Hugging Face's logo

paulh27
/
iwslt_aligned_smallT5_cont0

Text2Text Generation
Transformers
TensorBoard
Safetensors
German
English
mt5
Generated from Trainer
Eval Results
Inference Endpoints
Model card Files Metrics Community
paulh27 commited on
Commit
8933820
•
1 Parent(s): b2b573d

End of training

Browse files
Files changed (6) hide show
  1. README.md +20 -5
  2. all_results.json +15 -0
  3. eval_results.json +10 -0
  4. runs/May05_21-24-23_nlpg02.cs.washington.edu/events.out.tfevents.1715110790.nlpg02.cs.washington.edu.1176883.1 +3 -0
  5. train_results.json +8 -0
  6. trainer_state.json +4030 -0
README.md CHANGED
@@ -1,13 +1,28 @@
1
  ---
 
 
 
2
  license: apache-2.0
3
  base_model: google/mt5-small
4
  tags:
5
  - generated_from_trainer
 
 
6
  metrics:
7
  - bleu
8
  model-index:
9
  - name: iwslt_aligned_smallT5_cont0
10
- results: []
 
 
 
 
 
 
 
 
 
 
11
  ---
12
 
13
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
@@ -15,11 +30,11 @@ should probably proofread and complete it, then remove this comment. -->
15
 
16
  # iwslt_aligned_smallT5_cont0
17
 
18
- This model is a fine-tuned version of [google/mt5-small](https://huggingface.co/google/mt5-small) on an unknown dataset.
19
  It achieves the following results on the evaluation set:
20
- - Loss: 0.5573
21
- - Bleu: 65.0266
22
- - Gen Len: 28.7455
23
 
24
  ## Model description
25
 
 
1
  ---
2
+ language:
3
+ - de
4
+ - en
5
  license: apache-2.0
6
  base_model: google/mt5-small
7
  tags:
8
  - generated_from_trainer
9
+ datasets:
10
+ - paulh27/alignment_iwslt2017_de_en
11
  metrics:
12
  - bleu
13
  model-index:
14
  - name: iwslt_aligned_smallT5_cont0
15
+ results:
16
+ - task:
17
+ name: Translation
18
+ type: translation
19
+ dataset:
20
+ name: paulh27/alignment_iwslt2017_de_en
21
+ type: paulh27/alignment_iwslt2017_de_en
22
+ metrics:
23
+ - name: Bleu
24
+ type: bleu
25
+ value: 65.6358
26
  ---
27
 
28
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
 
30
 
31
  # iwslt_aligned_smallT5_cont0
32
 
33
+ This model is a fine-tuned version of [google/mt5-small](https://huggingface.co/google/mt5-small) on the paulh27/alignment_iwslt2017_de_en dataset.
34
  It achieves the following results on the evaluation set:
35
+ - Loss: 0.5612
36
+ - Bleu: 65.6358
37
+ - Gen Len: 28.7691
38
 
