"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-108894 → checkpoint-109516}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-108894 → checkpoint-109516}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-108894 → checkpoint-109516}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-108894 → checkpoint-109516}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-108894 → checkpoint-109516}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-108894 → checkpoint-109516}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-108894 → checkpoint-109516}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-108894 → checkpoint-109516}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-108894 → checkpoint-109516}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630038501.523142/events.out.tfevents.1630038501.52f5c7e305a3.886.41 +3 -0
- model-bin/finetune/base/log/1630038959.4217858/events.out.tfevents.1630038959.52f5c7e305a3.886.43 +3 -0
- model-bin/finetune/base/log/1630039419.2220693/events.out.tfevents.1630039419.52f5c7e305a3.886.45 +3 -0
- model-bin/finetune/base/log/1630039886.0656877/events.out.tfevents.1630039886.52f5c7e305a3.886.47 +3 -0
- model-bin/finetune/base/log/1630040347.1132815/events.out.tfevents.1630040347.52f5c7e305a3.886.49 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630038501.52f5c7e305a3.886.40 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630038959.52f5c7e305a3.886.42 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630039419.52f5c7e305a3.886.44 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630039886.52f5c7e305a3.886.46 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630040347.52f5c7e305a3.886.48 +3 -0
model-bin/finetune/base/{checkpoint-108894 → checkpoint-109516}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-108894 → checkpoint-109516}/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 sha256:449094c3a4589cfafc88ccf9c5aee1fad6b6fb1bc9fd7a1b26db1c8a887d5b97
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-108894 → checkpoint-109516}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-108894 → checkpoint-109516}/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 sha256:0c6db6a0631133291652de05f3704ed2b1bbc9427592a09cb7bb6f9bf638a798
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-108894 → checkpoint-109516}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e8d8bd78bcf53e365811adc8ff3da7658419c1c8f347160280fcdc767485f38f
|
| 3 |
+
size 14567
|
model-bin/finetune/base/{checkpoint-108894 → checkpoint-109516}/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 sha256:22ee433ca4bee95d3224c650687a9b377d4580c10728650c0f178ed2466c52ae
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-108894 → checkpoint-109516}/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:c18ec007f24c8f92d68d5a7078c003d85cec6eecdd191adcafcd27465e7379f9
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-108894 → checkpoint-109516}/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,
|
|
@@ -237909,11 +237909,806 @@
|
|
| 237909 |
"eval_steps_per_second": 0.657,
|
| 237910 |
"eval_wer": 0.17886361930496134,
|
| 237911 |
"step": 108894
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 237912 |
}
|
| 237913 |
],
|
| 237914 |
-
"max_steps":
|
| 237915 |
"num_train_epochs": 5000,
|
| 237916 |
-
"total_flos": 3.
|
| 237917 |
"trial_name": null,
|
| 237918 |
"trial_params": null
|
| 237919 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 876.0,
|
| 5 |
+
"global_step": 109516,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 237909 |
"eval_steps_per_second": 0.657,
|
| 237910 |
"eval_wer": 0.17886361930496134,
|
| 237911 |
"step": 108894
|
| 237912 |
+
},
|
| 237913 |
+
{
|
| 237914 |
+
"epoch": 878.01,
|
| 237915 |
+
"learning_rate": 8.257641357027465e-06,
|
| 237916 |
+
"loss": 0.4942,
|
