"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-59485 → checkpoint-60105}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-59485 → checkpoint-60105}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-59485 → checkpoint-60105}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-59485 → checkpoint-60105}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-59485 → checkpoint-60105}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-59485 → checkpoint-60105}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-59485 → checkpoint-60105}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-59485 → checkpoint-60105}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-59485 → checkpoint-60105}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629794989.6896856/events.out.tfevents.1629794989.c435e1c5ee04.920.91 +3 -0
- model-bin/finetune/base/log/1629795651.0910487/events.out.tfevents.1629795651.c435e1c5ee04.920.93 +3 -0
- model-bin/finetune/base/log/1629796295.0343266/events.out.tfevents.1629796295.c435e1c5ee04.920.95 +3 -0
- model-bin/finetune/base/log/1629796939.1229146/events.out.tfevents.1629796939.c435e1c5ee04.920.97 +3 -0
- model-bin/finetune/base/log/1629797588.2766123/events.out.tfevents.1629797588.c435e1c5ee04.920.99 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629794989.c435e1c5ee04.920.90 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629795651.c435e1c5ee04.920.92 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629796295.c435e1c5ee04.920.94 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629796939.c435e1c5ee04.920.96 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629797588.c435e1c5ee04.920.98 +3 -0
model-bin/finetune/base/{checkpoint-59485 → checkpoint-60105}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-59485 → checkpoint-60105}/optimizer.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 722165009
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:40b3314448fea63402a07aa602956be42c9e740284c01a33022761c3b6ed3b0c
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-59485 → checkpoint-60105}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-59485 → checkpoint-60105}/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 sha256:61f15a7e42709bb323434cf25df9eb7599246900a93a843147941594b17bad97
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-59485 → checkpoint-60105}/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 sha256:2928313b9be06a9303804ad370f60cd7c0e5a076b0b5be3d27103d3299ee392c
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-59485 → checkpoint-60105}/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 sha256:058991856934dabbf10711a0d6d47759e1ab7bf80455e28bbbd566855d88c31d
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-59485 → checkpoint-60105}/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:c1ac81a5988d8abd42a7dd17e61fda58638a7b0876ba21dfaf0d68642ddcb129
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-59485 → checkpoint-60105}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18631571186315712,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-55501",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -175050,11 +175050,800 @@
|
|
| 175050 |
"eval_steps_per_second": 0.692,
|
| 175051 |
"eval_wer": 0.1971501644135915,
|
| 175052 |
"step": 59485
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 175053 |
}
|
| 175054 |
],
|
| 175055 |
"max_steps": 620000,
|
| 175056 |
"num_train_epochs": 5000,
|
| 175057 |
-
"total_flos": 1.
|
| 175058 |
"trial_name": null,
|
| 175059 |
"trial_params": null
|
| 175060 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18631571186315712,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-55501",
|
| 4 |
+
"epoch": 483.99598393574297,
|
| 5 |
+
"global_step": 60105,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 175050 |
