Training in progress, epoch 0
Browse files- all_results.json +12 -0
- config.json +1 -1
- eval_results.json +6 -0
- model.safetensors +3 -0
- preprocessor_config.json +2 -1
- pytorch_model.bin +1 -1
- runs/Nov13_09-57-02_f45074b7ee61/events.out.tfevents.1699869426.f45074b7ee61.8807.2 +3 -0
- runs/Nov13_10-41-02_f45074b7ee61/events.out.tfevents.1699872065.f45074b7ee61.8807.3 +3 -0
- runs/Nov13_14-14-06_fa108b7cccfa/events.out.tfevents.1699884892.fa108b7cccfa.5895.0 +3 -0
- train_results.json +8 -0
- trainer_state.json +2544 -0
- training_args.bin +2 -2
all_results.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 145.45,
|
| 3 |
+
"eval_accuracy": 0.14285714285714285,
|
| 4 |
+
"eval_f1": 0.4186046511627907,
|
| 5 |
+
"eval_loss": 0.5066081285476685,
|
| 6 |
+
"eval_roc_auc": 0.7233436853002071,
|
| 7 |
+
"total_flos": 1.357171875053568e+17,
|
| 8 |
+
"train_loss": 0.5678796648979187,
|
| 9 |
+
"train_runtime": 1897.2513,
|
| 10 |
+
"train_samples_per_second": 17.499,
|
| 11 |
+
"train_steps_per_second": 0.105
|
| 12 |
+
}
|
config.json
CHANGED
|
@@ -99,6 +99,6 @@
|
|
| 99 |
"qkv_bias": true,
|
| 100 |
"semantic_loss_ignore_index": 255,
|
| 101 |
"torch_dtype": "float32",
|
| 102 |
-
"transformers_version": "4.
|
| 103 |
"width_multiplier": 1.0
|
| 104 |
}
|
|
|
|
| 99 |
"qkv_bias": true,
|
| 100 |
"semantic_loss_ignore_index": 255,
|
| 101 |
"torch_dtype": "float32",
|
| 102 |
+
"transformers_version": "4.35.0",
|
| 103 |
"width_multiplier": 1.0
|
| 104 |
}
|
eval_results.json
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"eval_accuracy": 0.14285714285714285,
|
| 3 |
+
"eval_f1": 0.4186046511627907,
|
| 4 |
+
"eval_loss": 0.5066081285476685,
|
| 5 |
+
"eval_roc_auc": 0.7233436853002071
|
| 6 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7f9de3abe73a345f27311b60137e096a542ffa01718b088edf20be1699015ed6
|
| 3 |
+
size 19136756
|
preprocessor_config.json
CHANGED
|
@@ -12,5 +12,6 @@
|
|
| 12 |
"rescale_factor": 0.00392156862745098,
|
| 13 |
"size": {
|
| 14 |
"shortest_edge": 288
|
| 15 |
-
}
|
|
|
|
| 16 |
}
|
|
|
|
| 12 |
"rescale_factor": 0.00392156862745098,
|
| 13 |
"size": {
|
| 14 |
"shortest_edge": 288
|
| 15 |
+
},
|
| 16 |
+
"use_square_size": false
|
| 17 |
}
|
pytorch_model.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 19215286
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:119eb357c684a114c2511f1ed43d57a402c096d4aa70ef71be7cfea5f4f1a732
|
| 3 |
size 19215286
|
runs/Nov13_09-57-02_f45074b7ee61/events.out.tfevents.1699869426.f45074b7ee61.8807.2
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1a4a05387616b49114c155608fb0572c9e2f7fb8dea7f3d645b06cdbbc61daf3
|
| 3 |
+
size 66851
|
runs/Nov13_10-41-02_f45074b7ee61/events.out.tfevents.1699872065.f45074b7ee61.8807.3
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f4f961f526cec2446b56f163f8ac3f5250431908d1c7dffe6f1bc78d0e163dbb
|
| 3 |
+
size 101777
|
runs/Nov13_14-14-06_fa108b7cccfa/events.out.tfevents.1699884892.fa108b7cccfa.5895.0
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1d12947925e52d3b9b30c8518f96309f37c738b6a925ade330211b0beda33a35
|
| 3 |
+
size 6409
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 145.45,
|
| 3 |
+
"total_flos": 1.357171875053568e+17,
|
| 4 |
+
"train_loss": 0.5678796648979187,
|
| 5 |
+
"train_runtime": 1897.2513,
|
| 6 |
+
"train_samples_per_second": 17.499,
|
| 7 |
+
"train_steps_per_second": 0.105
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,2544 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": 0.43965517241379304,
|
| 3 |
+
"best_model_checkpoint": "car_identified_model_7/checkpoint-165",
|
| 4 |
+
"epoch": 228.36363636363637,
|
| 5 |
+
"eval_steps": 500,
|
| 6 |
+
"global_step": 314,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.73,
|
| 13 |
+
"learning_rate": 1.9950000000000004e-05,
|
| 14 |
+
"loss": 0.6943,
|
| 15 |
+
"step": 1
|
| 16 |
+
},
|
| 17 |
+
{
|
| 18 |
+
"epoch": 0.73,
|
| 19 |
+
"eval_accuracy": 0.0,
|
| 20 |
+
"eval_f1": 0.09523809523809525,
|
| 21 |
+
"eval_loss": 0.6940220594406128,
|
| 22 |
+
"eval_roc_auc": 0.4290890269151138,
|
| 23 |
+
"eval_runtime": 2.3165,
|
| 24 |
+
"eval_samples_per_second": 18.131,
|
| 25 |
+
"eval_steps_per_second": 1.295,
|
| 26 |
+
"step": 1
|
| 27 |
+
},
|
| 28 |
+
{
|
| 29 |
+
"epoch": 1.45,
|
| 30 |
+
"eval_accuracy": 0.0,
|
| 31 |
+
"eval_f1": 0.09126984126984126,
|
| 32 |
+
"eval_loss": 0.6954004764556885,
|
| 33 |
+
"eval_roc_auc": 0.4314182194616977,
|
| 34 |
+
"eval_runtime": 2.1188,
|
| 35 |
+
"eval_samples_per_second": 19.823,
|
| 36 |
+
"eval_steps_per_second": 1.416,
|
| 37 |
+
"step": 2
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 2.91,
|
| 41 |
+
"eval_accuracy": 0.0,
|
| 42 |
+
"eval_f1": 0.10158730158730159,
|
| 43 |
+
"eval_loss": 0.7020308971405029,
|
| 44 |
+
"eval_roc_auc": 0.4244306418219462,
|
| 45 |
+
"eval_runtime": 2.0566,
|
| 46 |
+
"eval_samples_per_second": 20.422,
|
| 47 |
+
"eval_steps_per_second": 1.459,
|
| 48 |
+
"step": 4
|
| 49 |
+
},
|
| 50 |
+
{
|
| 51 |
+
"epoch": 3.64,
|
| 52 |
+
"eval_accuracy": 0.0,
|
| 53 |
+
"eval_f1": 0.1020408163265306,
|
| 54 |
+
"eval_loss": 0.7104746103286743,
|
| 55 |
+
"eval_roc_auc": 0.43322981366459634,
|
| 56 |
+
"eval_runtime": 2.0845,
|
| 57 |
+
"eval_samples_per_second": 20.149,
|
| 58 |
+
"eval_steps_per_second": 1.439,
|
| 59 |
+
"step": 5
|
| 60 |
+
},
|
| 61 |
+
{
|
| 62 |
+
"epoch": 4.36,
|
| 63 |
+
"eval_accuracy": 0.0,
|
| 64 |
+
"eval_f1": 0.15476190476190477,
|
| 65 |
+
"eval_loss": 0.7156671285629272,
|
| 66 |
+
"eval_roc_auc": 0.5320910973084887,
|
| 67 |
+
"eval_runtime": 2.0793,
|
| 68 |
+
"eval_samples_per_second": 20.199,
|
| 69 |
+
"eval_steps_per_second": 1.443,
|
| 70 |
+
"step": 6
|
| 71 |
+
},
|
| 72 |
+
{
|
| 73 |
+
"epoch": 5.82,
|
| 74 |
+
"eval_accuracy": 0.0,
|
| 75 |
+
"eval_f1": 0.1546218487394958,
|
| 76 |
+
"eval_loss": 0.6987839937210083,
|
| 77 |
+
"eval_roc_auc": 0.5331262939958592,
|
| 78 |
+
"eval_runtime": 2.1617,
|
| 79 |
+
"eval_samples_per_second": 19.429,
|
| 80 |
+
"eval_steps_per_second": 1.388,
|
| 81 |
+
"step": 8
|
| 82 |
+
},
|
| 83 |
+
{
|
| 84 |
+
"epoch": 6.55,
|
| 85 |
+
"eval_accuracy": 0.0,
|
| 86 |
+
"eval_f1": 0.15355805243445694,
|
| 87 |
+
"eval_loss": 0.6832717061042786,
|
| 88 |
+
"eval_roc_auc": 0.5323498964803313,
|
| 89 |
+
"eval_runtime": 2.117,
|
| 90 |
+
"eval_samples_per_second": 19.84,
|
| 91 |
+
"eval_steps_per_second": 1.417,
|
| 92 |
+
"step": 9
|
| 93 |
+
},
|
| 94 |
+
{
|
| 95 |
+
"epoch": 8.0,
|
| 96 |
+
"eval_accuracy": 0.0,
|
| 97 |
+
"eval_f1": 0.18217054263565893,
|
| 98 |
+
"eval_loss": 0.6745525598526001,
|
| 99 |
+
"eval_roc_auc": 0.5804865424430642,
|
| 100 |
+
"eval_runtime": 2.0693,
|
| 101 |
+
"eval_samples_per_second": 20.297,
|
| 102 |
+
"eval_steps_per_second": 1.45,
|
| 103 |
+
"step": 11
|
| 104 |
+
},
|
| 105 |
+
{
|
| 106 |
+
"epoch": 8.73,
|
| 107 |
+
"eval_accuracy": 0.0,
|
| 108 |
+
"eval_f1": 0.2058252427184466,
|
| 109 |
+
"eval_loss": 0.6750043034553528,
|
| 110 |
+
"eval_roc_auc": 0.619824016563147,
|
| 111 |
+
"eval_runtime": 2.0585,
|
| 112 |
+
"eval_samples_per_second": 20.403,
|
| 113 |
+
"eval_steps_per_second": 1.457,
|
| 114 |
+
"step": 12
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 9.45,
|
| 118 |
+
"eval_accuracy": 0.0,
|
| 119 |
+
"eval_f1": 0.2152641878669276,
|
| 120 |
+
"eval_loss": 0.673625648021698,
|
| 121 |
+
"eval_roc_auc": 0.6348343685300206,
|
| 122 |
+
"eval_runtime": 2.1768,
|
| 123 |
+
"eval_samples_per_second": 19.294,
|
| 124 |
+
"eval_steps_per_second": 1.378,
|
| 125 |
+
"step": 13
|
| 126 |
+
},
|
| 127 |
+
{
|
| 128 |
+
"epoch": 10.91,
|
| 129 |
+
"eval_accuracy": 0.0,
|
| 130 |
+
"eval_f1": 0.2222222222222222,
|
| 131 |
+
"eval_loss": 0.6704300045967102,
|
| 132 |
+
"eval_roc_auc": 0.644927536231884,
|
| 133 |
+
"eval_runtime": 2.2148,
|
| 134 |
+
"eval_samples_per_second": 18.963,
|
| 135 |
+
"eval_steps_per_second": 1.355,
|
| 136 |
+
"step": 15
|
| 137 |
+
},
|
| 138 |
+
{
|
| 139 |
+
"epoch": 11.64,
|
| 140 |
+
"eval_accuracy": 0.0,
|
| 141 |
+
"eval_f1": 0.2258064516129032,
|
| 142 |
+
"eval_loss": 0.6693084836006165,
|
| 143 |
+
"eval_roc_auc": 0.6490683229813665,
|
| 144 |
+
"eval_runtime": 2.0734,
|
| 145 |
+
"eval_samples_per_second": 20.257,
|
| 146 |
+
"eval_steps_per_second": 1.447,
|
| 147 |
+
"step": 16
|
| 148 |
+
},
|
| 149 |
+
{
|
| 150 |
+
"epoch": 12.36,
|
| 151 |
+
"eval_accuracy": 0.0,
|
| 152 |
+
"eval_f1": 0.23293172690763056,
|
| 153 |
+
"eval_loss": 0.6673960089683533,
|
| 154 |
+
"eval_roc_auc": 0.6609730848861284,
|
| 155 |
+
"eval_runtime": 2.1348,
|
| 156 |
+
"eval_samples_per_second": 19.674,
|
| 157 |
+
"eval_steps_per_second": 1.405,
|
| 158 |
+
"step": 17
|
| 159 |
+
},
|
| 160 |
+
{
|
| 161 |
+
"epoch": 13.82,
|
| 162 |
+
"eval_accuracy": 0.0,
|
| 163 |
+
"eval_f1": 0.2240325865580448,
|
| 164 |
+
"eval_loss": 0.6640763282775879,
|
| 165 |
+
"eval_roc_auc": 0.6451863354037266,
|
| 166 |
+
"eval_runtime": 2.1666,
|
| 167 |
+
"eval_samples_per_second": 19.385,
|
| 168 |
+
"eval_steps_per_second": 1.385,
|
| 169 |
+
"step": 19
|
| 170 |
+
},
|
| 171 |
+
{
|
| 172 |
+
"epoch": 14.55,
|
| 173 |
+
"eval_accuracy": 0.0,
|
| 174 |
+
"eval_f1": 0.22085889570552147,
|
| 175 |
+
"eval_loss": 0.6631330847740173,
|
| 176 |
+
"eval_roc_auc": 0.639751552795031,
|
| 177 |
+
"eval_runtime": 2.0574,
|
| 178 |
+
"eval_samples_per_second": 20.414,
|
| 179 |
+
"eval_steps_per_second": 1.458,
|
| 180 |
+
"step": 20
|
| 181 |
+
},
|
| 182 |
+
{
|
| 183 |
+
"epoch": 16.0,
|
| 184 |
+
"eval_accuracy": 0.0,
|
| 185 |
+
"eval_f1": 0.22406639004149378,
|
| 186 |
+
"eval_loss": 0.6598684191703796,
|
| 187 |
+
"eval_roc_auc": 0.6433747412008283,
|
| 188 |
+
"eval_runtime": 2.1618,
|
| 189 |
+
"eval_samples_per_second": 19.428,
|
| 190 |
+
"eval_steps_per_second": 1.388,
|
| 191 |
+
"step": 22
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 16.73,
|
| 195 |
+
"eval_accuracy": 0.0,
|
| 196 |
+
"eval_f1": 0.23454157782515989,
|
| 197 |
+
"eval_loss": 0.6574246287345886,
|
| 198 |
+
"eval_roc_auc": 0.6565734989648033,
|
| 199 |
+
"eval_runtime": 2.0709,
|
| 200 |
+
"eval_samples_per_second": 20.281,
|
| 201 |
+
"eval_steps_per_second": 1.449,
|
| 202 |
+
"step": 23
|
| 203 |
+
},
|
| 204 |
+
{
|
| 205 |
+
"epoch": 17.45,
|
| 206 |
+
"eval_accuracy": 0.0,
|
| 207 |
+
"eval_f1": 0.24017467248908297,
|
| 208 |
+
"eval_loss": 0.6559909582138062,
|
| 209 |
+
"eval_roc_auc": 0.6622670807453417,
|
| 210 |
+
"eval_runtime": 2.2938,
|
| 211 |
+
"eval_samples_per_second": 18.311,
|
| 212 |
+
"eval_steps_per_second": 1.308,
|
| 213 |
+
"step": 24
|
| 214 |
+
},
|
| 215 |
+
{
|
| 216 |
+
"epoch": 18.91,
|
| 217 |
+
"eval_accuracy": 0.0,
|
| 218 |
+
"eval_f1": 0.23399558498896247,
|
| 219 |
+
"eval_loss": 0.6524508595466614,
|
| 220 |
+
"eval_roc_auc": 0.6519151138716356,
|
| 221 |
+
"eval_runtime": 2.1408,
|
| 222 |
+
"eval_samples_per_second": 19.619,
|
| 223 |
+
"eval_steps_per_second": 1.401,
|
| 224 |
+
"step": 26
|
| 225 |
+
},
|
| 226 |
+
{
|
| 227 |
+
"epoch": 19.64,
|
| 228 |
+
"eval_accuracy": 0.0,
|
| 229 |
+
"eval_f1": 0.24608501118568235,
|
| 230 |
+
"eval_loss": 0.650789737701416,
|
| 231 |
+
"eval_roc_auc": 0.6679606625258798,
|
| 232 |
+
"eval_runtime": 2.1461,
|
| 233 |
+
"eval_samples_per_second": 19.571,
|
| 234 |
+
"eval_steps_per_second": 1.398,
|
| 235 |
+
"step": 27
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 20.36,
|
| 239 |
+
"eval_accuracy": 0.0,
|
| 240 |
+
"eval_f1": 0.24074074074074073,
|
| 241 |
+
"eval_loss": 0.6483226418495178,
|
| 242 |
+
"eval_roc_auc": 0.6563146997929606,
|
| 243 |
+
"eval_runtime": 2.2028,
|
| 244 |
+
"eval_samples_per_second": 19.067,
|
| 245 |
+
"eval_steps_per_second": 1.362,
|
| 246 |
+
"step": 28
|
| 247 |
+
},
|
| 248 |
+
{
|
| 249 |
+
"epoch": 21.82,
|
| 250 |
+
"eval_accuracy": 0.0,
|
| 251 |
+
"eval_f1": 0.24545454545454545,
