"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-114613 → checkpoint-115234}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-114613 → checkpoint-115234}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-114613 → checkpoint-115234}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-114613 → checkpoint-115234}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-114613 → checkpoint-115234}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-114613 → checkpoint-115234}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-114613 → checkpoint-115234}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-114613 → checkpoint-115234}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-114613 → checkpoint-115234}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630109431.8143556/events.out.tfevents.1630109431.86bb0ddabf9b.1042.11 +3 -0
- model-bin/finetune/base/log/1630109864.1808882/events.out.tfevents.1630109864.86bb0ddabf9b.1042.13 +3 -0
- model-bin/finetune/base/log/1630110291.1942098/events.out.tfevents.1630110291.86bb0ddabf9b.1042.15 +3 -0
- model-bin/finetune/base/log/1630110708.1539922/events.out.tfevents.1630110708.86bb0ddabf9b.1042.17 +3 -0
- model-bin/finetune/base/log/1630111124.0089736/events.out.tfevents.1630111124.86bb0ddabf9b.1042.19 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630109431.86bb0ddabf9b.1042.10 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630109864.86bb0ddabf9b.1042.12 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630110291.86bb0ddabf9b.1042.14 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630110708.86bb0ddabf9b.1042.16 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630111124.86bb0ddabf9b.1042.18 +3 -0
model-bin/finetune/base/{checkpoint-114613 → checkpoint-115234}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-114613 → checkpoint-115234}/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:a3bc213f8b1a88e0b2512f257ed2494755fd2795f55af3712d785a4772256d44
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-114613 → checkpoint-115234}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-114613 → checkpoint-115234}/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 sha256:845b93ea6994b83196fa59050206c61d3f4fc9ef2aa5ba6e72e95c173579ac5a
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-114613 → checkpoint-115234}/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 sha256:3c7fcf8524438e496a57b73e453b24b2519c6f58941a6fb8da174143dac50f8d
|
| 3 |
size 14567
|
model-bin/finetune/base/{checkpoint-114613 → checkpoint-115234}/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 sha256:0e28e0530f2551ac6de66de5fb95c14533593794a980009f90dec1f828fe4986
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-114613 → checkpoint-115234}/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:f6eac4ec54a649aef54843448fdae717a79bc1578521fca8e21301c3cddb0022
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-114613 → checkpoint-115234}/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,
|
|
@@ -245187,11 +245187,800 @@
|
|
| 245187 |
"eval_steps_per_second": 0.677,
|
| 245188 |
"eval_wer": 0.18582472364713531,
|
| 245189 |
"step": 114613
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 245190 |
}
|
| 245191 |
],
|
| 245192 |
"max_steps": 620000,
|
| 245193 |
"num_train_epochs": 5000,
|
| 245194 |
-
"total_flos": 3.
|
| 245195 |
"trial_name": null,
|
| 245196 |
"trial_params": null
|
| 245197 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 928.995983935743,
|
| 5 |
+
"global_step": 115234,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 245187 |
"eval_steps_per_second": 0.677,
|
| 245188 |
"eval_wer": 0.18582472364713531,
|
| 245189 |
"step": 114613
|
| 245190 |
+
},
|
| 245191 |
+
{
|
| 245192 |
+
"epoch": 924.02,
|
| 245193 |
+
"learning_rate": 8.179983974358975e-06,