"eval_steps_per_second": 0.692,
|
| 175051 |
"eval_wer": 0.1971501644135915,
|
| 175052 |
"step": 59485
|
| 175053 |
+
},
|
| 175054 |
+
{
|
| 175055 |
+
"epoch": 479.04,
|
| 175056 |
+
"learning_rate": 9.06301282051282e-06,
|
| 175057 |
+
"loss": 0.3356,
|
| 175058 |
+
"step": 59490
|
| 175059 |
+
},
|
| 175060 |
+
{
|
| 175061 |
+
"epoch": 479.08,
|
| 175062 |
+
"learning_rate": 9.062932692307694e-06,
|
| 175063 |
+
"loss": 0.323,
|
| 175064 |
+
"step": 59495
|
| 175065 |
+
},
|
| 175066 |
+
{
|
| 175067 |
+
"epoch": 479.12,
|
| 175068 |
+
"learning_rate": 9.062852564102565e-06,
|
| 175069 |
+
"loss": 0.315,
|
| 175070 |
+
"step": 59500
|
| 175071 |
+
},
|
| 175072 |
+
{
|
| 175073 |
+
"epoch": 479.16,
|
| 175074 |
+
"learning_rate": 9.062772435897436e-06,
|
| 175075 |
+
"loss": 0.4629,
|
| 175076 |
+
"step": 59505
|
| 175077 |
+
},
|
| 175078 |
+
{
|
| 175079 |
+
"epoch": 479.2,
|
| 175080 |
+
"learning_rate": 9.062692307692308e-06,
|
| 175081 |
+
"loss": 1.2657,
|
| 175082 |
+
"step": 59510
|
| 175083 |
+
},
|
| 175084 |
+
{
|
| 175085 |
+
"epoch": 479.24,
|
| 175086 |
+
"learning_rate": 9.06261217948718e-06,
|
| 175087 |
+
"loss": 0.4003,
|
| 175088 |
+
"step": 59515
|
| 175089 |
+
},
|
| 175090 |
+
{
|
| 175091 |
+
"epoch": 479.28,
|
| 175092 |
+
"learning_rate": 9.062532051282052e-06,
|
| 175093 |
+
"loss": 0.2867,
|
| 175094 |
+
"step": 59520
|
| 175095 |
+
},
|
| 175096 |
+
{
|
| 175097 |
+
"epoch": 479.32,
|
| 175098 |
+
"learning_rate": 9.062451923076924e-06,
|
| 175099 |
+
"loss": 0.4239,
|
| 175100 |
+
"step": 59525
|
| 175101 |
+
},
|
| 175102 |
+
{
|
| 175103 |
+
"epoch": 479.36,
|
| 175104 |
+
"learning_rate": 9.062371794871797e-06,
|
| 175105 |
+
"loss": 0.5055,
|
| 175106 |
+
"step": 59530
|
| 175107 |
+
},
|
| 175108 |
+
{
|
| 175109 |
+
"epoch": 479.4,
|
| 175110 |
+
"learning_rate": 9.062291666666668e-06,
|
| 175111 |
+
"loss": 1.3277,
|
| 175112 |
+
"step": 59535
|
| 175113 |
+
},
|
| 175114 |
+
{
|
| 175115 |
+
"epoch": 479.44,
|
| 175116 |
+
"learning_rate": 9.06221153846154e-06,
|
| 175117 |
+
"loss": 0.3746,
|
| 175118 |
+
"step": 59540
|
| 175119 |
+
},
|
| 175120 |
+
{
|
| 175121 |
+
"epoch": 479.48,
|
| 175122 |
+
"learning_rate": 9.06213141025641e-06,
|
| 175123 |
+
"loss": 0.3115,
|
| 175124 |
+
"step": 59545
|
| 175125 |
+
},
|
| 175126 |
+
{
|
| 175127 |
+
"epoch": 479.52,
|
| 175128 |
+
"learning_rate": 9.062051282051284e-06,
|
| 175129 |
+
"loss": 0.3166,
|
| 175130 |
+
"step": 59550
|
| 175131 |
+
},
|
| 175132 |
+
{
|
| 175133 |
+
"epoch": 479.56,
|
| 175134 |
+
"learning_rate": 9.061971153846153e-06,
|
| 175135 |
+
"loss": 0.4324,
|
| 175136 |
+
"step": 59555
|
| 175137 |
+
},
|
| 175138 |
+
{
|
| 175139 |
+
"epoch": 479.6,
|
| 175140 |
+
"learning_rate": 9.061891025641026e-06,
|
| 175141 |
+
"loss": 1.2776,
|
| 175142 |
+
"step": 59560
|
| 175143 |
+
},
|
| 175144 |
+
{
|
| 175145 |
+
"epoch": 479.65,
|
| 175146 |
+
"learning_rate": 9.0618108974359e-06,
|
| 175147 |
+
"loss": 0.2987,
|
| 175148 |
+
"step": 59565
|
| 175149 |
+
},
|
| 175150 |
+
{
|
| 175151 |
+
"epoch": 479.69,
|
| 175152 |
+
"learning_rate": 9.06173076923077e-06,
|
| 175153 |
+
"loss": 0.3367,
|
| 175154 |
+
"step": 59570
|
| 175155 |
+
},
|
| 175156 |
+
{
|
| 175157 |
+
"epoch": 479.73,
|
| 175158 |
+
"learning_rate": 9.061650641025642e-06,
|
| 175159 |
+
"loss": 0.3426,
|
| 175160 |
+
"step": 59575
|
| 175161 |
+
},
|
| 175162 |
+
{
|
| 175163 |
+
"epoch": 479.77,
|
| 175164 |
+
"learning_rate": 9.061570512820514e-06,
|
| 175165 |
+
"loss": 0.4634,
|
| 175166 |
+
"step": 59580
|
| 175167 |
+
},
|
| 175168 |
+
{
|
| 175169 |
+
"epoch": 479.81,
|
| 175170 |
+
"learning_rate": 9.061490384615385e-06,
|
| 175171 |
+
"loss": 1.2066,
|
| 175172 |
+
"step": 59585
|
| 175173 |
+
},
|
| 175174 |
+
{
|
| 175175 |
+
"epoch": 479.85,
|
