+
"epoch": 958.76,
|
| 250683 |
+
"learning_rate": 8.110769230769231e-06,
|
| 250684 |
+
"loss": 0.3622,
|
| 250685 |
+
"step": 118935
|
| 250686 |
+
},
|
| 250687 |
+
{
|
| 250688 |
+
"epoch": 958.8,
|
| 250689 |
+
"learning_rate": 8.110689102564103e-06,
|
| 250690 |
+
"loss": 0.9148,
|
| 250691 |
+
"step": 118940
|
| 250692 |
+
},
|
| 250693 |
+
{
|
| 250694 |
+
"epoch": 958.84,
|
| 250695 |
+
"learning_rate": 8.110608974358976e-06,
|
| 250696 |
+
"loss": 0.6741,
|
| 250697 |
+
"step": 118945
|
| 250698 |
+
},
|
| 250699 |
+
{
|
| 250700 |
+
"epoch": 958.88,
|
| 250701 |
+
"learning_rate": 8.110528846153847e-06,
|
| 250702 |
+
"loss": 0.2854,
|
| 250703 |
+
"step": 118950
|
| 250704 |
+
},
|
| 250705 |
+
{
|
| 250706 |
+
"epoch": 958.92,
|
| 250707 |
+
"learning_rate": 8.110448717948718e-06,
|
| 250708 |
+
"loss": 0.3468,
|
| 250709 |
+
"step": 118955
|
| 250710 |
+
},
|
| 250711 |
+
{
|
| 250712 |
+
"epoch": 958.96,
|
| 250713 |
+
"learning_rate": 8.11036858974359e-06,
|
| 250714 |
+
"loss": 0.4967,
|
| 250715 |
+
"step": 118960
|
| 250716 |
+
},
|
| 250717 |
+
{
|
| 250718 |
+
"epoch": 959.0,
|
| 250719 |
+
"learning_rate": 8.110288461538463e-06,
|
| 250720 |
+
"loss": 1.1176,
|
| 250721 |
+
"step": 118965
|
| 250722 |
+
},
|
| 250723 |
+
{
|
| 250724 |
+
"epoch": 959.0,
|
| 250725 |
+
"eval_loss": 0.40179431438446045,
|
| 250726 |
+
"eval_runtime": 35.7492,
|
| 250727 |
+
"eval_samples_per_second": 23.469,
|
| 250728 |
+
"eval_steps_per_second": 0.755,
|
| 250729 |
+
"eval_wer": 0.17739017152782804,
|
| 250730 |
+
"step": 118965
|
| 250731 |
+
},
|
| 250732 |
+
{
|
| 250733 |
+
"epoch": 951.04,
|
| 250734 |
+
"learning_rate": 8.110208333333334e-06,
|
| 250735 |
+
"loss": 0.3327,
|
| 250736 |
+
"step": 118970
|
| 250737 |
+
},
|
| 250738 |
+
{
|
| 250739 |
+
"epoch": 951.08,
|
| 250740 |
+
"learning_rate": 8.110128205128205e-06,
|
| 250741 |
+
"loss": 0.3402,
|
| 250742 |
+
"step": 118975
|
| 250743 |
+
},
|
| 250744 |
+
{
|
| 250745 |
+
"epoch": 951.12,
|
| 250746 |
+
"learning_rate": 8.110048076923077e-06,
|
| 250747 |
+
"loss": 0.2956,
|
| 250748 |
+
"step": 118980
|
| 250749 |
+
},
|
| 250750 |
+
{
|
| 250751 |
+
"epoch": 951.16,
|
| 250752 |
+
"learning_rate": 8.10996794871795e-06,
|
| 250753 |
+
"loss": 0.4449,
|
| 250754 |
+
"step": 118985
|
| 250755 |
+
},
|
| 250756 |
+
{
|
| 250757 |
+
"epoch": 951.2,
|
| 250758 |
+
"learning_rate": 8.109887820512821e-06,
|
| 250759 |
+
"loss": 1.1227,
|
| 250760 |
+
"step": 118990
|
| 250761 |
+
},
|
| 250762 |
+
{
|
| 250763 |
+
"epoch": 951.24,
|
| 250764 |
+
"learning_rate": 8.109807692307693e-06,
|
| 250765 |
+
"loss": 0.3309,
|
| 250766 |
+
"step": 118995
|
| 250767 |
+
},
|
| 250768 |
+
{
|
| 250769 |
+
"epoch": 951.28,
|
| 250770 |
+
