"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-111379 → checkpoint-112002}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-111379 → checkpoint-112002}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-111379 → checkpoint-112002}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-111379 → checkpoint-112002}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-111379 → checkpoint-112002}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-111379 → checkpoint-112002}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-111379 → checkpoint-112002}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-111379 → checkpoint-112002}/trainer_state.json +798 -3
- model-bin/finetune/base/{checkpoint-111379 → checkpoint-112002}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630048138.6581674/events.out.tfevents.1630048138.52f5c7e305a3.886.81 +3 -0
- model-bin/finetune/base/log/1630048619.9533648/events.out.tfevents.1630048619.52f5c7e305a3.886.83 +3 -0
- model-bin/finetune/base/log/1630049102.1531558/events.out.tfevents.1630049102.52f5c7e305a3.886.85 +3 -0
- model-bin/finetune/base/log/1630049572.6758537/events.out.tfevents.1630049572.52f5c7e305a3.886.87 +3 -0
- model-bin/finetune/base/log/1630050046.2035077/events.out.tfevents.1630050046.52f5c7e305a3.886.89 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630048138.52f5c7e305a3.886.80 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630048619.52f5c7e305a3.886.82 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630049102.52f5c7e305a3.886.84 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630049572.52f5c7e305a3.886.86 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630050046.52f5c7e305a3.886.88 +3 -0
model-bin/finetune/base/{checkpoint-111379 → checkpoint-112002}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-111379 → checkpoint-112002}/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 sha256:1573691ea7d0fc61236cc2759103b23e73e155f43b48959ab586bdb034de93db
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-111379 → checkpoint-112002}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-111379 → checkpoint-112002}/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 sha256:0ad107c141d3f7809507fa2921dcccdf0788b1defb6ed5e37e0ce2eaa7d50eec
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-111379 → checkpoint-112002}/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 sha256:0940cfeead080e2f3ccd94ee15dab324a13e5461273c03d52367d56baa29b599
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-111379 → checkpoint-112002}/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:bd0c1baf2f5398bf54715f4be73eadcf661402972348e07ee31ecfc7ef64bd75
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-111379 → checkpoint-112002}/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:dcf1b5511c9d2f60343a826d533e62e2cd2de37fa4f207684f8625a938a917e1
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-111379 → checkpoint-112002}/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,
|
|
@@ -241071,11 +241071,806 @@
|
|
| 241071 |
"eval_steps_per_second": 0.642,
|
| 241072 |
"eval_wer": 0.18447376051029285,
|
| 241073 |
"step": 111379
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 241074 |
}
|
| 241075 |
],
|
| 241076 |
"max_steps": 620000,
|
| 241077 |
"num_train_epochs": 5000,
|
| 241078 |
-
"total_flos": 3.
|
| 241079 |
"trial_name": null,
|
| 241080 |
"trial_params": null
|
| 241081 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 902.995983935743,
|
| 5 |
+
"global_step": 112002,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 241071 |
"eval_steps_per_second": 0.642,
|
| 241072 |
"eval_wer": 0.18447376051029285,
|
| 241073 |
"step": 111379
|
| 241074 |
+
},
|
| 241075 |
+
{
|
| 241076 |
+
"epoch": 898.01,
|
| 241077 |
+
"learning_rate": 8.217512116316641e-06,
|
| 241078 |