| 175176 |
+
"learning_rate": 9.061410256410256e-06,
|
| 175177 |
+
"loss": 0.3727,
|
| 175178 |
+
"step": 59590
|
| 175179 |
+
},
|
| 175180 |
+
{
|
| 175181 |
+
"epoch": 479.89,
|
| 175182 |
+
"learning_rate": 9.06133012820513e-06,
|
| 175183 |
+
"loss": 0.2684,
|
| 175184 |
+
"step": 59595
|
| 175185 |
+
},
|
| 175186 |
+
{
|
| 175187 |
+
"epoch": 479.93,
|
| 175188 |
+
"learning_rate": 9.06125e-06,
|
| 175189 |
+
"loss": 0.3481,
|
| 175190 |
+
"step": 59600
|
| 175191 |
+
},
|
| 175192 |
+
{
|
| 175193 |
+
"epoch": 479.97,
|
| 175194 |
+
"learning_rate": 9.061169871794872e-06,
|
| 175195 |
+
"loss": 0.6479,
|
| 175196 |
+
"step": 59605
|
| 175197 |
+
},
|
| 175198 |
+
{
|
| 175199 |
+
"epoch": 480.0,
|
| 175200 |
+
"eval_loss": 0.3956400752067566,
|
| 175201 |
+
"eval_runtime": 39.9229,
|
| 175202 |
+
"eval_samples_per_second": 20.99,
|
| 175203 |
+
"eval_steps_per_second": 0.676,
|
| 175204 |
+
"eval_wer": 0.2000587544065805,
|
| 175205 |
+
"step": 59609
|
| 175206 |
+
},
|
| 175207 |
+
{
|
| 175208 |
+
"epoch": 480.01,
|
| 175209 |
+
"learning_rate": 9.061089743589743e-06,
|
| 175210 |
+
"loss": 0.443,
|
| 175211 |
+
"step": 59610
|
| 175212 |
+
},
|
| 175213 |
+
{
|
| 175214 |
+
"epoch": 480.05,
|
| 175215 |
+
"learning_rate": 9.061009615384616e-06,
|
| 175216 |
+
"loss": 0.3685,
|
| 175217 |
+
"step": 59615
|
| 175218 |
+
},
|
| 175219 |
+
{
|
| 175220 |
+
"epoch": 480.09,
|
| 175221 |
+
"learning_rate": 9.060929487179488e-06,
|
| 175222 |
+
"loss": 0.3978,
|
| 175223 |
+
"step": 59620
|
| 175224 |
+
},
|
| 175225 |
+
{
|
| 175226 |
+
"epoch": 480.13,
|
| 175227 |
+
"learning_rate": 9.06084935897436e-06,
|
| 175228 |
+
"loss": 0.3446,
|
| 175229 |
+
"step": 59625
|
| 175230 |
+
},
|
| 175231 |
+
{
|
| 175232 |
+
"epoch": 480.17,
|
| 175233 |
+
"learning_rate": 9.060769230769232e-06,
|
| 175234 |
+
"loss": 0.5786,
|
| 175235 |
+
"step": 59630
|
| 175236 |
+
},
|
| 175237 |
+
{
|
| 175238 |
+
"epoch": 480.21,
|
| 175239 |
+
"learning_rate": 9.060689102564104e-06,
|
| 175240 |
+
"loss": 1.1501,
|
| 175241 |
+
"step": 59635
|
| 175242 |
+
},
|
| 175243 |
+
{
|
| 175244 |
+
"epoch": 480.25,
|
| 175245 |
+
"learning_rate": 9.060608974358975e-06,
|
| 175246 |
+
"loss": 0.3218,
|
| 175247 |
+
"step": 59640
|
| 175248 |
+
},
|
| 175249 |
+
{
|
| 175250 |
+
"epoch": 480.29,
|
| 175251 |
+
"learning_rate": 9.060528846153846e-06,
|
| 175252 |
+
"loss": 0.3555,
|
| 175253 |
+
"step": 59645
|
| 175254 |
+
},
|
| 175255 |
+
{
|
| 175256 |
+
"epoch": 480.33,
|
| 175257 |
+
"learning_rate": 9.06044871794872e-06,
|
| 175258 |
+
"loss": 0.3923,
|
| 175259 |
+
"step": 59650
|
| 175260 |
+
},
|
| 175261 |
+
{
|
| 175262 |
+
"epoch": 480.37,
|
| 175263 |
+
"learning_rate": 9.06036858974359e-06,
|
| 175264 |
+
"loss": 0.5356,
|
| 175265 |
+
"step": 59655
|
| 175266 |
+
},
|
| 175267 |
+
{
|
| 175268 |
+
"epoch": 480.41,
|
| 175269 |
+
"learning_rate": 9.060288461538462e-06,
|
| 175270 |
+
"loss": 1.3161,
|
| 175271 |
+
"step": 59660
|
| 175272 |
+
},
|
| 175273 |
+
{
|
| 175274 |
+
"epoch": 480.45,
|
| 175275 |
+
"learning_rate": 9.060208333333335e-06,
|
| 175276 |
+
"loss": 0.3142,
|
| 175277 |
+
"step": 59665
|
| 175278 |
+
},
|
| 175279 |
+
{
|
| 175280 |
+
"epoch": 480.49,
|
| 175281 |
+
"learning_rate": 9.060128205128206e-06,
|
| 175282 |
+
"loss": 0.3776,
|
| 175283 |
+
"step": 59670
|
| 175284 |
+
},
|
| 175285 |
+
{
|
| 175286 |
+
"epoch": 480.53,
|
| 175287 |
+
"learning_rate": 9.060048076923078e-06,
|
| 175288 |
+
"loss": 0.3986,
|
| 175289 |
+
"step": 59675
|
| 175290 |
+
},
|
| 175291 |
+
{
|
| 175292 |
+
"epoch": 480.57,
|
| 175293 |
+
"learning_rate": 9.05996794871795e-06,
|
| 175294 |
+
"loss": 0.6112,
|
| 175295 |
+
"step": 59680
|
| 175296 |
+
},
|
| 175297 |
+
{
|
| 175298 |
+
"epoch": 480.61,
|
| 175299 |
+
"learning_rate": 9.059887820512822e-06,
|
| 175300 |