| 237917 |
+
"step": 108895
|
| 237918 |
+
},
|
| 237919 |
+
{
|
| 237920 |
+
"epoch": 878.05,
|
| 237921 |
+
"learning_rate": 8.257560581583199e-06,
|
| 237922 |
+
"loss": 0.3136,
|
| 237923 |
+
"step": 108900
|
| 237924 |
+
},
|
| 237925 |
+
{
|
| 237926 |
+
"epoch": 878.09,
|
| 237927 |
+
"learning_rate": 8.257479806138935e-06,
|
| 237928 |
+
"loss": 0.2701,
|
| 237929 |
+
"step": 108905
|
| 237930 |
+
},
|
| 237931 |
+
{
|
| 237932 |
+
"epoch": 878.13,
|
| 237933 |
+
"learning_rate": 8.257399030694669e-06,
|
| 237934 |
+
"loss": 0.3846,
|
| 237935 |
+
"step": 108910
|
| 237936 |
+
},
|
| 237937 |
+
{
|
| 237938 |
+
"epoch": 878.17,
|
| 237939 |
+
"learning_rate": 8.257318255250404e-06,
|
| 237940 |
+
"loss": 0.5586,
|
| 237941 |
+
"step": 108915
|
| 237942 |
+
},
|
| 237943 |
+
{
|
| 237944 |
+
"epoch": 878.21,
|
| 237945 |
+
"learning_rate": 8.257237479806139e-06,
|
| 237946 |
+
"loss": 1.1794,
|
| 237947 |
+
"step": 108920
|
| 237948 |
+
},
|
| 237949 |
+
{
|
| 237950 |
+
"epoch": 878.25,
|
| 237951 |
+
"learning_rate": 8.257156704361874e-06,
|
| 237952 |
+
"loss": 0.2967,
|
| 237953 |
+
"step": 108925
|
| 237954 |
+
},
|
| 237955 |
+
{
|
| 237956 |
+
"epoch": 878.29,
|
| 237957 |
+
"learning_rate": 8.257075928917609e-06,
|
| 237958 |
+
"loss": 0.2592,
|
| 237959 |
+
"step": 108930
|
| 237960 |
+
},
|
| 237961 |
+
{
|
| 237962 |
+
"epoch": 878.33,
|
| 237963 |
+
"learning_rate": 8.256995153473344e-06,
|
| 237964 |
+
"loss": 0.7212,
|
| 237965 |
+
"step": 108935
|
| 237966 |
+
},
|
| 237967 |
+
{
|
| 237968 |
+
"epoch": 878.37,
|
| 237969 |
+
"learning_rate": 8.25691437802908e-06,
|
| 237970 |
+
"loss": 0.595,
|
| 237971 |
+
"step": 108940
|
| 237972 |
+
},
|
| 237973 |
+
{
|
| 237974 |
+
"epoch": 878.41,
|
| 237975 |
+
"learning_rate": 8.256833602584814e-06,
|
| 237976 |
+
"loss": 1.1496,
|
| 237977 |
+
"step": 108945
|
| 237978 |
+
},
|
| 237979 |
+
{
|
| 237980 |
+
"epoch": 878.45,
|
| 237981 |
+
"learning_rate": 8.25675282714055e-06,
|
| 237982 |
+
"loss": 0.2888,
|
| 237983 |
+
"step": 108950
|
| 237984 |
+
},
|
| 237985 |
+
{
|
| 237986 |
+
"epoch": 878.49,
|
| 237987 |
+
"learning_rate": 8.256672051696284e-06,
|
| 237988 |
+
"loss": 0.4588,
|
| 237989 |
+
"step": 108955
|
| 237990 |
+
},
|
| 237991 |
+
{
|
| 237992 |
+
"epoch": 878.53,
|
| 237993 |
+
"learning_rate": 8.25659127625202e-06,
|
| 237994 |
+
"loss": 0.3674,
|
| 237995 |
+
"step": 108960
|
| 237996 |
+
},
|
| 237997 |
+
{
|
| 237998 |
+
"epoch": 878.57,
|
| 237999 |
+
"learning_rate": 8.256510500807754e-06,
|
| 238000 |
+
"loss": 0.6112,
|
| 238001 |
+
"step": 108965
|
| 238002 |
+
},
|
| 238003 |
+
{
|
| 238004 |
+
"epoch": 878.61,
|
| 238005 |
+
"learning_rate": 8.25642972536349e-06,
|
| 238006 |
+
"loss": 1.0556,
|
| 238007 |
+
"step": 108970
|
| 238008 |
+
},
|
| 238009 |
+
{
|
| 238010 |
+
"epoch": 878.65,
|
| 238011 |
+
"learning_rate": 8.256348949919224e-06,
|
| 238012 |
+
"loss": 0.2855,
|
| 238013 |
+
"step": 108975
|
| 238014 |
+
},
|
| 238015 |
+
{
|
| 238016 |
+
"epoch": 878.69,
|
| 238017 |
+
"learning_rate": 8.25626817447496e-06,
|
| 238018 |
+
"loss": 0.2686,
|
| 238019 |
+
"step": 108980
|
| 238020 |
+
},
|
| 238021 |
+
{
|
| 238022 |
+
"epoch": 878.73,
|
| 238023 |
+
"learning_rate": 8.256187399030694e-06,
|
| 238024 |
+
"loss": 0.3335,
|
| 238025 |
+
"step": 108985
|
| 238026 |
+
},
|
| 238027 |
+
{
|
| 238028 |
+
"epoch": 878.77,
|
| 238029 |
+
"learning_rate": 8.25610662358643e-06,
|
| 238030 |
+
"loss": 0.534,
|
| 238031 |
+
"step": 108990
|
| 238032 |
+
},
|
| 238033 |
+
{
|
| 238034 |
+
"epoch": 878.81,
|
| 238035 |
+
"learning_rate": 8.256025848142164e-06,
|
| 238036 |
+
"loss": 1.018,
|
| 238037 |
+
"step": 108995
|
| 238038 |
+
},
|
| 238039 |
+
{
|
| 238040 |
+
"epoch": 878.85,
|
| 238041 |
+
"learning_rate": 8.2559450726979e-06,
|
| 238042 |
+
"loss": 0.3416,
|