|
| 252 |
+
"eval_loss": 0.6458417177200317,
|
| 253 |
+
"eval_roc_auc": 0.6651138716356108,
|
| 254 |
+
"eval_runtime": 2.2709,
|
| 255 |
+
"eval_samples_per_second": 18.495,
|
| 256 |
+
"eval_steps_per_second": 1.321,
|
| 257 |
+
"step": 30
|
| 258 |
+
},
|
| 259 |
+
{
|
| 260 |
+
"epoch": 22.55,
|
| 261 |
+
"eval_accuracy": 0.0,
|
| 262 |
+
"eval_f1": 0.24601366742596809,
|
| 263 |
+
"eval_loss": 0.644438624382019,
|
| 264 |
+
"eval_roc_auc": 0.665631469979296,
|
| 265 |
+
"eval_runtime": 2.1455,
|
| 266 |
+
"eval_samples_per_second": 19.576,
|
| 267 |
+
"eval_steps_per_second": 1.398,
|
| 268 |
+
"step": 31
|
| 269 |
+
},
|
| 270 |
+
{
|
| 271 |
+
"epoch": 24.0,
|
| 272 |
+
"eval_accuracy": 0.0,
|
| 273 |
+
"eval_f1": 0.25116279069767444,
|
| 274 |
+
"eval_loss": 0.6407680511474609,
|
| 275 |
+
"eval_roc_auc": 0.6702898550724637,
|
| 276 |
+
"eval_runtime": 2.3067,
|
| 277 |
+
"eval_samples_per_second": 18.208,
|
| 278 |
+
"eval_steps_per_second": 1.301,
|
| 279 |
+
"step": 33
|
| 280 |
+
},
|
| 281 |
+
{
|
| 282 |
+
"epoch": 24.73,
|
| 283 |
+
"eval_accuracy": 0.0,
|
| 284 |
+
"eval_f1": 0.2523809523809524,
|
| 285 |
+
"eval_loss": 0.6386719346046448,
|
| 286 |
+
"eval_roc_auc": 0.6689958592132504,
|
| 287 |
+
"eval_runtime": 2.1261,
|
| 288 |
+
"eval_samples_per_second": 19.754,
|
| 289 |
+
"eval_steps_per_second": 1.411,
|
| 290 |
+
"step": 34
|
| 291 |
+
},
|
| 292 |
+
{
|
| 293 |
+
"epoch": 25.45,
|
| 294 |
+
"eval_accuracy": 0.0,
|
| 295 |
+
"eval_f1": 0.2579075425790754,
|
| 296 |
+
"eval_loss": 0.6363871693611145,
|
| 297 |
+
"eval_roc_auc": 0.6736542443064182,
|
| 298 |
+
"eval_runtime": 2.0731,
|
| 299 |
+
"eval_samples_per_second": 20.259,
|
| 300 |
+
"eval_steps_per_second": 1.447,
|
| 301 |
+
"step": 35
|
| 302 |
+
},
|
| 303 |
+
{
|
| 304 |
+
"epoch": 26.91,
|
| 305 |
+
"eval_accuracy": 0.0,
|
| 306 |
+
"eval_f1": 0.2706766917293233,
|
| 307 |
+
"eval_loss": 0.633211076259613,
|
| 308 |
+
"eval_roc_auc": 0.686335403726708,
|
| 309 |
+
"eval_runtime": 2.1802,
|
| 310 |
+
"eval_samples_per_second": 19.264,
|
| 311 |
+
"eval_steps_per_second": 1.376,
|
| 312 |
+
"step": 37
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 27.64,
|
| 316 |
+
"eval_accuracy": 0.0,
|
| 317 |
+
"eval_f1": 0.26153846153846155,
|
| 318 |
+
"eval_loss": 0.6316937208175659,
|
| 319 |
+
"eval_roc_auc": 0.671583850931677,
|
| 320 |
+
"eval_runtime": 2.2588,
|
| 321 |
+
"eval_samples_per_second": 18.594,
|
| 322 |
+
"eval_steps_per_second": 1.328,
|
| 323 |
+
"step": 38
|
| 324 |
+
},
|
| 325 |
+
{
|
| 326 |
+
"epoch": 28.36,
|
| 327 |
+
"eval_accuracy": 0.0,
|
| 328 |
+
"eval_f1": 0.26666666666666666,
|
| 329 |
+
"eval_loss": 0.6295819878578186,
|
| 330 |
+
"eval_roc_auc": 0.6780538302277432,
|
| 331 |
+
"eval_runtime": 2.0803,
|
| 332 |
+
"eval_samples_per_second": 20.189,
|
| 333 |
+
"eval_steps_per_second": 1.442,
|
| 334 |
+
"step": 39
|
| 335 |
+
},
|
| 336 |
+
{
|
| 337 |
+
"epoch": 29.82,
|
| 338 |
+
"eval_accuracy": 0.0,
|
| 339 |
+
"eval_f1": 0.2819843342036553,
|
| 340 |
+
"eval_loss": 0.6249864101409912,
|
| 341 |
+
"eval_roc_auc": 0.6946169772256728,
|
| 342 |
+
"eval_runtime": 2.1933,
|
| 343 |
+
"eval_samples_per_second": 19.149,
|
| 344 |
+
"eval_steps_per_second": 1.368,
|
| 345 |
+
"step": 41
|
| 346 |
+
},
|
| 347 |
+
{
|
| 348 |
+
"epoch": 30.55,
|
| 349 |
+
"eval_accuracy": 0.0,
|
| 350 |
+
"eval_f1": 0.2872340425531915,
|
| 351 |
+
"eval_loss": 0.624562680721283,
|
| 352 |
+
"eval_roc_auc": 0.6982401656314701,
|
| 353 |
+
"eval_runtime": 2.1386,
|
| 354 |
+
"eval_samples_per_second": 19.639,
|
| 355 |
+
"eval_steps_per_second": 1.403,
|
| 356 |
+
"step": 42
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 32.0,
|
| 360 |
+
"eval_accuracy": 0.0,
|
| 361 |
+
"eval_f1": 0.2936288088642659,
|
| 362 |
+
"eval_loss": 0.6203415393829346,
|
| 363 |
+
"eval_roc_auc": 0.6995341614906833,
|
| 364 |
+
"eval_runtime": 2.0334,
|
| 365 |
+
"eval_samples_per_second": 20.655,
|
| 366 |
+
"eval_steps_per_second": 1.475,
|
| 367 |
+
"step": 44
|
| 368 |
+
},
|
| 369 |
+
{
|
| 370 |
+
"epoch": 32.73,
|
| 371 |
+
"eval_accuracy": 0.0,
|
| 372 |
+
"eval_f1": 0.27932960893854747,
|
| 373 |
+
"eval_loss": 0.619665265083313,
|
| 374 |
+
"eval_roc_auc": 0.6816770186335404,
|
| 375 |
+
"eval_runtime": 2.2329,
|
| 376 |
+
"eval_samples_per_second": 18.809,
|
| 377 |
+
"eval_steps_per_second": 1.344,
|
| 378 |
+
"step": 45
|
| 379 |
+
},
|
| 380 |
+
{
|
| 381 |
+
"epoch": 33.45,
|
| 382 |
+
"eval_accuracy": 0.0,
|
| 383 |
+
"eval_f1": 0.2832861189801699,
|
| 384 |
+
"eval_loss": 0.6191604733467102,
|
| 385 |
+
"eval_roc_auc": 0.6842650103519669,
|
| 386 |
+
"eval_runtime": 2.1287,
|
| 387 |
+
"eval_samples_per_second": 19.73,
|
| 388 |
+
"eval_steps_per_second": 1.409,
|
| 389 |
+
"step": 46
|
| 390 |
+
},
|
| 391 |
+
{
|
| 392 |
+
"epoch": 34.91,
|
| 393 |
+
"eval_accuracy": 0.0,
|
| 394 |
+
"eval_f1": 0.29059829059829057,
|
| 395 |
+
"eval_loss": 0.6188385486602783,
|
| 396 |
+
"eval_roc_auc": 0.6917701863354038,
|
| 397 |
+
"eval_runtime": 1.9704,
|
| 398 |
+
"eval_samples_per_second": 21.316,
|
| 399 |
+
"eval_steps_per_second": 1.523,
|
| 400 |
+
"step": 48
|
| 401 |
+
},
|
| 402 |
+
{
|
| 403 |
+
"epoch": 35.64,
|
| 404 |
+
"eval_accuracy": 0.0,
|
| 405 |
+
"eval_f1": 0.2994350282485876,
|
| 406 |
+
"eval_loss": 0.6187217831611633,
|
| 407 |
+
"eval_roc_auc": 0.7031573498964803,
|
| 408 |
+
"eval_runtime": 2.1103,
|
| 409 |
+
"eval_samples_per_second": 19.902,
|
| 410 |
+
"eval_steps_per_second": 1.422,
|
| 411 |
+
"step": 49
|
| 412 |
+
},
|
| 413 |
+
{
|
| 414 |
+
"epoch": 36.36,
|
| 415 |
+
"eval_accuracy": 0.0,
|
| 416 |
+
"eval_f1": 0.29394812680115273,
|
| 417 |
+
"eval_loss": 0.6172403693199158,
|
| 418 |
+
"eval_roc_auc": 0.6938405797101449,
|
| 419 |
+
"eval_runtime": 2.0512,
|
| 420 |
+
"eval_samples_per_second": 20.476,
|
| 421 |
+
"eval_steps_per_second": 1.463,
|
| 422 |
+
"step": 50
|
| 423 |
+
},
|
| 424 |
+
{
|
| 425 |
+
"epoch": 37.82,
|
| 426 |
+
"eval_accuracy": 0.0,
|
| 427 |
+
"eval_f1": 0.3072289156626506,
|
| 428 |
+
"eval_loss": 0.6139649748802185,
|
| 429 |
+
"eval_roc_auc": 0.7016045548654244,
|
| 430 |
+
"eval_runtime": 2.0138,
|
| 431 |
+
"eval_samples_per_second": 20.857,
|
| 432 |
+
"eval_steps_per_second": 1.49,
|
| 433 |
+
"step": 52
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 38.55,
|
| 437 |
+
"eval_accuracy": 0.0,
|
| 438 |
+
"eval_f1": 0.30538922155688625,
|
| 439 |
+
"eval_loss": 0.6133411526679993,
|
| 440 |
+
"eval_roc_auc": 0.7005693581780539,
|
| 441 |
+
"eval_runtime": 2.1136,
|
| 442 |
+
"eval_samples_per_second": 19.871,
|
| 443 |
+
"eval_steps_per_second": 1.419,
|
| 444 |
+
"step": 53
|
| 445 |
+
},
|
| 446 |
+
{
|
| 447 |
+
"epoch": 40.0,
|
| 448 |
+
"eval_accuracy": 0.0,
|
| 449 |
+
"eval_f1": 0.3081570996978852,
|
| 450 |
+
"eval_loss": 0.6092615723609924,
|
| 451 |
+
"eval_roc_auc": 0.7021221532091098,
|
| 452 |
+
"eval_runtime": 2.0132,
|
| 453 |
+
"eval_samples_per_second": 20.863,
|
| 454 |
+
"eval_steps_per_second": 1.49,
|
| 455 |
+
"step": 55
|
| 456 |
+
},
|
| 457 |
+
{
|
| 458 |
+
"epoch": 40.73,
|
| 459 |
+
"eval_accuracy": 0.0,
|
| 460 |
+
"eval_f1": 0.3151515151515151,
|
| 461 |
+
"eval_loss": 0.6076943278312683,
|
| 462 |
+
"eval_roc_auc": 0.7091097308488613,
|
| 463 |
+
"eval_runtime": 2.0499,
|
| 464 |
+
"eval_samples_per_second": 20.489,
|
| 465 |
+
"eval_steps_per_second": 1.463,
|
| 466 |
+
"step": 56
|
| 467 |
+
},
|
| 468 |
+
{
|
| 469 |
+
"epoch": 41.45,
|
| 470 |
+
"eval_accuracy": 0.0,
|
| 471 |
+
"eval_f1": 0.31097560975609756,
|
| 472 |
+
"eval_loss": 0.6047279238700867,
|
| 473 |
+
"eval_roc_auc": 0.7036749482401656,
|
| 474 |
+
"eval_runtime": 2.2835,
|
| 475 |
+
"eval_samples_per_second": 18.392,
|
| 476 |
+
"eval_steps_per_second": 1.314,
|
| 477 |
+
"step": 57
|
| 478 |
+
},
|
| 479 |
+
{
|
| 480 |
+
"epoch": 42.91,
|
| 481 |
+
"eval_accuracy": 0.0,
|
| 482 |
+
"eval_f1": 0.3164556962025316,
|
| 483 |
+
"eval_loss": 0.6036221385002136,
|
| 484 |
+
"eval_roc_auc": 0.7034161490683231,
|
| 485 |
+
"eval_runtime": 1.9891,
|
| 486 |
+
"eval_samples_per_second": 21.115,
|
| 487 |
+
"eval_steps_per_second": 1.508,
|
| 488 |
+
"step": 59
|
| 489 |
+
},
|
| 490 |
+
{
|
| 491 |
+
"epoch": 43.64,
|
| 492 |
+
"eval_accuracy": 0.0,
|
| 493 |
+
"eval_f1": 0.33116883116883117,
|
| 494 |
+
"eval_loss": 0.6039034128189087,
|
| 495 |
+
"eval_roc_auc": 0.7140269151138717,
|
| 496 |
+
"eval_runtime": 2.2035,
|
| 497 |
+
"eval_samples_per_second": 19.061,
|
| 498 |
+
"eval_steps_per_second": 1.361,
|
| 499 |
+
"step": 60
|
| 500 |
+
},
|
| 501 |
+
{
|
| 502 |
+
"epoch": 44.36,
|
| 503 |
+
"eval_accuracy": 0.0,
|
| 504 |
+
"eval_f1": 0.32258064516129037,
|
| 505 |
+
"eval_loss": 0.6021687984466553,
|
| 506 |
+
"eval_roc_auc": 0.7065217391304347,
|
| 507 |
+
"eval_runtime": 1.9925,
|
| 508 |
+
"eval_samples_per_second": 21.079,
|
| 509 |
+
"eval_steps_per_second": 1.506,
|
| 510 |
+
"step": 61
|
| 511 |
+
},
|
| 512 |
+
{
|
| 513 |
+
"epoch": 45.82,
|
| 514 |
+
"eval_accuracy": 0.0,
|
| 515 |
+
"eval_f1": 0.326797385620915,
|
| 516 |
+
"eval_loss": 0.6014130711555481,
|
| 517 |
+
"eval_roc_auc": 0.708592132505176,
|
| 518 |
+
"eval_runtime": 2.0419,
|
| 519 |
+
"eval_samples_per_second": 20.569,
|
| 520 |
+
"eval_steps_per_second": 1.469,
|
| 521 |
+
"step": 63
|
| 522 |
+
},
|
| 523 |
+
{
|
| 524 |
+
"epoch": 46.55,
|
| 525 |
+
"eval_accuracy": 0.0,
|
| 526 |
+
"eval_f1": 0.3130990415335463,
|
| 527 |
+
"eval_loss": 0.6009289622306824,
|
| 528 |
+
"eval_roc_auc": 0.6984989648033127,
|
| 529 |
+
"eval_runtime": 1.9923,
|
| 530 |
+
"eval_samples_per_second": 21.081,
|
| 531 |
+
"eval_steps_per_second": 1.506,
|
| 532 |
+
"step": 64
|
| 533 |
+
},
|
| 534 |
+
{
|
| 535 |
+
"epoch": 48.0,
|
| 536 |
+
"eval_accuracy": 0.0,
|
| 537 |
+
"eval_f1": 0.3411371237458194,
|
| 538 |
+
"eval_loss": 0.5962069034576416,
|
| 539 |
+
"eval_roc_auc": 0.7186853002070394,
|
| 540 |
+
"eval_runtime": 2.2383,
|
| 541 |
+
"eval_samples_per_second": 18.764,
|
| 542 |
+
"eval_steps_per_second": 1.34,
|
| 543 |
+
"step": 66
|
| 544 |
+
},
|
| 545 |
+
{
|
| 546 |
+
"epoch": 48.73,
|
| 547 |
+
"eval_accuracy": 0.0,
|
| 548 |
+
"eval_f1": 0.34576271186440677,
|
| 549 |
+
"eval_loss": 0.5965468287467957,
|
| 550 |
+
"eval_roc_auc": 0.7207556935817806,
|
| 551 |
+
"eval_runtime": 2.0677,
|
| 552 |
+
"eval_samples_per_second": 20.312,
|
| 553 |
+
"eval_steps_per_second": 1.451,
|
| 554 |
+
"step": 67
|
| 555 |
+
},
|
| 556 |
+
{
|
| 557 |
+
"epoch": 49.45,
|
| 558 |
+
"eval_accuracy": 0.0,
|
| 559 |
+
"eval_f1": 0.341296928327645,
|
| 560 |
+
"eval_loss": 0.5968855023384094,
|
| 561 |
+
"eval_roc_auc": 0.7153209109730849,
|
| 562 |
+
"eval_runtime": 2.0241,
|
| 563 |
+
"eval_samples_per_second": 20.75,
|
| 564 |
+
"eval_steps_per_second": 1.482,
|
| 565 |
+
"step": 68
|
| 566 |
+
},
|
| 567 |
+
{
|
| 568 |
+
"epoch": 50.91,
|
| 569 |
+
"eval_accuracy": 0.0,
|
| 570 |
+
"eval_f1": 0.34,
|
| 571 |
+
"eval_loss": 0.593564510345459,
|
| 572 |
+
"eval_roc_auc": 0.7181677018633541,
|
| 573 |
+
"eval_runtime": 2.0433,
|
| 574 |
+
"eval_samples_per_second": 20.555,
|
| 575 |
+
"eval_steps_per_second": 1.468,
|
| 576 |
+
"step": 70
|
| 577 |
+
},
|
| 578 |
+
{
|
| 579 |
+
"epoch": 51.64,
|
| 580 |
+
"eval_accuracy": 0.0,
|
| 581 |
+
"eval_f1": 0.3411371237458194,
|
| 582 |
+
"eval_loss": 0.591184139251709,
|
| 583 |
+
"eval_roc_auc": 0.7186853002070394,
|
| 584 |
+
"eval_runtime": 2.0663,
|
| 585 |
+
"eval_samples_per_second": 20.326,
|
| 586 |
+
"eval_steps_per_second": 1.452,
|
| 587 |
+
"step": 71
|
| 588 |
+
},
|
| 589 |
+
{
|
| 590 |
+
"epoch": 52.36,
|
| 591 |
+
"eval_accuracy": 0.0,
|
| 592 |
+
"eval_f1": 0.3310344827586207,
|
| 593 |
+
"eval_loss": 0.592108428478241,
|
| 594 |
+
"eval_roc_auc": 0.7039337474120083,
|
| 595 |
+
"eval_runtime": 2.2694,
|
| 596 |
+