+
"loss": 1.047,
|
| 175301 |
+
"step": 59685
|
| 175302 |
+
},
|
| 175303 |
+
{
|
| 175304 |
+
"epoch": 480.65,
|
| 175305 |
+
"learning_rate": 9.059807692307692e-06,
|
| 175306 |
+
"loss": 0.3826,
|
| 175307 |
+
"step": 59690
|
| 175308 |
+
},
|
| 175309 |
+
{
|
| 175310 |
+
"epoch": 480.69,
|
| 175311 |
+
"learning_rate": 9.059727564102565e-06,
|
| 175312 |
+
"loss": 0.3081,
|
| 175313 |
+
"step": 59695
|
| 175314 |
+
},
|
| 175315 |
+
{
|
| 175316 |
+
"epoch": 480.73,
|
| 175317 |
+
"learning_rate": 9.059647435897436e-06,
|
| 175318 |
+
"loss": 0.36,
|
| 175319 |
+
"step": 59700
|
| 175320 |
+
},
|
| 175321 |
+
{
|
| 175322 |
+
"epoch": 480.77,
|
| 175323 |
+
"learning_rate": 9.059567307692308e-06,
|
| 175324 |
+
"loss": 0.8625,
|
| 175325 |
+
"step": 59705
|
| 175326 |
+
},
|
| 175327 |
+
{
|
| 175328 |
+
"epoch": 480.81,
|
| 175329 |
+
"learning_rate": 9.059487179487179e-06,
|
| 175330 |
+
"loss": 1.2067,
|
| 175331 |
+
"step": 59710
|
| 175332 |
+
},
|
| 175333 |
+
{
|
| 175334 |
+
"epoch": 480.85,
|
| 175335 |
+
"learning_rate": 9.059407051282052e-06,
|
| 175336 |
+
"loss": 0.3791,
|
| 175337 |
+
"step": 59715
|
| 175338 |
+
},
|
| 175339 |
+
{
|
| 175340 |
+
"epoch": 480.89,
|
| 175341 |
+
"learning_rate": 9.059326923076923e-06,
|
| 175342 |
+
"loss": 0.3461,
|
| 175343 |
+
"step": 59720
|
| 175344 |
+
},
|
| 175345 |
+
{
|
| 175346 |
+
"epoch": 480.93,
|
| 175347 |
+
"learning_rate": 9.059246794871795e-06,
|
| 175348 |
+
"loss": 0.3396,
|
| 175349 |
+
"step": 59725
|
| 175350 |
+
},
|
| 175351 |
+
{
|
| 175352 |
+
"epoch": 480.97,
|
| 175353 |
+
"learning_rate": 9.059166666666668e-06,
|
| 175354 |
+
"loss": 0.6423,
|
| 175355 |
+
"step": 59730
|
| 175356 |
+
},
|
| 175357 |
+
{
|
| 175358 |
+
"epoch": 481.0,
|
| 175359 |
+
"eval_loss": 0.41703173518180847,
|
| 175360 |
+
"eval_runtime": 38.9195,
|
| 175361 |
+
"eval_samples_per_second": 21.532,
|
| 175362 |
+
"eval_steps_per_second": 0.694,
|
| 175363 |
+
"eval_wer": 0.18889541715628672,
|
| 175364 |
+
"step": 59733
|
| 175365 |
+
},
|
| 175366 |
+
{
|
| 175367 |
+
"epoch": 481.02,
|
| 175368 |
+
"learning_rate": 9.05908653846154e-06,
|
| 175369 |
+
"loss": 0.3694,
|
| 175370 |
+
"step": 59735
|
| 175371 |
+
},
|
| 175372 |
+
{
|
| 175373 |
+
"epoch": 481.06,
|
| 175374 |
+
"learning_rate": 9.05900641025641e-06,
|
| 175375 |
+
"loss": 0.3529,
|
| 175376 |
+
"step": 59740
|
| 175377 |
+
},
|
| 175378 |
+
{
|
| 175379 |
+
"epoch": 481.1,
|
| 175380 |
+
"learning_rate": 9.058926282051282e-06,
|
| 175381 |
+
"loss": 0.3505,
|
| 175382 |
+
"step": 59745
|
| 175383 |
+
},
|
| 175384 |
+
{
|
| 175385 |
+
"epoch": 481.14,
|
| 175386 |
+
"learning_rate": 9.058846153846155e-06,
|
| 175387 |
+
"loss": 0.3736,
|
| 175388 |
+
"step": 59750
|
| 175389 |
+
},
|
| 175390 |
+
{
|
| 175391 |
+
"epoch": 481.18,
|
| 175392 |
+
"learning_rate": 9.058766025641026e-06,
|
| 175393 |
+
"loss": 0.7267,
|
| 175394 |
+
"step": 59755
|
| 175395 |
+
},
|
| 175396 |
+
{
|
| 175397 |
+
"epoch": 481.22,
|
| 175398 |
+
"learning_rate": 9.058685897435898e-06,
|
| 175399 |
+
"loss": 1.1875,
|
| 175400 |
+
"step": 59760
|
| 175401 |
+
},
|
| 175402 |
+
{
|
| 175403 |
+
"epoch": 481.26,
|
| 175404 |
+
"learning_rate": 9.05860576923077e-06,
|
| 175405 |
+
"loss": 0.3363,
|
| 175406 |
+
"step": 59765
|
| 175407 |
+
},
|
| 175408 |
+
{
|
| 175409 |
+
"epoch": 481.3,
|
| 175410 |
+
"learning_rate": 9.058525641025642e-06,
|
| 175411 |
+
"loss": 0.3393,
|
| 175412 |
+
"step": 59770
|
| 175413 |
+
},
|
| 175414 |
+
{
|
| 175415 |
+
"epoch": 481.34,
|
| 175416 |
+
"learning_rate": 9.058445512820513e-06,
|
| 175417 |
+
"loss": 0.4142,
|
| 175418 |
+
"step": 59775
|
| 175419 |
+
},
|
| 175420 |
+
{
|
| 175421 |
+
"epoch": 481.38,
|
| 175422 |
+
"learning_rate": 9.058365384615385e-06,
|
| 175423 |
+
"loss": 0.7035,
|
| 175424 |
+
"step": 59780