+
"loss": 0.3168,
|
| 241079 |
+
"step": 111380
|
| 241080 |
+
},
|
| 241081 |
+
{
|
| 241082 |
+
"epoch": 898.05,
|
| 241083 |
+
"learning_rate": 8.217431340872375e-06,
|
| 241084 |
+
"loss": 0.2956,
|
| 241085 |
+
"step": 111385
|
| 241086 |
+
},
|
| 241087 |
+
{
|
| 241088 |
+
"epoch": 898.09,
|
| 241089 |
+
"learning_rate": 8.217350565428111e-06,
|
| 241090 |
+
"loss": 0.2882,
|
| 241091 |
+
"step": 111390
|
| 241092 |
+
},
|
| 241093 |
+
{
|
| 241094 |
+
"epoch": 898.13,
|
| 241095 |
+
"learning_rate": 8.217269789983845e-06,
|
| 241096 |
+
"loss": 0.2941,
|
| 241097 |
+
"step": 111395
|
| 241098 |
+
},
|
| 241099 |
+
{
|
| 241100 |
+
"epoch": 898.17,
|
| 241101 |
+
"learning_rate": 8.217189014539581e-06,
|
| 241102 |
+
"loss": 0.5359,
|
| 241103 |
+
"step": 111400
|
| 241104 |
+
},
|
| 241105 |
+
{
|
| 241106 |
+
"epoch": 898.21,
|
| 241107 |
+
"learning_rate": 8.217108239095315e-06,
|
| 241108 |
+
"loss": 1.293,
|
| 241109 |
+
"step": 111405
|
| 241110 |
+
},
|
| 241111 |
+
{
|
| 241112 |
+
"epoch": 898.25,
|
| 241113 |
+
"learning_rate": 8.217027463651051e-06,
|
| 241114 |
+
"loss": 0.2586,
|
| 241115 |
+
"step": 111410
|
| 241116 |
+
},
|
| 241117 |
+
{
|
| 241118 |
+
"epoch": 898.29,
|
| 241119 |
+
"learning_rate": 8.216946688206785e-06,
|
| 241120 |
+
"loss": 0.2983,
|
| 241121 |
+
"step": 111415
|
| 241122 |
+
},
|
| 241123 |
+
{
|
| 241124 |
+
"epoch": 898.33,
|
| 241125 |
+
"learning_rate": 8.216865912762521e-06,
|
| 241126 |
+
"loss": 0.3252,
|
| 241127 |
+
"step": 111420
|
| 241128 |
+
},
|
| 241129 |
+
{
|
| 241130 |
+
"epoch": 898.37,
|
| 241131 |
+
"learning_rate": 8.216785137318255e-06,
|
| 241132 |
+
"loss": 0.4947,
|
| 241133 |
+
"step": 111425
|
| 241134 |
+
},
|
| 241135 |
+
{
|
| 241136 |
+
"epoch": 898.41,
|
| 241137 |
+
"learning_rate": 8.216704361873991e-06,
|
| 241138 |
+
"loss": 1.025,
|
| 241139 |
+
"step": 111430
|
| 241140 |
+
},
|
| 241141 |
+
{
|
| 241142 |
+
"epoch": 898.45,
|
| 241143 |
+
"learning_rate": 8.216623586429727e-06,
|
| 241144 |
+
"loss": 0.3136,
|
| 241145 |
+
"step": 111435
|
| 241146 |
+
},
|
| 241147 |
+
{
|
| 241148 |
+
"epoch": 898.49,
|
| 241149 |
+
"learning_rate": 8.216542810985461e-06,
|
| 241150 |
+
"loss": 0.2458,
|
| 241151 |
+
"step": 111440
|
| 241152 |
+
},
|
| 241153 |
+
{
|
| 241154 |
+
"epoch": 898.53,
|
| 241155 |
+
"learning_rate": 8.216462035541197e-06,
|
| 241156 |
+
"loss": 0.3212,
|
| 241157 |
+
"step": 111445
|
| 241158 |
+
},
|
| 241159 |
+
{
|
| 241160 |
+
"epoch": 898.57,
|
| 241161 |
+
"learning_rate": 8.216381260096931e-06,
|
| 241162 |
+
"loss": 0.5968,
|
| 241163 |
+
"step": 111450
|
| 241164 |
+
},
|
| 241165 |
+
{
|
| 241166 |
+
"epoch": 898.61,
|
| 241167 |
+
"learning_rate": 8.216300484652667e-06,
|
| 241168 |
+
"loss": 1.074,
|
| 241169 |
+
"step": 111455
|
| 241170 |
+
},
|
| 241171 |
+
{
|
| 241172 |
+
"epoch": 898.65,
|
| 241173 |
+
"learning_rate": 8.216219709208401e-06,
|
| 241174 |
+
"loss": 0.2987,
|
| 241175 |
+
"step": 111460
|
| 241176 |
+
},
|
| 241177 |
+
{
|
| 241178 |
+
"epoch": 898.69,
|
| 241179 |
+
"learning_rate": 8.216138933764137e-06,
|
| 241180 |
+
"loss": 0.3227,
|
| 241181 |
+
"step": 111465
|
| 241182 |
+
},
|
| 241183 |
+
{
|
| 241184 |
+
"epoch": 898.73,
|
| 241185 |
+
"learning_rate": 8.216058158319871e-06,
|
| 241186 |
+
"loss": 0.5044,
|
| 241187 |
+
"step": 111470
|
| 241188 |
+
},
|
| 241189 |
+
{
|
| 241190 |
+
"epoch": 898.77,
|
| 241191 |
+
"learning_rate": 8.215977382875607e-06,
|
| 241192 |
+
"loss": 0.5049,
|
| 241193 |
+
"step": 111475
|
| 241194 |
+
},
|
| 241195 |
+
{
|
| 241196 |
+
"epoch": 898.81,
|
| 241197 |
+
"learning_rate": 8.215896607431341e-06,
|
| 241198 |
+
"loss": 1.1103,
|
| 241199 |
+
"step": 111480
|
| 241200 |
+
},
|
| 241201 |
+
{
|
| 241202 |
+
"epoch": 898.85,
|
| 241203 |
+
"learning_rate": 8.215815831987077e-06,
|