|
| 245194 |
+
"loss": 0.3841,
|
| 245195 |
+
"step": 114615
|
| 245196 |
+
},
|
| 245197 |
+
{
|
| 245198 |
+
"epoch": 924.06,
|
| 245199 |
+
"learning_rate": 8.179903846153847e-06,
|
| 245200 |
+
"loss": 0.3366,
|
| 245201 |
+
"step": 114620
|
| 245202 |
+
},
|
| 245203 |
+
{
|
| 245204 |
+
"epoch": 924.1,
|
| 245205 |
+
"learning_rate": 8.179823717948718e-06,
|
| 245206 |
+
"loss": 0.2559,
|
| 245207 |
+
"step": 114625
|
| 245208 |
+
},
|
| 245209 |
+
{
|
| 245210 |
+
"epoch": 924.14,
|
| 245211 |
+
"learning_rate": 8.179743589743591e-06,
|
| 245212 |
+
"loss": 0.3411,
|
| 245213 |
+
"step": 114630
|
| 245214 |
+
},
|
| 245215 |
+
{
|
| 245216 |
+
"epoch": 924.18,
|
| 245217 |
+
"learning_rate": 8.179663461538462e-06,
|
| 245218 |
+
"loss": 0.6367,
|
| 245219 |
+
"step": 114635
|
| 245220 |
+
},
|
| 245221 |
+
{
|
| 245222 |
+
"epoch": 924.22,
|
| 245223 |
+
"learning_rate": 8.179583333333334e-06,
|
| 245224 |
+
"loss": 1.0442,
|
| 245225 |
+
"step": 114640
|
| 245226 |
+
},
|
| 245227 |
+
{
|
| 245228 |
+
"epoch": 924.26,
|
| 245229 |
+
"learning_rate": 8.179503205128205e-06,
|
| 245230 |
+
"loss": 0.3043,
|
| 245231 |
+
"step": 114645
|
| 245232 |
+
},
|
| 245233 |
+
{
|
| 245234 |
+
"epoch": 924.3,
|
| 245235 |
+
"learning_rate": 8.179423076923078e-06,
|
| 245236 |
+
"loss": 0.2539,
|
| 245237 |
+
"step": 114650
|
| 245238 |
+
},
|
| 245239 |
+
{
|
| 245240 |
+
"epoch": 924.34,
|
| 245241 |
+
"learning_rate": 8.17934294871795e-06,
|
| 245242 |
+
"loss": 0.3802,
|
| 245243 |
+
"step": 114655
|
| 245244 |
+
},
|
| 245245 |
+
{
|
| 245246 |
+
"epoch": 924.38,
|
| 245247 |
+
"learning_rate": 8.179262820512821e-06,
|
| 245248 |
+
"loss": 0.6318,
|
| 245249 |
+
"step": 114660
|
| 245250 |
+
},
|
| 245251 |
+
{
|
| 245252 |
+
"epoch": 924.42,
|
| 245253 |
+
"learning_rate": 8.179182692307694e-06,
|
| 245254 |
+
"loss": 0.9742,
|
| 245255 |
+
"step": 114665
|
| 245256 |
+
},
|
| 245257 |
+
{
|
| 245258 |
+
"epoch": 924.46,
|
| 245259 |
+
"learning_rate": 8.179102564102565e-06,
|
| 245260 |
+
"loss": 0.2686,
|
| 245261 |
+
"step": 114670
|
| 245262 |
+
},
|
| 245263 |
+
{
|
| 245264 |
+
"epoch": 924.5,
|
| 245265 |
+
"learning_rate": 8.179022435897437e-06,
|
| 245266 |
+
"loss": 0.3525,
|
| 245267 |
+
"step": 114675
|
| 245268 |
+
},
|
| 245269 |
+
{
|
| 245270 |
+
"epoch": 924.54,
|
| 245271 |
+
"learning_rate": 8.178942307692308e-06,
|
| 245272 |
+
"loss": 0.3878,
|
| 245273 |
+
"step": 114680
|
| 245274 |
+
},
|
| 245275 |
+
{
|
| 245276 |
+
"epoch": 924.58,
|
| 245277 |
+
"learning_rate": 8.178862179487181e-06,
|
| 245278 |
+
"loss": 0.6801,
|
| 245279 |
+
"step": 114685
|
| 245280 |
+
},
|
| 245281 |
+
{
|
| 245282 |
+
"epoch": 924.62,
|
| 245283 |
+
"learning_rate": 8.178798076923078e-06,
|
| 245284 |
+
"loss": 0.976,
|
| 245285 |
+
"step": 114690
|
| 245286 |
+
},
|
| 245287 |
+
{
|
| 245288 |
+
"epoch": 924.66,
|
| 245289 |
+
"learning_rate": 8.17871794871795e-06,
|
| 245290 |
+
"loss": 0.2602,
|
| 245291 |
+
"step": 114695
|
| 245292 |
+
},
|
| 245293 |
+
{
|
| 245294 |
+
"epoch": 924.7,
|
| 245295 |
+
"learning_rate": 8.178637820512821e-06,
|
| 245296 |
+
"loss": 0.2754,
|
| 245297 |
+
"step": 114700
|
| 245298 |
+
},
|
| 245299 |
+
{
|
| 245300 |
+
"epoch": 924.74,
|
| 245301 |
+
"learning_rate": 8.178557692307693e-06,
|
| 245302 |
+
"loss": 0.3819,
|
| 245303 |
+
"step": 114705
|
| 245304 |
+
},
|
| 245305 |
+
{
|
| 245306 |
+
"epoch": 924.78,
|
| 245307 |
+
"learning_rate": 8.178477564102566e-06,
|
| 245308 |
+
"loss": 0.5863,
|
| 245309 |
+
"step": 114710
|
| 245310 |
+
},
|
| 245311 |
+
{
|
| 245312 |
+
"epoch": 924.82,
|
| 245313 |
+
"learning_rate": 8.178397435897435e-06,
|
| 245314 |
+
"loss": 1.1332,
|
| 245315 |
+
"step": 114715
|
| 245316 |
+
},
|
| 245317 |
+
{
|
| 245318 |
+
"epoch": 924.86,
|
| 245319 |
+