"learning_rate": 8.109727564102566e-06,
|
| 250771 |
+
"loss": 0.288,
|
| 250772 |
+
"step": 119000
|
| 250773 |
+
},
|
| 250774 |
+
{
|
| 250775 |
+
"epoch": 951.32,
|
| 250776 |
+
"learning_rate": 8.109647435897437e-06,
|
| 250777 |
+
"loss": 0.3384,
|
| 250778 |
+
"step": 119005
|
| 250779 |
+
},
|
| 250780 |
+
{
|
| 250781 |
+
"epoch": 951.36,
|
| 250782 |
+
"learning_rate": 8.109567307692308e-06,
|
| 250783 |
+
"loss": 0.3956,
|
| 250784 |
+
"step": 119010
|
| 250785 |
+
},
|
| 250786 |
+
{
|
| 250787 |
+
"epoch": 951.4,
|
| 250788 |
+
"learning_rate": 8.10948717948718e-06,
|
| 250789 |
+
"loss": 1.2679,
|
| 250790 |
+
"step": 119015
|
| 250791 |
+
},
|
| 250792 |
+
{
|
| 250793 |
+
"epoch": 951.44,
|
| 250794 |
+
"learning_rate": 8.109407051282053e-06,
|
| 250795 |
+
"loss": 0.3574,
|
| 250796 |
+
"step": 119020
|
| 250797 |
+
},
|
| 250798 |
+
{
|
| 250799 |
+
"epoch": 951.48,
|
| 250800 |
+
"learning_rate": 8.109326923076924e-06,
|
| 250801 |
+
"loss": 0.2528,
|
| 250802 |
+
"step": 119025
|
| 250803 |
+
},
|
| 250804 |
+
{
|
| 250805 |
+
"epoch": 951.52,
|
| 250806 |
+
"learning_rate": 8.109246794871795e-06,
|
| 250807 |
+
"loss": 0.2937,
|
| 250808 |
+
"step": 119030
|
| 250809 |
+
},
|
| 250810 |
+
{
|
| 250811 |
+
"epoch": 951.56,
|
| 250812 |
+
"learning_rate": 8.109166666666667e-06,
|
| 250813 |
+
"loss": 0.4535,
|
| 250814 |
+
"step": 119035
|
| 250815 |
+
},
|
| 250816 |
+
{
|
| 250817 |
+
"epoch": 951.6,
|
| 250818 |
+
"learning_rate": 8.10908653846154e-06,
|
| 250819 |
+
"loss": 1.2491,
|
| 250820 |
+
"step": 119040
|
| 250821 |
+
},
|
| 250822 |
+
{
|
| 250823 |
+
"epoch": 951.64,
|
| 250824 |
+
"learning_rate": 8.109006410256411e-06,
|
| 250825 |
+
"loss": 0.3057,
|
| 250826 |
+
"step": 119045
|
| 250827 |
+
},
|
| 250828 |
+
{
|
| 250829 |
+
"epoch": 951.68,
|
| 250830 |
+
"learning_rate": 8.108926282051283e-06,
|
| 250831 |
+
"loss": 0.2951,
|
| 250832 |
+
"step": 119050
|
| 250833 |
+
},
|
| 250834 |
+
{
|
| 250835 |
+
"epoch": 951.72,
|
| 250836 |
+
"learning_rate": 8.108846153846156e-06,
|
| 250837 |
+
"loss": 0.3247,
|
| 250838 |
+
"step": 119055
|
| 250839 |
+
},
|
| 250840 |
+
{
|
| 250841 |
+
"epoch": 951.76,
|
| 250842 |
+
"learning_rate": 8.108766025641025e-06,
|
| 250843 |
+
"loss": 0.4163,
|
| 250844 |
+
"step": 119060
|
| 250845 |
+
},
|
| 250846 |
+
{
|
| 250847 |
+
"epoch": 951.8,
|
| 250848 |
+
"learning_rate": 8.108685897435898e-06,
|
| 250849 |
+
"loss": 0.9172,
|
| 250850 |
+
"step": 119065
|
| 250851 |
+
},
|
| 250852 |
+
{
|
| 250853 |
+
"epoch": 951.84,
|
| 250854 |
+
"learning_rate": 8.10860576923077e-06,
|
| 250855 |
+
"loss": 0.3262,
|
| 250856 |
+
"step": 119070
|
| 250857 |
+
},
|
| 250858 |
+
{
|
| 250859 |
+
"epoch": 951.88,
|
| 250860 |
+