|
| 175425 |
+
},
|
| 175426 |
+
{
|
| 175427 |
+
"epoch": 481.42,
|
| 175428 |
+
"learning_rate": 9.058285256410258e-06,
|
| 175429 |
+
"loss": 1.1652,
|
| 175430 |
+
"step": 59785
|
| 175431 |
+
},
|
| 175432 |
+
{
|
| 175433 |
+
"epoch": 481.46,
|
| 175434 |
+
"learning_rate": 9.05820512820513e-06,
|
| 175435 |
+
"loss": 0.3585,
|
| 175436 |
+
"step": 59790
|
| 175437 |
+
},
|
| 175438 |
+
{
|
| 175439 |
+
"epoch": 481.5,
|
| 175440 |
+
"learning_rate": 9.058125e-06,
|
| 175441 |
+
"loss": 0.3514,
|
| 175442 |
+
"step": 59795
|
| 175443 |
+
},
|
| 175444 |
+
{
|
| 175445 |
+
"epoch": 481.54,
|
| 175446 |
+
"learning_rate": 9.058044871794872e-06,
|
| 175447 |
+
"loss": 0.4453,
|
| 175448 |
+
"step": 59800
|
| 175449 |
+
},
|
| 175450 |
+
{
|
| 175451 |
+
"epoch": 481.58,
|
| 175452 |
+
"learning_rate": 9.057964743589745e-06,
|
| 175453 |
+
"loss": 0.6278,
|
| 175454 |
+
"step": 59805
|
| 175455 |
+
},
|
| 175456 |
+
{
|
| 175457 |
+
"epoch": 481.62,
|
| 175458 |
+
"learning_rate": 9.057884615384616e-06,
|
| 175459 |
+
"loss": 1.0347,
|
| 175460 |
+
"step": 59810
|
| 175461 |
+
},
|
| 175462 |
+
{
|
| 175463 |
+
"epoch": 481.66,
|
| 175464 |
+
"learning_rate": 9.057804487179488e-06,
|
| 175465 |
+
"loss": 0.3393,
|
| 175466 |
+
"step": 59815
|
| 175467 |
+
},
|
| 175468 |
+
{
|
| 175469 |
+
"epoch": 481.7,
|
| 175470 |
+
"learning_rate": 9.05772435897436e-06,
|
| 175471 |
+
"loss": 0.3891,
|
| 175472 |
+
"step": 59820
|
| 175473 |
+
},
|
| 175474 |
+
{
|
| 175475 |
+
"epoch": 481.74,
|
| 175476 |
+
"learning_rate": 9.057644230769232e-06,
|
| 175477 |
+
"loss": 0.3516,
|
| 175478 |
+
"step": 59825
|
| 175479 |
+
},
|
| 175480 |
+
{
|
| 175481 |
+
"epoch": 481.78,
|
| 175482 |
+
"learning_rate": 9.057564102564104e-06,
|
| 175483 |
+
"loss": 0.6628,
|
| 175484 |
+
"step": 59830
|
| 175485 |
+
},
|
| 175486 |
+
{
|
| 175487 |
+
"epoch": 481.82,
|
| 175488 |
+
"learning_rate": 9.057483974358975e-06,
|
| 175489 |
+
"loss": 1.1032,
|
| 175490 |
+
"step": 59835
|
| 175491 |
+
},
|
| 175492 |
+
{
|
| 175493 |
+
"epoch": 481.86,
|
| 175494 |
+
"learning_rate": 9.057403846153848e-06,
|
| 175495 |
+
"loss": 0.3249,
|
| 175496 |
+
"step": 59840
|
| 175497 |
+
},
|
| 175498 |
+
{
|
| 175499 |
+
"epoch": 481.9,
|
| 175500 |
+
"learning_rate": 9.057323717948718e-06,
|
| 175501 |
+
"loss": 0.3207,
|
| 175502 |
+
"step": 59845
|
| 175503 |
+
},
|
| 175504 |
+
{
|
| 175505 |
+
"epoch": 481.94,
|
| 175506 |
+
"learning_rate": 9.05724358974359e-06,
|
| 175507 |
+
"loss": 0.4562,
|
| 175508 |
+
"step": 59850
|
| 175509 |
+
},
|
| 175510 |
+
{
|
| 175511 |
+
"epoch": 481.98,
|
| 175512 |
+
"learning_rate": 9.057163461538462e-06,
|
| 175513 |
+
"loss": 0.8911,
|
| 175514 |
+
"step": 59855
|
| 175515 |
+
},
|
| 175516 |
+
{
|
| 175517 |
+
"epoch": 482.0,
|
| 175518 |
+
"eval_loss": 0.3883751332759857,
|
| 175519 |
+
"eval_runtime": 39.1577,
|
| 175520 |
+
"eval_samples_per_second": 21.401,
|
| 175521 |
+
"eval_steps_per_second": 0.69,
|
| 175522 |
+
"eval_wer": 0.1955703532136653,
|
| 175523 |
+
"step": 59857
|
| 175524 |
+
},
|
| 175525 |
+
{
|
| 175526 |
+
"epoch": 482.02,
|
| 175527 |
+
"learning_rate": 9.057083333333333e-06,
|
| 175528 |
+
"loss": 0.3851,
|
| 175529 |
+
"step": 59860
|
| 175530 |
+
},
|
| 175531 |
+
{
|
| 175532 |
+
"epoch": 482.06,
|
| 175533 |
+
"learning_rate": 9.057003205128206e-06,
|
| 175534 |
+
"loss": 0.3364,
|
| 175535 |
+
"step": 59865
|
| 175536 |
+
},
|
| 175537 |
+
{
|
| 175538 |
+
"epoch": 482.1,
|
| 175539 |
+
"learning_rate": 9.056923076923078e-06,
|
| 175540 |
+
"loss": 0.3155,
|
| 175541 |
+
"step": 59870
|
| 175542 |
+
},
|
| 175543 |
+
{
|
| 175544 |
+
"epoch": 482.14,
|
| 175545 |
+
"learning_rate": 9.056842948717949e-06,
|
| 175546 |
+
"loss": 0.3957,
|
| 175547 |
+
"step": 59875
|
| 175548 |
+
},
|
| 175549 |
+
{
|
| 175550 |
+