"learning_rate": 8.178317307692308e-06,
|
| 245320 |
+
"loss": 0.2884,
|
| 245321 |
+
"step": 114720
|
| 245322 |
+
},
|
| 245323 |
+
{
|
| 245324 |
+
"epoch": 924.9,
|
| 245325 |
+
"learning_rate": 8.178237179487181e-06,
|
| 245326 |
+
"loss": 0.2688,
|
| 245327 |
+
"step": 114725
|
| 245328 |
+
},
|
| 245329 |
+
{
|
| 245330 |
+
"epoch": 924.94,
|
| 245331 |
+
"learning_rate": 8.178157051282051e-06,
|
| 245332 |
+
"loss": 0.4118,
|
| 245333 |
+
"step": 114730
|
| 245334 |
+
},
|
| 245335 |
+
{
|
| 245336 |
+
"epoch": 924.98,
|
| 245337 |
+
"learning_rate": 8.178076923076924e-06,
|
| 245338 |
+
"loss": 0.608,
|
| 245339 |
+
"step": 114735
|
| 245340 |
+
},
|
| 245341 |
+
{
|
| 245342 |
+
"epoch": 925.0,
|
| 245343 |
+
"eval_loss": 0.3686845004558563,
|
| 245344 |
+
"eval_runtime": 40.4476,
|
| 245345 |
+
"eval_samples_per_second": 20.842,
|
| 245346 |
+
"eval_steps_per_second": 0.668,
|
| 245347 |
+
"eval_wer": 0.18588031473327077,
|
| 245348 |
+
"step": 114737
|
| 245349 |
+
},
|
| 245350 |
+
{
|
| 245351 |
+
"epoch": 925.02,
|
| 245352 |
+
"learning_rate": 8.177996794871795e-06,
|
| 245353 |
+
"loss": 0.4001,
|
| 245354 |
+
"step": 114740
|
| 245355 |
+
},
|
| 245356 |
+
{
|
| 245357 |
+
"epoch": 925.06,
|
| 245358 |
+
"learning_rate": 8.177916666666667e-06,
|
| 245359 |
+
"loss": 0.2602,
|
| 245360 |
+
"step": 114745
|
| 245361 |
+
},
|
| 245362 |
+
{
|
| 245363 |
+
"epoch": 925.1,
|
| 245364 |
+
"learning_rate": 8.177836538461538e-06,
|
| 245365 |
+
"loss": 0.3255,
|
| 245366 |
+
"step": 114750
|
| 245367 |
+
},
|
| 245368 |
+
{
|
| 245369 |
+
"epoch": 925.14,
|
| 245370 |
+
"learning_rate": 8.177756410256411e-06,
|
| 245371 |
+
"loss": 0.3547,
|
| 245372 |
+
"step": 114755
|
| 245373 |
+
},
|
| 245374 |
+
{
|
| 245375 |
+
"epoch": 925.18,
|
| 245376 |
+
"learning_rate": 8.177676282051283e-06,
|
| 245377 |
+
"loss": 0.7224,
|
| 245378 |
+
"step": 114760
|
| 245379 |
+
},
|
| 245380 |
+
{
|
| 245381 |
+
"epoch": 925.22,
|
| 245382 |
+
"learning_rate": 8.177596153846154e-06,
|
| 245383 |
+
"loss": 0.8328,
|
| 245384 |
+
"step": 114765
|
| 245385 |
+
},
|
| 245386 |
+
{
|
| 245387 |
+
"epoch": 925.27,
|
| 245388 |
+
"learning_rate": 8.177516025641025e-06,
|
| 245389 |
+
"loss": 0.2706,
|
| 245390 |
+
"step": 114770
|
| 245391 |
+
},
|
| 245392 |
+
{
|
| 245393 |
+
"epoch": 925.31,
|
| 245394 |
+
"learning_rate": 8.177435897435898e-06,
|
| 245395 |
+
"loss": 0.2727,
|
| 245396 |
+
"step": 114775
|
| 245397 |
+
},
|
| 245398 |
+
{
|
| 245399 |
+
"epoch": 925.35,
|
| 245400 |
+
"learning_rate": 8.17735576923077e-06,
|
| 245401 |
+
"loss": 0.3431,
|
| 245402 |
+
"step": 114780
|
| 245403 |
+
},
|
| 245404 |
+
{
|
| 245405 |
+
"epoch": 925.39,
|
| 245406 |
+
"learning_rate": 8.177275641025641e-06,
|
| 245407 |
+
"loss": 0.7878,
|
| 245408 |
+
"step": 114785
|
| 245409 |
+
},
|
| 245410 |
+
{
|
| 245411 |
+
"epoch": 925.43,
|
| 245412 |
+
"learning_rate": 8.177195512820514e-06,
|
| 245413 |
+
"loss": 1.0417,
|
| 245414 |
+
"step": 114790
|
| 245415 |
+
},
|
| 245416 |
+
{
|
| 245417 |
+
"epoch": 925.47,
|
| 245418 |
+
"learning_rate": 8.177115384615385e-06,
|
| 245419 |
+
"loss": 0.3444,
|
| 245420 |
+
"step": 114795
|
| 245421 |
+
},
|
| 245422 |
+
{
|
| 245423 |
+
"epoch": 925.51,
|
| 245424 |
+
"learning_rate": 8.177035256410257e-06,
|
| 245425 |
+
"loss": 0.2683,
|
| 245426 |
+
"step": 114800
|
| 245427 |
+
},
|
| 245428 |
+
{
|
| 245429 |
+
"epoch": 925.55,
|
| 245430 |
+
"learning_rate": 8.176955128205128e-06,
|
| 245431 |
+
"loss": 0.3586,
|
| 245432 |
+
"step": 114805
|
| 245433 |
+
},
|
| 245434 |
+
{
|
| 245435 |
+
"epoch": 925.59,
|
| 245436 |
+
"learning_rate": 8.176875000000001e-06,
|
| 245437 |
+
"loss": 0.7673,
|
| 245438 |
+
"step": 114810
|
| 245439 |
+
},
|
| 245440 |
+
{
|
| 245441 |
+
"epoch": 925.63,
|
| 245442 |
+
"learning_rate": 8.176794871794873e-06,