"eval_samples_per_second": 18.507,
|
| 597 |
+
"eval_steps_per_second": 1.322,
|
| 598 |
+
"step": 72
|
| 599 |
+
},
|
| 600 |
+
{
|
| 601 |
+
"epoch": 53.82,
|
| 602 |
+
"eval_accuracy": 0.0,
|
| 603 |
+
"eval_f1": 0.34657039711191334,
|
| 604 |
+
"eval_loss": 0.5902323722839355,
|
| 605 |
+
"eval_roc_auc": 0.7106625258799172,
|
| 606 |
+
"eval_runtime": 2.0722,
|
| 607 |
+
"eval_samples_per_second": 20.269,
|
| 608 |
+
"eval_steps_per_second": 1.448,
|
| 609 |
+
"step": 74
|
| 610 |
+
},
|
| 611 |
+
{
|
| 612 |
+
"epoch": 54.55,
|
| 613 |
+
"eval_accuracy": 0.0,
|
| 614 |
+
"eval_f1": 0.34306569343065696,
|
| 615 |
+
"eval_loss": 0.5893351435661316,
|
| 616 |
+
"eval_roc_auc": 0.7057453416149069,
|
| 617 |
+
"eval_runtime": 2.055,
|
| 618 |
+
"eval_samples_per_second": 20.438,
|
| 619 |
+
"eval_steps_per_second": 1.46,
|
| 620 |
+
"step": 75
|
| 621 |
+
},
|
| 622 |
+
{
|
| 623 |
+
"epoch": 56.0,
|
| 624 |
+
"eval_accuracy": 0.0,
|
| 625 |
+
"eval_f1": 0.3404255319148936,
|
| 626 |
+
"eval_loss": 0.5895388722419739,
|
| 627 |
+
"eval_roc_auc": 0.7080745341614907,
|
| 628 |
+
"eval_runtime": 2.0467,
|
| 629 |
+
"eval_samples_per_second": 20.52,
|
| 630 |
+
"eval_steps_per_second": 1.466,
|
| 631 |
+
"step": 77
|
| 632 |
+
},
|
| 633 |
+
{
|
| 634 |
+
"epoch": 56.73,
|
| 635 |
+
"eval_accuracy": 0.0,
|
| 636 |
+
"eval_f1": 0.3252595155709343,
|
| 637 |
+
"eval_loss": 0.5894604921340942,
|
| 638 |
+
"eval_roc_auc": 0.6979813664596274,
|
| 639 |
+
"eval_runtime": 2.1145,
|
| 640 |
+
"eval_samples_per_second": 19.863,
|
| 641 |
+
"eval_steps_per_second": 1.419,
|
| 642 |
+
"step": 78
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 57.45,
|
| 646 |
+
"eval_accuracy": 0.0,
|
| 647 |
+
"eval_f1": 0.35294117647058826,
|
| 648 |
+
"eval_loss": 0.5903995037078857,
|
| 649 |
+
"eval_roc_auc": 0.7238612836438924,
|
| 650 |
+
"eval_runtime": 2.1187,
|
| 651 |
+
"eval_samples_per_second": 19.824,
|
| 652 |
+
"eval_steps_per_second": 1.416,
|
| 653 |
+
"step": 79
|
| 654 |
+
},
|
| 655 |
+
{
|
| 656 |
+
"epoch": 58.91,
|
| 657 |
+
"eval_accuracy": 0.0,
|
| 658 |
+
"eval_f1": 0.33793103448275863,
|
| 659 |
+
"eval_loss": 0.5916685461997986,
|
| 660 |
+
"eval_roc_auc": 0.7104037267080746,
|
| 661 |
+
"eval_runtime": 2.1241,
|
| 662 |
+
"eval_samples_per_second": 19.773,
|
| 663 |
+
"eval_steps_per_second": 1.412,
|
| 664 |
+
"step": 81
|
| 665 |
+
},
|
| 666 |
+
{
|
| 667 |
+
"epoch": 59.64,
|
| 668 |
+
"eval_accuracy": 0.0,
|
| 669 |
+
"eval_f1": 0.3298245614035088,
|
| 670 |
+
"eval_loss": 0.5893025994300842,
|
| 671 |
+
"eval_roc_auc": 0.7000517598343685,
|
| 672 |
+
"eval_runtime": 2.1145,
|
| 673 |
+
"eval_samples_per_second": 19.863,
|
| 674 |
+
"eval_steps_per_second": 1.419,
|
| 675 |
+
"step": 82
|
| 676 |
+
},
|
| 677 |
+
{
|
| 678 |
+
"epoch": 60.36,
|
| 679 |
+
"eval_accuracy": 0.0,
|
| 680 |
+
"eval_f1": 0.36823104693140796,
|
| 681 |
+
"eval_loss": 0.5839800238609314,
|
| 682 |
+
"eval_roc_auc": 0.7300724637681159,
|
| 683 |
+
"eval_runtime": 2.0295,
|
| 684 |
+
"eval_samples_per_second": 20.695,
|
| 685 |
+
"eval_steps_per_second": 1.478,
|
| 686 |
+
"step": 83
|
| 687 |
+
},
|
| 688 |
+
{
|
| 689 |
+
"epoch": 61.82,
|
| 690 |
+
"eval_accuracy": 0.023809523809523808,
|
| 691 |
+
"eval_f1": 0.3636363636363636,
|
| 692 |
+
"eval_loss": 0.5821632146835327,
|
| 693 |
+
"eval_roc_auc": 0.7246376811594203,
|
| 694 |
+
"eval_runtime": 2.1312,
|
| 695 |
+
"eval_samples_per_second": 19.707,
|
| 696 |
+
"eval_steps_per_second": 1.408,
|
| 697 |
+
"step": 85
|
| 698 |
+
},
|
| 699 |
+
{
|
| 700 |
+
"epoch": 62.55,
|
| 701 |
+
"eval_accuracy": 0.023809523809523808,
|
| 702 |
+
"eval_f1": 0.35507246376811596,
|
| 703 |
+
"eval_loss": 0.5816187858581543,
|
| 704 |
+
"eval_roc_auc": 0.7176501035196688,
|
| 705 |
+
"eval_runtime": 2.1966,
|
| 706 |
+
"eval_samples_per_second": 19.121,
|
| 707 |
+
"eval_steps_per_second": 1.366,
|
| 708 |
+
"step": 86
|
| 709 |
+
},
|
| 710 |
+
{
|
| 711 |
+
"epoch": 64.0,
|
| 712 |
+
"eval_accuracy": 0.0,
|
| 713 |
+
"eval_f1": 0.34532374100719426,
|
| 714 |
+
"eval_loss": 0.5823355317115784,
|
| 715 |
+
"eval_roc_auc": 0.7101449275362319,
|
| 716 |
+
"eval_runtime": 2.4993,
|
| 717 |
+
"eval_samples_per_second": 16.804,
|
| 718 |
+
"eval_steps_per_second": 1.2,
|
| 719 |
+
"step": 88
|
| 720 |
+
},
|
| 721 |
+
{
|
| 722 |
+
"epoch": 64.73,
|
| 723 |
+
"eval_accuracy": 0.0,
|
| 724 |
+
"eval_f1": 0.3537906137184116,
|
| 725 |
+
"eval_loss": 0.5836206674575806,
|
| 726 |
+
"eval_roc_auc": 0.7171325051759834,
|
| 727 |
+
"eval_runtime": 1.9876,
|
| 728 |
+
"eval_samples_per_second": 21.131,
|
| 729 |
+
"eval_steps_per_second": 1.509,
|
| 730 |
+
"step": 89
|
| 731 |
+
},
|
| 732 |
+
{
|
| 733 |
+
"epoch": 65.45,
|
| 734 |
+
"eval_accuracy": 0.0,
|
| 735 |
+
"eval_f1": 0.36823104693140796,
|
| 736 |
+
"eval_loss": 0.5811671018600464,
|
| 737 |
+
"eval_roc_auc": 0.7300724637681159,
|
| 738 |
+
"eval_runtime": 2.2487,
|
| 739 |
+
"eval_samples_per_second": 18.677,
|
| 740 |
+
"eval_steps_per_second": 1.334,
|
| 741 |
+
"step": 90
|
| 742 |
+
},
|
| 743 |
+
{
|
| 744 |
+
"epoch": 66.91,
|
| 745 |
+
"eval_accuracy": 0.023809523809523808,
|
| 746 |
+
"eval_f1": 0.35294117647058826,
|
| 747 |
+
"eval_loss": 0.5822634100914001,
|
| 748 |
+
"eval_roc_auc": 0.7132505175983437,
|
| 749 |
+
"eval_runtime": 2.1189,
|
| 750 |
+
"eval_samples_per_second": 19.821,
|
| 751 |
+
"eval_steps_per_second": 1.416,
|
| 752 |
+
"step": 92
|
| 753 |
+
},
|
| 754 |
+
{
|
| 755 |
+
"epoch": 67.64,
|
| 756 |
+
"eval_accuracy": 0.023809523809523808,
|
| 757 |
+
"eval_f1": 0.3468634686346863,
|
| 758 |
+
"eval_loss": 0.5832495093345642,
|
| 759 |
+
"eval_roc_auc": 0.7072981366459627,
|
| 760 |
+
"eval_runtime": 2.1564,
|
| 761 |
+
"eval_samples_per_second": 19.477,
|
| 762 |
+
"eval_steps_per_second": 1.391,
|
| 763 |
+
"step": 93
|
| 764 |
+
},
|
| 765 |
+
{
|
| 766 |
+
"epoch": 68.36,
|
| 767 |
+
"eval_accuracy": 0.023809523809523808,
|
| 768 |
+
"eval_f1": 0.34558823529411764,
|
| 769 |
+
"eval_loss": 0.5849612355232239,
|
| 770 |
+
"eval_roc_auc": 0.7067805383022774,
|
| 771 |
+
"eval_runtime": 2.2328,
|
| 772 |
+
"eval_samples_per_second": 18.811,
|
| 773 |
+
"eval_steps_per_second": 1.344,
|
| 774 |
+
"step": 94
|
| 775 |
+
},
|
| 776 |
+
{
|
| 777 |
+
"epoch": 69.82,
|
| 778 |
+
"eval_accuracy": 0.0,
|
| 779 |
+
"eval_f1": 0.35125448028673834,
|
| 780 |
+
"eval_loss": 0.5786746144294739,
|
| 781 |
+
"eval_roc_auc": 0.7160973084886129,
|
| 782 |
+
"eval_runtime": 2.0885,
|
| 783 |
+
"eval_samples_per_second": 20.11,
|
| 784 |
+
"eval_steps_per_second": 1.436,
|
| 785 |
+
"step": 96
|
| 786 |
+
},
|
| 787 |
+
{
|
| 788 |
+
"epoch": 70.55,
|
| 789 |
+
"eval_accuracy": 0.0,
|
| 790 |
+
"eval_f1": 0.3695652173913044,
|
| 791 |
+
"eval_loss": 0.5768818855285645,
|
| 792 |
+
"eval_roc_auc": 0.7305900621118012,
|
| 793 |
+
"eval_runtime": 2.1063,
|
| 794 |
+
"eval_samples_per_second": 19.94,
|
| 795 |
+
"eval_steps_per_second": 1.424,
|
| 796 |
+
"step": 97
|
| 797 |
+
},
|
| 798 |
+
{
|
| 799 |
+
"epoch": 72.0,
|
| 800 |
+
"eval_accuracy": 0.0,
|
| 801 |
+
"eval_f1": 0.3684210526315789,
|
| 802 |
+
"eval_loss": 0.5726543068885803,
|
| 803 |
+
"eval_roc_auc": 0.7228260869565218,
|
| 804 |
+
"eval_runtime": 2.2057,
|
| 805 |
+
"eval_samples_per_second": 19.042,
|
| 806 |
+
"eval_steps_per_second": 1.36,
|
| 807 |
+
"step": 99
|
| 808 |
+
},
|
| 809 |
+
{
|
| 810 |
+
"epoch": 72.73,
|
| 811 |
+
"eval_accuracy": 0.0,
|
| 812 |
+
"eval_f1": 0.3863636363636363,
|
| 813 |
+
"eval_loss": 0.5710762739181519,
|
| 814 |
+
"eval_roc_auc": 0.736801242236025,
|
| 815 |
+
"eval_runtime": 2.1542,
|
| 816 |
+
"eval_samples_per_second": 19.496,
|
| 817 |
+
"eval_steps_per_second": 1.393,
|
| 818 |
+
"step": 100
|
| 819 |
+
},
|
| 820 |
+
{
|
| 821 |
+
"epoch": 73.45,
|
| 822 |
+
"eval_accuracy": 0.0,
|
| 823 |
+
"eval_f1": 0.37037037037037035,
|
| 824 |
+
"eval_loss": 0.5743962526321411,
|
| 825 |
+
"eval_roc_auc": 0.7272256728778468,
|
| 826 |
+
"eval_runtime": 2.0601,
|
| 827 |
+
"eval_samples_per_second": 20.387,
|
| 828 |
+
"eval_steps_per_second": 1.456,
|
| 829 |
+
"step": 101
|
| 830 |
+
},
|
| 831 |
+
{
|
| 832 |
+
"epoch": 74.91,
|
| 833 |
+
"eval_accuracy": 0.0,
|
| 834 |
+
"eval_f1": 0.36363636363636365,
|
| 835 |
+
"eval_loss": 0.5718241333961487,
|
| 836 |
+
"eval_roc_auc": 0.7173913043478259,
|
| 837 |
+
"eval_runtime": 2.3822,
|
| 838 |
+
"eval_samples_per_second": 17.631,
|
| 839 |
+
"eval_steps_per_second": 1.259,
|
| 840 |
+
"step": 103
|
| 841 |
+
},
|
| 842 |
+
{
|
| 843 |
+
"epoch": 75.64,
|
| 844 |
+
"eval_accuracy": 0.0,
|
| 845 |
+
"eval_f1": 0.36296296296296293,
|
| 846 |
+
"eval_loss": 0.5745264887809753,
|
| 847 |
+
"eval_roc_auc": 0.7207556935817806,
|
| 848 |
+
"eval_runtime": 2.1941,
|
| 849 |
+
"eval_samples_per_second": 19.143,
|
| 850 |
+
"eval_steps_per_second": 1.367,
|
| 851 |
+
"step": 104
|
| 852 |
+
},
|
| 853 |
+
{
|
| 854 |
+
"epoch": 76.36,
|
| 855 |
+
"eval_accuracy": 0.0,
|
| 856 |
+
"eval_f1": 0.3676470588235294,
|
| 857 |
+
"eval_loss": 0.5776281952857971,
|
| 858 |
+
"eval_roc_auc": 0.7261904761904763,
|
| 859 |
+
"eval_runtime": 2.4317,
|
| 860 |
+
"eval_samples_per_second": 17.272,
|
| 861 |
+
"eval_steps_per_second": 1.234,
|
| 862 |
+
"step": 105
|
| 863 |
+
},
|
| 864 |
+
{
|
| 865 |
+
"epoch": 77.82,
|
| 866 |
+
"eval_accuracy": 0.023809523809523808,
|
| 867 |
+
"eval_f1": 0.36296296296296293,
|
| 868 |
+
"eval_loss": 0.5703898668289185,
|
| 869 |
+
"eval_roc_auc": 0.7207556935817806,
|
| 870 |
+
"eval_runtime": 2.0787,
|
| 871 |
+
"eval_samples_per_second": 20.205,
|
| 872 |
+
"eval_steps_per_second": 1.443,
|
| 873 |
+
"step": 107
|
| 874 |
+
},
|
| 875 |
+
{
|
| 876 |
+
"epoch": 78.55,
|
| 877 |
+
"eval_accuracy": 0.0,
|
| 878 |
+
"eval_f1": 0.37547892720306514,
|
| 879 |
+
"eval_loss": 0.5683363676071167,
|
| 880 |
+
"eval_roc_auc": 0.7254140786749483,
|
| 881 |
+
"eval_runtime": 2.2741,
|
| 882 |
+
"eval_samples_per_second": 18.469,
|
| 883 |
+
"eval_steps_per_second": 1.319,
|
| 884 |
+
"step": 108
|
| 885 |
+
},
|
| 886 |
+
{
|
| 887 |
+
"epoch": 80.0,
|
| 888 |
+
"eval_accuracy": 0.023809523809523808,
|
| 889 |
+
"eval_f1": 0.41600000000000004,
|
| 890 |
+
"eval_loss": 0.5640491247177124,
|
| 891 |
+
"eval_roc_auc": 0.7505175983436853,
|
| 892 |
+
"eval_runtime": 2.3409,
|
| 893 |
+
"eval_samples_per_second": 17.941,
|
| 894 |
+
"eval_steps_per_second": 1.282,
|
| 895 |
+
"step": 110
|
| 896 |
+
},
|
| 897 |
+
{
|
| 898 |
+
"epoch": 80.73,
|
| 899 |
+
"eval_accuracy": 0.023809523809523808,
|
| 900 |
+
"eval_f1": 0.38095238095238093,
|
| 901 |
+
"eval_loss": 0.5674084424972534,
|
| 902 |
+
"eval_roc_auc": 0.7236024844720497,
|
| 903 |
+
"eval_runtime": 2.19,
|
| 904 |
+
"eval_samples_per_second": 19.178,
|
| 905 |
+
"eval_steps_per_second": 1.37,
|
| 906 |
+
"step": 111
|
| 907 |
+
},
|
| 908 |
+
{
|
| 909 |
+
"epoch": 81.45,
|
| 910 |
+
"eval_accuracy": 0.023809523809523808,
|
| 911 |
+
"eval_f1": 0.3730158730158731,
|
| 912 |
+
"eval_loss": 0.564328670501709,
|
| 913 |
+
"eval_roc_auc": 0.7171325051759834,
|
| 914 |
+
"eval_runtime": 2.4742,
|
| 915 |
+
"eval_samples_per_second": 16.975,
|
| 916 |
+
"eval_steps_per_second": 1.213,
|
| 917 |
+
"step": 112
|
| 918 |
+
},
|
| 919 |
+
{
|
| 920 |
+
"epoch": 82.91,
|
| 921 |
+
"eval_accuracy": 0.023809523809523808,
|
| 922 |
+
"eval_f1": 0.3813229571984436,
|
| 923 |
+
"eval_loss": 0.5711435675621033,
|
| 924 |
+
"eval_roc_auc": 0.7274844720496896,
|
| 925 |
+
"eval_runtime": 2.1624,
|
| 926 |
+
"eval_samples_per_second": 19.423,
|
| 927 |
+
"eval_steps_per_second": 1.387,
|
| 928 |
+
"step": 114
|
| 929 |
+
},
|
| 930 |
+
{
|
| 931 |
+
"epoch": 83.64,
|
| 932 |
+
"eval_accuracy": 0.023809523809523808,
|
| 933 |
+
"eval_f1": 0.366412213740458,
|
| 934 |
+
"eval_loss": 0.5771781206130981,
|
| 935 |
+
"eval_roc_auc": 0.7184265010351967,
|
| 936 |