"epoch": 482.18,
|
| 175551 |
+
"learning_rate": 9.05676282051282e-06,
|
| 175552 |
+
"loss": 0.7534,
|
| 175553 |
+
"step": 59880
|
| 175554 |
+
},
|
| 175555 |
+
{
|
| 175556 |
+
"epoch": 482.22,
|
| 175557 |
+
"learning_rate": 9.056682692307694e-06,
|
| 175558 |
+
"loss": 0.8385,
|
| 175559 |
+
"step": 59885
|
| 175560 |
+
},
|
| 175561 |
+
{
|
| 175562 |
+
"epoch": 482.27,
|
| 175563 |
+
"learning_rate": 9.056602564102565e-06,
|
| 175564 |
+
"loss": 0.3278,
|
| 175565 |
+
"step": 59890
|
| 175566 |
+
},
|
| 175567 |
+
{
|
| 175568 |
+
"epoch": 482.31,
|
| 175569 |
+
"learning_rate": 9.056522435897436e-06,
|
| 175570 |
+
"loss": 0.3314,
|
| 175571 |
+
"step": 59895
|
| 175572 |
+
},
|
| 175573 |
+
{
|
| 175574 |
+
"epoch": 482.35,
|
| 175575 |
+
"learning_rate": 9.056442307692308e-06,
|
| 175576 |
+
"loss": 0.4132,
|
| 175577 |
+
"step": 59900
|
| 175578 |
+
},
|
| 175579 |
+
{
|
| 175580 |
+
"epoch": 482.39,
|
| 175581 |
+
"learning_rate": 9.05636217948718e-06,
|
| 175582 |
+
"loss": 0.8778,
|
| 175583 |
+
"step": 59905
|
| 175584 |
+
},
|
| 175585 |
+
{
|
| 175586 |
+
"epoch": 482.43,
|
| 175587 |
+
"learning_rate": 9.056282051282052e-06,
|
| 175588 |
+
"loss": 0.9713,
|
| 175589 |
+
"step": 59910
|
| 175590 |
+
},
|
| 175591 |
+
{
|
| 175592 |
+
"epoch": 482.47,
|
| 175593 |
+
"learning_rate": 9.056201923076923e-06,
|
| 175594 |
+
"loss": 0.314,
|
| 175595 |
+
"step": 59915
|
| 175596 |
+
},
|
| 175597 |
+
{
|
| 175598 |
+
"epoch": 482.51,
|
| 175599 |
+
"learning_rate": 9.056121794871796e-06,
|
| 175600 |
+
"loss": 0.3701,
|
| 175601 |
+
"step": 59920
|
| 175602 |
+
},
|
| 175603 |
+
{
|
| 175604 |
+
"epoch": 482.55,
|
| 175605 |
+
"learning_rate": 9.056041666666668e-06,
|
| 175606 |
+
"loss": 0.3668,
|
| 175607 |
+
"step": 59925
|
| 175608 |
+
},
|
| 175609 |
+
{
|
| 175610 |
+
"epoch": 482.59,
|
| 175611 |
+
"learning_rate": 9.05596153846154e-06,
|
| 175612 |
+
"loss": 0.8761,
|
| 175613 |
+
"step": 59930
|
| 175614 |
+
},
|
| 175615 |
+
{
|
| 175616 |
+
"epoch": 482.63,
|
| 175617 |
+
"learning_rate": 9.05588141025641e-06,
|
| 175618 |
+
"loss": 0.9086,
|
| 175619 |
+
"step": 59935
|
| 175620 |
+
},
|
| 175621 |
+
{
|
| 175622 |
+
"epoch": 482.67,
|
| 175623 |
+
"learning_rate": 9.055801282051284e-06,
|
| 175624 |
+
"loss": 0.3652,
|
| 175625 |
+
"step": 59940
|
| 175626 |
+
},
|
| 175627 |
+
{
|
| 175628 |
+
"epoch": 482.71,
|
| 175629 |
+
"learning_rate": 9.055721153846155e-06,
|
| 175630 |
+
"loss": 0.3231,
|
| 175631 |
+
"step": 59945
|
| 175632 |
+
},
|
| 175633 |
+
{
|
| 175634 |
+
"epoch": 482.75,
|
| 175635 |
+
"learning_rate": 9.055641025641026e-06,
|
| 175636 |
+
"loss": 0.3422,
|
| 175637 |
+
"step": 59950
|
| 175638 |
+
},
|
| 175639 |
+
{
|
| 175640 |
+
"epoch": 482.79,
|
| 175641 |
+
"learning_rate": 9.055560897435898e-06,
|
| 175642 |
+
"loss": 0.8809,
|
| 175643 |
+
"step": 59955
|
| 175644 |
+
},
|
| 175645 |
+
{
|
| 175646 |
+
"epoch": 482.83,
|
| 175647 |
+
"learning_rate": 9.05548076923077e-06,
|
| 175648 |
+
"loss": 1.0296,
|
| 175649 |
+
"step": 59960
|
| 175650 |
+
},
|
| 175651 |
+
{
|
| 175652 |
+
"epoch": 482.87,
|
| 175653 |
+
"learning_rate": 9.055400641025642e-06,
|
| 175654 |
+
"loss": 0.2951,
|
| 175655 |
+
"step": 59965
|
| 175656 |
+
},
|
| 175657 |
+
{
|
| 175658 |
+
"epoch": 482.91,
|
| 175659 |
+
"learning_rate": 9.055320512820513e-06,
|
| 175660 |
+
"loss": 0.3178,
|
| 175661 |
+
"step": 59970
|
| 175662 |
+
},
|
| 175663 |
+
{
|
| 175664 |
+
"epoch": 482.95,
|
| 175665 |
+
"learning_rate": 9.055240384615386e-06,
|
| 175666 |
+
"loss": 0.5058,
|
| 175667 |
+
"step": 59975
|
| 175668 |
+
},
|
| 175669 |
+
{
|
| 175670 |
+
"epoch": 482.99,
|
| 175671 |
+
"learning_rate": 9.055160256410256e-06,
|
| 175672 |
+
"loss": 0.9992,
|
| 175673 |
+
"step": 59980
|
| 175674 |
+
},
|
| 175675 |
+
{
|
| 175676 |
+
"epoch": 483.0,