| 241204 |
+
"loss": 0.292,
|
| 241205 |
+
"step": 111485
|
| 241206 |
+
},
|
| 241207 |
+
{
|
| 241208 |
+
"epoch": 898.89,
|
| 241209 |
+
"learning_rate": 8.215735056542811e-06,
|
| 241210 |
+
"loss": 0.2973,
|
| 241211 |
+
"step": 111490
|
| 241212 |
+
},
|
| 241213 |
+
{
|
| 241214 |
+
"epoch": 898.93,
|
| 241215 |
+
"learning_rate": 8.215654281098547e-06,
|
| 241216 |
+
"loss": 0.3051,
|
| 241217 |
+
"step": 111495
|
| 241218 |
+
},
|
| 241219 |
+
{
|
| 241220 |
+
"epoch": 898.97,
|
| 241221 |
+
"learning_rate": 8.215573505654283e-06,
|
| 241222 |
+
"loss": 0.5463,
|
| 241223 |
+
"step": 111500
|
| 241224 |
+
},
|
| 241225 |
+
{
|
| 241226 |
+
"epoch": 899.0,
|
| 241227 |
+
"eval_loss": 0.45212462544441223,
|
| 241228 |
+
"eval_runtime": 41.8965,
|
| 241229 |
+
"eval_samples_per_second": 19.954,
|
| 241230 |
+
"eval_steps_per_second": 0.644,
|
| 241231 |
+
"eval_wer": 0.18223921193854298,
|
| 241232 |
+
"step": 111503
|
| 241233 |
+
},
|
| 241234 |
+
{
|
| 241235 |
+
"epoch": 892.02,
|
| 241236 |
+
"learning_rate": 8.215492730210017e-06,
|
| 241237 |
+
"loss": 0.5205,
|
| 241238 |
+
"step": 111505
|
| 241239 |
+
},
|
| 241240 |
+
{
|
| 241241 |
+
"epoch": 892.06,
|
| 241242 |
+
"learning_rate": 8.215411954765753e-06,
|
| 241243 |
+
"loss": 0.2519,
|
| 241244 |
+
"step": 111510
|
| 241245 |
+
},
|
| 241246 |
+
{
|
| 241247 |
+
"epoch": 892.1,
|
| 241248 |
+
"learning_rate": 8.215331179321487e-06,
|
| 241249 |
+
"loss": 0.2812,
|
| 241250 |
+
"step": 111515
|
| 241251 |
+
},
|
| 241252 |
+
{
|
| 241253 |
+
"epoch": 892.14,
|
| 241254 |
+
"learning_rate": 8.215250403877223e-06,
|
| 241255 |
+
"loss": 0.3625,
|
| 241256 |
+
"step": 111520
|
| 241257 |
+
},
|
| 241258 |
+
{
|
| 241259 |
+
"epoch": 892.18,
|
| 241260 |
+
"learning_rate": 8.215169628432957e-06,
|
| 241261 |
+
"loss": 0.6445,
|
| 241262 |
+
"step": 111525
|
| 241263 |
+
},
|
| 241264 |
+
{
|
| 241265 |
+
"epoch": 892.22,
|
| 241266 |
+
"learning_rate": 8.215088852988693e-06,
|
| 241267 |
+
"loss": 0.9554,
|
| 241268 |
+
"step": 111530
|
| 241269 |
+
},
|
| 241270 |
+
{
|
| 241271 |
+
"epoch": 892.25,
|
| 241272 |
+
"learning_rate": 8.215008077544427e-06,
|
| 241273 |
+
"loss": 0.2955,
|
| 241274 |
+
"step": 111535
|
| 241275 |
+
},
|
| 241276 |
+
{
|
| 241277 |
+
"epoch": 892.29,
|
| 241278 |
+
"learning_rate": 8.214927302100163e-06,
|
| 241279 |
+
"loss": 0.2583,
|
| 241280 |
+
"step": 111540
|
| 241281 |
+
},
|
| 241282 |
+
{
|
| 241283 |
+
"epoch": 892.33,
|
| 241284 |
+
"learning_rate": 8.214846526655897e-06,
|
| 241285 |
+
"loss": 0.3004,
|
| 241286 |
+
"step": 111545
|
| 241287 |
+
},
|
| 241288 |
+
{
|
| 241289 |
+
"epoch": 892.37,
|
| 241290 |
+
"learning_rate": 8.214765751211633e-06,
|
| 241291 |
+
"loss": 0.6362,
|
| 241292 |
+
"step": 111550
|
| 241293 |
+
},
|
| 241294 |
+
{
|
| 241295 |
+
"epoch": 892.41,
|
| 241296 |
+
"learning_rate": 8.214684975767367e-06,
|
| 241297 |
+
"loss": 1.0824,
|
| 241298 |
+
"step": 111555
|
| 241299 |
+
},
|
| 241300 |
+
{
|
| 241301 |
+
"epoch": 892.45,
|
| 241302 |
+
"learning_rate": 8.214604200323102e-06,
|
| 241303 |
+
"loss": 0.2687,
|
| 241304 |
+
"step": 111560
|
| 241305 |
+
},
|
| 241306 |
+
{
|
| 241307 |
+
"epoch": 892.49,
|
| 241308 |
+
"learning_rate": 8.214523424878837e-06,
|
| 241309 |
+
"loss": 0.3912,
|
| 241310 |
+
"step": 111565
|
| 241311 |
+
},
|
| 241312 |
+
{
|
| 241313 |
+
"epoch": 892.53,
|
| 241314 |
+
"learning_rate": 8.214442649434572e-06,
|
| 241315 |
+
"loss": 0.4064,
|
| 241316 |
+
"step": 111570
|
| 241317 |
+
},
|
| 241318 |
+
{
|
| 241319 |
+
"epoch": 892.57,
|
| 241320 |
+
"learning_rate": 8.214361873990308e-06,
|
| 241321 |
+
"loss": 0.6121,
|
| 241322 |
+
"step": 111575
|
| 241323 |
+
},
|
| 241324 |
+
{
|
| 241325 |
+
"epoch": 892.61,
|
| 241326 |
+
"learning_rate": 8.214281098546042e-06,
|
| 241327 |
+
"loss": 1.0386,
|
| 241328 |