| 238043 |
+
"step": 109000
|
| 238044 |
+
},
|
| 238045 |
+
{
|
| 238046 |
+
"epoch": 878.89,
|
| 238047 |
+
"learning_rate": 8.255864297253636e-06,
|
| 238048 |
+
"loss": 0.2572,
|
| 238049 |
+
"step": 109005
|
| 238050 |
+
},
|
| 238051 |
+
{
|
| 238052 |
+
"epoch": 878.93,
|
| 238053 |
+
"learning_rate": 8.25578352180937e-06,
|
| 238054 |
+
"loss": 0.3844,
|
| 238055 |
+
"step": 109010
|
| 238056 |
+
},
|
| 238057 |
+
{
|
| 238058 |
+
"epoch": 878.97,
|
| 238059 |
+
"learning_rate": 8.255702746365106e-06,
|
| 238060 |
+
"loss": 0.5158,
|
| 238061 |
+
"step": 109015
|
| 238062 |
+
},
|
| 238063 |
+
{
|
| 238064 |
+
"epoch": 879.0,
|
| 238065 |
+
"eval_loss": 0.3908926248550415,
|
| 238066 |
+
"eval_runtime": 42.6009,
|
| 238067 |
+
"eval_samples_per_second": 19.647,
|
| 238068 |
+
"eval_steps_per_second": 0.634,
|
| 238069 |
+
"eval_wer": 0.1916660648515924,
|
| 238070 |
+
"step": 109018
|
| 238071 |
+
},
|
| 238072 |
+
{
|
| 238073 |
+
"epoch": 879.02,
|
| 238074 |
+
"learning_rate": 8.25562197092084e-06,
|
| 238075 |
+
"loss": 0.4281,
|
| 238076 |
+
"step": 109020
|
| 238077 |
+
},
|
| 238078 |
+
{
|
| 238079 |
+
"epoch": 879.06,
|
| 238080 |
+
"learning_rate": 8.255541195476576e-06,
|
| 238081 |
+
"loss": 0.2807,
|
| 238082 |
+
"step": 109025
|
| 238083 |
+
},
|
| 238084 |
+
{
|
| 238085 |
+
"epoch": 879.1,
|
| 238086 |
+
"learning_rate": 8.25546042003231e-06,
|
| 238087 |
+
"loss": 0.2722,
|
| 238088 |
+
"step": 109030
|
| 238089 |
+
},
|
| 238090 |
+
{
|
| 238091 |
+
"epoch": 879.14,
|
| 238092 |
+
"learning_rate": 8.255379644588046e-06,
|
| 238093 |
+
"loss": 0.3234,
|
| 238094 |
+
"step": 109035
|
| 238095 |
+
},
|
| 238096 |
+
{
|
| 238097 |
+
"epoch": 879.18,
|
| 238098 |
+
"learning_rate": 8.25529886914378e-06,
|
| 238099 |
+
"loss": 0.6337,
|
| 238100 |
+
"step": 109040
|
| 238101 |
+
},
|
| 238102 |
+
{
|
| 238103 |
+
"epoch": 879.22,
|
| 238104 |
+
"learning_rate": 8.255218093699516e-06,
|
| 238105 |
+
"loss": 0.9453,
|
| 238106 |
+
"step": 109045
|
| 238107 |
+
},
|
| 238108 |
+
{
|
| 238109 |
+
"epoch": 879.26,
|
| 238110 |
+
"learning_rate": 8.25513731825525e-06,
|
| 238111 |
+
"loss": 0.3167,
|
| 238112 |
+
"step": 109050
|
| 238113 |
+
},
|
| 238114 |
+
{
|
| 238115 |
+
"epoch": 879.3,
|
| 238116 |
+
"learning_rate": 8.255056542810986e-06,
|
| 238117 |
+
"loss": 0.3273,
|
| 238118 |
+
"step": 109055
|
| 238119 |
+
},
|
| 238120 |
+
{
|
| 238121 |
+
"epoch": 879.34,
|
| 238122 |
+
"learning_rate": 8.25497576736672e-06,
|
| 238123 |
+
"loss": 0.3399,
|
| 238124 |
+
"step": 109060
|
| 238125 |
+
},
|
| 238126 |
+
{
|
| 238127 |
+
"epoch": 879.38,
|
| 238128 |
+
"learning_rate": 8.254894991922456e-06,
|
| 238129 |
+
"loss": 0.6348,
|
| 238130 |
+
"step": 109065
|
| 238131 |
+
},
|
| 238132 |
+
{
|
| 238133 |
+
"epoch": 879.42,
|
| 238134 |
+
"learning_rate": 8.254814216478192e-06,
|
| 238135 |
+
"loss": 1.0596,
|
| 238136 |
+
"step": 109070
|
| 238137 |
+
},
|
| 238138 |
+
{
|
| 238139 |
+
"epoch": 879.46,
|
| 238140 |
+
"learning_rate": 8.254733441033926e-06,
|
| 238141 |
+
"loss": 0.2522,
|
| 238142 |
+
"step": 109075
|
| 238143 |
+
},
|
| 238144 |
+
{
|
| 238145 |
+
"epoch": 879.5,
|
| 238146 |
+
"learning_rate": 8.254652665589662e-06,
|
| 238147 |
+
"loss": 0.2868,
|
| 238148 |
+
"step": 109080
|
| 238149 |
+
},
|
| 238150 |
+
{
|
| 238151 |
+
"epoch": 879.54,
|
| 238152 |
+
"learning_rate": 8.254571890145396e-06,
|
| 238153 |
+
"loss": 0.3137,
|
| 238154 |
+
"step": 109085
|
| 238155 |
+
},
|
| 238156 |
+
{
|
| 238157 |
+
"epoch": 879.58,
|
| 238158 |
+
"learning_rate": 8.254491114701132e-06,
|
| 238159 |
+
"loss": 0.6219,
|
| 238160 |
+
"step": 109090
|
| 238161 |
+
},
|
| 238162 |
+
{
|
| 238163 |
+
"epoch": 879.62,
|
| 238164 |
+
"learning_rate": 8.254410339256866e-06,
|
| 238165 |
+
"loss": 0.9126,
|
| 238166 |
+
"step": 109095
|
| 238167 |
+
},