+
"eval_runtime": 2.2132,
|
| 937 |
+
"eval_samples_per_second": 18.977,
|
| 938 |
+
"eval_steps_per_second": 1.356,
|
| 939 |
+
"step": 115
|
| 940 |
+
},
|
| 941 |
+
{
|
| 942 |
+
"epoch": 84.36,
|
| 943 |
+
"eval_accuracy": 0.023809523809523808,
|
| 944 |
+
"eval_f1": 0.3582089552238805,
|
| 945 |
+
"eval_loss": 0.5758051872253418,
|
| 946 |
+
"eval_roc_auc": 0.7153209109730849,
|
| 947 |
+
"eval_runtime": 2.2313,
|
| 948 |
+
"eval_samples_per_second": 18.823,
|
| 949 |
+
"eval_steps_per_second": 1.345,
|
| 950 |
+
"step": 116
|
| 951 |
+
},
|
| 952 |
+
{
|
| 953 |
+
"epoch": 85.82,
|
| 954 |
+
"eval_accuracy": 0.0,
|
| 955 |
+
"eval_f1": 0.37209302325581395,
|
| 956 |
+
"eval_loss": 0.5687234401702881,
|
| 957 |
+
"eval_roc_auc": 0.7204968944099378,
|
| 958 |
+
"eval_runtime": 2.117,
|
| 959 |
+
"eval_samples_per_second": 19.839,
|
| 960 |
+
"eval_steps_per_second": 1.417,
|
| 961 |
+
"step": 118
|
| 962 |
+
},
|
| 963 |
+
{
|
| 964 |
+
"epoch": 86.55,
|
| 965 |
+
"eval_accuracy": 0.0,
|
| 966 |
+
"eval_f1": 0.3904382470119522,
|
| 967 |
+
"eval_loss": 0.565062403678894,
|
| 968 |
+
"eval_roc_auc": 0.7305900621118013,
|
| 969 |
+
"eval_runtime": 2.145,
|
| 970 |
+
"eval_samples_per_second": 19.58,
|
| 971 |
+
"eval_steps_per_second": 1.399,
|
| 972 |
+
"step": 119
|
| 973 |
+
},
|
| 974 |
+
{
|
| 975 |
+
"epoch": 88.0,
|
| 976 |
+
"eval_accuracy": 0.0,
|
| 977 |
+
"eval_f1": 0.4,
|
| 978 |
+
"eval_loss": 0.5594702959060669,
|
| 979 |
+
"eval_roc_auc": 0.7375776397515528,
|
| 980 |
+
"eval_runtime": 2.141,
|
| 981 |
+
"eval_samples_per_second": 19.617,
|
| 982 |
+
"eval_steps_per_second": 1.401,
|
| 983 |
+
"step": 121
|
| 984 |
+
},
|
| 985 |
+
{
|
| 986 |
+
"epoch": 88.73,
|
| 987 |
+
"eval_accuracy": 0.0,
|
| 988 |
+
"eval_f1": 0.421875,
|
| 989 |
+
"eval_loss": 0.5597259998321533,
|
| 990 |
+
"eval_roc_auc": 0.760351966873706,
|
| 991 |
+
"eval_runtime": 2.2421,
|
| 992 |
+
"eval_samples_per_second": 18.733,
|
| 993 |
+
"eval_steps_per_second": 1.338,
|
| 994 |
+
"step": 122
|
| 995 |
+
},
|
| 996 |
+
{
|
| 997 |
+
"epoch": 89.45,
|
| 998 |
+
"eval_accuracy": 0.0,
|
| 999 |
+
"eval_f1": 0.4156862745098039,
|
| 1000 |
+
"eval_loss": 0.5603341460227966,
|
| 1001 |
+
"eval_roc_auc": 0.754399585921325,
|
| 1002 |
+
"eval_runtime": 2.1509,
|
| 1003 |
+
"eval_samples_per_second": 19.527,
|
| 1004 |
+
"eval_steps_per_second": 1.395,
|
| 1005 |
+
"step": 123
|
| 1006 |
+
},
|
| 1007 |
+
{
|
| 1008 |
+
"epoch": 90.91,
|
| 1009 |
+
"eval_accuracy": 0.0,
|
| 1010 |
+
"eval_f1": 0.4108527131782946,
|
| 1011 |
+
"eval_loss": 0.554810643196106,
|
| 1012 |
+
"eval_roc_auc": 0.7528467908902691,
|
| 1013 |
+
"eval_runtime": 2.0949,
|
| 1014 |
+
"eval_samples_per_second": 20.049,
|
| 1015 |
+
"eval_steps_per_second": 1.432,
|
| 1016 |
+
"step": 125
|
| 1017 |
+
},
|
| 1018 |
+
{
|
| 1019 |
+
"epoch": 91.64,
|
| 1020 |
+
"eval_accuracy": 0.0,
|
| 1021 |
+
"eval_f1": 0.41406249999999994,
|
| 1022 |
+
"eval_loss": 0.55705326795578,
|
| 1023 |
+
"eval_roc_auc": 0.7538819875776397,
|
| 1024 |
+
"eval_runtime": 2.1818,
|
| 1025 |
+
"eval_samples_per_second": 19.25,
|
| 1026 |
+
"eval_steps_per_second": 1.375,
|
| 1027 |
+
"step": 126
|
| 1028 |
+
},
|
| 1029 |
+
{
|
| 1030 |
+
"epoch": 92.36,
|
| 1031 |
+
"eval_accuracy": 0.023809523809523808,
|
| 1032 |
+
"eval_f1": 0.40476190476190477,
|
| 1033 |
+
"eval_loss": 0.5568869709968567,
|
| 1034 |
+
"eval_roc_auc": 0.7430124223602484,
|
| 1035 |
+
"eval_runtime": 2.1783,
|
| 1036 |
+
"eval_samples_per_second": 19.281,
|
| 1037 |
+
"eval_steps_per_second": 1.377,
|
| 1038 |
+
"step": 127
|
| 1039 |
+
},
|
| 1040 |
+
{
|
| 1041 |
+
"epoch": 93.82,
|
| 1042 |
+
"eval_accuracy": 0.023809523809523808,
|
| 1043 |
+
"eval_f1": 0.41732283464566927,
|
| 1044 |
+
"eval_loss": 0.5582275986671448,
|
| 1045 |
+
"eval_roc_auc": 0.7549171842650103,
|
| 1046 |
+
"eval_runtime": 2.1395,
|
| 1047 |
+
"eval_samples_per_second": 19.63,
|
| 1048 |
+
"eval_steps_per_second": 1.402,
|
| 1049 |
+
"step": 129
|
| 1050 |
+
},
|
| 1051 |
+
{
|
| 1052 |
+
"epoch": 94.55,
|
| 1053 |
+
"eval_accuracy": 0.023809523809523808,
|
| 1054 |
+
"eval_f1": 0.40310077519379844,
|
| 1055 |
+
"eval_loss": 0.561271071434021,
|
| 1056 |
+
"eval_roc_auc": 0.7463768115942029,
|
| 1057 |
+
"eval_runtime": 2.1505,
|
| 1058 |
+
"eval_samples_per_second": 19.53,
|
| 1059 |
+
"eval_steps_per_second": 1.395,
|
| 1060 |
+
"step": 130
|
| 1061 |
+
},
|
| 1062 |
+
{
|
| 1063 |
+
"epoch": 96.0,
|
| 1064 |
+
"eval_accuracy": 0.023809523809523808,
|
| 1065 |
+
"eval_f1": 0.40800000000000003,
|
| 1066 |
+
"eval_loss": 0.5574062466621399,
|
| 1067 |
+
"eval_roc_auc": 0.7440476190476191,
|
| 1068 |
+
"eval_runtime": 2.2873,
|
| 1069 |
+
"eval_samples_per_second": 18.362,
|
| 1070 |
+
"eval_steps_per_second": 1.312,
|
| 1071 |
+
"step": 132
|
| 1072 |
+
},
|
| 1073 |
+
{
|
| 1074 |
+
"epoch": 96.73,
|
| 1075 |
+
"eval_accuracy": 0.0,
|
| 1076 |
+
"eval_f1": 0.421875,
|
| 1077 |
+
"eval_loss": 0.5570902228355408,
|
| 1078 |
+
"eval_roc_auc": 0.760351966873706,
|
| 1079 |
+
"eval_runtime": 2.2328,
|
| 1080 |
+
"eval_samples_per_second": 18.81,
|
| 1081 |
+
"eval_steps_per_second": 1.344,
|
| 1082 |
+
"step": 133
|
| 1083 |
+
},
|
| 1084 |
+
{
|
| 1085 |
+
"epoch": 97.45,
|
| 1086 |
+
"eval_accuracy": 0.023809523809523808,
|
| 1087 |
+
"eval_f1": 0.41732283464566927,
|
| 1088 |
+
"eval_loss": 0.5585192441940308,
|
| 1089 |
+
"eval_roc_auc": 0.7549171842650103,
|
| 1090 |
+
"eval_runtime": 2.1462,
|
| 1091 |
+
"eval_samples_per_second": 19.569,
|
| 1092 |
+
"eval_steps_per_second": 1.398,
|
| 1093 |
+
"step": 134
|
| 1094 |
+
},
|
| 1095 |
+
{
|
| 1096 |
+
"epoch": 98.91,
|
| 1097 |
+
"eval_accuracy": 0.0,
|
| 1098 |
+
"eval_f1": 0.40625,
|
| 1099 |
+
"eval_loss": 0.5598769783973694,
|
| 1100 |
+
"eval_roc_auc": 0.7474120082815735,
|
| 1101 |
+
"eval_runtime": 2.1253,
|
| 1102 |
+
"eval_samples_per_second": 19.762,
|
| 1103 |
+
"eval_steps_per_second": 1.412,
|
| 1104 |
+
"step": 136
|
| 1105 |
+
},
|
| 1106 |
+
{
|
| 1107 |
+
"epoch": 99.64,
|
| 1108 |
+
"eval_accuracy": 0.023809523809523808,
|
| 1109 |
+
"eval_f1": 0.40476190476190477,
|
| 1110 |
+
"eval_loss": 0.5586134195327759,
|
| 1111 |
+
"eval_roc_auc": 0.7430124223602484,
|
| 1112 |
+
"eval_runtime": 2.1887,
|
| 1113 |
+
"eval_samples_per_second": 19.189,
|
| 1114 |
+
"eval_steps_per_second": 1.371,
|
| 1115 |
+
"step": 137
|
| 1116 |
+
},
|
| 1117 |
+
{
|
| 1118 |
+
"epoch": 100.36,
|
| 1119 |
+
"eval_accuracy": 0.047619047619047616,
|
| 1120 |
+
"eval_f1": 0.4112903225806452,
|
| 1121 |
+
"eval_loss": 0.5581173300743103,
|
| 1122 |
+
"eval_roc_auc": 0.7450828157349897,
|
| 1123 |
+
"eval_runtime": 2.1216,
|
| 1124 |
+
"eval_samples_per_second": 19.796,
|
| 1125 |
+
"eval_steps_per_second": 1.414,
|
| 1126 |
+
"step": 138
|
| 1127 |
+
},
|
| 1128 |
+
{
|
| 1129 |
+
"epoch": 101.82,
|
| 1130 |
+
"eval_accuracy": 0.047619047619047616,
|
| 1131 |
+
"eval_f1": 0.4117647058823529,
|
| 1132 |
+
"eval_loss": 0.554703414440155,
|
| 1133 |
+
"eval_roc_auc": 0.7373188405797102,
|
| 1134 |
+
"eval_runtime": 2.199,
|
| 1135 |
+
"eval_samples_per_second": 19.1,
|
| 1136 |
+
"eval_steps_per_second": 1.364,
|
| 1137 |
+
"step": 140
|
| 1138 |
+
},
|
| 1139 |
+
{
|
| 1140 |
+
"epoch": 102.55,
|
| 1141 |
+
"eval_accuracy": 0.047619047619047616,
|
| 1142 |
+
"eval_f1": 0.430379746835443,
|
| 1143 |
+
"eval_loss": 0.5479743480682373,
|
| 1144 |
+
"eval_roc_auc": 0.7507763975155279,
|
| 1145 |
+
"eval_runtime": 2.2214,
|
| 1146 |
+
"eval_samples_per_second": 18.907,
|
| 1147 |
+
"eval_steps_per_second": 1.35,
|
| 1148 |
+
"step": 141
|
| 1149 |
+
},
|
| 1150 |
+
{
|
| 1151 |
+
"epoch": 104.0,
|
| 1152 |
+
"eval_accuracy": 0.023809523809523808,
|
| 1153 |
+
"eval_f1": 0.43333333333333335,
|
| 1154 |
+
"eval_loss": 0.5521491169929504,
|
| 1155 |
+
"eval_roc_auc": 0.7556935817805382,
|
| 1156 |
+
"eval_runtime": 2.0733,
|
| 1157 |
+
"eval_samples_per_second": 20.258,
|
| 1158 |
+
"eval_steps_per_second": 1.447,
|
| 1159 |
+
"step": 143
|
| 1160 |
+
},
|
| 1161 |
+
{
|
| 1162 |
+
"epoch": 104.73,
|
| 1163 |
+
"eval_accuracy": 0.047619047619047616,
|
| 1164 |
+
"eval_f1": 0.4193548387096774,
|
| 1165 |
+
"eval_loss": 0.5539843440055847,
|
| 1166 |
+
"eval_roc_auc": 0.7515527950310559,
|
| 1167 |
+
"eval_runtime": 2.1533,
|
| 1168 |
+
"eval_samples_per_second": 19.505,
|
| 1169 |
+
"eval_steps_per_second": 1.393,
|
| 1170 |
+
"step": 144
|
| 1171 |
+
},
|
| 1172 |
+
{
|
| 1173 |
+
"epoch": 105.45,
|
| 1174 |
+
"eval_accuracy": 0.047619047619047616,
|
| 1175 |
+
"eval_f1": 0.4032921810699588,
|
| 1176 |
+
"eval_loss": 0.5572003126144409,
|
| 1177 |
+
"eval_roc_auc": 0.7347308488612837,
|
| 1178 |
+
"eval_runtime": 2.1589,
|
| 1179 |
+
"eval_samples_per_second": 19.454,
|
| 1180 |
+
"eval_steps_per_second": 1.39,
|
| 1181 |
+
"step": 145
|
| 1182 |
+
},
|
| 1183 |
+
{
|
| 1184 |
+
"epoch": 106.91,
|
| 1185 |
+
"eval_accuracy": 0.023809523809523808,
|
| 1186 |
+
"eval_f1": 0.40322580645161293,
|
| 1187 |
+
"eval_loss": 0.5563521981239319,
|
| 1188 |
+
"eval_roc_auc": 0.7386128364389234,
|
| 1189 |
+
"eval_runtime": 2.1472,
|
| 1190 |
+
"eval_samples_per_second": 19.561,
|
| 1191 |
+
"eval_steps_per_second": 1.397,
|
| 1192 |
+
"step": 147
|
| 1193 |
+
},
|
| 1194 |
+
{
|
| 1195 |
+
"epoch": 107.64,
|
| 1196 |
+
"eval_accuracy": 0.047619047619047616,
|
| 1197 |
+
"eval_f1": 0.39669421487603307,
|
| 1198 |
+
"eval_loss": 0.5568532943725586,
|
| 1199 |
+
"eval_roc_auc": 0.7287784679089027,
|
| 1200 |
+
"eval_runtime": 2.2259,
|
| 1201 |
+
"eval_samples_per_second": 18.869,
|
| 1202 |
+
"eval_steps_per_second": 1.348,
|
| 1203 |
+
"step": 148
|
| 1204 |
+
},
|
| 1205 |
+
{
|
| 1206 |
+
"epoch": 108.36,
|
| 1207 |
+
"eval_accuracy": 0.07142857142857142,
|
| 1208 |
+
"eval_f1": 0.41201716738197425,
|
| 1209 |
+
"eval_loss": 0.557677149772644,
|
| 1210 |
+
"eval_roc_auc": 0.7334368530020704,
|
| 1211 |
+
"eval_runtime": 2.0947,
|
| 1212 |
+
"eval_samples_per_second": 20.051,
|
| 1213 |
+
"eval_steps_per_second": 1.432,
|
| 1214 |
+
"step": 149
|
| 1215 |
+
},
|
| 1216 |
+
{
|
| 1217 |
+
"epoch": 109.82,
|
| 1218 |
+
"eval_accuracy": 0.0,
|
| 1219 |
+
"eval_f1": 0.39004149377593367,
|
| 1220 |
+
"eval_loss": 0.552203357219696,
|
| 1221 |
+
"eval_roc_auc": 0.7228260869565217,
|
| 1222 |
+
"eval_runtime": 2.2966,
|
| 1223 |
+
"eval_samples_per_second": 18.288,
|
| 1224 |
+
"eval_steps_per_second": 1.306,
|
| 1225 |
+
"step": 151
|
| 1226 |
+
},
|
| 1227 |
+
{
|
| 1228 |
+
"epoch": 110.55,
|
| 1229 |
+
"eval_accuracy": 0.0,
|
| 1230 |
+
"eval_f1": 0.3949579831932773,
|
| 1231 |
+
"eval_loss": 0.5478394031524658,
|
| 1232 |
+
"eval_roc_auc": 0.7243788819875777,
|
| 1233 |
+
"eval_runtime": 2.2413,
|
| 1234 |
+
"eval_samples_per_second": 18.739,
|
| 1235 |
+
"eval_steps_per_second": 1.339,
|
| 1236 |
+
"step": 152
|
| 1237 |
+
},
|
| 1238 |
+
{
|
| 1239 |
+
"epoch": 112.0,
|
| 1240 |
+
"eval_accuracy": 0.023809523809523808,
|
| 1241 |
+
"eval_f1": 0.41841004184100417,
|
| 1242 |
+
"eval_loss": 0.5458911061286926,
|
| 1243 |
+
"eval_roc_auc": 0.7432712215320911,
|
| 1244 |
+
"eval_runtime": 2.2404,
|
| 1245 |
+
"eval_samples_per_second": 18.747,
|
| 1246 |
+
"eval_steps_per_second": 1.339,
|
| 1247 |
+
"step": 154
|
| 1248 |
+
},
|
| 1249 |
+
{
|
| 1250 |
+
"epoch": 112.73,
|
| 1251 |
+
"eval_accuracy": 0.047619047619047616,
|
| 1252 |
+
"eval_f1": 0.40506329113924044,
|
| 1253 |
+
"eval_loss": 0.5490586161613464,
|
| 1254 |
+
"eval_roc_auc": 0.731366459627329,
|
| 1255 |
+
"eval_runtime": 2.2886,
|
| 1256 |
+
"eval_samples_per_second": 18.352,
|
| 1257 |
+
"eval_steps_per_second": 1.311,
|
| 1258 |
+
"step": 155
|
| 1259 |
+
},
|
| 1260 |
+
{
|
| 1261 |
+
"epoch": 113.45,
|
| 1262 |
+
"eval_accuracy": 0.023809523809523808,
|
| 1263 |
+
"eval_f1": 0.41841004184100417,
|
| 1264 |
+
"eval_loss": 0.5519034266471863,
|
| 1265 |