+
"step": 111580
|
| 241329 |
+
},
|
| 241330 |
+
{
|
| 241331 |
+
"epoch": 892.65,
|
| 241332 |
+
"learning_rate": 8.214200323101778e-06,
|
| 241333 |
+
"loss": 0.2815,
|
| 241334 |
+
"step": 111585
|
| 241335 |
+
},
|
| 241336 |
+
{
|
| 241337 |
+
"epoch": 892.69,
|
| 241338 |
+
"learning_rate": 8.214119547657512e-06,
|
| 241339 |
+
"loss": 0.3376,
|
| 241340 |
+
"step": 111590
|
| 241341 |
+
},
|
| 241342 |
+
{
|
| 241343 |
+
"epoch": 892.73,
|
| 241344 |
+
"learning_rate": 8.214038772213248e-06,
|
| 241345 |
+
"loss": 0.3432,
|
| 241346 |
+
"step": 111595
|
| 241347 |
+
},
|
| 241348 |
+
{
|
| 241349 |
+
"epoch": 892.77,
|
| 241350 |
+
"learning_rate": 8.213957996768982e-06,
|
| 241351 |
+
"loss": 0.6253,
|
| 241352 |
+
"step": 111600
|
| 241353 |
+
},
|
| 241354 |
+
{
|
| 241355 |
+
"epoch": 892.81,
|
| 241356 |
+
"learning_rate": 8.213877221324718e-06,
|
| 241357 |
+
"loss": 0.8157,
|
| 241358 |
+
"step": 111605
|
| 241359 |
+
},
|
| 241360 |
+
{
|
| 241361 |
+
"epoch": 892.85,
|
| 241362 |
+
"learning_rate": 8.213796445880452e-06,
|
| 241363 |
+
"loss": 0.282,
|
| 241364 |
+
"step": 111610
|
| 241365 |
+
},
|
| 241366 |
+
{
|
| 241367 |
+
"epoch": 892.89,
|
| 241368 |
+
"learning_rate": 8.213715670436188e-06,
|
| 241369 |
+
"loss": 0.2759,
|
| 241370 |
+
"step": 111615
|
| 241371 |
+
},
|
| 241372 |
+
{
|
| 241373 |
+
"epoch": 892.93,
|
| 241374 |
+
"learning_rate": 8.213634894991922e-06,
|
| 241375 |
+
"loss": 0.3045,
|
| 241376 |
+
"step": 111620
|
| 241377 |
+
},
|
| 241378 |
+
{
|
| 241379 |
+
"epoch": 892.97,
|
| 241380 |
+
"learning_rate": 8.213554119547658e-06,
|
| 241381 |
+
"loss": 0.5921,
|
| 241382 |
+
"step": 111625
|
| 241383 |
+
},
|
| 241384 |
+
{
|
| 241385 |
+
"epoch": 893.0,
|
| 241386 |
+
"eval_loss": 0.4683041274547577,
|
| 241387 |
+
"eval_runtime": 42.5438,
|
| 241388 |
+
"eval_samples_per_second": 19.65,
|
| 241389 |
+
"eval_steps_per_second": 0.635,
|
| 241390 |
+
"eval_wer": 0.18959383338274532,
|
| 241391 |
+
"step": 111628
|
| 241392 |
+
},
|
| 241393 |
+
{
|
| 241394 |
+
"epoch": 893.02,
|
| 241395 |
+
"learning_rate": 8.213473344103392e-06,
|
| 241396 |
+
"loss": 0.3051,
|
| 241397 |
+
"step": 111630
|
| 241398 |
+
},
|
| 241399 |
+
{
|
| 241400 |
+
"epoch": 893.06,
|
| 241401 |
+
"learning_rate": 8.213392568659128e-06,
|
| 241402 |
+
"loss": 0.2988,
|
| 241403 |
+
"step": 111635
|
| 241404 |
+
},
|
| 241405 |
+
{
|
| 241406 |
+
"epoch": 893.1,
|
| 241407 |
+
"learning_rate": 8.213311793214864e-06,
|
| 241408 |
+
"loss": 0.3022,
|
| 241409 |
+
"step": 111640
|
| 241410 |
+
},
|
| 241411 |
+
{
|
| 241412 |
+
"epoch": 893.14,
|
| 241413 |
+
"learning_rate": 8.213231017770598e-06,
|
| 241414 |
+
"loss": 0.3361,
|
| 241415 |
+
"step": 111645
|
| 241416 |
+
},
|
| 241417 |
+
{
|
| 241418 |
+
"epoch": 893.18,
|
| 241419 |
+
"learning_rate": 8.213150242326334e-06,
|
| 241420 |
+
"loss": 0.6711,
|
| 241421 |
+
"step": 111650
|
| 241422 |
+
},
|
| 241423 |
+
{
|
| 241424 |
+
"epoch": 893.22,
|
| 241425 |
+
"learning_rate": 8.213069466882068e-06,
|
| 241426 |
+
"loss": 1.1263,
|
| 241427 |
+
"step": 111655
|
| 241428 |
+
},
|
| 241429 |
+
{
|
| 241430 |
+
"epoch": 893.26,
|
| 241431 |
+
"learning_rate": 8.212988691437804e-06,
|
| 241432 |
+
"loss": 0.3122,
|
| 241433 |
+
"step": 111660
|
| 241434 |
+
},
|
| 241435 |
+
{
|
| 241436 |
+
"epoch": 893.3,
|
| 241437 |
+
"learning_rate": 8.212907915993538e-06,
|
| 241438 |
+
"loss": 0.2739,
|
| 241439 |
+
"step": 111665
|
| 241440 |
+
},
|
| 241441 |
+
{
|
| 241442 |
+
"epoch": 893.34,
|
| 241443 |
+
"learning_rate": 8.212827140549274e-06,
|
| 241444 |
+
"loss": 0.3829,
|
| 241445 |
+
"step": 111670
|
| 241446 |
+
},
|
| 241447 |
+
{
|
| 241448 |
+
"epoch": 893.38,
|
| 241449 |
+
"learning_rate": 8.212746365105008e-06,
|
| 241450 |
+
"loss": 0.6704,
|
| 241451 |
+
"step": 111675
|
| 241452 |
+