|
| 238168 |
+
{
|
| 238169 |
+
"epoch": 879.66,
|
| 238170 |
+
"learning_rate": 8.254329563812602e-06,
|
| 238171 |
+
"loss": 0.3179,
|
| 238172 |
+
"step": 109100
|
| 238173 |
+
},
|
| 238174 |
+
{
|
| 238175 |
+
"epoch": 879.7,
|
| 238176 |
+
"learning_rate": 8.254248788368336e-06,
|
| 238177 |
+
"loss": 0.2617,
|
| 238178 |
+
"step": 109105
|
| 238179 |
+
},
|
| 238180 |
+
{
|
| 238181 |
+
"epoch": 879.74,
|
| 238182 |
+
"learning_rate": 8.254168012924072e-06,
|
| 238183 |
+
"loss": 0.3084,
|
| 238184 |
+
"step": 109110
|
| 238185 |
+
},
|
| 238186 |
+
{
|
| 238187 |
+
"epoch": 879.78,
|
| 238188 |
+
"learning_rate": 8.254087237479806e-06,
|
| 238189 |
+
"loss": 0.5737,
|
| 238190 |
+
"step": 109115
|
| 238191 |
+
},
|
| 238192 |
+
{
|
| 238193 |
+
"epoch": 879.82,
|
| 238194 |
+
"learning_rate": 8.254006462035542e-06,
|
| 238195 |
+
"loss": 1.0034,
|
| 238196 |
+
"step": 109120
|
| 238197 |
+
},
|
| 238198 |
+
{
|
| 238199 |
+
"epoch": 879.86,
|
| 238200 |
+
"learning_rate": 8.253925686591276e-06,
|
| 238201 |
+
"loss": 0.2856,
|
| 238202 |
+
"step": 109125
|
| 238203 |
+
},
|
| 238204 |
+
{
|
| 238205 |
+
"epoch": 879.9,
|
| 238206 |
+
"learning_rate": 8.253844911147011e-06,
|
| 238207 |
+
"loss": 0.2544,
|
| 238208 |
+
"step": 109130
|
| 238209 |
+
},
|
| 238210 |
+
{
|
| 238211 |
+
"epoch": 879.94,
|
| 238212 |
+
"learning_rate": 8.253764135702747e-06,
|
| 238213 |
+
"loss": 0.3562,
|
| 238214 |
+
"step": 109135
|
| 238215 |
+
},
|
| 238216 |
+
{
|
| 238217 |
+
"epoch": 879.98,
|
| 238218 |
+
"learning_rate": 8.253683360258481e-06,
|
| 238219 |
+
"loss": 0.6451,
|
| 238220 |
+
"step": 109140
|
| 238221 |
+
},
|
| 238222 |
+
{
|
| 238223 |
+
"epoch": 880.0,
|
| 238224 |
+
"eval_loss": 0.4516902267932892,
|
| 238225 |
+
"eval_runtime": 42.7375,
|
| 238226 |
+
"eval_samples_per_second": 19.608,
|
| 238227 |
+
"eval_steps_per_second": 0.632,
|
| 238228 |
+
"eval_wer": 0.1786863662960796,
|
| 238229 |
+
"step": 109142
|
| 238230 |
+
},
|
| 238231 |
+
{
|
| 238232 |
+
"epoch": 873.02,
|
| 238233 |
+
"learning_rate": 8.253602584814217e-06,
|
| 238234 |
+
"loss": 0.3657,
|
| 238235 |
+
"step": 109145
|
| 238236 |
+
},
|
| 238237 |
+
{
|
| 238238 |
+
"epoch": 873.06,
|
| 238239 |
+
"learning_rate": 8.253521809369951e-06,
|
| 238240 |
+
"loss": 0.2349,
|
| 238241 |
+
"step": 109150
|
| 238242 |
+
},
|
| 238243 |
+
{
|
| 238244 |
+
"epoch": 873.1,
|
| 238245 |
+
"learning_rate": 8.253441033925687e-06,
|
| 238246 |
+
"loss": 0.3169,
|
| 238247 |
+
"step": 109155
|
| 238248 |
+
},
|
| 238249 |
+
{
|
| 238250 |
+
"epoch": 873.14,
|
| 238251 |
+
"learning_rate": 8.253360258481421e-06,
|
| 238252 |
+
"loss": 0.3745,
|
| 238253 |
+
"step": 109160
|
| 238254 |
+
},
|
| 238255 |
+
{
|
| 238256 |
+
"epoch": 873.18,
|
| 238257 |
+
"learning_rate": 8.253279483037157e-06,
|
| 238258 |
+
"loss": 0.7199,
|
| 238259 |
+
"step": 109165
|
| 238260 |
+
},
|
| 238261 |
+
{
|
| 238262 |
+
"epoch": 873.22,
|
| 238263 |
+
"learning_rate": 8.253198707592891e-06,
|
| 238264 |
+
"loss": 1.048,
|
| 238265 |
+
"step": 109170
|
| 238266 |
+
},
|
| 238267 |
+
{
|
| 238268 |
+
"epoch": 873.26,
|
| 238269 |
+
"learning_rate": 8.253117932148627e-06,
|
| 238270 |
+
"loss": 0.2758,
|
| 238271 |
+
"step": 109175
|
| 238272 |
+
},
|
| 238273 |
+
{
|
| 238274 |
+
"epoch": 873.3,
|
| 238275 |
+
"learning_rate": 8.253037156704361e-06,
|
| 238276 |
+
"loss": 0.2872,
|
| 238277 |
+
"step": 109180
|
| 238278 |
+
},
|
| 238279 |
+
{
|
| 238280 |
+
"epoch": 873.34,
|
| 238281 |
+
"learning_rate": 8.252956381260097e-06,
|
| 238282 |
+
"loss": 0.3778,
|
| 238283 |
+
"step": 109185
|
| 238284 |
+
},
|
| 238285 |
+
{
|
| 238286 |
+
"epoch": 873.38,
|
| 238287 |
+
"learning_rate": 8.252875605815831e-06,
|
| 238288 |
+
"loss": 0.8357,
|
| 238289 |
+
"step": 109190
|
| 238290 |
+
},
|
| 238291 |
+
{
|
| 238292 |
+