|
| 175677 |
+
"eval_loss": 0.38944196701049805,
|
| 175678 |
+
"eval_runtime": 38.8473,
|
| 175679 |
+
"eval_samples_per_second": 21.572,
|
| 175680 |
+
"eval_steps_per_second": 0.695,
|
| 175681 |
+
"eval_wer": 0.2015458655388654,
|
| 175682 |
+
"step": 59981
|
| 175683 |
+
},
|
| 175684 |
+
{
|
| 175685 |
+
"epoch": 483.03,
|
| 175686 |
+
"learning_rate": 9.05508012820513e-06,
|
| 175687 |
+
"loss": 0.3545,
|
| 175688 |
+
"step": 59985
|
| 175689 |
+
},
|
| 175690 |
+
{
|
| 175691 |
+
"epoch": 483.07,
|
| 175692 |
+
"learning_rate": 9.055e-06,
|
| 175693 |
+
"loss": 0.3728,
|
| 175694 |
+
"step": 59990
|
| 175695 |
+
},
|
| 175696 |
+
{
|
| 175697 |
+
"epoch": 483.11,
|
| 175698 |
+
"learning_rate": 9.054919871794872e-06,
|
| 175699 |
+
"loss": 0.3861,
|
| 175700 |
+
"step": 59995
|
| 175701 |
+
},
|
| 175702 |
+
{
|
| 175703 |
+
"epoch": 483.15,
|
| 175704 |
+
"learning_rate": 9.054839743589743e-06,
|
| 175705 |
+
"loss": 0.4201,
|
| 175706 |
+
"step": 60000
|
| 175707 |
+
},
|
| 175708 |
+
{
|
| 175709 |
+
"epoch": 483.19,
|
| 175710 |
+
"learning_rate": 9.054759615384616e-06,
|
| 175711 |
+
"loss": 0.9784,
|
| 175712 |
+
"step": 60005
|
| 175713 |
+
},
|
| 175714 |
+
{
|
| 175715 |
+
"epoch": 483.23,
|
| 175716 |
+
"learning_rate": 9.054679487179488e-06,
|
| 175717 |
+
"loss": 0.7261,
|
| 175718 |
+
"step": 60010
|
| 175719 |
+
},
|
| 175720 |
+
{
|
| 175721 |
+
"epoch": 483.27,
|
| 175722 |
+
"learning_rate": 9.054599358974359e-06,
|
| 175723 |
+
"loss": 0.3255,
|
| 175724 |
+
"step": 60015
|
| 175725 |
+
},
|
| 175726 |
+
{
|
| 175727 |
+
"epoch": 483.31,
|
| 175728 |
+
"learning_rate": 9.054519230769232e-06,
|
| 175729 |
+
"loss": 0.3087,
|
| 175730 |
+
"step": 60020
|
| 175731 |
+
},
|
| 175732 |
+
{
|
| 175733 |
+
"epoch": 483.35,
|
| 175734 |
+
"learning_rate": 9.054439102564103e-06,
|
| 175735 |
+
"loss": 0.4424,
|
| 175736 |
+
"step": 60025
|
| 175737 |
+
},
|
| 175738 |
+
{
|
| 175739 |
+
"epoch": 483.39,
|
| 175740 |
+
"learning_rate": 9.054358974358975e-06,
|
| 175741 |
+
"loss": 0.997,
|
| 175742 |
+
"step": 60030
|
| 175743 |
+
},
|
| 175744 |
+
{
|
| 175745 |
+
"epoch": 483.43,
|
| 175746 |
+
"learning_rate": 9.054278846153846e-06,
|
| 175747 |
+
"loss": 0.6698,
|
| 175748 |
+
"step": 60035
|
| 175749 |
+
},
|
| 175750 |
+
{
|
| 175751 |
+
"epoch": 483.47,
|
| 175752 |
+
"learning_rate": 9.05419871794872e-06,
|
| 175753 |
+
"loss": 0.3005,
|
| 175754 |
+
"step": 60040
|
| 175755 |
+
},
|
| 175756 |
+
{
|
| 175757 |
+
"epoch": 483.51,
|
| 175758 |
+
"learning_rate": 9.05411858974359e-06,
|
| 175759 |
+
"loss": 0.3648,
|
| 175760 |
+
"step": 60045
|
| 175761 |
+
},
|
| 175762 |
+
{
|
| 175763 |
+
"epoch": 483.55,
|
| 175764 |
+
"learning_rate": 9.054038461538462e-06,
|
| 175765 |
+
"loss": 0.4496,
|
| 175766 |
+
"step": 60050
|
| 175767 |
+
},
|
| 175768 |
+
{
|
| 175769 |
+
"epoch": 483.59,
|
| 175770 |
+
"learning_rate": 9.053958333333333e-06,
|
| 175771 |
+
"loss": 0.8764,
|
| 175772 |
+
"step": 60055
|
| 175773 |
+
},
|
| 175774 |
+
{
|
| 175775 |
+
"epoch": 483.63,
|
| 175776 |
+
"learning_rate": 9.053878205128206e-06,
|
| 175777 |
+
"loss": 0.7098,
|
| 175778 |
+
"step": 60060
|
| 175779 |
+
},
|
| 175780 |
+
{
|
| 175781 |
+
"epoch": 483.67,
|
| 175782 |
+
"learning_rate": 9.053798076923078e-06,
|
| 175783 |
+
"loss": 0.5857,
|
| 175784 |
+
"step": 60065
|
| 175785 |
+
},
|
| 175786 |
+
{
|
| 175787 |
+
"epoch": 483.71,
|
| 175788 |
+
"learning_rate": 9.053717948717949e-06,
|
| 175789 |
+
"loss": 0.3443,
|
| 175790 |
+
"step": 60070
|
| 175791 |
+
},
|
| 175792 |
+
{
|
| 175793 |
+
"epoch": 483.76,
|
| 175794 |
+
"learning_rate": 9.053637820512822e-06,
|
| 175795 |
+
"loss": 0.4854,
|
| 175796 |
+
"step": 60075
|
| 175797 |
+
},
|
| 175798 |
+
{
|
| 175799 |
+
"epoch": 483.8,
|
| 175800 |
+
"learning_rate": 9.053557692307693e-06,