},
|
| 241453 |
+
{
|
| 241454 |
+
"epoch": 893.42,
|
| 241455 |
+
"learning_rate": 8.212665589660744e-06,
|
| 241456 |
+
"loss": 1.0724,
|
| 241457 |
+
"step": 111680
|
| 241458 |
+
},
|
| 241459 |
+
{
|
| 241460 |
+
"epoch": 893.46,
|
| 241461 |
+
"learning_rate": 8.212584814216478e-06,
|
| 241462 |
+
"loss": 0.2658,
|
| 241463 |
+
"step": 111685
|
| 241464 |
+
},
|
| 241465 |
+
{
|
| 241466 |
+
"epoch": 893.5,
|
| 241467 |
+
"learning_rate": 8.212504038772214e-06,
|
| 241468 |
+
"loss": 0.3109,
|
| 241469 |
+
"step": 111690
|
| 241470 |
+
},
|
| 241471 |
+
{
|
| 241472 |
+
"epoch": 893.54,
|
| 241473 |
+
"learning_rate": 8.212423263327948e-06,
|
| 241474 |
+
"loss": 0.3202,
|
| 241475 |
+
"step": 111695
|
| 241476 |
+
},
|
| 241477 |
+
{
|
| 241478 |
+
"epoch": 893.58,
|
| 241479 |
+
"learning_rate": 8.212342487883684e-06,
|
| 241480 |
+
"loss": 0.6632,
|
| 241481 |
+
"step": 111700
|
| 241482 |
+
},
|
| 241483 |
+
{
|
| 241484 |
+
"epoch": 893.62,
|
| 241485 |
+
"learning_rate": 8.21226171243942e-06,
|
| 241486 |
+
"loss": 1.0617,
|
| 241487 |
+
"step": 111705
|
| 241488 |
+
},
|
| 241489 |
+
{
|
| 241490 |
+
"epoch": 893.66,
|
| 241491 |
+
"learning_rate": 8.212180936995154e-06,
|
| 241492 |
+
"loss": 0.2693,
|
| 241493 |
+
"step": 111710
|
| 241494 |
+
},
|
| 241495 |
+
{
|
| 241496 |
+
"epoch": 893.7,
|
| 241497 |
+
"learning_rate": 8.21210016155089e-06,
|
| 241498 |
+
"loss": 0.32,
|
| 241499 |
+
"step": 111715
|
| 241500 |
+
},
|
| 241501 |
+
{
|
| 241502 |
+
"epoch": 893.74,
|
| 241503 |
+
"learning_rate": 8.212019386106624e-06,
|
| 241504 |
+
"loss": 0.4235,
|
| 241505 |
+
"step": 111720
|
| 241506 |
+
},
|
| 241507 |
+
{
|
| 241508 |
+
"epoch": 893.78,
|
| 241509 |
+
"learning_rate": 8.21193861066236e-06,
|
| 241510 |
+
"loss": 0.6357,
|
| 241511 |
+
"step": 111725
|
| 241512 |
+
},
|
| 241513 |
+
{
|
| 241514 |
+
"epoch": 893.82,
|
| 241515 |
+
"learning_rate": 8.211857835218094e-06,
|
| 241516 |
+
"loss": 0.9449,
|
| 241517 |
+
"step": 111730
|
| 241518 |
+
},
|
| 241519 |
+
{
|
| 241520 |
+
"epoch": 893.86,
|
| 241521 |
+
"learning_rate": 8.21177705977383e-06,
|
| 241522 |
+
"loss": 0.2906,
|
| 241523 |
+
"step": 111735
|
| 241524 |
+
},
|
| 241525 |
+
{
|
| 241526 |
+
"epoch": 893.9,
|
| 241527 |
+
"learning_rate": 8.211696284329564e-06,
|
| 241528 |
+
"loss": 0.298,
|
| 241529 |
+
"step": 111740
|
| 241530 |
+
},
|
| 241531 |
+
{
|
| 241532 |
+
"epoch": 893.94,
|
| 241533 |
+
"learning_rate": 8.2116155088853e-06,
|
| 241534 |
+
"loss": 0.3205,
|
| 241535 |
+
"step": 111745
|
| 241536 |
+
},
|
| 241537 |
+
{
|
| 241538 |
+
"epoch": 893.98,
|
| 241539 |
+
"learning_rate": 8.211534733441034e-06,
|
| 241540 |
+
"loss": 0.602,
|
| 241541 |
+
"step": 111750
|
| 241542 |
+
},
|
| 241543 |
+
{
|
| 241544 |
+
"epoch": 894.0,
|
| 241545 |
+
"eval_loss": 0.3670633137226105,
|
| 241546 |
+
"eval_runtime": 42.7136,
|
| 241547 |
+
"eval_samples_per_second": 19.572,
|
| 241548 |
+
"eval_steps_per_second": 0.632,
|
| 241549 |
+
"eval_wer": 0.18261776982707215,
|
| 241550 |
+
"step": 111753
|
| 241551 |
+
},
|
| 241552 |
+
{
|
| 241553 |
+
"epoch": 894.02,
|
| 241554 |
+
"learning_rate": 8.21145395799677e-06,
|
| 241555 |
+
"loss": 0.3963,
|
| 241556 |
+
"step": 111755
|
| 241557 |
+
},
|
| 241558 |
+
{
|
| 241559 |
+
"epoch": 894.06,
|
| 241560 |
+
"learning_rate": 8.211373182552504e-06,
|
| 241561 |
+
"loss": 0.3368,
|
| 241562 |
+
"step": 111760
|
| 241563 |
+
},
|
| 241564 |
+
{
|
| 241565 |
+
"epoch": 894.1,
|
| 241566 |
+
"learning_rate": 8.21129240710824e-06,
|
| 241567 |
+
"loss": 0.2995,
|
| 241568 |
+
"step": 111765
|
| 241569 |
+
},
|
| 241570 |
+
{
|
| 241571 |
+
"epoch": 894.14,
|
| 241572 |
+
"learning_rate": 8.211211631663974e-06,
|
| 241573 |
+
"loss": 0.3304,
|
| 241574 |
+
"step": 111770
|
| 241575 |
+
},
|
| 241576 |
+
{
|
| 241577 |
+
"epoch": 894.18,