"epoch": 873.42,
|
| 238293 |
+
"learning_rate": 8.252794830371567e-06,
|
| 238294 |
+
"loss": 0.9124,
|
| 238295 |
+
"step": 109195
|
| 238296 |
+
},
|
| 238297 |
+
{
|
| 238298 |
+
"epoch": 873.46,
|
| 238299 |
+
"learning_rate": 8.252714054927303e-06,
|
| 238300 |
+
"loss": 0.2622,
|
| 238301 |
+
"step": 109200
|
| 238302 |
+
},
|
| 238303 |
+
{
|
| 238304 |
+
"epoch": 873.5,
|
| 238305 |
+
"learning_rate": 8.252633279483037e-06,
|
| 238306 |
+
"loss": 0.2493,
|
| 238307 |
+
"step": 109205
|
| 238308 |
+
},
|
| 238309 |
+
{
|
| 238310 |
+
"epoch": 873.54,
|
| 238311 |
+
"learning_rate": 8.252552504038773e-06,
|
| 238312 |
+
"loss": 0.3982,
|
| 238313 |
+
"step": 109210
|
| 238314 |
+
},
|
| 238315 |
+
{
|
| 238316 |
+
"epoch": 873.58,
|
| 238317 |
+
"learning_rate": 8.252471728594507e-06,
|
| 238318 |
+
"loss": 0.7649,
|
| 238319 |
+
"step": 109215
|
| 238320 |
+
},
|
| 238321 |
+
{
|
| 238322 |
+
"epoch": 873.62,
|
| 238323 |
+
"learning_rate": 8.252390953150243e-06,
|
| 238324 |
+
"loss": 0.9403,
|
| 238325 |
+
"step": 109220
|
| 238326 |
+
},
|
| 238327 |
+
{
|
| 238328 |
+
"epoch": 873.66,
|
| 238329 |
+
"learning_rate": 8.252310177705977e-06,
|
| 238330 |
+
"loss": 0.2741,
|
| 238331 |
+
"step": 109225
|
| 238332 |
+
},
|
| 238333 |
+
{
|
| 238334 |
+
"epoch": 873.7,
|
| 238335 |
+
"learning_rate": 8.252229402261713e-06,
|
| 238336 |
+
"loss": 0.2522,
|
| 238337 |
+
"step": 109230
|
| 238338 |
+
},
|
| 238339 |
+
{
|
| 238340 |
+
"epoch": 873.74,
|
| 238341 |
+
"learning_rate": 8.252148626817447e-06,
|
| 238342 |
+
"loss": 0.3822,
|
| 238343 |
+
"step": 109235
|
| 238344 |
+
},
|
| 238345 |
+
{
|
| 238346 |
+
"epoch": 873.78,
|
| 238347 |
+
"learning_rate": 8.252067851373183e-06,
|
| 238348 |
+
"loss": 0.7637,
|
| 238349 |
+
"step": 109240
|
| 238350 |
+
},
|
| 238351 |
+
{
|
| 238352 |
+
"epoch": 873.82,
|
| 238353 |
+
"learning_rate": 8.251987075928917e-06,
|
| 238354 |
+
"loss": 0.8275,
|
| 238355 |
+
"step": 109245
|
| 238356 |
+
},
|
| 238357 |
+
{
|
| 238358 |
+
"epoch": 873.86,
|
| 238359 |
+
"learning_rate": 8.251906300484653e-06,
|
| 238360 |
+
"loss": 0.3128,
|
| 238361 |
+
"step": 109250
|
| 238362 |
+
},
|
| 238363 |
+
{
|
| 238364 |
+
"epoch": 873.9,
|
| 238365 |
+
"learning_rate": 8.251825525040389e-06,
|
| 238366 |
+
"loss": 0.3143,
|
| 238367 |
+
"step": 109255
|
| 238368 |
+
},
|
| 238369 |
+
{
|
| 238370 |
+
"epoch": 873.94,
|
| 238371 |
+
"learning_rate": 8.251744749596123e-06,
|
| 238372 |
+
"loss": 0.4088,
|
| 238373 |
+
"step": 109260
|
| 238374 |
+
},
|
| 238375 |
+
{
|
| 238376 |
+
"epoch": 873.98,
|
| 238377 |
+
"learning_rate": 8.251663974151859e-06,
|
| 238378 |
+
"loss": 0.7688,
|
| 238379 |
+
"step": 109265
|
| 238380 |
+
},
|
| 238381 |
+
{
|
| 238382 |
+
"epoch": 874.0,
|
| 238383 |
+
"eval_loss": 0.4127758741378784,
|
| 238384 |
+
"eval_runtime": 42.9711,
|
| 238385 |
+
"eval_samples_per_second": 19.501,
|
| 238386 |
+
"eval_steps_per_second": 0.628,
|
| 238387 |
+
"eval_wer": 0.18809031877213694,
|
| 238388 |
+
"step": 109267
|
| 238389 |
+
},
|
| 238390 |
+
{
|
| 238391 |
+
"epoch": 881.02,
|
| 238392 |
+
"learning_rate": 8.251583198707593e-06,
|
| 238393 |
+
"loss": 0.3105,
|
| 238394 |
+
"step": 109270
|
| 238395 |
+
},
|
| 238396 |
+
{
|
| 238397 |
+
"epoch": 881.06,
|
| 238398 |
+
"learning_rate": 8.251502423263329e-06,
|
| 238399 |
+
"loss": 0.4217,
|
| 238400 |
+
"step": 109275
|
| 238401 |
+
},
|
| 238402 |
+
{
|
| 238403 |
+
"epoch": 881.1,
|
| 238404 |
+
"learning_rate": 8.251421647819063e-06,
|
| 238405 |
+
"loss": 0.2634,
|
| 238406 |
+
"step": 109280
|
| 238407 |
+
},
|
| 238408 |
+
{
|
| 238409 |
+
"epoch": 881.14,
|
| 238410 |
+
"learning_rate": 8.251340872374799e-06,
|
| 238411 |
+
"loss": 0.3549,
|
| 238412 |
+
"step": 109285
|
| 238413 |
+
},
|
| 238414 |
+
{
|
| 238415 |
+
"epoch": 881.18,
|
| 238416 |
+