"learning_rate": 8.108525641025641e-06,
|
| 250861 |
+
"loss": 0.2451,
|
| 250862 |
+
"step": 119075
|
| 250863 |
+
},
|
| 250864 |
+
{
|
| 250865 |
+
"epoch": 951.92,
|
| 250866 |
+
"learning_rate": 8.108445512820512e-06,
|
| 250867 |
+
"loss": 0.4066,
|
| 250868 |
+
"step": 119080
|
| 250869 |
+
},
|
| 250870 |
+
{
|
| 250871 |
+
"epoch": 951.96,
|
| 250872 |
+
"learning_rate": 8.108365384615385e-06,
|
| 250873 |
+
"loss": 0.4462,
|
| 250874 |
+
"step": 119085
|
| 250875 |
+
},
|
| 250876 |
+
{
|
| 250877 |
+
"epoch": 952.0,
|
| 250878 |
+
"learning_rate": 8.108285256410257e-06,
|
| 250879 |
+
"loss": 1.2865,
|
| 250880 |
+
"step": 119090
|
| 250881 |
+
},
|
| 250882 |
+
{
|
| 250883 |
+
"epoch": 952.0,
|
| 250884 |
+
"eval_loss": 0.4650250971317291,
|
| 250885 |
+
"eval_runtime": 35.8362,
|
| 250886 |
+
"eval_samples_per_second": 23.412,
|
| 250887 |
+
"eval_steps_per_second": 0.753,
|
| 250888 |
+
"eval_wer": 0.1926915284760261,
|
| 250889 |
+
"step": 119090
|
| 250890 |
+
},
|
| 250891 |
+
{
|
| 250892 |
+
"epoch": 960.04,
|
| 250893 |
+
"learning_rate": 8.108205128205128e-06,
|
| 250894 |
+
"loss": 0.3843,
|
| 250895 |
+
"step": 119095
|
| 250896 |
+
},
|
| 250897 |
+
{
|
| 250898 |
+
"epoch": 960.08,
|
| 250899 |
+
"learning_rate": 8.108125000000001e-06,
|
| 250900 |
+
"loss": 0.2742,
|
| 250901 |
+
"step": 119100
|
| 250902 |
+
},
|
| 250903 |
+
{
|
| 250904 |
+
"epoch": 960.12,
|
| 250905 |
+
"learning_rate": 8.108044871794873e-06,
|
| 250906 |
+
"loss": 0.2903,
|
| 250907 |
+
"step": 119105
|
| 250908 |
+
},
|
| 250909 |
+
{
|
| 250910 |
+
"epoch": 960.16,
|
| 250911 |
+
"learning_rate": 8.107964743589744e-06,
|
| 250912 |
+
"loss": 0.4582,
|
| 250913 |
+
"step": 119110
|
| 250914 |
+
},
|
| 250915 |
+
{
|
| 250916 |
+
"epoch": 960.2,
|
| 250917 |
+
"learning_rate": 8.107884615384615e-06,
|
| 250918 |
+
"loss": 1.0597,
|
| 250919 |
+
"step": 119115
|
| 250920 |
+
},
|
| 250921 |
+
{
|
| 250922 |
+
"epoch": 960.24,
|
| 250923 |
+
"learning_rate": 8.107804487179488e-06,
|
| 250924 |
+
"loss": 0.302,
|
| 250925 |
+
"step": 119120
|
| 250926 |
+
},
|
| 250927 |
+
{
|
| 250928 |
+
"epoch": 960.28,
|
| 250929 |
+
"learning_rate": 8.10772435897436e-06,
|
| 250930 |
+
"loss": 0.3508,
|
| 250931 |
+
"step": 119125
|
| 250932 |
+
},
|
| 250933 |
+
{
|
| 250934 |
+
"epoch": 960.32,
|
| 250935 |
+
"learning_rate": 8.107644230769231e-06,
|
| 250936 |
+
"loss": 0.3365,
|
| 250937 |
+
"step": 119130
|
| 250938 |
+
},
|
| 250939 |
+
{
|
| 250940 |
+
"epoch": 960.36,
|
| 250941 |
+
"learning_rate": 8.107564102564102e-06,
|
| 250942 |
+
"loss": 0.3898,
|
| 250943 |
+
"step": 119135
|
| 250944 |
+
},
|
| 250945 |
+
{
|
| 250946 |
+
"epoch": 960.4,
|
| 250947 |
+
"learning_rate": 8.107483974358976e-06,
|
| 250948 |