+
"eval_roc_auc": 0.7432712215320911,
|
| 1266 |
+
"eval_runtime": 2.1727,
|
| 1267 |
+
"eval_samples_per_second": 19.331,
|
| 1268 |
+
"eval_steps_per_second": 1.381,
|
| 1269 |
+
"step": 156
|
| 1270 |
+
},
|
| 1271 |
+
{
|
| 1272 |
+
"epoch": 114.91,
|
| 1273 |
+
"eval_accuracy": 0.023809523809523808,
|
| 1274 |
+
"eval_f1": 0.3950617283950617,
|
| 1275 |
+
"eval_loss": 0.550848662853241,
|
| 1276 |
+
"eval_roc_auc": 0.7282608695652175,
|
| 1277 |
+
"eval_runtime": 2.5249,
|
| 1278 |
+
"eval_samples_per_second": 16.635,
|
| 1279 |
+
"eval_steps_per_second": 1.188,
|
| 1280 |
+
"step": 158
|
| 1281 |
+
},
|
| 1282 |
+
{
|
| 1283 |
+
"epoch": 115.64,
|
| 1284 |
+
"eval_accuracy": 0.0,
|
| 1285 |
+
"eval_f1": 0.4032921810699588,
|
| 1286 |
+
"eval_loss": 0.5499736070632935,
|
| 1287 |
+
"eval_roc_auc": 0.7347308488612837,
|
| 1288 |
+
"eval_runtime": 2.203,
|
| 1289 |
+
"eval_samples_per_second": 19.065,
|
| 1290 |
+
"eval_steps_per_second": 1.362,
|
| 1291 |
+
"step": 159
|
| 1292 |
+
},
|
| 1293 |
+
{
|
| 1294 |
+
"epoch": 116.36,
|
| 1295 |
+
"eval_accuracy": 0.023809523809523808,
|
| 1296 |
+
"eval_f1": 0.38866396761133604,
|
| 1297 |
+
"eval_loss": 0.5513898134231567,
|
| 1298 |
+
"eval_roc_auc": 0.7261904761904762,
|
| 1299 |
+
"eval_runtime": 2.1649,
|
| 1300 |
+
"eval_samples_per_second": 19.4,
|
| 1301 |
+
"eval_steps_per_second": 1.386,
|
| 1302 |
+
"step": 160
|
| 1303 |
+
},
|
| 1304 |
+
{
|
| 1305 |
+
"epoch": 117.82,
|
| 1306 |
+
"eval_accuracy": 0.047619047619047616,
|
| 1307 |
+
"eval_f1": 0.4255319148936171,
|
| 1308 |
+
"eval_loss": 0.544844388961792,
|
| 1309 |
+
"eval_roc_auc": 0.7453416149068324,
|
| 1310 |
+
"eval_runtime": 2.2226,
|
| 1311 |
+
"eval_samples_per_second": 18.897,
|
| 1312 |
+
"eval_steps_per_second": 1.35,
|
| 1313 |
+
"step": 162
|
| 1314 |
+
},
|
| 1315 |
+
{
|
| 1316 |
+
"epoch": 118.55,
|
| 1317 |
+
"eval_accuracy": 0.047619047619047616,
|
| 1318 |
+
"eval_f1": 0.430379746835443,
|
| 1319 |
+
"eval_loss": 0.5451788306236267,
|
| 1320 |
+
"eval_roc_auc": 0.7507763975155279,
|
| 1321 |
+
"eval_runtime": 2.2158,
|
| 1322 |
+
"eval_samples_per_second": 18.955,
|
| 1323 |
+
"eval_steps_per_second": 1.354,
|
| 1324 |
+
"step": 163
|
| 1325 |
+
},
|
| 1326 |
+
{
|
| 1327 |
+
"epoch": 120.0,
|
| 1328 |
+
"eval_accuracy": 0.07142857142857142,
|
| 1329 |
+
"eval_f1": 0.43965517241379304,
|
| 1330 |
+
"eval_loss": 0.5420504212379456,
|
| 1331 |
+
"eval_roc_auc": 0.7533643892339544,
|
| 1332 |
+
"eval_runtime": 2.2713,
|
| 1333 |
+
"eval_samples_per_second": 18.492,
|
| 1334 |
+
"eval_steps_per_second": 1.321,
|
| 1335 |
+
"step": 165
|
| 1336 |
+
},
|
| 1337 |
+
{
|
| 1338 |
+
"epoch": 120.73,
|
| 1339 |
+
"eval_accuracy": 0.07142857142857142,
|
| 1340 |
+
"eval_f1": 0.43478260869565216,
|
| 1341 |
+
"eval_loss": 0.5413345098495483,
|
| 1342 |
+
"eval_roc_auc": 0.7479296066252589,
|
| 1343 |
+
"eval_runtime": 2.1593,
|
| 1344 |
+
"eval_samples_per_second": 19.451,
|
| 1345 |
+
"eval_steps_per_second": 1.389,
|
| 1346 |
+
"step": 166
|
| 1347 |
+
},
|
| 1348 |
+
{
|
| 1349 |
+
"epoch": 121.45,
|
| 1350 |
+
"eval_accuracy": 0.07142857142857142,
|
| 1351 |
+
"eval_f1": 0.42735042735042733,
|
| 1352 |
+
"eval_loss": 0.5442317724227905,
|
| 1353 |
+
"eval_roc_auc": 0.7458592132505176,
|
| 1354 |
+
"eval_runtime": 2.22,
|
| 1355 |
+
"eval_samples_per_second": 18.919,
|
| 1356 |
+
"eval_steps_per_second": 1.351,
|
| 1357 |
+
"step": 167
|
| 1358 |
+
},
|
| 1359 |
+
{
|
| 1360 |
+
"epoch": 122.91,
|
| 1361 |
+
"eval_accuracy": 0.047619047619047616,
|
| 1362 |
+
"eval_f1": 0.430379746835443,
|
| 1363 |
+
"eval_loss": 0.5420621037483215,
|
| 1364 |
+
"eval_roc_auc": 0.7507763975155279,
|
| 1365 |
+
"eval_runtime": 2.2358,
|
| 1366 |
+
"eval_samples_per_second": 18.785,
|
| 1367 |
+
"eval_steps_per_second": 1.342,
|
| 1368 |
+
"step": 169
|
| 1369 |
+
},
|
| 1370 |
+
{
|
| 1371 |
+
"epoch": 123.64,
|
| 1372 |
+
"eval_accuracy": 0.07142857142857142,
|
| 1373 |
+
"eval_f1": 0.42372881355932207,
|
| 1374 |
+
"eval_loss": 0.5403344035148621,
|
| 1375 |
+
"eval_roc_auc": 0.7448240165631471,
|
| 1376 |
+
"eval_runtime": 2.1592,
|
| 1377 |
+
"eval_samples_per_second": 19.451,
|
| 1378 |
+
"eval_steps_per_second": 1.389,
|
| 1379 |
+
"step": 170
|
| 1380 |
+
},
|
| 1381 |
+
{
|
| 1382 |
+
"epoch": 124.36,
|
| 1383 |
+
"eval_accuracy": 0.09523809523809523,
|
| 1384 |
+
"eval_f1": 0.43478260869565216,
|
| 1385 |
+
"eval_loss": 0.5386914014816284,
|
| 1386 |
+
"eval_roc_auc": 0.7479296066252589,
|
| 1387 |
+
"eval_runtime": 2.1762,
|
| 1388 |
+
"eval_samples_per_second": 19.3,
|
| 1389 |
+
"eval_steps_per_second": 1.379,
|
| 1390 |
+
"step": 171
|
| 1391 |
+
},
|
| 1392 |
+
{
|
| 1393 |
+
"epoch": 125.82,
|
| 1394 |
+
"eval_accuracy": 0.047619047619047616,
|
| 1395 |
+
"eval_f1": 0.42241379310344834,
|
| 1396 |
+
"eval_loss": 0.5383176207542419,
|
| 1397 |
+
"eval_roc_auc": 0.740424430641822,
|
| 1398 |
+
"eval_runtime": 2.3145,
|
| 1399 |
+
"eval_samples_per_second": 18.147,
|
| 1400 |
+
"eval_steps_per_second": 1.296,
|
| 1401 |
+
"step": 173
|
| 1402 |
+
},
|
| 1403 |
+
{
|
| 1404 |
+
"epoch": 126.55,
|
| 1405 |
+
"eval_accuracy": 0.07142857142857142,
|
| 1406 |
+
"eval_f1": 0.42290748898678415,
|
| 1407 |
+
"eval_loss": 0.5346717238426208,
|
| 1408 |
+
"eval_roc_auc": 0.7365424430641822,
|
| 1409 |
+
"eval_runtime": 2.2566,
|
| 1410 |
+
"eval_samples_per_second": 18.612,
|
| 1411 |
+
"eval_steps_per_second": 1.329,
|
| 1412 |
+
"step": 174
|
| 1413 |
+
},
|
| 1414 |
+
{
|
| 1415 |
+
"epoch": 128.0,
|
| 1416 |
+
"eval_accuracy": 0.07142857142857142,
|
| 1417 |
+
"eval_f1": 0.43478260869565216,
|
| 1418 |
+
"eval_loss": 0.5410609245300293,
|
| 1419 |
+
"eval_roc_auc": 0.7479296066252589,
|
| 1420 |
+
"eval_runtime": 2.2479,
|
| 1421 |
+
"eval_samples_per_second": 18.684,
|
| 1422 |
+
"eval_steps_per_second": 1.335,
|
| 1423 |
+
"step": 176
|
| 1424 |
+
},
|
| 1425 |
+
{
|
| 1426 |
+
"epoch": 128.73,
|
| 1427 |
+
"eval_accuracy": 0.07142857142857142,
|
| 1428 |
+
"eval_f1": 0.40869565217391307,
|
| 1429 |
+
"eval_loss": 0.5448868870735168,
|
| 1430 |
+
"eval_roc_auc": 0.72851966873706,
|
| 1431 |
+
"eval_runtime": 2.2593,
|
| 1432 |
+
"eval_samples_per_second": 18.59,
|
| 1433 |
+
"eval_steps_per_second": 1.328,
|
| 1434 |
+
"step": 177
|
| 1435 |
+
},
|
| 1436 |
+
{
|
| 1437 |
+
"epoch": 129.45,
|
| 1438 |
+
"eval_accuracy": 0.07142857142857142,
|
| 1439 |
+
"eval_f1": 0.4188034188034188,
|
| 1440 |
+
"eval_loss": 0.5426321029663086,
|
| 1441 |
+
"eval_roc_auc": 0.7393892339544514,
|
| 1442 |
+
"eval_runtime": 2.1827,
|
| 1443 |
+
"eval_samples_per_second": 19.242,
|
| 1444 |
+
"eval_steps_per_second": 1.374,
|
| 1445 |
+
"step": 178
|
| 1446 |
+
},
|
| 1447 |
+
{
|
| 1448 |
+
"epoch": 130.91,
|
| 1449 |
+
"eval_accuracy": 0.07142857142857142,
|
| 1450 |
+
"eval_f1": 0.4034334763948498,
|
| 1451 |
+
"eval_loss": 0.5429388880729675,
|
| 1452 |
+
"eval_roc_auc": 0.7269668737060041,
|
| 1453 |
+
"eval_runtime": 2.1992,
|
| 1454 |
+
"eval_samples_per_second": 19.098,
|
| 1455 |
+
"eval_steps_per_second": 1.364,
|
| 1456 |
+
"step": 180
|
| 1457 |
+
},
|
| 1458 |
+
{
|
| 1459 |
+
"epoch": 131.64,
|
| 1460 |
+
"eval_accuracy": 0.09523809523809523,
|
| 1461 |
+
"eval_f1": 0.4255319148936171,
|
| 1462 |
+
"eval_loss": 0.5380597710609436,
|
| 1463 |
+
"eval_roc_auc": 0.7453416149068324,
|
| 1464 |
+
"eval_runtime": 2.2546,
|
| 1465 |
+
"eval_samples_per_second": 18.628,
|
| 1466 |
+
"eval_steps_per_second": 1.331,
|
| 1467 |
+
"step": 181
|
| 1468 |
+
},
|
| 1469 |
+
{
|
| 1470 |
+
"epoch": 132.36,
|
| 1471 |
+
"eval_accuracy": 0.07142857142857142,
|
| 1472 |
+
"eval_f1": 0.4279475982532751,
|
| 1473 |
+
"eval_loss": 0.5344164967536926,
|
| 1474 |
+
"eval_roc_auc": 0.7419772256728779,
|
| 1475 |
+
"eval_runtime": 2.2146,
|
| 1476 |
+
"eval_samples_per_second": 18.965,
|
| 1477 |
+
"eval_steps_per_second": 1.355,
|
| 1478 |
+
"step": 182
|
| 1479 |
+
},
|
| 1480 |
+
{
|
| 1481 |
+
"epoch": 133.82,
|
| 1482 |
+
"eval_accuracy": 0.047619047619047616,
|
| 1483 |
+
"eval_f1": 0.39316239316239315,
|
| 1484 |
+
"eval_loss": 0.5364577174186707,
|
| 1485 |
+
"eval_roc_auc": 0.7199792960662525,
|
| 1486 |
+
"eval_runtime": 2.1956,
|
| 1487 |
+
"eval_samples_per_second": 19.129,
|
| 1488 |
+
"eval_steps_per_second": 1.366,
|
| 1489 |
+
"step": 184
|
| 1490 |
+
},
|
| 1491 |
+
{
|
| 1492 |
+
"epoch": 134.55,
|
| 1493 |
+
"eval_accuracy": 0.07142857142857142,
|
| 1494 |
+
"eval_f1": 0.40869565217391307,
|
| 1495 |
+
"eval_loss": 0.5361512303352356,
|
| 1496 |
+
"eval_roc_auc": 0.72851966873706,
|
| 1497 |
+
"eval_runtime": 2.2016,
|
| 1498 |
+
"eval_samples_per_second": 19.077,
|
| 1499 |
+
"eval_steps_per_second": 1.363,
|
| 1500 |
+
"step": 185
|
| 1501 |
+
},
|
| 1502 |
+
{
|
| 1503 |
+
"epoch": 136.0,
|
| 1504 |
+
"eval_accuracy": 0.023809523809523808,
|
| 1505 |
+
"eval_f1": 0.39999999999999997,
|
| 1506 |
+
"eval_loss": 0.5336142182350159,
|
| 1507 |
+
"eval_roc_auc": 0.7142857142857142,
|
| 1508 |
+
"eval_runtime": 2.176,
|
| 1509 |
+
"eval_samples_per_second": 19.302,
|
| 1510 |
+
"eval_steps_per_second": 1.379,
|
| 1511 |
+
"step": 187
|
| 1512 |
+
},
|
| 1513 |
+
{
|
| 1514 |
+
"epoch": 136.73,
|
| 1515 |
+
"eval_accuracy": 0.047619047619047616,
|
| 1516 |
+
"eval_f1": 0.42533936651583715,
|
| 1517 |
+
"eval_loss": 0.5356618762016296,
|
| 1518 |
+
"eval_roc_auc": 0.7331780538302277,
|
| 1519 |
+
"eval_runtime": 2.3066,
|
| 1520 |
+
"eval_samples_per_second": 18.209,
|
| 1521 |
+
"eval_steps_per_second": 1.301,
|
| 1522 |
+
"step": 188
|
| 1523 |
+
},
|
| 1524 |
+
{
|
| 1525 |
+
"epoch": 137.45,
|
| 1526 |
+
"eval_accuracy": 0.047619047619047616,
|
| 1527 |
+
"eval_f1": 0.4266666666666666,
|
| 1528 |
+
"eval_loss": 0.5409244894981384,
|
| 1529 |
+
"eval_roc_auc": 0.7375776397515528,
|
| 1530 |
+
"eval_runtime": 2.269,
|
| 1531 |
+
"eval_samples_per_second": 18.51,
|
| 1532 |
+
"eval_steps_per_second": 1.322,
|
| 1533 |
+
"step": 189
|
| 1534 |
+
},
|
| 1535 |
+
{
|
| 1536 |
+
"epoch": 138.91,
|
| 1537 |
+
"eval_accuracy": 0.023809523809523808,
|
| 1538 |
+
"eval_f1": 0.41558441558441556,
|
| 1539 |
+
"eval_loss": 0.5381691455841064,
|
| 1540 |
+
"eval_roc_auc": 0.7344720496894409,
|
| 1541 |
+
"eval_runtime": 2.2181,
|
| 1542 |
+
"eval_samples_per_second": 18.935,
|
| 1543 |
+
"eval_steps_per_second": 1.353,
|
| 1544 |
+
"step": 191
|
| 1545 |
+
},
|
| 1546 |
+
{
|
| 1547 |
+
"epoch": 139.64,
|
| 1548 |
+
"eval_accuracy": 0.047619047619047616,
|
| 1549 |
+
"eval_f1": 0.41558441558441556,
|
| 1550 |
+
"eval_loss": 0.5408413410186768,
|
| 1551 |
+
"eval_roc_auc": 0.7344720496894409,
|
| 1552 |
+
"eval_runtime": 2.1767,
|
| 1553 |
+
"eval_samples_per_second": 19.296,
|
| 1554 |
+
"eval_steps_per_second": 1.378,
|
| 1555 |
+
"step": 192
|
| 1556 |
+
},
|
| 1557 |
+
{
|
| 1558 |
+
"epoch": 140.36,
|
| 1559 |
+
"eval_accuracy": 0.047619047619047616,
|
| 1560 |
+
"eval_f1": 0.39473684210526316,
|
| 1561 |
+
"eval_loss": 0.54255610704422,
|
| 1562 |
+
"eval_roc_auc": 0.716614906832298,
|
| 1563 |
+
"eval_runtime": 2.3421,
|
| 1564 |
+
"eval_samples_per_second": 17.933,
|
| 1565 |
+
"eval_steps_per_second": 1.281,
|
| 1566 |
+
"step": 193
|
| 1567 |
+
},
|
| 1568 |
+
{
|
| 1569 |
+
"epoch": 141.82,
|
| 1570 |
+
"eval_accuracy": 0.023809523809523808,
|
| 1571 |
+
"eval_f1": 0.38260869565217387,
|
| 1572 |
+
"eval_loss": 0.5436768531799316,
|
| 1573 |
+
"eval_roc_auc": 0.7091097308488613,
|
| 1574 |
+
"eval_runtime": 2.2168,
|
| 1575 |
+
"eval_samples_per_second": 18.946,
|
| 1576 |
+
"eval_steps_per_second": 1.353,
|
| 1577 |
+
"step": 195
|
| 1578 |
+
},
|
| 1579 |
+
{
|
| 1580 |
+
"epoch": 142.55,
|
| 1581 |
+
"eval_accuracy": 0.07142857142857142,
|
| 1582 |
+
"eval_f1": 0.3876651982378855,
|
| 1583 |
+
"eval_loss": 0.540955662727356,
|
| 1584 |
+
"eval_roc_auc": 0.7106625258799172,
|
| 1585 |
+
"eval_runtime": 2.1746,