"learning_rate": 8.251260096930533e-06,
|
| 238417 |
+
"loss": 0.7516,
|
| 238418 |
+
"step": 109290
|
| 238419 |
+
},
|
| 238420 |
+
{
|
| 238421 |
+
"epoch": 881.22,
|
| 238422 |
+
"learning_rate": 8.251179321486269e-06,
|
| 238423 |
+
"loss": 0.8462,
|
| 238424 |
+
"step": 109295
|
| 238425 |
+
},
|
| 238426 |
+
{
|
| 238427 |
+
"epoch": 881.27,
|
| 238428 |
+
"learning_rate": 8.251098546042003e-06,
|
| 238429 |
+
"loss": 0.2915,
|
| 238430 |
+
"step": 109300
|
| 238431 |
+
},
|
| 238432 |
+
{
|
| 238433 |
+
"epoch": 881.31,
|
| 238434 |
+
"learning_rate": 8.251017770597739e-06,
|
| 238435 |
+
"loss": 0.2822,
|
| 238436 |
+
"step": 109305
|
| 238437 |
+
},
|
| 238438 |
+
{
|
| 238439 |
+
"epoch": 881.35,
|
| 238440 |
+
"learning_rate": 8.250936995153474e-06,
|
| 238441 |
+
"loss": 0.3588,
|
| 238442 |
+
"step": 109310
|
| 238443 |
+
},
|
| 238444 |
+
{
|
| 238445 |
+
"epoch": 881.39,
|
| 238446 |
+
"learning_rate": 8.250856219709209e-06,
|
| 238447 |
+
"loss": 0.772,
|
| 238448 |
+
"step": 109315
|
| 238449 |
+
},
|
| 238450 |
+
{
|
| 238451 |
+
"epoch": 881.43,
|
| 238452 |
+
"learning_rate": 8.250775444264944e-06,
|
| 238453 |
+
"loss": 0.8989,
|
| 238454 |
+
"step": 109320
|
| 238455 |
+
},
|
| 238456 |
+
{
|
| 238457 |
+
"epoch": 881.47,
|
| 238458 |
+
"learning_rate": 8.250694668820679e-06,
|
| 238459 |
+
"loss": 0.2993,
|
| 238460 |
+
"step": 109325
|
| 238461 |
+
},
|
| 238462 |
+
{
|
| 238463 |
+
"epoch": 881.51,
|
| 238464 |
+
"learning_rate": 8.250613893376414e-06,
|
| 238465 |
+
"loss": 0.2748,
|
| 238466 |
+
"step": 109330
|
| 238467 |
+
},
|
| 238468 |
+
{
|
| 238469 |
+
"epoch": 881.55,
|
| 238470 |
+
"learning_rate": 8.250533117932149e-06,
|
| 238471 |
+
"loss": 0.3367,
|
| 238472 |
+
"step": 109335
|
| 238473 |
+
},
|
| 238474 |
+
{
|
| 238475 |
+
"epoch": 881.59,
|
| 238476 |
+
"learning_rate": 8.250452342487884e-06,
|
| 238477 |
+
"loss": 0.6681,
|
| 238478 |
+
"step": 109340
|
| 238479 |
+
},
|
| 238480 |
+
{
|
| 238481 |
+
"epoch": 881.63,
|
| 238482 |
+
"learning_rate": 8.250371567043619e-06,
|
| 238483 |
+
"loss": 0.7778,
|
| 238484 |
+
"step": 109345
|
| 238485 |
+
},
|
| 238486 |
+
{
|
| 238487 |
+
"epoch": 881.67,
|
| 238488 |
+
"learning_rate": 8.250290791599354e-06,
|
| 238489 |
+
"loss": 0.2912,
|
| 238490 |
+
"step": 109350
|
| 238491 |
+
},
|
| 238492 |
+
{
|
| 238493 |
+
"epoch": 881.71,
|
| 238494 |
+
"learning_rate": 8.250210016155088e-06,
|
| 238495 |
+
"loss": 0.2896,
|
| 238496 |
+
"step": 109355
|
| 238497 |
+
},
|
| 238498 |
+
{
|
| 238499 |
+
"epoch": 881.75,
|
| 238500 |
+
"learning_rate": 8.250129240710824e-06,
|
| 238501 |
+
"loss": 0.3983,
|
| 238502 |
+
"step": 109360
|
| 238503 |
+
},
|
| 238504 |
+
{
|
| 238505 |
+
"epoch": 881.79,
|
| 238506 |
+
"learning_rate": 8.250048465266558e-06,
|
| 238507 |
+
"loss": 0.762,
|
| 238508 |
+
"step": 109365
|
| 238509 |
+
},
|
| 238510 |
+
{
|
| 238511 |
+
"epoch": 881.83,
|
| 238512 |
+
"learning_rate": 8.249967689822294e-06,
|
| 238513 |
+
"loss": 1.0731,
|
| 238514 |
+
"step": 109370
|
| 238515 |
+
},
|
| 238516 |
+
{
|
| 238517 |
+
"epoch": 881.87,
|
| 238518 |
+
"learning_rate": 8.24988691437803e-06,
|
| 238519 |
+
"loss": 0.2488,
|
| 238520 |
+
"step": 109375
|
| 238521 |
+
},
|
| 238522 |
+
{
|
| 238523 |
+
"epoch": 881.91,
|
| 238524 |
+
"learning_rate": 8.249806138933764e-06,
|
| 238525 |
+
"loss": 0.4398,
|
| 238526 |
+
"step": 109380
|
| 238527 |
+
},
|
| 238528 |
+
{
|
| 238529 |
+
"epoch": 881.95,
|
| 238530 |
+
"learning_rate": 8.2497253634895e-06,
|
| 238531 |
+
"loss": 0.3585,
|
| 238532 |
+
"step": 109385
|
| 238533 |
+
},
|
| 238534 |
+
{
|
| 238535 |
+
"epoch": 881.99,
|
| 238536 |
+
"learning_rate": 8.249644588045234e-06,
|
| 238537 |
+
"loss": 0.8093,
|
| 238538 |
+
"step": 109390
|
| 238539 |
+
},
|
| 238540 |
+
{
|
| 238541 |
+
"epoch": 882.0,
|
| 238542 |
+