|
| 175801 |
+
"loss": 0.9756,
|
| 175802 |
+
"step": 60080
|
| 175803 |
+
},
|
| 175804 |
+
{
|
| 175805 |
+
"epoch": 483.84,
|
| 175806 |
+
"learning_rate": 9.053477564102565e-06,
|
| 175807 |
+
"loss": 0.8236,
|
| 175808 |
+
"step": 60085
|
| 175809 |
+
},
|
| 175810 |
+
{
|
| 175811 |
+
"epoch": 483.88,
|
| 175812 |
+
"learning_rate": 9.053397435897436e-06,
|
| 175813 |
+
"loss": 0.3184,
|
| 175814 |
+
"step": 60090
|
| 175815 |
+
},
|
| 175816 |
+
{
|
| 175817 |
+
"epoch": 483.92,
|
| 175818 |
+
"learning_rate": 9.05331730769231e-06,
|
| 175819 |
+
"loss": 0.3342,
|
| 175820 |
+
"step": 60095
|
| 175821 |
+
},
|
| 175822 |
+
{
|
| 175823 |
+
"epoch": 483.96,
|
| 175824 |
+
"learning_rate": 9.05323717948718e-06,
|
| 175825 |
+
"loss": 0.4978,
|
| 175826 |
+
"step": 60100
|
| 175827 |
+
},
|
| 175828 |
+
{
|
| 175829 |
+
"epoch": 484.0,
|
| 175830 |
+
"learning_rate": 9.053157051282052e-06,
|
| 175831 |
+
"loss": 1.0842,
|
| 175832 |
+
"step": 60105
|
| 175833 |
+
},
|
| 175834 |
+
{
|
| 175835 |
+
"epoch": 484.0,
|
| 175836 |
+
"eval_loss": 0.3729969561100006,
|
| 175837 |
+
"eval_runtime": 39.728,
|
| 175838 |
+
"eval_samples_per_second": 21.119,
|
| 175839 |
+
"eval_steps_per_second": 0.68,
|
| 175840 |
+
"eval_wer": 0.19872743415211602,
|
| 175841 |
+
"step": 60105
|
| 175842 |
}
|
| 175843 |
],
|
| 175844 |
"max_steps": 620000,
|
| 175845 |
"num_train_epochs": 5000,
|
| 175846 |
+
"total_flos": 1.691543804876753e+20,
|
| 175847 |
"trial_name": null,
|
| 175848 |
"trial_params": null
|
| 175849 |
}
|
model-bin/finetune/base/{checkpoint-59485 → checkpoint-60105}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629794989.6896856/events.out.tfevents.1629794989.c435e1c5ee04.920.91
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:715dea97204d2c23c97888e2f4ea02ef7bbded30b8192fb5827ce68ef84a0e95
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629795651.0910487/events.out.tfevents.1629795651.c435e1c5ee04.920.93
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:344d2f6f8cb75265ec53925887c37b45392384698caad841e1c4f69dc593e8e6
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629796295.0343266/events.out.tfevents.1629796295.c435e1c5ee04.920.95
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:84fb7a7560f0df4a624378f4ca451fb75f1614612b37f08f0019b338918b8243
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629796939.1229146/events.out.tfevents.1629796939.c435e1c5ee04.920.97
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6ee7ffcc3306df14b63be9e17a9271b437e0aa65c675c35d0337ab6f3394206e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629797588.2766123/events.out.tfevents.1629797588.c435e1c5ee04.920.99
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8eff3e26cd882420a9b8948d65e51741724ad21352ee9a8934d819c14ac6ba63
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629794989.c435e1c5ee04.920.90
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:768931c6cc3b5c0121091290d40b99cc9a5ee99ce2bc552135915a96f5904609
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629795651.c435e1c5ee04.920.92
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5c33271ea372b104aa8a8e91c6e3d09d946035695d90aea752c6630e44d6d6f2
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629796295.c435e1c5ee04.920.94
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7dbc73f79f8a5c7a7731dda032937f0cb2b989e00e7a398b47033457395b7cb7
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629796939.c435e1c5ee04.920.96
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:007322ed4c8cfce68a4ec75dec6e0e905baf10a5ccfc4a337644dcfa0639734b
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629797588.c435e1c5ee04.920.98
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4ddaff9a42fa8cbf421a60242f954536346fd1245e43fa2c2d9bb18f804484b3
|
| 3 |
+
size 8622
|