|
| 241578 |
+
"learning_rate": 8.21113085621971e-06,
|
| 241579 |
+
"loss": 0.5743,
|
| 241580 |
+
"step": 111775
|
| 241581 |
+
},
|
| 241582 |
+
{
|
| 241583 |
+
"epoch": 894.22,
|
| 241584 |
+
"learning_rate": 8.211050080775445e-06,
|
| 241585 |
+
"loss": 0.9688,
|
| 241586 |
+
"step": 111780
|
| 241587 |
+
},
|
| 241588 |
+
{
|
| 241589 |
+
"epoch": 894.26,
|
| 241590 |
+
"learning_rate": 8.21096930533118e-06,
|
| 241591 |
+
"loss": 0.2755,
|
| 241592 |
+
"step": 111785
|
| 241593 |
+
},
|
| 241594 |
+
{
|
| 241595 |
+
"epoch": 894.3,
|
| 241596 |
+
"learning_rate": 8.210888529886915e-06,
|
| 241597 |
+
"loss": 0.3825,
|
| 241598 |
+
"step": 111790
|
| 241599 |
+
},
|
| 241600 |
+
{
|
| 241601 |
+
"epoch": 894.34,
|
| 241602 |
+
"learning_rate": 8.21080775444265e-06,
|
| 241603 |
+
"loss": 0.3727,
|
| 241604 |
+
"step": 111795
|
| 241605 |
+
},
|
| 241606 |
+
{
|
| 241607 |
+
"epoch": 894.38,
|
| 241608 |
+
"learning_rate": 8.210726978998385e-06,
|
| 241609 |
+
"loss": 0.6055,
|
| 241610 |
+
"step": 111800
|
| 241611 |
+
},
|
| 241612 |
+
{
|
| 241613 |
+
"epoch": 894.42,
|
| 241614 |
+
"learning_rate": 8.21064620355412e-06,
|
| 241615 |
+
"loss": 0.949,
|
| 241616 |
+
"step": 111805
|
| 241617 |
+
},
|
| 241618 |
+
{
|
| 241619 |
+
"epoch": 894.46,
|
| 241620 |
+
"learning_rate": 8.210565428109855e-06,
|
| 241621 |
+
"loss": 0.3053,
|
| 241622 |
+
"step": 111810
|
| 241623 |
+
},
|
| 241624 |
+
{
|
| 241625 |
+
"epoch": 894.5,
|
| 241626 |
+
"learning_rate": 8.21048465266559e-06,
|
| 241627 |
+
"loss": 0.4324,
|
| 241628 |
+
"step": 111815
|
| 241629 |
+
},
|
| 241630 |
+
{
|
| 241631 |
+
"epoch": 894.54,
|
| 241632 |
+
"learning_rate": 8.210403877221325e-06,
|
| 241633 |
+
"loss": 0.4127,
|
| 241634 |
+
"step": 111820
|
| 241635 |
+
},
|
| 241636 |
+
{
|
| 241637 |
+
"epoch": 894.58,
|
| 241638 |
+
"learning_rate": 8.21032310177706e-06,
|
| 241639 |
+
"loss": 0.5997,
|
| 241640 |
+
"step": 111825
|
| 241641 |
+
},
|
| 241642 |
+
{
|
| 241643 |
+
"epoch": 894.62,
|
| 241644 |
+
"learning_rate": 8.210242326332795e-06,
|
| 241645 |
+
"loss": 1.0255,
|
| 241646 |
+
"step": 111830
|
| 241647 |
+
},
|
| 241648 |
+
{
|
| 241649 |
+
"epoch": 894.66,
|
| 241650 |
+
"learning_rate": 8.21016155088853e-06,
|
| 241651 |
+
"loss": 0.2515,
|
| 241652 |
+
"step": 111835
|
| 241653 |
+
},
|
| 241654 |
+
{
|
| 241655 |
+
"epoch": 894.7,
|
| 241656 |
+
"learning_rate": 8.210080775444265e-06,
|
| 241657 |
+
"loss": 0.3091,
|
| 241658 |
+
"step": 111840
|
| 241659 |
+
},
|
| 241660 |
+
{
|
| 241661 |
+
"epoch": 894.74,
|
| 241662 |
+
"learning_rate": 8.210000000000001e-06,
|
| 241663 |
+
"loss": 0.3108,
|
| 241664 |
+
"step": 111845
|
| 241665 |
+
},
|
| 241666 |
+
{
|
| 241667 |
+
"epoch": 894.78,
|
| 241668 |
+
"learning_rate": 8.209919224555735e-06,
|
| 241669 |
+
"loss": 0.6145,
|
| 241670 |
+
"step": 111850
|
| 241671 |
+
},
|
| 241672 |
+
{
|
| 241673 |
+
"epoch": 894.82,
|
| 241674 |
+
"learning_rate": 8.209838449111471e-06,
|
| 241675 |
+
"loss": 1.1353,
|
| 241676 |
+
"step": 111855
|
| 241677 |
+
},
|
| 241678 |
+
{
|
| 241679 |
+
"epoch": 894.86,
|
| 241680 |
+
"learning_rate": 8.209757673667205e-06,
|
| 241681 |
+
"loss": 0.3686,
|
| 241682 |
+
"step": 111860
|
| 241683 |
+
},
|
| 241684 |
+
{
|
| 241685 |
+
"epoch": 894.9,
|
| 241686 |
+
"learning_rate": 8.209676898222941e-06,
|
| 241687 |
+
"loss": 0.281,
|
| 241688 |
+
"step": 111865
|
| 241689 |
+
},
|
| 241690 |
+
{
|
| 241691 |
+
"epoch": 894.94,
|
| 241692 |
+
"learning_rate": 8.209596122778675e-06,
|
| 241693 |
+
"loss": 0.4096,
|
| 241694 |
+
"step": 111870
|
| 241695 |
+
},
|
| 241696 |
+
{
|
| 241697 |
+
"epoch": 894.98,
|
| 241698 |
+
"learning_rate": 8.209515347334411e-06,
|
| 241699 |
+
"loss": 0.6653,
|
| 241700 |
+
"step": 111875
|
| 241701 |
+
},
|
| 241702 |
+
{
|
| 241703 |
+
"epoch": 895.0,
|