|
| 245443 |
+
"loss": 0.8954,
|
| 245444 |
+
"step": 114815
|
| 245445 |
+
},
|
| 245446 |
+
{
|
| 245447 |
+
"epoch": 925.67,
|
| 245448 |
+
"learning_rate": 8.176714743589744e-06,
|
| 245449 |
+
"loss": 0.3164,
|
| 245450 |
+
"step": 114820
|
| 245451 |
+
},
|
| 245452 |
+
{
|
| 245453 |
+
"epoch": 925.71,
|
| 245454 |
+
"learning_rate": 8.176634615384617e-06,
|
| 245455 |
+
"loss": 0.3331,
|
| 245456 |
+
"step": 114825
|
| 245457 |
+
},
|
| 245458 |
+
{
|
| 245459 |
+
"epoch": 925.75,
|
| 245460 |
+
"learning_rate": 8.176554487179488e-06,
|
| 245461 |
+
"loss": 0.3298,
|
| 245462 |
+
"step": 114830
|
| 245463 |
+
},
|
| 245464 |
+
{
|
| 245465 |
+
"epoch": 925.79,
|
| 245466 |
+
"learning_rate": 8.17647435897436e-06,
|
| 245467 |
+
"loss": 0.6546,
|
| 245468 |
+
"step": 114835
|
| 245469 |
+
},
|
| 245470 |
+
{
|
| 245471 |
+
"epoch": 925.83,
|
| 245472 |
+
"learning_rate": 8.176394230769231e-06,
|
| 245473 |
+
"loss": 0.8006,
|
| 245474 |
+
"step": 114840
|
| 245475 |
+
},
|
| 245476 |
+
{
|
| 245477 |
+
"epoch": 925.87,
|
| 245478 |
+
"learning_rate": 8.176314102564104e-06,
|
| 245479 |
+
"loss": 0.3202,
|
| 245480 |
+
"step": 114845
|
| 245481 |
+
},
|
| 245482 |
+
{
|
| 245483 |
+
"epoch": 925.91,
|
| 245484 |
+
"learning_rate": 8.176233974358975e-06,
|
| 245485 |
+
"loss": 0.3295,
|
| 245486 |
+
"step": 114850
|
| 245487 |
+
},
|
| 245488 |
+
{
|
| 245489 |
+
"epoch": 925.95,
|
| 245490 |
+
"learning_rate": 8.176153846153847e-06,
|
| 245491 |
+
"loss": 0.3493,
|
| 245492 |
+
"step": 114855
|
| 245493 |
+
},
|
| 245494 |
+
{
|
| 245495 |
+
"epoch": 925.99,
|
| 245496 |
+
"learning_rate": 8.176073717948718e-06,
|
| 245497 |
+
"loss": 0.7909,
|
| 245498 |
+
"step": 114860
|
| 245499 |
+
},
|
| 245500 |
+
{
|
| 245501 |
+
"epoch": 926.0,
|
| 245502 |
+
"eval_loss": 0.38164782524108887,
|
| 245503 |
+
"eval_runtime": 38.7323,
|
| 245504 |
+
"eval_samples_per_second": 21.765,
|
| 245505 |
+
"eval_steps_per_second": 0.697,
|
| 245506 |
+
"eval_wer": 0.180527833222444,
|
| 245507 |
+
"step": 114861
|
| 245508 |
+
},
|
| 245509 |
+
{
|
| 245510 |
+
"epoch": 926.03,
|
| 245511 |
+
"learning_rate": 8.175993589743591e-06,
|
| 245512 |
+
"loss": 0.3478,
|
| 245513 |
+
"step": 114865
|
| 245514 |
+
},
|
| 245515 |
+
{
|
| 245516 |
+
"epoch": 926.07,
|
| 245517 |
+
"learning_rate": 8.175913461538461e-06,
|
| 245518 |
+
"loss": 0.3354,
|
| 245519 |
+
"step": 114870
|
| 245520 |
+
},
|
| 245521 |
+
{
|
| 245522 |
+
"epoch": 926.11,
|
| 245523 |
+
"learning_rate": 8.175833333333334e-06,
|
| 245524 |
+
"loss": 0.2706,
|
| 245525 |
+
"step": 114875
|
| 245526 |
+
},
|
| 245527 |
+
{
|
| 245528 |
+
"epoch": 926.15,
|
| 245529 |
+
"learning_rate": 8.175753205128207e-06,
|
| 245530 |
+
"loss": 0.3976,
|
| 245531 |
+
"step": 114880
|
| 245532 |
+
},
|
| 245533 |
+
{
|
| 245534 |
+
"epoch": 926.19,
|
| 245535 |
+
"learning_rate": 8.175673076923077e-06,
|
| 245536 |
+
"loss": 0.9176,
|
| 245537 |
+
"step": 114885
|
| 245538 |
+
},
|
| 245539 |
+
{
|
| 245540 |
+
"epoch": 926.23,
|
| 245541 |
+
"learning_rate": 8.17559294871795e-06,
|
| 245542 |
+
"loss": 0.6236,
|
| 245543 |
+
"step": 114890
|
| 245544 |
+
},
|
| 245545 |
+
{
|
| 245546 |
+
"epoch": 926.27,
|
| 245547 |
+
"learning_rate": 8.175512820512821e-06,
|
| 245548 |
+
"loss": 0.2932,
|
| 245549 |
+
"step": 114895
|
| 245550 |
+
},
|
| 245551 |
+
{
|
| 245552 |
+
"epoch": 926.31,
|
| 245553 |
+
"learning_rate": 8.175432692307692e-06,
|
| 245554 |
+
"loss": 0.3139,
|
| 245555 |
+
"step": 114900
|
| 245556 |
+
},
|
| 245557 |
+
{
|
| 245558 |
+
"epoch": 926.35,
|
| 245559 |
+
"learning_rate": 8.175352564102564e-06,
|
| 245560 |
+
"loss": 0.4103,
|
| 245561 |
+
"step": 114905
|
| 245562 |
+
},
|
| 245563 |
+
{
|
| 245564 |
+
"epoch": 926.39,
|
| 245565 |
+
"learning_rate": 8.175272435897437e-06,