|
| 1586 |
+
"eval_samples_per_second": 19.314,
|
| 1587 |
+
"eval_steps_per_second": 1.38,
|
| 1588 |
+
"step": 196
|
| 1589 |
+
},
|
| 1590 |
+
{
|
| 1591 |
+
"epoch": 144.0,
|
| 1592 |
+
"eval_accuracy": 0.07142857142857142,
|
| 1593 |
+
"eval_f1": 0.3982683982683983,
|
| 1594 |
+
"eval_loss": 0.5399672389030457,
|
| 1595 |
+
"eval_roc_auc": 0.7215320910973085,
|
| 1596 |
+
"eval_runtime": 2.2745,
|
| 1597 |
+
"eval_samples_per_second": 18.466,
|
| 1598 |
+
"eval_steps_per_second": 1.319,
|
| 1599 |
+
"step": 198
|
| 1600 |
+
},
|
| 1601 |
+
{
|
| 1602 |
+
"epoch": 144.73,
|
| 1603 |
+
"eval_accuracy": 0.047619047619047616,
|
| 1604 |
+
"eval_f1": 0.3859649122807018,
|
| 1605 |
+
"eval_loss": 0.5390969514846802,
|
| 1606 |
+
"eval_roc_auc": 0.7101449275362318,
|
| 1607 |
+
"eval_runtime": 2.3353,
|
| 1608 |
+
"eval_samples_per_second": 17.984,
|
| 1609 |
+
"eval_steps_per_second": 1.285,
|
| 1610 |
+
"step": 199
|
| 1611 |
+
},
|
| 1612 |
+
{
|
| 1613 |
+
"epoch": 145.45,
|
| 1614 |
+
"eval_accuracy": 0.07142857142857142,
|
| 1615 |
+
"eval_f1": 0.42290748898678415,
|
| 1616 |
+
"eval_loss": 0.5356051325798035,
|
| 1617 |
+
"eval_roc_auc": 0.7365424430641822,
|
| 1618 |
+
"eval_runtime": 2.238,
|
| 1619 |
+
"eval_samples_per_second": 18.767,
|
| 1620 |
+
"eval_steps_per_second": 1.34,
|
| 1621 |
+
"step": 200
|
| 1622 |
+
},
|
| 1623 |
+
{
|
| 1624 |
+
"epoch": 146.91,
|
| 1625 |
+
"eval_accuracy": 0.07142857142857142,
|
| 1626 |
+
"eval_f1": 0.41441441441441446,
|
| 1627 |
+
"eval_loss": 0.5343185663223267,
|
| 1628 |
+
"eval_roc_auc": 0.7261904761904762,
|
| 1629 |
+
"eval_runtime": 2.2389,
|
| 1630 |
+
"eval_samples_per_second": 18.759,
|
| 1631 |
+
"eval_steps_per_second": 1.34,
|
| 1632 |
+
"step": 202
|
| 1633 |
+
},
|
| 1634 |
+
{
|
| 1635 |
+
"epoch": 147.64,
|
| 1636 |
+
"eval_accuracy": 0.09523809523809523,
|
| 1637 |
+
"eval_f1": 0.42857142857142855,
|
| 1638 |
+
"eval_loss": 0.5286682844161987,
|
| 1639 |
+
"eval_roc_auc": 0.7380952380952381,
|
| 1640 |
+
"eval_runtime": 2.2757,
|
| 1641 |
+
"eval_samples_per_second": 18.456,
|
| 1642 |
+
"eval_steps_per_second": 1.318,
|
| 1643 |
+
"step": 203
|
| 1644 |
+
},
|
| 1645 |
+
{
|
| 1646 |
+
"epoch": 148.36,
|
| 1647 |
+
"eval_accuracy": 0.07142857142857142,
|
| 1648 |
+
"eval_f1": 0.41071428571428575,
|
| 1649 |
+
"eval_loss": 0.5306946635246277,
|
| 1650 |
+
"eval_roc_auc": 0.7251552795031057,
|
| 1651 |
+
"eval_runtime": 2.1874,
|
| 1652 |
+
"eval_samples_per_second": 19.201,
|
| 1653 |
+
"eval_steps_per_second": 1.372,
|
| 1654 |
+
"step": 204
|
| 1655 |
+
},
|
| 1656 |
+
{
|
| 1657 |
+
"epoch": 149.82,
|
| 1658 |
+
"eval_accuracy": 0.09523809523809523,
|
| 1659 |
+
"eval_f1": 0.4177777777777778,
|
| 1660 |
+
"eval_loss": 0.5335448384284973,
|
| 1661 |
+
"eval_roc_auc": 0.7311076604554866,
|
| 1662 |
+
"eval_runtime": 2.3633,
|
| 1663 |
+
"eval_samples_per_second": 17.772,
|
| 1664 |
+
"eval_steps_per_second": 1.269,
|
| 1665 |
+
"step": 206
|
| 1666 |
+
},
|
| 1667 |
+
{
|
| 1668 |
+
"epoch": 150.55,
|
| 1669 |
+
"eval_accuracy": 0.07142857142857142,
|
| 1670 |
+
"eval_f1": 0.4000000000000001,
|
| 1671 |
+
"eval_loss": 0.535821795463562,
|
| 1672 |
+
"eval_roc_auc": 0.7220496894409938,
|
| 1673 |
+
"eval_runtime": 2.1476,
|
| 1674 |
+
"eval_samples_per_second": 19.557,
|
| 1675 |
+
"eval_steps_per_second": 1.397,
|
| 1676 |
+
"step": 207
|
| 1677 |
+
},
|
| 1678 |
+
{
|
| 1679 |
+
"epoch": 152.0,
|
| 1680 |
+
"eval_accuracy": 0.07142857142857142,
|
| 1681 |
+
"eval_f1": 0.4052863436123348,
|
| 1682 |
+
"eval_loss": 0.5356825590133667,
|
| 1683 |
+
"eval_roc_auc": 0.7236024844720497,
|
| 1684 |
+
"eval_runtime": 2.2782,
|
| 1685 |
+
"eval_samples_per_second": 18.436,
|
| 1686 |
+
"eval_steps_per_second": 1.317,
|
| 1687 |
+
"step": 209
|
| 1688 |
+
},
|
| 1689 |
+
{
|
| 1690 |
+
"epoch": 152.73,
|
| 1691 |
+
"eval_accuracy": 0.09523809523809523,
|
| 1692 |
+
"eval_f1": 0.4069264069264069,
|
| 1693 |
+
"eval_loss": 0.5324029922485352,
|
| 1694 |
+
"eval_roc_auc": 0.7280020703933747,
|
| 1695 |
+
"eval_runtime": 2.2465,
|
| 1696 |
+
"eval_samples_per_second": 18.696,
|
| 1697 |
+
"eval_steps_per_second": 1.335,
|
| 1698 |
+
"step": 210
|
| 1699 |
+
},
|
| 1700 |
+
{
|
| 1701 |
+
"epoch": 153.45,
|
| 1702 |
+
"eval_accuracy": 0.07142857142857142,
|
| 1703 |
+
"eval_f1": 0.39316239316239315,
|
| 1704 |
+
"eval_loss": 0.5348923802375793,
|
| 1705 |
+
"eval_roc_auc": 0.7199792960662525,
|
| 1706 |
+
"eval_runtime": 2.1592,
|
| 1707 |
+
"eval_samples_per_second": 19.452,
|
| 1708 |
+
"eval_steps_per_second": 1.389,
|
| 1709 |
+
"step": 211
|
| 1710 |
+
},
|
| 1711 |
+
{
|
| 1712 |
+
"epoch": 154.91,
|
| 1713 |
+
"eval_accuracy": 0.07142857142857142,
|
| 1714 |
+
"eval_f1": 0.41379310344827586,
|
| 1715 |
+
"eval_loss": 0.5328140258789062,
|
| 1716 |
+
"eval_roc_auc": 0.7339544513457557,
|
| 1717 |
+
"eval_runtime": 2.1367,
|
| 1718 |
+
"eval_samples_per_second": 19.657,
|
| 1719 |
+
"eval_steps_per_second": 1.404,
|
| 1720 |
+
"step": 213
|
| 1721 |
+
},
|
| 1722 |
+
{
|
| 1723 |
+
"epoch": 155.64,
|
| 1724 |
+
"eval_accuracy": 0.07142857142857142,
|
| 1725 |
+
"eval_f1": 0.412280701754386,
|
| 1726 |
+
"eval_loss": 0.533903956413269,
|
| 1727 |
+
"eval_roc_auc": 0.7295548654244306,
|
| 1728 |
+
"eval_runtime": 2.1392,
|
| 1729 |
+
"eval_samples_per_second": 19.634,
|
| 1730 |
+
"eval_steps_per_second": 1.402,
|
| 1731 |
+
"step": 214
|
| 1732 |
+
},
|
| 1733 |
+
{
|
| 1734 |
+
"epoch": 156.36,
|
| 1735 |
+
"eval_accuracy": 0.09523809523809523,
|
| 1736 |
+
"eval_f1": 0.43049327354260086,
|
| 1737 |
+
"eval_loss": 0.5287569761276245,
|
| 1738 |
+
"eval_roc_auc": 0.7386128364389234,
|
| 1739 |
+
"eval_runtime": 2.3564,
|
| 1740 |
+
"eval_samples_per_second": 17.824,
|
| 1741 |
+
"eval_steps_per_second": 1.273,
|
| 1742 |
+
"step": 215
|
| 1743 |
+
},
|
| 1744 |
+
{
|
| 1745 |
+
"epoch": 157.82,
|
| 1746 |
+
"eval_accuracy": 0.07142857142857142,
|
| 1747 |
+
"eval_f1": 0.4272727272727273,
|
| 1748 |
+
"eval_loss": 0.5261039137840271,
|
| 1749 |
+
"eval_roc_auc": 0.733695652173913,
|
| 1750 |
+
"eval_runtime": 2.1573,
|
| 1751 |
+
"eval_samples_per_second": 19.469,
|
| 1752 |
+
"eval_steps_per_second": 1.391,
|
| 1753 |
+
"step": 217
|
| 1754 |
+
},
|
| 1755 |
+
{
|
| 1756 |
+
"epoch": 158.55,
|
| 1757 |
+
"eval_accuracy": 0.07142857142857142,
|
| 1758 |
+
"eval_f1": 0.42290748898678415,
|
| 1759 |
+
"eval_loss": 0.5312274694442749,
|
| 1760 |
+
"eval_roc_auc": 0.7365424430641822,
|
| 1761 |
+
"eval_runtime": 2.2337,
|
| 1762 |
+
"eval_samples_per_second": 18.803,
|
| 1763 |
+
"eval_steps_per_second": 1.343,
|
| 1764 |
+
"step": 218
|
| 1765 |
+
},
|
| 1766 |
+
{
|
| 1767 |
+
"epoch": 160.0,
|
| 1768 |
+
"eval_accuracy": 0.09523809523809523,
|
| 1769 |
+
"eval_f1": 0.40358744394618834,
|
| 1770 |
+
"eval_loss": 0.5341694951057434,
|
| 1771 |
+
"eval_roc_auc": 0.7192028985507245,
|
| 1772 |
+
"eval_runtime": 2.2456,
|
| 1773 |
+
"eval_samples_per_second": 18.703,
|
| 1774 |
+
"eval_steps_per_second": 1.336,
|
| 1775 |
+
"step": 220
|
| 1776 |
+
},
|
| 1777 |
+
{
|
| 1778 |
+
"epoch": 160.73,
|
| 1779 |
+
"eval_accuracy": 0.09523809523809523,
|
| 1780 |
+
"eval_f1": 0.3981900452488688,
|
| 1781 |
+
"eval_loss": 0.5362971425056458,
|
| 1782 |
+
"eval_roc_auc": 0.7137681159420289,
|
| 1783 |
+
"eval_runtime": 2.212,
|
| 1784 |
+
"eval_samples_per_second": 18.988,
|
| 1785 |
+
"eval_steps_per_second": 1.356,
|
| 1786 |
+
"step": 221
|
| 1787 |
+
},
|
| 1788 |
+
{
|
| 1789 |
+
"epoch": 161.45,
|
| 1790 |
+
"eval_accuracy": 0.09523809523809523,
|
| 1791 |
+
"eval_f1": 0.41409691629955947,
|
| 1792 |
+
"eval_loss": 0.532296895980835,
|
| 1793 |
+
"eval_roc_auc": 0.730072463768116,
|
| 1794 |
+
"eval_runtime": 2.3611,
|
| 1795 |
+
"eval_samples_per_second": 17.788,
|
| 1796 |
+
"eval_steps_per_second": 1.271,
|
| 1797 |
+
"step": 222
|
| 1798 |
+
},
|
| 1799 |
+
{
|
| 1800 |
+
"epoch": 162.91,
|
| 1801 |
+
"eval_accuracy": 0.09523809523809523,
|
| 1802 |
+
"eval_f1": 0.41071428571428575,
|
| 1803 |
+
"eval_loss": 0.5253220796585083,
|
| 1804 |
+
"eval_roc_auc": 0.7251552795031057,
|
| 1805 |
+
"eval_runtime": 2.1376,
|
| 1806 |
+
"eval_samples_per_second": 19.648,
|
| 1807 |
+
"eval_steps_per_second": 1.403,
|
| 1808 |
+
"step": 224
|
| 1809 |
+
},
|
| 1810 |
+
{
|
| 1811 |
+
"epoch": 163.64,
|
| 1812 |
+
"eval_accuracy": 0.09523809523809523,
|
| 1813 |
+
"eval_f1": 0.4234234234234235,
|
| 1814 |
+
"eval_loss": 0.5222101807594299,
|
| 1815 |
+
"eval_roc_auc": 0.7326604554865425,
|
| 1816 |
+
"eval_runtime": 2.2698,
|
| 1817 |
+
"eval_samples_per_second": 18.504,
|
| 1818 |
+
"eval_steps_per_second": 1.322,
|
| 1819 |
+
"step": 225
|
| 1820 |
+
},
|
| 1821 |
+
{
|
| 1822 |
+
"epoch": 164.36,
|
| 1823 |
+
"eval_accuracy": 0.07142857142857142,
|
| 1824 |
+
"eval_f1": 0.412280701754386,
|
| 1825 |
+
"eval_loss": 0.5278355479240417,
|
| 1826 |
+
"eval_roc_auc": 0.7295548654244306,
|
| 1827 |
+
"eval_runtime": 2.2788,
|
| 1828 |
+
"eval_samples_per_second": 18.431,
|
| 1829 |
+
"eval_steps_per_second": 1.316,
|
| 1830 |
+
"step": 226
|
| 1831 |
+
},
|
| 1832 |
+
{
|
| 1833 |
+
"epoch": 165.82,
|
| 1834 |
+
"eval_accuracy": 0.11904761904761904,
|
| 1835 |
+
"eval_f1": 0.415929203539823,
|
| 1836 |
+
"eval_loss": 0.5236978530883789,
|
| 1837 |
+
"eval_roc_auc": 0.7305900621118012,
|
| 1838 |
+
"eval_runtime": 2.3596,
|
| 1839 |
+
"eval_samples_per_second": 17.8,
|
| 1840 |
+
"eval_steps_per_second": 1.271,
|
| 1841 |
+
"step": 228
|
| 1842 |
+
},
|
| 1843 |
+
{
|
| 1844 |
+
"epoch": 166.55,
|
| 1845 |
+
"eval_accuracy": 0.11904761904761904,
|
| 1846 |
+
"eval_f1": 0.42290748898678415,
|
| 1847 |
+
"eval_loss": 0.5267902612686157,
|
| 1848 |
+
"eval_roc_auc": 0.7365424430641822,
|
| 1849 |
+
"eval_runtime": 2.2183,
|
| 1850 |
+
"eval_samples_per_second": 18.933,
|
| 1851 |
+
"eval_steps_per_second": 1.352,
|
| 1852 |
+
"step": 229
|
| 1853 |
+
},
|
| 1854 |
+
{
|
| 1855 |
+
"epoch": 168.0,
|
| 1856 |
+
"eval_accuracy": 0.09523809523809523,
|
| 1857 |
+
"eval_f1": 0.39639639639639634,
|
| 1858 |
+
"eval_loss": 0.5275822877883911,
|
| 1859 |
+
"eval_roc_auc": 0.7132505175983436,
|
| 1860 |
+
"eval_runtime": 2.1555,
|
| 1861 |
+
"eval_samples_per_second": 19.485,
|
| 1862 |
+
"eval_steps_per_second": 1.392,
|
| 1863 |
+
"step": 231
|
| 1864 |
+
},
|
| 1865 |
+
{
|
| 1866 |
+
"epoch": 168.73,
|
| 1867 |
+
"eval_accuracy": 0.11904761904761904,
|
| 1868 |
+
"eval_f1": 0.41409691629955947,
|
| 1869 |
+
"eval_loss": 0.522292971611023,
|
| 1870 |
+
"eval_roc_auc": 0.730072463768116,
|
| 1871 |
+
"eval_runtime": 2.5244,
|
| 1872 |
+
"eval_samples_per_second": 16.638,
|
| 1873 |
+
"eval_steps_per_second": 1.188,
|
| 1874 |
+
"step": 232
|
| 1875 |
+
},
|
| 1876 |
+
{
|
| 1877 |
+
"epoch": 169.45,
|
| 1878 |
+
"eval_accuracy": 0.11904761904761904,
|
| 1879 |
+
"eval_f1": 0.4177777777777778,
|
| 1880 |
+
"eval_loss": 0.5228198170661926,
|
| 1881 |
+
"eval_roc_auc": 0.7311076604554866,
|
| 1882 |
+
"eval_runtime": 2.2486,
|
| 1883 |
+
"eval_samples_per_second": 18.678,
|
| 1884 |
+
"eval_steps_per_second": 1.334,
|
| 1885 |
+
"step": 233
|
| 1886 |
+
},
|
| 1887 |
+
{
|
| 1888 |
+
"epoch": 170.91,
|
| 1889 |
+
"eval_accuracy": 0.11904761904761904,
|
| 1890 |
+
"eval_f1": 0.4234234234234235,
|
| 1891 |
+
"eval_loss": 0.5201071500778198,
|
| 1892 |
+
"eval_roc_auc": 0.7326604554865425,
|
| 1893 |
+
"eval_runtime": 2.2664,
|
| 1894 |
+
"eval_samples_per_second": 18.531,
|
| 1895 |
+
"eval_steps_per_second": 1.324,
|
| 1896 |
+
"step": 235
|
| 1897 |
+
},
|
| 1898 |
+
{
|
| 1899 |
+
"epoch": 171.64,
|
| 1900 |
+
"eval_accuracy": 0.11904761904761904,
|
| 1901 |
+
"eval_f1": 0.42152466367713004,
|
| 1902 |
+
"eval_loss": 0.5248059034347534,
|
| 1903 |
+
"eval_roc_auc": 0.7321428571428572,
|
| 1904 |
+
"eval_runtime": 2.1597,
|
| 1905 |
+
"eval_samples_per_second": 19.447,
|
| 1906 |
+