+
"loss": 1.1342,
|
| 250949 |
+
"step": 119140
|
| 250950 |
+
},
|
| 250951 |
+
{
|
| 250952 |
+
"epoch": 960.44,
|
| 250953 |
+
"learning_rate": 8.107403846153847e-06,
|
| 250954 |
+
"loss": 0.3327,
|
| 250955 |
+
"step": 119145
|
| 250956 |
+
},
|
| 250957 |
+
{
|
| 250958 |
+
"epoch": 960.48,
|
| 250959 |
+
"learning_rate": 8.107323717948718e-06,
|
| 250960 |
+
"loss": 0.2717,
|
| 250961 |
+
"step": 119150
|
| 250962 |
+
},
|
| 250963 |
+
{
|
| 250964 |
+
"epoch": 960.52,
|
| 250965 |
+
"learning_rate": 8.107243589743591e-06,
|
| 250966 |
+
"loss": 0.3667,
|
| 250967 |
+
"step": 119155
|
| 250968 |
+
},
|
| 250969 |
+
{
|
| 250970 |
+
"epoch": 960.56,
|
| 250971 |
+
"learning_rate": 8.107163461538463e-06,
|
| 250972 |
+
"loss": 0.4429,
|
| 250973 |
+
"step": 119160
|
| 250974 |
+
},
|
| 250975 |
+
{
|
| 250976 |
+
"epoch": 960.6,
|
| 250977 |
+
"learning_rate": 8.107083333333334e-06,
|
| 250978 |
+
"loss": 1.2418,
|
| 250979 |
+
"step": 119165
|
| 250980 |
+
},
|
| 250981 |
+
{
|
| 250982 |
+
"epoch": 960.64,
|
| 250983 |
+
"learning_rate": 8.107003205128205e-06,
|
| 250984 |
+
"loss": 0.3244,
|
| 250985 |
+
"step": 119170
|
| 250986 |
+
},
|
| 250987 |
+
{
|
| 250988 |
+
"epoch": 960.68,
|
| 250989 |
+
"learning_rate": 8.106923076923078e-06,
|
| 250990 |
+
"loss": 0.2825,
|
| 250991 |
+
"step": 119175
|
| 250992 |
+
},
|
| 250993 |
+
{
|
| 250994 |
+
"epoch": 960.72,
|
| 250995 |
+
"learning_rate": 8.10684294871795e-06,
|
| 250996 |
+
"loss": 0.2662,
|
| 250997 |
+
"step": 119180
|
| 250998 |
+
},
|
| 250999 |
+
{
|
| 251000 |
+
"epoch": 960.76,
|
| 251001 |
+
"learning_rate": 8.106762820512821e-06,
|
| 251002 |
+
"loss": 0.4766,
|
| 251003 |
+
"step": 119185
|
| 251004 |
+
},
|
| 251005 |
+
{
|
| 251006 |
+
"epoch": 960.8,
|
| 251007 |
+
"learning_rate": 8.106682692307692e-06,
|
| 251008 |
+
"loss": 1.1543,
|
| 251009 |
+
"step": 119190
|
| 251010 |
+
},
|
| 251011 |
+
{
|
| 251012 |
+
"epoch": 960.84,
|
| 251013 |
+
"learning_rate": 8.106602564102566e-06,
|
| 251014 |
+
"loss": 0.3048,
|
| 251015 |
+
"step": 119195
|
| 251016 |
+
},
|
| 251017 |
+
{
|
| 251018 |
+
"epoch": 960.88,
|
| 251019 |
+
"learning_rate": 8.106522435897437e-06,
|
| 251020 |
+
"loss": 0.2547,
|
| 251021 |
+
"step": 119200
|
| 251022 |
+
},
|
| 251023 |
+
{
|
| 251024 |
+
"epoch": 960.92,
|
| 251025 |
+
"learning_rate": 8.106442307692308e-06,
|
| 251026 |
+
"loss": 0.3489,
|
| 251027 |
+
"step": 119205
|
| 251028 |
+
},
|
| 251029 |
+
{
|
| 251030 |
+
"epoch": 960.96,
|
| 251031 |
+
"learning_rate": 8.106362179487181e-06,
|
| 251032 |
+
"loss": 0.4812,
|
| 251033 |
+
"step": 119210
|
| 251034 |
+
},
|
| 251035 |
+
{
|
| 251036 |
+
"epoch": 961.0,
|
| 251037 |
+
"eval_loss": 0.38714075088500977,
|
| 251038 |
+