|
| 245566 |
+
"loss": 0.8761,
|
| 245567 |
+
"step": 114910
|
| 245568 |
+
},
|
| 245569 |
+
{
|
| 245570 |
+
"epoch": 926.43,
|
| 245571 |
+
"learning_rate": 8.175192307692308e-06,
|
| 245572 |
+
"loss": 0.6826,
|
| 245573 |
+
"step": 114915
|
| 245574 |
+
},
|
| 245575 |
+
{
|
| 245576 |
+
"epoch": 926.47,
|
| 245577 |
+
"learning_rate": 8.17511217948718e-06,
|
| 245578 |
+
"loss": 0.268,
|
| 245579 |
+
"step": 114920
|
| 245580 |
+
},
|
| 245581 |
+
{
|
| 245582 |
+
"epoch": 926.51,
|
| 245583 |
+
"learning_rate": 8.175032051282053e-06,
|
| 245584 |
+
"loss": 0.2962,
|
| 245585 |
+
"step": 114925
|
| 245586 |
+
},
|
| 245587 |
+
{
|
| 245588 |
+
"epoch": 926.55,
|
| 245589 |
+
"learning_rate": 8.174951923076924e-06,
|
| 245590 |
+
"loss": 0.4022,
|
| 245591 |
+
"step": 114930
|
| 245592 |
+
},
|
| 245593 |
+
{
|
| 245594 |
+
"epoch": 926.59,
|
| 245595 |
+
"learning_rate": 8.174871794871795e-06,
|
| 245596 |
+
"loss": 0.8668,
|
| 245597 |
+
"step": 114935
|
| 245598 |
+
},
|
| 245599 |
+
{
|
| 245600 |
+
"epoch": 926.63,
|
| 245601 |
+
"learning_rate": 8.174791666666667e-06,
|
| 245602 |
+
"loss": 0.6838,
|
| 245603 |
+
"step": 114940
|
| 245604 |
+
},
|
| 245605 |
+
{
|
| 245606 |
+
"epoch": 926.67,
|
| 245607 |
+
"learning_rate": 8.17471153846154e-06,
|
| 245608 |
+
"loss": 0.2747,
|
| 245609 |
+
"step": 114945
|
| 245610 |
+
},
|
| 245611 |
+
{
|
| 245612 |
+
"epoch": 926.71,
|
| 245613 |
+
"learning_rate": 8.174631410256411e-06,
|
| 245614 |
+
"loss": 0.364,
|
| 245615 |
+
"step": 114950
|
| 245616 |
+
},
|
| 245617 |
+
{
|
| 245618 |
+
"epoch": 926.76,
|
| 245619 |
+
"learning_rate": 8.174551282051282e-06,
|
| 245620 |
+
"loss": 0.3743,
|
| 245621 |
+
"step": 114955
|
| 245622 |
+
},
|
| 245623 |
+
{
|
| 245624 |
+
"epoch": 926.8,
|
| 245625 |
+
"learning_rate": 8.174471153846154e-06,
|
| 245626 |
+
"loss": 0.9189,
|
| 245627 |
+
"step": 114960
|
| 245628 |
+
},
|
| 245629 |
+
{
|
| 245630 |
+
"epoch": 926.84,
|
| 245631 |
+
"learning_rate": 8.174391025641027e-06,
|
| 245632 |
+
"loss": 0.6867,
|
| 245633 |
+
"step": 114965
|
| 245634 |
+
},
|
| 245635 |
+
{
|
| 245636 |
+
"epoch": 926.88,
|
| 245637 |
+
"learning_rate": 8.174310897435898e-06,
|
| 245638 |
+
"loss": 0.3038,
|
| 245639 |
+
"step": 114970
|
| 245640 |
+
},
|
| 245641 |
+
{
|
| 245642 |
+
"epoch": 926.92,
|
| 245643 |
+
"learning_rate": 8.17423076923077e-06,
|
| 245644 |
+
"loss": 0.2775,
|
| 245645 |
+
"step": 114975
|
| 245646 |
+
},
|
| 245647 |
+
{
|
| 245648 |
+
"epoch": 926.96,
|
| 245649 |
+
"learning_rate": 8.174150641025643e-06,
|
| 245650 |
+
"loss": 0.4685,
|
| 245651 |
+
"step": 114980
|
| 245652 |
+
},
|
| 245653 |
+
{
|
| 245654 |
+
"epoch": 927.0,
|
| 245655 |
+
"learning_rate": 8.174070512820514e-06,
|
| 245656 |
+
"loss": 1.0971,
|
| 245657 |
+
"step": 114985
|
| 245658 |
+
},
|
| 245659 |
+
{
|
| 245660 |
+
"epoch": 927.0,
|
| 245661 |
+
"eval_loss": 0.3635014295578003,
|
| 245662 |
+
"eval_runtime": 39.7417,
|
| 245663 |
+
"eval_samples_per_second": 21.212,
|
| 245664 |
+
"eval_steps_per_second": 0.679,
|
| 245665 |
+
"eval_wer": 0.18649314104818854,
|
| 245666 |
+
"step": 114985
|
| 245667 |
+
},
|
| 245668 |
+
{
|
| 245669 |
+
"epoch": 919.04,
|
| 245670 |
+
"learning_rate": 8.173990384615385e-06,
|
| 245671 |
+
"loss": 0.302,
|
| 245672 |
+
"step": 114990
|
| 245673 |
+
},
|
| 245674 |
+
{
|
| 245675 |
+
"epoch": 919.08,
|
| 245676 |
+
"learning_rate": 8.173910256410257e-06,
|
| 245677 |
+
"loss": 0.3467,
|
| 245678 |
+
"step": 114995
|
| 245679 |
+
},
|
| 245680 |
+
{
|
| 245681 |
+
"epoch": 919.12,
|
| 245682 |
+
"learning_rate": 8.17383012820513e-06,
|
| 245683 |
+
"loss": 0.339,
|
| 245684 |
+
"step": 115000
|
| 245685 |
+
},
|
| 245686 |
+
{
|
| 245687 |
+
"epoch": 919.16,
|
| 245688 |
+
"learning_rate": 8.17375e-06,
|
| 245689 |
+
"loss": 0.4937,