| 241704 |
+
"eval_loss": 0.3755977153778076,
|
| 241705 |
+
"eval_runtime": 42.4892,
|
| 241706 |
+
"eval_samples_per_second": 19.676,
|
| 241707 |
+
"eval_steps_per_second": 0.635,
|
| 241708 |
+
"eval_wer": 0.18443218123729305,
|
| 241709 |
+
"step": 111878
|
| 241710 |
+
},
|
| 241711 |
+
{
|
| 241712 |
+
"epoch": 902.02,
|
| 241713 |
+
"learning_rate": 8.209434571890145e-06,
|
| 241714 |
+
"loss": 0.3006,
|
| 241715 |
+
"step": 111880
|
| 241716 |
+
},
|
| 241717 |
+
{
|
| 241718 |
+
"epoch": 902.06,
|
| 241719 |
+
"learning_rate": 8.209353796445881e-06,
|
| 241720 |
+
"loss": 0.2412,
|
| 241721 |
+
"step": 111885
|
| 241722 |
+
},
|
| 241723 |
+
{
|
| 241724 |
+
"epoch": 902.1,
|
| 241725 |
+
"learning_rate": 8.209273021001615e-06,
|
| 241726 |
+
"loss": 0.3399,
|
| 241727 |
+
"step": 111890
|
| 241728 |
+
},
|
| 241729 |
+
{
|
| 241730 |
+
"epoch": 902.14,
|
| 241731 |
+
"learning_rate": 8.209192245557351e-06,
|
| 241732 |
+
"loss": 0.3268,
|
| 241733 |
+
"step": 111895
|
| 241734 |
+
},
|
| 241735 |
+
{
|
| 241736 |
+
"epoch": 902.18,
|
| 241737 |
+
"learning_rate": 8.209111470113085e-06,
|
| 241738 |
+
"loss": 0.6156,
|
| 241739 |
+
"step": 111900
|
| 241740 |
+
},
|
| 241741 |
+
{
|
| 241742 |
+
"epoch": 902.22,
|
| 241743 |
+
"learning_rate": 8.209030694668821e-06,
|
| 241744 |
+
"loss": 1.1313,
|
| 241745 |
+
"step": 111905
|
| 241746 |
+
},
|
| 241747 |
+
{
|
| 241748 |
+
"epoch": 902.26,
|
| 241749 |
+
"learning_rate": 8.208949919224557e-06,
|
| 241750 |
+
"loss": 0.3566,
|
| 241751 |
+
"step": 111910
|
| 241752 |
+
},
|
| 241753 |
+
{
|
| 241754 |
+
"epoch": 902.3,
|
| 241755 |
+
"learning_rate": 8.20886914378029e-06,
|
| 241756 |
+
"loss": 0.3318,
|
| 241757 |
+
"step": 111915
|
| 241758 |
+
},
|
| 241759 |
+
{
|
| 241760 |
+
"epoch": 902.34,
|
| 241761 |
+
"learning_rate": 8.208788368336027e-06,
|
| 241762 |
+
"loss": 0.3042,
|
| 241763 |
+
"step": 111920
|
| 241764 |
+
},
|
| 241765 |
+
{
|
| 241766 |
+
"epoch": 902.38,
|
| 241767 |
+
"learning_rate": 8.20870759289176e-06,
|
| 241768 |
+
"loss": 0.5518,
|
| 241769 |
+
"step": 111925
|
| 241770 |
+
},
|
| 241771 |
+
{
|
| 241772 |
+
"epoch": 902.42,
|
| 241773 |
+
"learning_rate": 8.208626817447497e-06,
|
| 241774 |
+
"loss": 0.9917,
|
| 241775 |
+
"step": 111930
|
| 241776 |
+
},
|
| 241777 |
+
{
|
| 241778 |
+
"epoch": 902.46,
|
| 241779 |
+
"learning_rate": 8.20854604200323e-06,
|
| 241780 |
+
"loss": 0.3356,
|
| 241781 |
+
"step": 111935
|
| 241782 |
+
},
|
| 241783 |
+
{
|
| 241784 |
+
"epoch": 902.5,
|
| 241785 |
+
"learning_rate": 8.208465266558967e-06,
|
| 241786 |
+
"loss": 0.286,
|
| 241787 |
+
"step": 111940
|
| 241788 |
+
},
|
| 241789 |
+
{
|
| 241790 |
+
"epoch": 902.54,
|
| 241791 |
+
"learning_rate": 8.2083844911147e-06,
|
| 241792 |
+
"loss": 0.3935,
|
| 241793 |
+
"step": 111945
|
| 241794 |
+
},
|
| 241795 |
+
{
|
| 241796 |
+
"epoch": 902.58,
|
| 241797 |
+
"learning_rate": 8.208303715670437e-06,
|
| 241798 |
+
"loss": 0.6371,
|
| 241799 |
+
"step": 111950
|
| 241800 |
+
},
|
| 241801 |
+
{
|
| 241802 |
+
"epoch": 902.62,
|
| 241803 |
+
"learning_rate": 8.20822294022617e-06,
|
| 241804 |
+
"loss": 0.9316,
|
| 241805 |
+
"step": 111955
|
| 241806 |
+
},
|
| 241807 |
+
{
|
| 241808 |
+
"epoch": 902.66,
|
| 241809 |
+
"learning_rate": 8.208142164781907e-06,
|
| 241810 |
+
"loss": 0.3458,
|
| 241811 |
+
"step": 111960
|
| 241812 |
+
},
|
| 241813 |
+
{
|
| 241814 |
+
"epoch": 902.7,
|
| 241815 |
+
"learning_rate": 8.20806138933764e-06,
|
| 241816 |
+
"loss": 0.3009,
|
| 241817 |
+
"step": 111965
|
| 241818 |
+
},
|
| 241819 |
+
{
|
| 241820 |
+
"epoch": 902.74,
|
| 241821 |
+
"learning_rate": 8.207980613893377e-06,
|
| 241822 |
+
"loss": 0.3419,
|
| 241823 |
+
"step": 111970
|
| 241824 |
+
},
|
| 241825 |
+
{
|
| 241826 |
+
"epoch": 902.78,
|
| 241827 |
+
"learning_rate": 8.207899838449112e-06,