"eval_loss": 0.36528652906417847,
|
| 238543 |
+
"eval_runtime": 41.9391,
|
| 238544 |
+
"eval_samples_per_second": 19.981,
|
| 238545 |
+
"eval_steps_per_second": 0.644,
|
| 238546 |
+
"eval_wer": 0.1787843536425767,
|
| 238547 |
+
"step": 109391
|
| 238548 |
+
},
|
| 238549 |
+
{
|
| 238550 |
+
"epoch": 875.03,
|
| 238551 |
+
"learning_rate": 8.24956381260097e-06,
|
| 238552 |
+
"loss": 0.3458,
|
| 238553 |
+
"step": 109395
|
| 238554 |
+
},
|
| 238555 |
+
{
|
| 238556 |
+
"epoch": 875.07,
|
| 238557 |
+
"learning_rate": 8.249483037156704e-06,
|
| 238558 |
+
"loss": 0.2919,
|
| 238559 |
+
"step": 109400
|
| 238560 |
+
},
|
| 238561 |
+
{
|
| 238562 |
+
"epoch": 875.11,
|
| 238563 |
+
"learning_rate": 8.24940226171244e-06,
|
| 238564 |
+
"loss": 0.2769,
|
| 238565 |
+
"step": 109405
|
| 238566 |
+
},
|
| 238567 |
+
{
|
| 238568 |
+
"epoch": 875.15,
|
| 238569 |
+
"learning_rate": 8.249321486268174e-06,
|
| 238570 |
+
"loss": 0.3908,
|
| 238571 |
+
"step": 109410
|
| 238572 |
+
},
|
| 238573 |
+
{
|
| 238574 |
+
"epoch": 875.19,
|
| 238575 |
+
"learning_rate": 8.24924071082391e-06,
|
| 238576 |
+
"loss": 0.9022,
|
| 238577 |
+
"step": 109415
|
| 238578 |
+
},
|
| 238579 |
+
{
|
| 238580 |
+
"epoch": 875.23,
|
| 238581 |
+
"learning_rate": 8.249159935379644e-06,
|
| 238582 |
+
"loss": 0.6427,
|
| 238583 |
+
"step": 109420
|
| 238584 |
+
},
|
| 238585 |
+
{
|
| 238586 |
+
"epoch": 875.27,
|
| 238587 |
+
"learning_rate": 8.24907915993538e-06,
|
| 238588 |
+
"loss": 0.2571,
|
| 238589 |
+
"step": 109425
|
| 238590 |
+
},
|
| 238591 |
+
{
|
| 238592 |
+
"epoch": 875.31,
|
| 238593 |
+
"learning_rate": 8.248998384491116e-06,
|
| 238594 |
+
"loss": 0.3107,
|
| 238595 |
+
"step": 109430
|
| 238596 |
+
},
|
| 238597 |
+
{
|
| 238598 |
+
"epoch": 875.35,
|
| 238599 |
+
"learning_rate": 8.24891760904685e-06,
|
| 238600 |
+
"loss": 0.4479,
|
| 238601 |
+
"step": 109435
|
| 238602 |
+
},
|
| 238603 |
+
{
|
| 238604 |
+
"epoch": 875.39,
|
| 238605 |
+
"learning_rate": 8.248836833602586e-06,
|
| 238606 |
+
"loss": 0.9061,
|
| 238607 |
+
"step": 109440
|
| 238608 |
+
},
|
| 238609 |
+
{
|
| 238610 |
+
"epoch": 875.43,
|
| 238611 |
+
"learning_rate": 8.24875605815832e-06,
|
| 238612 |
+
"loss": 0.5396,
|
| 238613 |
+
"step": 109445
|
| 238614 |
+
},
|
| 238615 |
+
{
|
| 238616 |
+
"epoch": 875.47,
|
| 238617 |
+
"learning_rate": 8.248675282714056e-06,
|
| 238618 |
+
"loss": 0.2692,
|
| 238619 |
+
"step": 109450
|
| 238620 |
+
},
|
| 238621 |
+
{
|
| 238622 |
+
"epoch": 875.51,
|
| 238623 |
+
"learning_rate": 8.24859450726979e-06,
|
| 238624 |
+
"loss": 0.3434,
|
| 238625 |
+
"step": 109455
|
| 238626 |
+
},
|
| 238627 |
+
{
|
| 238628 |
+
"epoch": 875.55,
|
| 238629 |
+
"learning_rate": 8.248513731825526e-06,
|
| 238630 |
+
"loss": 0.3502,
|
| 238631 |
+
"step": 109460
|
| 238632 |
+
},
|
| 238633 |
+
{
|
| 238634 |
+
"epoch": 875.59,
|
| 238635 |
+
"learning_rate": 8.24843295638126e-06,
|
| 238636 |
+
"loss": 0.9187,
|
| 238637 |
+
"step": 109465
|
| 238638 |
+
},
|
| 238639 |
+
{
|
| 238640 |
+
"epoch": 875.63,
|
| 238641 |
+
"learning_rate": 8.248352180936996e-06,
|
| 238642 |
+
"loss": 0.7583,
|
| 238643 |
+
"step": 109470
|
| 238644 |
+
},
|
| 238645 |
+
{
|
| 238646 |
+
"epoch": 875.67,
|
| 238647 |
+
"learning_rate": 8.24827140549273e-06,
|
| 238648 |
+
"loss": 0.2532,
|
| 238649 |
+
"step": 109475
|
| 238650 |
+
},
|
| 238651 |
+
{
|
| 238652 |
+
"epoch": 875.71,
|
| 238653 |
+
"learning_rate": 8.248190630048466e-06,
|
| 238654 |
+
"loss": 0.2976,
|
| 238655 |
+
"step": 109480
|
| 238656 |
+
},
|
| 238657 |
+
{
|
| 238658 |
+
"epoch": 875.75,
|
| 238659 |
+
"learning_rate": 8.248109854604202e-06,
|
| 238660 |
+
"loss": 0.3528,
|
| 238661 |
+
"step": 109485
|
| 238662 |
+
},
|
| 238663 |
+
{
|
| 238664 |
+
"epoch": 875.79,
|
| 238665 |
+
"learning_rate": 8.248029079159936e-06,
|