|
| 245690 |
+
"step": 115005
|
| 245691 |
+
},
|
| 245692 |
+
{
|
| 245693 |
+
"epoch": 919.2,
|
| 245694 |
+
"learning_rate": 8.173669871794873e-06,
|
| 245695 |
+
"loss": 1.1532,
|
| 245696 |
+
"step": 115010
|
| 245697 |
+
},
|
| 245698 |
+
{
|
| 245699 |
+
"epoch": 919.24,
|
| 245700 |
+
"learning_rate": 8.173589743589744e-06,
|
| 245701 |
+
"loss": 0.3483,
|
| 245702 |
+
"step": 115015
|
| 245703 |
+
},
|
| 245704 |
+
{
|
| 245705 |
+
"epoch": 919.28,
|
| 245706 |
+
"learning_rate": 8.173509615384615e-06,
|
| 245707 |
+
"loss": 0.3049,
|
| 245708 |
+
"step": 115020
|
| 245709 |
+
},
|
| 245710 |
+
{
|
| 245711 |
+
"epoch": 919.32,
|
| 245712 |
+
"learning_rate": 8.173429487179488e-06,
|
| 245713 |
+
"loss": 0.3562,
|
| 245714 |
+
"step": 115025
|
| 245715 |
+
},
|
| 245716 |
+
{
|
| 245717 |
+
"epoch": 919.36,
|
| 245718 |
+
"learning_rate": 8.17334935897436e-06,
|
| 245719 |
+
"loss": 0.4262,
|
| 245720 |
+
"step": 115030
|
| 245721 |
+
},
|
| 245722 |
+
{
|
| 245723 |
+
"epoch": 919.4,
|
| 245724 |
+
"learning_rate": 8.173269230769233e-06,
|
| 245725 |
+
"loss": 1.2422,
|
| 245726 |
+
"step": 115035
|
| 245727 |
+
},
|
| 245728 |
+
{
|
| 245729 |
+
"epoch": 919.44,
|
| 245730 |
+
"learning_rate": 8.173189102564102e-06,
|
| 245731 |
+
"loss": 0.341,
|
| 245732 |
+
"step": 115040
|
| 245733 |
+
},
|
| 245734 |
+
{
|
| 245735 |
+
"epoch": 919.48,
|
| 245736 |
+
"learning_rate": 8.173108974358975e-06,
|
| 245737 |
+
"loss": 0.2614,
|
| 245738 |
+
"step": 115045
|
| 245739 |
+
},
|
| 245740 |
+
{
|
| 245741 |
+
"epoch": 919.52,
|
| 245742 |
+
"learning_rate": 8.173028846153847e-06,
|
| 245743 |
+
"loss": 0.2816,
|
| 245744 |
+
"step": 115050
|
| 245745 |
+
},
|
| 245746 |
+
{
|
| 245747 |
+
"epoch": 919.56,
|
| 245748 |
+
"learning_rate": 8.172948717948718e-06,
|
| 245749 |
+
"loss": 0.3867,
|
| 245750 |
+
"step": 115055
|
| 245751 |
+
},
|
| 245752 |
+
{
|
| 245753 |
+
"epoch": 919.6,
|
| 245754 |
+
"learning_rate": 8.17286858974359e-06,
|
| 245755 |
+
"loss": 1.1399,
|
| 245756 |
+
"step": 115060
|
| 245757 |
+
},
|
| 245758 |
+
{
|
| 245759 |
+
"epoch": 919.64,
|
| 245760 |
+
"learning_rate": 8.172788461538463e-06,
|
| 245761 |
+
"loss": 0.278,
|
| 245762 |
+
"step": 115065
|
| 245763 |
+
},
|
| 245764 |
+
{
|
| 245765 |
+
"epoch": 919.68,
|
| 245766 |
+
"learning_rate": 8.172708333333334e-06,
|
| 245767 |
+
"loss": 0.2704,
|
| 245768 |
+
"step": 115070
|
| 245769 |
+
},
|
| 245770 |
+
{
|
| 245771 |
+
"epoch": 919.72,
|
| 245772 |
+
"learning_rate": 8.172628205128205e-06,
|
| 245773 |
+
"loss": 0.2866,
|
| 245774 |
+
"step": 115075
|
| 245775 |
+
},
|
| 245776 |
+
{
|
| 245777 |
+
"epoch": 919.76,
|
| 245778 |
+
"learning_rate": 8.172548076923078e-06,
|
| 245779 |
+
"loss": 0.4154,
|
| 245780 |
+
"step": 115080
|
| 245781 |
+
},
|
| 245782 |
+
{
|
| 245783 |
+
"epoch": 919.8,
|
| 245784 |
+
"learning_rate": 8.17246794871795e-06,
|
| 245785 |
+
"loss": 1.0377,
|
| 245786 |
+
"step": 115085
|
| 245787 |
+
},
|
| 245788 |
+
{
|
| 245789 |
+
"epoch": 919.84,
|
| 245790 |
+
"learning_rate": 8.172387820512821e-06,
|
| 245791 |
+
"loss": 0.3569,
|
| 245792 |
+
"step": 115090
|
| 245793 |
+
},
|
| 245794 |
+
{
|
| 245795 |
+
"epoch": 919.88,
|
| 245796 |
+
"learning_rate": 8.172307692307692e-06,
|
| 245797 |
+
"loss": 0.2438,
|
| 245798 |
+
"step": 115095
|
| 245799 |
+
},
|
| 245800 |
+
{
|
| 245801 |
+
"epoch": 919.92,
|
| 245802 |
+
"learning_rate": 8.172227564102565e-06,
|
| 245803 |
+
"loss": 0.3913,
|
| 245804 |
+
"step": 115100
|
| 245805 |
+
},
|
| 245806 |
+
{
|
| 245807 |
+
"epoch": 919.96,
|
| 245808 |
+
"learning_rate": 8.172147435897437e-06,
|
| 245809 |
+
"loss": 0.4283,
|
| 245810 |
+
"step": 115105
|
| 245811 |
+
},
|
| 245812 |
+
{
|
| 245813 |
+
"epoch": 920.0,
|
| 245814 |
+
"learning_rate": 8.172067307692308e-06,
|
| 245815 |