"eval_steps_per_second": 1.389,
|
| 1907 |
+
"step": 236
|
| 1908 |
+
},
|
| 1909 |
+
{
|
| 1910 |
+
"epoch": 172.36,
|
| 1911 |
+
"eval_accuracy": 0.11904761904761904,
|
| 1912 |
+
"eval_f1": 0.4125560538116592,
|
| 1913 |
+
"eval_loss": 0.5245119333267212,
|
| 1914 |
+
"eval_roc_auc": 0.7256728778467909,
|
| 1915 |
+
"eval_runtime": 2.2008,
|
| 1916 |
+
"eval_samples_per_second": 19.084,
|
| 1917 |
+
"eval_steps_per_second": 1.363,
|
| 1918 |
+
"step": 237
|
| 1919 |
+
},
|
| 1920 |
+
{
|
| 1921 |
+
"epoch": 173.82,
|
| 1922 |
+
"eval_accuracy": 0.11904761904761904,
|
| 1923 |
+
"eval_f1": 0.4177777777777778,
|
| 1924 |
+
"eval_loss": 0.5223248600959778,
|
| 1925 |
+
"eval_roc_auc": 0.7311076604554866,
|
| 1926 |
+
"eval_runtime": 2.2306,
|
| 1927 |
+
"eval_samples_per_second": 18.829,
|
| 1928 |
+
"eval_steps_per_second": 1.345,
|
| 1929 |
+
"step": 239
|
| 1930 |
+
},
|
| 1931 |
+
{
|
| 1932 |
+
"epoch": 174.55,
|
| 1933 |
+
"eval_accuracy": 0.11904761904761904,
|
| 1934 |
+
"eval_f1": 0.409090909090909,
|
| 1935 |
+
"eval_loss": 0.5221878886222839,
|
| 1936 |
+
"eval_roc_auc": 0.7207556935817805,
|
| 1937 |
+
"eval_runtime": 2.2812,
|
| 1938 |
+
"eval_samples_per_second": 18.411,
|
| 1939 |
+
"eval_steps_per_second": 1.315,
|
| 1940 |
+
"step": 240
|
| 1941 |
+
},
|
| 1942 |
+
{
|
| 1943 |
+
"epoch": 176.0,
|
| 1944 |
+
"eval_accuracy": 0.11904761904761904,
|
| 1945 |
+
"eval_f1": 0.41628959276018096,
|
| 1946 |
+
"eval_loss": 0.5217894315719604,
|
| 1947 |
+
"eval_roc_auc": 0.7267080745341615,
|
| 1948 |
+
"eval_runtime": 2.3942,
|
| 1949 |
+
"eval_samples_per_second": 17.542,
|
| 1950 |
+
"eval_steps_per_second": 1.253,
|
| 1951 |
+
"step": 242
|
| 1952 |
+
},
|
| 1953 |
+
{
|
| 1954 |
+
"epoch": 176.73,
|
| 1955 |
+
"eval_accuracy": 0.09523809523809523,
|
| 1956 |
+
"eval_f1": 0.40540540540540543,
|
| 1957 |
+
"eval_loss": 0.5226414203643799,
|
| 1958 |
+
"eval_roc_auc": 0.7197204968944099,
|
| 1959 |
+
"eval_runtime": 2.1776,
|
| 1960 |
+
"eval_samples_per_second": 19.287,
|
| 1961 |
+
"eval_steps_per_second": 1.378,
|
| 1962 |
+
"step": 243
|
| 1963 |
+
},
|
| 1964 |
+
{
|
| 1965 |
+
"epoch": 177.45,
|
| 1966 |
+
"eval_accuracy": 0.09523809523809523,
|
| 1967 |
+
"eval_f1": 0.39823008849557523,
|
| 1968 |
+
"eval_loss": 0.5248770713806152,
|
| 1969 |
+
"eval_roc_auc": 0.7176501035196686,
|
| 1970 |
+
"eval_runtime": 2.26,
|
| 1971 |
+
"eval_samples_per_second": 18.584,
|
| 1972 |
+
"eval_steps_per_second": 1.327,
|
| 1973 |
+
"step": 244
|
| 1974 |
+
},
|
| 1975 |
+
{
|
| 1976 |
+
"epoch": 178.91,
|
| 1977 |
+
"eval_accuracy": 0.09523809523809523,
|
| 1978 |
+
"eval_f1": 0.4017857142857143,
|
| 1979 |
+
"eval_loss": 0.5241516828536987,
|
| 1980 |
+
"eval_roc_auc": 0.7186853002070392,
|
| 1981 |
+
"eval_runtime": 2.1743,
|
| 1982 |
+
"eval_samples_per_second": 19.317,
|
| 1983 |
+
"eval_steps_per_second": 1.38,
|
| 1984 |
+
"step": 246
|
| 1985 |
+
},
|
| 1986 |
+
{
|
| 1987 |
+
"epoch": 179.64,
|
| 1988 |
+
"eval_accuracy": 0.09523809523809523,
|
| 1989 |
+
"eval_f1": 0.38914027149321273,
|
| 1990 |
+
"eval_loss": 0.5245065093040466,
|
| 1991 |
+
"eval_roc_auc": 0.7072981366459626,
|
| 1992 |
+
"eval_runtime": 2.2384,
|
| 1993 |
+
"eval_samples_per_second": 18.763,
|
| 1994 |
+
"eval_steps_per_second": 1.34,
|
| 1995 |
+
"step": 247
|
| 1996 |
+
},
|
| 1997 |
+
{
|
| 1998 |
+
"epoch": 180.36,
|
| 1999 |
+
"eval_accuracy": 0.11904761904761904,
|
| 2000 |
+
"eval_f1": 0.39449541284403666,
|
| 2001 |
+
"eval_loss": 0.5249381065368652,
|
| 2002 |
+
"eval_roc_auc": 0.7088509316770186,
|
| 2003 |
+
"eval_runtime": 2.2201,
|
| 2004 |
+
"eval_samples_per_second": 18.918,
|
| 2005 |
+
"eval_steps_per_second": 1.351,
|
| 2006 |
+
"step": 248
|
| 2007 |
+
},
|
| 2008 |
+
{
|
| 2009 |
+
"epoch": 181.82,
|
| 2010 |
+
"eval_accuracy": 0.09523809523809523,
|
| 2011 |
+
"eval_f1": 0.4234234234234235,
|
| 2012 |
+
"eval_loss": 0.52437824010849,
|
| 2013 |
+
"eval_roc_auc": 0.7326604554865425,
|
| 2014 |
+
"eval_runtime": 2.2705,
|
| 2015 |
+
"eval_samples_per_second": 18.498,
|
| 2016 |
+
"eval_steps_per_second": 1.321,
|
| 2017 |
+
"step": 250
|
| 2018 |
+
},
|
| 2019 |
+
{
|
| 2020 |
+
"epoch": 182.55,
|
| 2021 |
+
"eval_accuracy": 0.07142857142857142,
|
| 2022 |
+
"eval_f1": 0.409090909090909,
|
| 2023 |
+
"eval_loss": 0.5216240286827087,
|
| 2024 |
+
"eval_roc_auc": 0.7207556935817805,
|
| 2025 |
+
"eval_runtime": 2.238,
|
| 2026 |
+
"eval_samples_per_second": 18.767,
|
| 2027 |
+
"eval_steps_per_second": 1.34,
|
| 2028 |
+
"step": 251
|
| 2029 |
+
},
|
| 2030 |
+
{
|
| 2031 |
+
"epoch": 184.0,
|
| 2032 |
+
"eval_accuracy": 0.09523809523809523,
|
| 2033 |
+
"eval_f1": 0.40358744394618834,
|
| 2034 |
+
"eval_loss": 0.5211586952209473,
|
| 2035 |
+
"eval_roc_auc": 0.7192028985507245,
|
| 2036 |
+
"eval_runtime": 2.244,
|
| 2037 |
+
"eval_samples_per_second": 18.716,
|
| 2038 |
+
"eval_steps_per_second": 1.337,
|
| 2039 |
+
"step": 253
|
| 2040 |
+
},
|
| 2041 |
+
{
|
| 2042 |
+
"epoch": 184.73,
|
| 2043 |
+
"eval_accuracy": 0.09523809523809523,
|
| 2044 |
+
"eval_f1": 0.41628959276018096,
|
| 2045 |
+
"eval_loss": 0.5235975384712219,
|
| 2046 |
+
"eval_roc_auc": 0.7267080745341615,
|
| 2047 |
+
"eval_runtime": 2.1767,
|
| 2048 |
+
"eval_samples_per_second": 19.295,
|
| 2049 |
+
"eval_steps_per_second": 1.378,
|
| 2050 |
+
"step": 254
|
| 2051 |
+
},
|
| 2052 |
+
{
|
| 2053 |
+
"epoch": 185.45,
|
| 2054 |
+
"eval_accuracy": 0.11904761904761904,
|
| 2055 |
+
"eval_f1": 0.41441441441441446,
|
| 2056 |
+
"eval_loss": 0.5226503610610962,
|
| 2057 |
+
"eval_roc_auc": 0.7261904761904762,
|
| 2058 |
+
"eval_runtime": 2.3845,
|
| 2059 |
+
"eval_samples_per_second": 17.614,
|
| 2060 |
+
"eval_steps_per_second": 1.258,
|
| 2061 |
+
"step": 255
|
| 2062 |
+
},
|
| 2063 |
+
{
|
| 2064 |
+
"epoch": 186.91,
|
| 2065 |
+
"eval_accuracy": 0.11904761904761904,
|
| 2066 |
+
"eval_f1": 0.40540540540540543,
|
| 2067 |
+
"eval_loss": 0.5200766324996948,
|
| 2068 |
+
"eval_roc_auc": 0.7197204968944099,
|
| 2069 |
+
"eval_runtime": 2.2501,
|
| 2070 |
+
"eval_samples_per_second": 18.666,
|
| 2071 |
+
"eval_steps_per_second": 1.333,
|
| 2072 |
+
"step": 257
|
| 2073 |
+
},
|
| 2074 |
+
{
|
| 2075 |
+
"epoch": 187.64,
|
| 2076 |
+
"eval_accuracy": 0.14285714285714285,
|
| 2077 |
+
"eval_f1": 0.4181818181818182,
|
| 2078 |
+
"eval_loss": 0.5195832848548889,
|
| 2079 |
+
"eval_roc_auc": 0.7272256728778468,
|
| 2080 |
+
"eval_runtime": 2.2154,
|
| 2081 |
+
"eval_samples_per_second": 18.958,
|
| 2082 |
+
"eval_steps_per_second": 1.354,
|
| 2083 |
+
"step": 258
|
| 2084 |
+
},
|
| 2085 |
+
{
|
| 2086 |
+
"epoch": 188.36,
|
| 2087 |
+
"eval_accuracy": 0.14285714285714285,
|
| 2088 |
+
"eval_f1": 0.4177777777777778,
|
| 2089 |
+
"eval_loss": 0.5187467336654663,
|
| 2090 |
+
"eval_roc_auc": 0.7311076604554866,
|
| 2091 |
+
"eval_runtime": 2.2882,
|
| 2092 |
+
"eval_samples_per_second": 18.355,
|
| 2093 |
+
"eval_steps_per_second": 1.311,
|
| 2094 |
+
"step": 259
|
| 2095 |
+
},
|
| 2096 |
+
{
|
| 2097 |
+
"epoch": 189.82,
|
| 2098 |
+
"eval_accuracy": 0.11904761904761904,
|
| 2099 |
+
"eval_f1": 0.4125560538116592,
|
| 2100 |
+
"eval_loss": 0.5165961384773254,
|
| 2101 |
+
"eval_roc_auc": 0.7256728778467909,
|
| 2102 |
+
"eval_runtime": 2.2388,
|
| 2103 |
+
"eval_samples_per_second": 18.76,
|
| 2104 |
+
"eval_steps_per_second": 1.34,
|
| 2105 |
+
"step": 261
|
| 2106 |
+
},
|
| 2107 |
+
{
|
| 2108 |
+
"epoch": 190.55,
|
| 2109 |
+
"eval_accuracy": 0.11904761904761904,
|
| 2110 |
+
"eval_f1": 0.3873873873873873,
|
| 2111 |
+
"eval_loss": 0.5241953730583191,
|
| 2112 |
+
"eval_roc_auc": 0.7067805383022774,
|
| 2113 |
+
"eval_runtime": 2.4908,
|
| 2114 |
+
"eval_samples_per_second": 16.862,
|
| 2115 |
+
"eval_steps_per_second": 1.204,
|
| 2116 |
+
"step": 262
|
| 2117 |
+
},
|
| 2118 |
+
{
|
| 2119 |
+
"epoch": 192.0,
|
| 2120 |
+
"eval_accuracy": 0.11904761904761904,
|
| 2121 |
+
"eval_f1": 0.42533936651583715,
|
| 2122 |
+
"eval_loss": 0.5164801478385925,
|
| 2123 |
+
"eval_roc_auc": 0.7331780538302277,
|
| 2124 |
+
"eval_runtime": 2.2626,
|
| 2125 |
+
"eval_samples_per_second": 18.563,
|
| 2126 |
+
"eval_steps_per_second": 1.326,
|
| 2127 |
+
"step": 264
|
| 2128 |
+
},
|
| 2129 |
+
{
|
| 2130 |
+
"epoch": 192.73,
|
| 2131 |
+
"eval_accuracy": 0.11904761904761904,
|
| 2132 |
+
"eval_f1": 0.43518518518518523,
|
| 2133 |
+
"eval_loss": 0.5178862810134888,
|
| 2134 |
+
"eval_roc_auc": 0.7357660455486543,
|
| 2135 |
+
"eval_runtime": 2.5128,
|
| 2136 |
+
"eval_samples_per_second": 16.714,
|
| 2137 |
+
"eval_steps_per_second": 1.194,
|
| 2138 |
+
"step": 265
|
| 2139 |
+
},
|
| 2140 |
+
{
|
| 2141 |
+
"epoch": 193.45,
|
| 2142 |
+
"eval_accuracy": 0.11904761904761904,
|
| 2143 |
+
"eval_f1": 0.42452830188679247,
|
| 2144 |
+
"eval_loss": 0.5133179426193237,
|
| 2145 |
+
"eval_roc_auc": 0.7248964803312629,
|
| 2146 |
+
"eval_runtime": 2.5865,
|
| 2147 |
+
"eval_samples_per_second": 16.238,
|
| 2148 |
+
"eval_steps_per_second": 1.16,
|
| 2149 |
+
"step": 266
|
| 2150 |
+
},
|
| 2151 |
+
{
|
| 2152 |
+
"epoch": 194.91,
|
| 2153 |
+
"eval_accuracy": 0.11904761904761904,
|
| 2154 |
+
"eval_f1": 0.42790697674418604,
|
| 2155 |
+
"eval_loss": 0.5165557861328125,
|
| 2156 |
+
"eval_roc_auc": 0.7298136645962733,
|
| 2157 |
+
"eval_runtime": 2.2181,
|
| 2158 |
+
"eval_samples_per_second": 18.935,
|
| 2159 |
+
"eval_steps_per_second": 1.353,
|
| 2160 |
+
"step": 268
|
| 2161 |
+
},
|
| 2162 |
+
{
|
| 2163 |
+
"epoch": 195.64,
|
| 2164 |
+
"eval_accuracy": 0.09523809523809523,
|
| 2165 |
+
"eval_f1": 0.4125560538116592,
|
| 2166 |
+
"eval_loss": 0.5179272294044495,
|
| 2167 |
+
"eval_roc_auc": 0.7256728778467909,
|
| 2168 |
+
"eval_runtime": 2.1768,
|
| 2169 |
+
"eval_samples_per_second": 19.294,
|
| 2170 |
+
"eval_steps_per_second": 1.378,
|
| 2171 |
+
"step": 269
|
| 2172 |
+
},
|
| 2173 |
+
{
|
| 2174 |
+
"epoch": 196.36,
|
| 2175 |
+
"eval_accuracy": 0.11904761904761904,
|
| 2176 |
+
"eval_f1": 0.4311926605504587,
|
| 2177 |
+
"eval_loss": 0.5189908146858215,
|
| 2178 |
+
"eval_roc_auc": 0.7347308488612837,
|
| 2179 |
+
"eval_runtime": 2.1519,
|
| 2180 |
+
"eval_samples_per_second": 19.517,
|
| 2181 |
+
"eval_steps_per_second": 1.394,
|
| 2182 |
+
"step": 270
|
| 2183 |
+
},
|
| 2184 |
+
{
|
| 2185 |
+
"epoch": 197.82,
|
| 2186 |
+
"eval_accuracy": 0.11904761904761904,
|
| 2187 |
+
"eval_f1": 0.40552995391705066,
|
| 2188 |
+
"eval_loss": 0.5177509784698486,
|
| 2189 |
+
"eval_roc_auc": 0.7158385093167701,
|
| 2190 |
+
"eval_runtime": 2.24,
|
| 2191 |
+
"eval_samples_per_second": 18.75,
|
| 2192 |
+
"eval_steps_per_second": 1.339,
|
| 2193 |
+
"step": 272
|
| 2194 |
+
},
|
| 2195 |
+
{
|
| 2196 |
+
"epoch": 198.55,
|
| 2197 |
+
"eval_accuracy": 0.11904761904761904,
|
| 2198 |
+
"eval_f1": 0.40740740740740733,
|
| 2199 |
+
"eval_loss": 0.5172959566116333,
|
| 2200 |
+
"eval_roc_auc": 0.7163561076604554,
|
| 2201 |
+
"eval_runtime": 2.2769,
|
| 2202 |
+
"eval_samples_per_second": 18.447,
|
| 2203 |
+
"eval_steps_per_second": 1.318,
|
| 2204 |
+
"step": 273
|
| 2205 |
+
},
|
| 2206 |
+
{
|
| 2207 |
+
"epoch": 200.0,
|
| 2208 |
+
"eval_accuracy": 0.11904761904761904,
|
| 2209 |
+
"eval_f1": 0.39639639639639634,
|
| 2210 |
+
"eval_loss": 0.5173150897026062,
|
| 2211 |
+
"eval_roc_auc": 0.7132505175983436,
|
| 2212 |
+
"eval_runtime": 2.4473,
|
| 2213 |
+
"eval_samples_per_second": 17.162,
|
| 2214 |
+
"eval_steps_per_second": 1.226,
|
| 2215 |
+
"step": 275
|
| 2216 |
+
},
|
| 2217 |
+
{
|
| 2218 |
+
"epoch": 200.73,
|
| 2219 |
+
"eval_accuracy": 0.11904761904761904,
|
| 2220 |
+
"eval_f1": 0.41666666666666663,
|
| 2221 |
+
"eval_loss": 0.5198870897293091,
|
| 2222 |
+
"eval_roc_auc": 0.7228260869565216,
|
| 2223 |
+
"eval_runtime": 2.2209,
|
| 2224 |
+
"eval_samples_per_second": 18.911,
|
| 2225 |
+
"eval_steps_per_second": 1.351,
|
| 2226 |
+
"step": 276
|
| 2227 |