"eval_runtime": 35.5909,
|
| 251039 |
+
"eval_samples_per_second": 23.602,
|
| 251040 |
+
"eval_steps_per_second": 0.759,
|
| 251041 |
+
"eval_wer": 0.18514659919630896,
|
| 251042 |
+
"step": 119214
|
| 251043 |
}
|
| 251044 |
],
|
| 251045 |
"max_steps": 620000,
|
| 251046 |
"num_train_epochs": 5000,
|
| 251047 |
+
"total_flos": 3.354967342927919e+20,
|
| 251048 |
"trial_name": null,
|
| 251049 |
"trial_params": null
|
| 251050 |
}
|
model-bin/finetune/base/{checkpoint-118592 → checkpoint-119214}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630124544.4129226/events.out.tfevents.1630124544.86bb0ddabf9b.4092.11
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7f004fe7f9cee9954167156d889f01efc5987f1e8da5cdc3c8aa709bc4b1edcb
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630124931.7639797/events.out.tfevents.1630124931.86bb0ddabf9b.4092.13
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6a416fb225ee312e8141a843fc4b176c786569afc43e8ffedb4ce72a9f0263a7
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630125329.5501342/events.out.tfevents.1630125329.86bb0ddabf9b.4092.15
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:805403b9875eaa53afec83cc42d8a11281566e073d09e098f10172a96d7a91bf
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630125717.7793531/events.out.tfevents.1630125717.86bb0ddabf9b.4092.17
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c8f445f44d3a080ebc303ba24d31d50b2660813abc56fc19e36473d7fe991598
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630126111.5479174/events.out.tfevents.1630126111.86bb0ddabf9b.4092.19
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ec529abb8decfaa4fd2b5b4c63f2ca6594320d785899a614e6b6f9dabac627b1
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630124544.86bb0ddabf9b.4092.10
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:40ae13004300d705bbdb677978d943f22eaa51e41a4ef645aaf6640d7d7a4d0c
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630124931.86bb0ddabf9b.4092.12
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e3255fdd44d311b7a362d64b27f9df701729230e526fa2acfe4ffccc4f13fc03
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630125329.86bb0ddabf9b.4092.14
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a1ff750124db9cf75c833d148205f161d7a38bcc890a258e79f310df579b11f5
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630125717.86bb0ddabf9b.4092.16
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1254c1a2fe5c3d17b99f13619f4f3816e8cd16042c3e100e77bd8e20d0f28c00
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630126111.86bb0ddabf9b.4092.18
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:865ad5c0ef2a011873beb77ea480e3117dd44eca9f844678b12ed2dbc4dd25a6
|
| 3 |
+
size 8462
|