+
"loss": 1.2519,
|
| 245816 |
+
"step": 115110
|
| 245817 |
+
},
|
| 245818 |
+
{
|
| 245819 |
+
"epoch": 920.0,
|
| 245820 |
+
"eval_loss": 0.36768418550491333,
|
| 245821 |
+
"eval_runtime": 39.8638,
|
| 245822 |
+
"eval_samples_per_second": 21.147,
|
| 245823 |
+
"eval_steps_per_second": 0.677,
|
| 245824 |
+
"eval_wer": 0.18435032304379037,
|
| 245825 |
+
"step": 115110
|
| 245826 |
+
},
|
| 245827 |
+
{
|
| 245828 |
+
"epoch": 928.04,
|
| 245829 |
+
"learning_rate": 8.17198717948718e-06,
|
| 245830 |
+
"loss": 0.3685,
|
| 245831 |
+
"step": 115115
|
| 245832 |
+
},
|
| 245833 |
+
{
|
| 245834 |
+
"epoch": 928.08,
|
| 245835 |
+
"learning_rate": 8.171907051282053e-06,
|
| 245836 |
+
"loss": 0.3035,
|
| 245837 |
+
"step": 115120
|
| 245838 |
+
},
|
| 245839 |
+
{
|
| 245840 |
+
"epoch": 928.12,
|
| 245841 |
+
"learning_rate": 8.171826923076924e-06,
|
| 245842 |
+
"loss": 0.3222,
|
| 245843 |
+
"step": 115125
|
| 245844 |
+
},
|
| 245845 |
+
{
|
| 245846 |
+
"epoch": 928.16,
|
| 245847 |
+
"learning_rate": 8.171746794871795e-06,
|
| 245848 |
+
"loss": 0.4358,
|
| 245849 |
+
"step": 115130
|
| 245850 |
+
},
|
| 245851 |
+
{
|
| 245852 |
+
"epoch": 928.2,
|
| 245853 |
+
"learning_rate": 8.171666666666668e-06,
|
| 245854 |
+
"loss": 1.1252,
|
| 245855 |
+
"step": 115135
|
| 245856 |
+
},
|
| 245857 |
+
{
|
| 245858 |
+
"epoch": 928.24,
|
| 245859 |
+
"learning_rate": 8.17158653846154e-06,
|
| 245860 |
+
"loss": 0.3178,
|
| 245861 |
+
"step": 115140
|
| 245862 |
+
},
|
| 245863 |
+
{
|
| 245864 |
+
"epoch": 928.28,
|
| 245865 |
+
"learning_rate": 8.171506410256411e-06,
|
| 245866 |
+
"loss": 0.3091,
|
| 245867 |
+
"step": 115145
|
| 245868 |
+
},
|
| 245869 |
+
{
|
| 245870 |
+
"epoch": 928.32,
|
| 245871 |
+
"learning_rate": 8.171426282051282e-06,
|
| 245872 |
+
"loss": 0.3167,
|
| 245873 |
+
"step": 115150
|
| 245874 |
+
},
|
| 245875 |
+
{
|
| 245876 |
+
"epoch": 928.36,
|
| 245877 |
+
"learning_rate": 8.171346153846155e-06,
|
| 245878 |
+
"loss": 0.4127,
|
| 245879 |
+
"step": 115155
|
| 245880 |
+
},
|
| 245881 |
+
{
|
| 245882 |
+
"epoch": 928.4,
|
| 245883 |
+
"learning_rate": 8.171266025641025e-06,
|
| 245884 |
+
"loss": 1.0843,
|
| 245885 |
+
"step": 115160
|
| 245886 |
+
},
|
| 245887 |
+
{
|
| 245888 |
+
"epoch": 928.44,
|
| 245889 |
+
"learning_rate": 8.171185897435898e-06,
|
| 245890 |
+
"loss": 0.3319,
|
| 245891 |
+
"step": 115165
|
| 245892 |
+
},
|
| 245893 |
+
{
|
| 245894 |
+
"epoch": 928.48,
|
| 245895 |
+
"learning_rate": 8.17110576923077e-06,
|
| 245896 |
+
"loss": 0.2568,
|
| 245897 |
+
"step": 115170
|
| 245898 |
+
},
|
| 245899 |
+
{
|
| 245900 |
+
"epoch": 928.52,
|
| 245901 |
+
"learning_rate": 8.171025641025641e-06,
|
| 245902 |
+
"loss": 0.3007,
|
| 245903 |
+
"step": 115175
|
| 245904 |
+
},
|
| 245905 |
+
{
|
| 245906 |
+
"epoch": 928.56,
|
| 245907 |
+
"learning_rate": 8.170945512820514e-06,
|
| 245908 |
+
"loss": 0.4638,
|
| 245909 |
+
"step": 115180
|
| 245910 |
+
},
|
| 245911 |
+
{
|
| 245912 |
+
"epoch": 928.6,
|
| 245913 |
+
"learning_rate": 8.170865384615385e-06,
|
| 245914 |
+
"loss": 1.3013,
|
| 245915 |
+
"step": 115185
|
| 245916 |
+
},
|
| 245917 |
+
{
|
| 245918 |
+
"epoch": 928.64,
|
| 245919 |
+
"learning_rate": 8.170785256410257e-06,
|
| 245920 |
+
"loss": 0.3193,
|
| 245921 |
+
"step": 115190
|
| 245922 |
+
},
|
| 245923 |
+
{
|
| 245924 |
+
"epoch": 928.68,
|
| 245925 |
+
"learning_rate": 8.170705128205128e-06,
|
| 245926 |
+
"loss": 0.273,
|
| 245927 |
+
"step": 115195
|
| 245928 |
+
},
|
| 245929 |
+
{
|
| 245930 |
+
"epoch": 928.72,
|
| 245931 |
+
"learning_rate": 8.170625000000001e-06,
|
| 245932 |
+
"loss": 0.3355,
|
| 245933 |
+
"step": 115200
|
| 245934 |
+
},
|
| 245935 |
+
{
|
| 245936 |
+
"epoch": 928.76,
|
| 245937 |
+
"learning_rate": 8.170544871794872e-06,
|
| 245938 |
+
"loss": 0.4521,