| 238666 |
+
"loss": 0.9945,
|
| 238667 |
+
"step": 109490
|
| 238668 |
+
},
|
| 238669 |
+
{
|
| 238670 |
+
"epoch": 875.83,
|
| 238671 |
+
"learning_rate": 8.247948303715672e-06,
|
| 238672 |
+
"loss": 0.7041,
|
| 238673 |
+
"step": 109495
|
| 238674 |
+
},
|
| 238675 |
+
{
|
| 238676 |
+
"epoch": 875.87,
|
| 238677 |
+
"learning_rate": 8.247867528271406e-06,
|
| 238678 |
+
"loss": 0.3011,
|
| 238679 |
+
"step": 109500
|
| 238680 |
+
},
|
| 238681 |
+
{
|
| 238682 |
+
"epoch": 875.91,
|
| 238683 |
+
"learning_rate": 8.247786752827141e-06,
|
| 238684 |
+
"loss": 0.3035,
|
| 238685 |
+
"step": 109505
|
| 238686 |
+
},
|
| 238687 |
+
{
|
| 238688 |
+
"epoch": 875.95,
|
| 238689 |
+
"learning_rate": 8.247705977382876e-06,
|
| 238690 |
+
"loss": 0.4184,
|
| 238691 |
+
"step": 109510
|
| 238692 |
+
},
|
| 238693 |
+
{
|
| 238694 |
+
"epoch": 875.99,
|
| 238695 |
+
"learning_rate": 8.247625201938611e-06,
|
| 238696 |
+
"loss": 0.9791,
|
| 238697 |
+
"step": 109515
|
| 238698 |
+
},
|
| 238699 |
+
{
|
| 238700 |
+
"epoch": 876.0,
|
| 238701 |
+
"eval_loss": 0.4340699315071106,
|
| 238702 |
+
"eval_runtime": 41.6362,
|
| 238703 |
+
"eval_samples_per_second": 20.127,
|
| 238704 |
+
"eval_steps_per_second": 0.648,
|
| 238705 |
+
"eval_wer": 0.18675573627650305,
|
| 238706 |
+
"step": 109516
|
| 238707 |
}
|
| 238708 |
],
|
| 238709 |
+
"max_steps": 625000,
|
| 238710 |
"num_train_epochs": 5000,
|
| 238711 |
+
"total_flos": 3.082023996552078e+20,
|
| 238712 |
"trial_name": null,
|
| 238713 |
"trial_params": null
|
| 238714 |
}
|
model-bin/finetune/base/{checkpoint-108894 → checkpoint-109516}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630038501.523142/events.out.tfevents.1630038501.52f5c7e305a3.886.41
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3c401ed62b484604a23848e19989c75b32520117d9550bddd70b4ae6902f2d01
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630038959.4217858/events.out.tfevents.1630038959.52f5c7e305a3.886.43
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e9babea275954112e093ee713694ed7972e862d7f606a9dbe28299c5046ec1e2
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630039419.2220693/events.out.tfevents.1630039419.52f5c7e305a3.886.45
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:134c9b3b7493265d59b258af820043544de2279a77cfd089fe5c8af3bfb0ff50
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630039886.0656877/events.out.tfevents.1630039886.52f5c7e305a3.886.47
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6c5fbd6ba4ce85628c95c09a0c3f2fd73354a8dcd2ef4877c286639d2dbf9b73
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630040347.1132815/events.out.tfevents.1630040347.52f5c7e305a3.886.49
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6e7199ab5167f975832a728d4180880640c77ba4410e03f83aec4b85ed6ab745
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630038501.52f5c7e305a3.886.40
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:07f45d8b7f1f11c9b5196500c64d74f13d34c269482330d6da6e4254ea9f646e
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630038959.52f5c7e305a3.886.42
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:466b30ec82d7b2d6b26d0774df616401eff4867121ff67c4082db2e7602dedf9
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630039419.52f5c7e305a3.886.44
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:50405b6d7beae2d9e44615d56dd9bf3572a9ca311f0f9c514052115386d878e3
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630039886.52f5c7e305a3.886.46
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:aa3da51855c3b23c2474cd23ddfca3a9e8a1afc730e7c04014131f8e82e6fdeb
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630040347.52f5c7e305a3.886.48
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4263db218c3486300f69f56c5ca2c4db0a06692c6fe89cb1de0d5cd7171f9baa
|
| 3 |
+
size 8622
|