|
| 241828 |
+
"loss": 0.5863,
|
| 241829 |
+
"step": 111975
|
| 241830 |
+
},
|
| 241831 |
+
{
|
| 241832 |
+
"epoch": 902.82,
|
| 241833 |
+
"learning_rate": 8.207819063004847e-06,
|
| 241834 |
+
"loss": 1.0017,
|
| 241835 |
+
"step": 111980
|
| 241836 |
+
},
|
| 241837 |
+
{
|
| 241838 |
+
"epoch": 902.86,
|
| 241839 |
+
"learning_rate": 8.207738287560582e-06,
|
| 241840 |
+
"loss": 0.2961,
|
| 241841 |
+
"step": 111985
|
| 241842 |
+
},
|
| 241843 |
+
{
|
| 241844 |
+
"epoch": 902.9,
|
| 241845 |
+
"learning_rate": 8.207657512116317e-06,
|
| 241846 |
+
"loss": 0.2509,
|
| 241847 |
+
"step": 111990
|
| 241848 |
+
},
|
| 241849 |
+
{
|
| 241850 |
+
"epoch": 902.94,
|
| 241851 |
+
"learning_rate": 8.207576736672052e-06,
|
| 241852 |
+
"loss": 0.33,
|
| 241853 |
+
"step": 111995
|
| 241854 |
+
},
|
| 241855 |
+
{
|
| 241856 |
+
"epoch": 902.98,
|
| 241857 |
+
"learning_rate": 8.207495961227786e-06,
|
| 241858 |
+
"loss": 0.6344,
|
| 241859 |
+
"step": 112000
|
| 241860 |
+
},
|
| 241861 |
+
{
|
| 241862 |
+
"epoch": 903.0,
|
| 241863 |
+
"eval_loss": 0.4465380311012268,
|
| 241864 |
+
"eval_runtime": 43.2217,
|
| 241865 |
+
"eval_samples_per_second": 19.342,
|
| 241866 |
+
"eval_steps_per_second": 0.625,
|
| 241867 |
+
"eval_wer": 0.1886928860613071,
|
| 241868 |
+
"step": 112002
|
| 241869 |
}
|
| 241870 |
],
|
| 241871 |
"max_steps": 620000,
|
| 241872 |
"num_train_epochs": 5000,
|
| 241873 |
+
"total_flos": 3.1520264569981986e+20,
|
| 241874 |
"trial_name": null,
|
| 241875 |
"trial_params": null
|
| 241876 |
}
|
model-bin/finetune/base/{checkpoint-111379 → checkpoint-112002}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630048138.6581674/events.out.tfevents.1630048138.52f5c7e305a3.886.81
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6fd565bfc4df649dc0f4e5aadcf1eead11a12fb058bf7d9f5390ea0410bd88c0
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630048619.9533648/events.out.tfevents.1630048619.52f5c7e305a3.886.83
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dd30dccc3acb9b02a74544425cb00b04f97859f9d1abf3d1eb655dec856c326c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630049102.1531558/events.out.tfevents.1630049102.52f5c7e305a3.886.85
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ab07217fdffdd042e7a95682d29c5c31f2248936ef812bffa701cd9f704a18f3
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630049572.6758537/events.out.tfevents.1630049572.52f5c7e305a3.886.87
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8d3d43cf4345ea0f605d5516e05ca68007bcf57f75ba4819cc09f69f91395822
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630050046.2035077/events.out.tfevents.1630050046.52f5c7e305a3.886.89
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:709b11cfce27e86178a99bf7524ea96c43320f71e12f0b0b627dfbda8169b128
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630048138.52f5c7e305a3.886.80
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9c9452f94cea99f7d845a7fce5fd6275f9cb7a4fc00b5141501e09163cefe720
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630048619.52f5c7e305a3.886.82
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cc002d98a39a9740f6a94d265791543e732e073653e36f3504b9b7813d90f821
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630049102.52f5c7e305a3.886.84
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7bac776b58ce1ed85f4af1b4fa33016f75c4b5c31fe37a4fa301132e607e7928
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630049572.52f5c7e305a3.886.86
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:298c644a488ca1e5101e65db1c62c62a2a587a66960116c40947d0643f9cc208
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630050046.52f5c7e305a3.886.88
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:afc1ed3a9e82f43336874596bd326ec16c14ee316e9cf41c1c14df3165618864
|
| 3 |
+
size 8622
|