|
| 245939 |
+
"step": 115205
|
| 245940 |
+
},
|
| 245941 |
+
{
|
| 245942 |
+
"epoch": 928.8,
|
| 245943 |
+
"learning_rate": 8.170464743589744e-06,
|
| 245944 |
+
"loss": 1.1896,
|
| 245945 |
+
"step": 115210
|
| 245946 |
+
},
|
| 245947 |
+
{
|
| 245948 |
+
"epoch": 928.84,
|
| 245949 |
+
"learning_rate": 8.170384615384615e-06,
|
| 245950 |
+
"loss": 0.2887,
|
| 245951 |
+
"step": 115215
|
| 245952 |
+
},
|
| 245953 |
+
{
|
| 245954 |
+
"epoch": 928.88,
|
| 245955 |
+
"learning_rate": 8.170304487179488e-06,
|
| 245956 |
+
"loss": 0.27,
|
| 245957 |
+
"step": 115220
|
| 245958 |
+
},
|
| 245959 |
+
{
|
| 245960 |
+
"epoch": 928.92,
|
| 245961 |
+
"learning_rate": 8.17022435897436e-06,
|
| 245962 |
+
"loss": 0.3528,
|
| 245963 |
+
"step": 115225
|
| 245964 |
+
},
|
| 245965 |
+
{
|
| 245966 |
+
"epoch": 928.96,
|
| 245967 |
+
"learning_rate": 8.170144230769231e-06,
|
| 245968 |
+
"loss": 0.4421,
|
| 245969 |
+
"step": 115230
|
| 245970 |
+
},
|
| 245971 |
+
{
|
| 245972 |
+
"epoch": 929.0,
|
| 245973 |
+
"eval_loss": 0.38033702969551086,
|
| 245974 |
+
"eval_runtime": 39.6678,
|
| 245975 |
+
"eval_samples_per_second": 21.277,
|
| 245976 |
+
"eval_steps_per_second": 0.681,
|
| 245977 |
+
"eval_wer": 0.18934867596075827,
|
| 245978 |
+
"step": 115234
|
| 245979 |
}
|
| 245980 |
],
|
| 245981 |
"max_steps": 620000,
|
| 245982 |
"num_train_epochs": 5000,
|
| 245983 |
+
"total_flos": 3.243039409882843e+20,
|
| 245984 |
"trial_name": null,
|
| 245985 |
"trial_params": null
|
| 245986 |
}
|
model-bin/finetune/base/{checkpoint-114613 → checkpoint-115234}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630109431.8143556/events.out.tfevents.1630109431.86bb0ddabf9b.1042.11
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b1752755bd47a1298986b325a70db8b95c0108718e4f7caea9973b49ff20223e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630109864.1808882/events.out.tfevents.1630109864.86bb0ddabf9b.1042.13
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:492fcb0f04e16405c50032f9b9f601fbf6ccb916bc8c6f9c82b6cbf6072c2f51
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630110291.1942098/events.out.tfevents.1630110291.86bb0ddabf9b.1042.15
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9678335599331e6c8a5e95ac8ae114b537ca30f5aa61f90121c6ad0909316186
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630110708.1539922/events.out.tfevents.1630110708.86bb0ddabf9b.1042.17
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a8789ebd4a0d049bcc6e7a6cfb131f3de072740ef62692b320e8d3b2457ea1f9
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630111124.0089736/events.out.tfevents.1630111124.86bb0ddabf9b.1042.19
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:50220645914c1340a01a0938428ab7312dd5ae9ee446dcb2cf3f29d82df09f88
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630109431.86bb0ddabf9b.1042.10
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e2f69acf780d6b84f0085f54961d3041b1ec46c0d7af1f152bfabb4b960ce1e9
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630109864.86bb0ddabf9b.1042.12
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dbf005686af30527fa466f8f955ccb86be67ba4c19c6e2a7a603c53d9ffe89c7
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630110291.86bb0ddabf9b.1042.14
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b0f307121ff5f987c2125b3494b029ead7320a9107cbbe1f662acbfb27ad23fa
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630110708.86bb0ddabf9b.1042.16
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b8c9cc1c211c48db928e1b7c27c884da18dd9b5870bc8b09f9fbb05f683bdcf5
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630111124.86bb0ddabf9b.1042.18
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b51ed23240af71ded7356cae20da843c27a40e6f060bd5a294754f7398910155
|
| 3 |
+
size 8462
|