+
},
|
| 2228 |
+
{
|
| 2229 |
+
"epoch": 201.45,
|
| 2230 |
+
"eval_accuracy": 0.11904761904761904,
|
| 2231 |
+
"eval_f1": 0.3981900452488688,
|
| 2232 |
+
"eval_loss": 0.5189902186393738,
|
| 2233 |
+
"eval_roc_auc": 0.7137681159420289,
|
| 2234 |
+
"eval_runtime": 2.2872,
|
| 2235 |
+
"eval_samples_per_second": 18.363,
|
| 2236 |
+
"eval_steps_per_second": 1.312,
|
| 2237 |
+
"step": 277
|
| 2238 |
+
},
|
| 2239 |
+
{
|
| 2240 |
+
"epoch": 202.91,
|
| 2241 |
+
"eval_accuracy": 0.14285714285714285,
|
| 2242 |
+
"eval_f1": 0.42452830188679247,
|
| 2243 |
+
"eval_loss": 0.5163199305534363,
|
| 2244 |
+
"eval_roc_auc": 0.7248964803312629,
|
| 2245 |
+
"eval_runtime": 2.2888,
|
| 2246 |
+
"eval_samples_per_second": 18.35,
|
| 2247 |
+
"eval_steps_per_second": 1.311,
|
| 2248 |
+
"step": 279
|
| 2249 |
+
},
|
| 2250 |
+
{
|
| 2251 |
+
"epoch": 203.64,
|
| 2252 |
+
"eval_accuracy": 0.11904761904761904,
|
| 2253 |
+
"eval_f1": 0.4170616113744076,
|
| 2254 |
+
"eval_loss": 0.5172598958015442,
|
| 2255 |
+
"eval_roc_auc": 0.7189440993788819,
|
| 2256 |
+
"eval_runtime": 2.2409,
|
| 2257 |
+
"eval_samples_per_second": 18.743,
|
| 2258 |
+
"eval_steps_per_second": 1.339,
|
| 2259 |
+
"step": 280
|
| 2260 |
+
},
|
| 2261 |
+
{
|
| 2262 |
+
"epoch": 204.36,
|
| 2263 |
+
"eval_accuracy": 0.14285714285714285,
|
| 2264 |
+
"eval_f1": 0.41314553990610325,
|
| 2265 |
+
"eval_loss": 0.5206424593925476,
|
| 2266 |
+
"eval_roc_auc": 0.7179089026915114,
|
| 2267 |
+
"eval_runtime": 2.2107,
|
| 2268 |
+
"eval_samples_per_second": 18.999,
|
| 2269 |
+
"eval_steps_per_second": 1.357,
|
| 2270 |
+
"step": 281
|
| 2271 |
+
},
|
| 2272 |
+
{
|
| 2273 |
+
"epoch": 205.82,
|
| 2274 |
+
"eval_accuracy": 0.11904761904761904,
|
| 2275 |
+
"eval_f1": 0.411214953271028,
|
| 2276 |
+
"eval_loss": 0.5185695886611938,
|
| 2277 |
+
"eval_roc_auc": 0.717391304347826,
|
| 2278 |
+
"eval_runtime": 2.2741,
|
| 2279 |
+
"eval_samples_per_second": 18.468,
|
| 2280 |
+
"eval_steps_per_second": 1.319,
|
| 2281 |
+
"step": 283
|
| 2282 |
+
},
|
| 2283 |
+
{
|
| 2284 |
+
"epoch": 206.55,
|
| 2285 |
+
"eval_accuracy": 0.11904761904761904,
|
| 2286 |
+
"eval_f1": 0.43317972350230416,
|
| 2287 |
+
"eval_loss": 0.5186063647270203,
|
| 2288 |
+
"eval_roc_auc": 0.735248447204969,
|
| 2289 |
+
"eval_runtime": 2.2252,
|
| 2290 |
+
"eval_samples_per_second": 18.875,
|
| 2291 |
+
"eval_steps_per_second": 1.348,
|
| 2292 |
+
"step": 284
|
| 2293 |
+
},
|
| 2294 |
+
{
|
| 2295 |
+
"epoch": 208.0,
|
| 2296 |
+
"eval_accuracy": 0.09523809523809523,
|
| 2297 |
+
"eval_f1": 0.39639639639639634,
|
| 2298 |
+
"eval_loss": 0.5183053016662598,
|
| 2299 |
+
"eval_roc_auc": 0.7132505175983436,
|
| 2300 |
+
"eval_runtime": 2.2662,
|
| 2301 |
+
"eval_samples_per_second": 18.533,
|
| 2302 |
+
"eval_steps_per_second": 1.324,
|
| 2303 |
+
"step": 286
|
| 2304 |
+
},
|
| 2305 |
+
{
|
| 2306 |
+
"epoch": 208.73,
|
| 2307 |
+
"eval_accuracy": 0.09523809523809523,
|
| 2308 |
+
"eval_f1": 0.3946188340807175,
|
| 2309 |
+
"eval_loss": 0.5180411338806152,
|
| 2310 |
+
"eval_roc_auc": 0.7127329192546583,
|
| 2311 |
+
"eval_runtime": 2.1744,
|
| 2312 |
+
"eval_samples_per_second": 19.316,
|
| 2313 |
+
"eval_steps_per_second": 1.38,
|
| 2314 |
+
"step": 287
|
| 2315 |
+
},
|
| 2316 |
+
{
|
| 2317 |
+
"epoch": 209.45,
|
| 2318 |
+
"eval_accuracy": 0.09523809523809523,
|
| 2319 |
+
"eval_f1": 0.40930232558139534,
|
| 2320 |
+
"eval_loss": 0.5190138220787048,
|
| 2321 |
+
"eval_roc_auc": 0.7168737060041408,
|
| 2322 |
+
"eval_runtime": 2.2594,
|
| 2323 |
+
"eval_samples_per_second": 18.589,
|
| 2324 |
+
"eval_steps_per_second": 1.328,
|
| 2325 |
+
"step": 288
|
| 2326 |
+
},
|
| 2327 |
+
{
|
| 2328 |
+
"epoch": 210.91,
|
| 2329 |
+
"eval_accuracy": 0.09523809523809523,
|
| 2330 |
+
"eval_f1": 0.3761467889908257,
|
| 2331 |
+
"eval_loss": 0.5196405053138733,
|
| 2332 |
+
"eval_roc_auc": 0.6959109730848861,
|
| 2333 |
+
"eval_runtime": 2.2692,
|
| 2334 |
+
"eval_samples_per_second": 18.509,
|
| 2335 |
+
"eval_steps_per_second": 1.322,
|
| 2336 |
+
"step": 290
|
| 2337 |
+
},
|
| 2338 |
+
{
|
| 2339 |
+
"epoch": 211.64,
|
| 2340 |
+
"eval_accuracy": 0.09523809523809523,
|
| 2341 |
+
"eval_f1": 0.3981900452488688,
|
| 2342 |
+
"eval_loss": 0.5149042010307312,
|
| 2343 |
+
"eval_roc_auc": 0.7137681159420289,
|
| 2344 |
+
"eval_runtime": 2.2833,
|
| 2345 |
+
"eval_samples_per_second": 18.395,
|
| 2346 |
+
"eval_steps_per_second": 1.314,
|
| 2347 |
+
"step": 291
|
| 2348 |
+
},
|
| 2349 |
+
{
|
| 2350 |
+
"epoch": 212.36,
|
| 2351 |
+
"eval_accuracy": 0.09523809523809523,
|
| 2352 |
+
"eval_f1": 0.39285714285714285,
|
| 2353 |
+
"eval_loss": 0.5198233723640442,
|
| 2354 |
+
"eval_roc_auc": 0.7122153209109731,
|
| 2355 |
+
"eval_runtime": 2.5159,
|
| 2356 |
+
"eval_samples_per_second": 16.694,
|
| 2357 |
+
"eval_steps_per_second": 1.192,
|
| 2358 |
+
"step": 292
|
| 2359 |
+
},
|
| 2360 |
+
{
|
| 2361 |
+
"epoch": 213.82,
|
| 2362 |
+
"eval_accuracy": 0.09523809523809523,
|
| 2363 |
+
"eval_f1": 0.3888888888888889,
|
| 2364 |
+
"eval_loss": 0.5179316401481628,
|
| 2365 |
+
"eval_roc_auc": 0.703416149068323,
|
| 2366 |
+
"eval_runtime": 2.3615,
|
| 2367 |
+
"eval_samples_per_second": 17.786,
|
| 2368 |
+
"eval_steps_per_second": 1.27,
|
| 2369 |
+
"step": 294
|
| 2370 |
+
},
|
| 2371 |
+
{
|
| 2372 |
+
"epoch": 214.55,
|
| 2373 |
+
"eval_accuracy": 0.09523809523809523,
|
| 2374 |
+
"eval_f1": 0.3981900452488688,
|
| 2375 |
+
"eval_loss": 0.5147101879119873,
|
| 2376 |
+
"eval_roc_auc": 0.7137681159420289,
|
| 2377 |
+
"eval_runtime": 2.2382,
|
| 2378 |
+
"eval_samples_per_second": 18.765,
|
| 2379 |
+
"eval_steps_per_second": 1.34,
|
| 2380 |
+
"step": 295
|
| 2381 |
+
},
|
| 2382 |
+
{
|
| 2383 |
+
"epoch": 216.0,
|
| 2384 |
+
"eval_accuracy": 0.09523809523809523,
|
| 2385 |
+
"eval_f1": 0.3870967741935484,
|
| 2386 |
+
"eval_loss": 0.5126305818557739,
|
| 2387 |
+
"eval_roc_auc": 0.7028985507246377,
|
| 2388 |
+
"eval_runtime": 2.2182,
|
| 2389 |
+
"eval_samples_per_second": 18.934,
|
| 2390 |
+
"eval_steps_per_second": 1.352,
|
| 2391 |
+
"step": 297
|
| 2392 |
+
},
|
| 2393 |
+
{
|
| 2394 |
+
"epoch": 216.73,
|
| 2395 |
+
"eval_accuracy": 0.09523809523809523,
|
| 2396 |
+
"eval_f1": 0.40740740740740733,
|
| 2397 |
+
"eval_loss": 0.5115792751312256,
|
| 2398 |
+
"eval_roc_auc": 0.7163561076604554,
|
| 2399 |
+
"eval_runtime": 2.1407,
|
| 2400 |
+
"eval_samples_per_second": 19.619,
|
| 2401 |
+
"eval_steps_per_second": 1.401,
|
| 2402 |
+
"step": 298
|
| 2403 |
+
},
|
| 2404 |
+
{
|
| 2405 |
+
"epoch": 217.45,
|
| 2406 |
+
"eval_accuracy": 0.11904761904761904,
|
| 2407 |
+
"eval_f1": 0.4170616113744076,
|
| 2408 |
+
"eval_loss": 0.5107331275939941,
|
| 2409 |
+
"eval_roc_auc": 0.7189440993788819,
|
| 2410 |
+
"eval_runtime": 2.5127,
|
| 2411 |
+
"eval_samples_per_second": 16.715,
|
| 2412 |
+
"eval_steps_per_second": 1.194,
|
| 2413 |
+
"step": 299
|
| 2414 |
+
},
|
| 2415 |
+
{
|
| 2416 |
+
"epoch": 218.91,
|
| 2417 |
+
"eval_accuracy": 0.14285714285714285,
|
| 2418 |
+
"eval_f1": 0.41095890410958896,
|
| 2419 |
+
"eval_loss": 0.511676549911499,
|
| 2420 |
+
"eval_roc_auc": 0.7212732919254659,
|
| 2421 |
+
"eval_runtime": 2.3542,
|
| 2422 |
+
"eval_samples_per_second": 17.84,
|
| 2423 |
+
"eval_steps_per_second": 1.274,
|
| 2424 |
+
"step": 301
|
| 2425 |
+
},
|
| 2426 |
+
{
|
| 2427 |
+
"epoch": 219.64,
|
| 2428 |
+
"eval_accuracy": 0.14285714285714285,
|
| 2429 |
+
"eval_f1": 0.41666666666666663,
|
| 2430 |
+
"eval_loss": 0.5091940760612488,
|
| 2431 |
+
"eval_roc_auc": 0.7228260869565216,
|
| 2432 |
+
"eval_runtime": 2.1376,
|
| 2433 |
+
"eval_samples_per_second": 19.648,
|
| 2434 |
+
"eval_steps_per_second": 1.403,
|
| 2435 |
+
"step": 302
|
| 2436 |
+
},
|
| 2437 |
+
{
|
| 2438 |
+
"epoch": 220.36,
|
| 2439 |
+
"eval_accuracy": 0.09523809523809523,
|
| 2440 |
+
"eval_f1": 0.3873873873873873,
|
| 2441 |
+
"eval_loss": 0.5167934894561768,
|
| 2442 |
+
"eval_roc_auc": 0.7067805383022774,
|
| 2443 |
+
"eval_runtime": 2.2065,
|
| 2444 |
+
"eval_samples_per_second": 19.035,
|
| 2445 |
+
"eval_steps_per_second": 1.36,
|
| 2446 |
+
"step": 303
|
| 2447 |
+
},
|
| 2448 |
+
{
|
| 2449 |
+
"epoch": 221.82,
|
| 2450 |
+
"eval_accuracy": 0.09523809523809523,
|
| 2451 |
+
"eval_f1": 0.39069767441860465,
|
| 2452 |
+
"eval_loss": 0.5202212333679199,
|
| 2453 |
+
"eval_roc_auc": 0.7039337474120083,
|
| 2454 |
+
"eval_runtime": 2.483,
|
| 2455 |
+
"eval_samples_per_second": 16.915,
|
| 2456 |
+
"eval_steps_per_second": 1.208,
|
| 2457 |
+
"step": 305
|
| 2458 |
+
},
|
| 2459 |
+
{
|
| 2460 |
+
"epoch": 222.55,
|
| 2461 |
+
"eval_accuracy": 0.09523809523809523,
|
| 2462 |
+
"eval_f1": 0.40740740740740733,
|
| 2463 |
+
"eval_loss": 0.5138672590255737,
|
| 2464 |
+
"eval_roc_auc": 0.7163561076604554,
|
| 2465 |
+
"eval_runtime": 2.2543,
|
| 2466 |
+
"eval_samples_per_second": 18.631,
|
| 2467 |
+
"eval_steps_per_second": 1.331,
|
| 2468 |
+
"step": 306
|
| 2469 |
+
},
|
| 2470 |
+
{
|
| 2471 |
+
"epoch": 224.0,
|
| 2472 |
+
"eval_accuracy": 0.11904761904761904,
|
| 2473 |
+
"eval_f1": 0.40552995391705066,
|
| 2474 |
+
"eval_loss": 0.5207112431526184,
|
| 2475 |
+
"eval_roc_auc": 0.7158385093167701,
|
| 2476 |
+
"eval_runtime": 2.3791,
|
| 2477 |
+
"eval_samples_per_second": 17.654,
|
| 2478 |
+
"eval_steps_per_second": 1.261,
|
| 2479 |
+
"step": 308
|
| 2480 |
+
},
|
| 2481 |
+
{
|
| 2482 |
+
"epoch": 224.73,
|
| 2483 |
+
"eval_accuracy": 0.11904761904761904,
|
| 2484 |
+
"eval_f1": 0.4292237442922375,
|
| 2485 |
+
"eval_loss": 0.5156222581863403,
|
| 2486 |
+
"eval_roc_auc": 0.7342132505175983,
|
| 2487 |
+
"eval_runtime": 2.6317,
|
| 2488 |
+
"eval_samples_per_second": 15.96,
|
| 2489 |
+
"eval_steps_per_second": 1.14,
|
| 2490 |
+
"step": 309
|
| 2491 |
+
},
|
| 2492 |
+
{
|
| 2493 |
+
"epoch": 225.45,
|
| 2494 |
+
"eval_accuracy": 0.11904761904761904,
|
| 2495 |
+
"eval_f1": 0.4181818181818182,
|
| 2496 |
+
"eval_loss": 0.5122308731079102,
|
| 2497 |
+
"eval_roc_auc": 0.7272256728778468,
|
| 2498 |
+
"eval_runtime": 2.6565,
|
| 2499 |
+
"eval_samples_per_second": 15.81,
|
| 2500 |
+
"eval_steps_per_second": 1.129,
|
| 2501 |
+
"step": 310
|
| 2502 |
+
},
|
| 2503 |
+
{
|
| 2504 |
+
"epoch": 226.91,
|
| 2505 |
+
"eval_accuracy": 0.11904761904761904,
|
| 2506 |
+
"eval_f1": 0.40540540540540543,
|
| 2507 |
+
"eval_loss": 0.511935830116272,
|
| 2508 |
+
"eval_roc_auc": 0.7197204968944099,
|
| 2509 |
+
"eval_runtime": 2.4688,
|
| 2510 |
+
"eval_samples_per_second": 17.012,
|
| 2511 |
+
"eval_steps_per_second": 1.215,
|
| 2512 |
+
"step": 312
|
| 2513 |
+
},
|
| 2514 |
+
{
|
| 2515 |
+
"epoch": 227.64,
|
| 2516 |
+
"eval_accuracy": 0.11904761904761904,
|
| 2517 |
+
"eval_f1": 0.41441441441441446,
|
| 2518 |
+
"eval_loss": 0.510082483291626,
|
| 2519 |
+
"eval_roc_auc": 0.7261904761904762,
|
| 2520 |
+
"eval_runtime": 2.2331,
|
| 2521 |
+
"eval_samples_per_second": 18.808,
|
| 2522 |
+
"eval_steps_per_second": 1.343,
|
| 2523 |
+
"step": 313
|
| 2524 |
+
},
|
| 2525 |
+
{
|
| 2526 |
+
"epoch": 228.36,
|
| 2527 |
+
"eval_accuracy": 0.14285714285714285,
|
| 2528 |
+
"eval_f1": 0.4186046511627907,
|
| 2529 |
+
"eval_loss": 0.5066081285476685,
|
| 2530 |
+
"eval_roc_auc": 0.7233436853002071,
|
| 2531 |
+
"eval_runtime": 2.3108,
|
| 2532 |
+
"eval_samples_per_second": 18.176,
|
| 2533 |
+
"eval_steps_per_second": 1.298,
|
| 2534 |
+
"step": 314
|
| 2535 |
+
}
|
| 2536 |
+
],
|
| 2537 |
+
"logging_steps": 500,
|
| 2538 |
+
"max_steps": 400,
|
| 2539 |
+
"num_train_epochs": 400,
|
| 2540 |
+
"save_steps": 500,
|
| 2541 |
+
"total_flos": 2.136295306761339e+17,
|
| 2542 |
+
"trial_name": null,
|
| 2543 |
+
"trial_params": null
|
| 2544 |
+
}
|
training_args.bin
CHANGED
|
@@ -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 sha256:6a0439b7aa74ea8bd7b8746ca7cc1e3d17e1b72726426ebb115b329e3fb6bced
|
| 3 |
+
size 4600
|