39
  ## Model description
40
 
all_results.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 38.81,
3
+ "eval_bleu": 65.6358,
4
+ "eval_gen_len": 28.7691,
5
+ "eval_loss": 0.5611980557441711,
6
+ "eval_runtime": 99.4148,
7
+ "eval_samples": 888,
8
+ "eval_samples_per_second": 8.932,
9
+ "eval_steps_per_second": 1.117,
10
+ "train_loss": 0.40471466763305664,
11
+ "train_runtime": 141171.7116,
12
+ "train_samples": 206112,
13
+ "train_samples_per_second": 56.669,
14
+ "train_steps_per_second": 3.542
15
+ }
eval_results.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 38.81,
3
+ "eval_bleu": 65.6358,
4
+ "eval_gen_len": 28.7691,
5
+ "eval_loss": 0.5611980557441711,
6
+ "eval_runtime": 99.4148,
7
+ "eval_samples": 888,
8
+ "eval_samples_per_second": 8.932,
9
+ "eval_steps_per_second": 1.117
10
+ }
runs/May05_21-24-23_nlpg02.cs.washington.edu/events.out.tfevents.1715110790.nlpg02.cs.washington.edu.1176883.1 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1c3472ca42a0847a54326cbcce566ea93092a7d02eaac4ceaa565b7b1694f6f8
3
+ size 465
train_results.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 38.81,
3
+ "train_loss": 0.40471466763305664,
4
+ "train_runtime": 141171.7116,
5
+ "train_samples": 206112,
6
+ "train_samples_per_second": 56.669,
7
+ "train_steps_per_second": 3.542
8
+ }
trainer_state.json ADDED
@@ -0,0 +1,4030 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 65.6358,
3
+ "best_model_checkpoint": "/tmp/finetuned_models/iwslt_aligned_smallT5_cont0/checkpoint-490000",
4
+ "epoch": 38.81384878124515,
5
+ "eval_steps": 10000,
6
+ "global_step": 500000,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.08,
13
+ "grad_norm": 3.0081684589385986,
14
+ "learning_rate": 0.0002,
15
+ "loss": 3.3371,
16
+ "step": 1000
17
+ },
18
+ {
19
+ "epoch": 0.16,
20
+ "grad_norm": 2.347439765930176,
21
+ "learning_rate": 0.0002,
22
+ "loss": 1.9825,
23
+ "step": 2000
24
+ },
25
+ {
26
+ "epoch": 0.23,
27
+ "grad_norm": 3.2636098861694336,
28
+ "learning_rate": 0.0002,
29
+ "loss": 1.7503,
30
+ "step": 3000
31
+ },
32
+ {
33
+ "epoch": 0.31,
34
+ "grad_norm": 2.1704177856445312,
35
+ "learning_rate": 0.0002,
36
+ "loss": 1.6063,
37
+ "step": 4000
38
+ },
39
+ {
40
+ "epoch": 0.39,
41
+ "grad_norm": 1.8257893323898315,
42
+ "learning_rate": 0.0002,
43
+ "loss": 1.5193,
44
+ "step": 5000
45
+ },
46
+ {
47
+ "epoch": 0.47,
48
+ "grad_norm": 2.1923136711120605,
49
+ "learning_rate": 0.0002,
50
+ "loss": 1.4439,
51
+ "step": 6000
52
+ },
53
+ {
54
+ "epoch": 0.54,
55
+ "grad_norm": 2.198930501937866,
56
+ "learning_rate": 0.0002,
57
+ "loss": 1.3742,
58
+ "step": 7000
59
+ },
60
+ {
61
+ "epoch": 0.62,
62
+ "grad_norm": 1.8519206047058105,
63
+ "learning_rate": 0.0002,
64
+ "loss": 1.3327,
65
+ "step": 8000
66
+ },
67
+ {
68
+ "epoch": 0.7,
69
+ "grad_norm": 2.371457576751709,
70
+ "learning_rate": 0.0002,
71
+ "loss": 1.2872,
72
+ "step": 9000
73
+ },
74
+ {
75
+ "epoch": 0.78,
76
+ "grad_norm": 2.2528538703918457,
77
+ "learning_rate": 0.0002,
78
+ "loss": 1.2426,
79
+ "step": 10000
80
+ },
81
+ {
82
+ "epoch": 0.78,
83
+ "eval_bleu": 46.2793,
84
+ "eval_gen_len": 28.6532,
85
+ "eval_loss": 0.8300400972366333,
86
+ "eval_runtime": 100.4495,
87
+ "eval_samples_per_second": 8.84,
88
+ "eval_steps_per_second": 1.105,
89
+ "step": 10000
90
+ },
91
+ {
92
+ "epoch": 0.85,
93
+ "grad_norm": 1.8507146835327148,
94
+ "learning_rate": 0.0002,
95
+ "loss": 1.2154,
96
+ "step": 11000
97
+ },
98
+ {
99
+ "epoch": 0.93,
100
+ "grad_norm": 1.629703164100647,
101
+ "learning_rate": 0.0002,
102
+ "loss": 1.1822,
103
+ "step": 12000
104
+ },
105
+ {
106
+ "epoch": 1.01,
107
+ "grad_norm": 1.9470340013504028,
108
+ "learning_rate": 0.0002,
109
+ "loss": 1.1505,
110
+ "step": 13000
111
+ },
112
+ {
113
+ "epoch": 1.09,
114
+ "grad_norm": 1.733299970626831,
115
+ "learning_rate": 0.0002,
116
+ "loss": 1.0877,
117
+ "step": 14000
118
+ },
119
+ {
120
+ "epoch": 1.16,
121
+ "grad_norm": 1.5334522724151611,
122
+ "learning_rate": 0.0002,
123
+ "loss": 1.0647,
124
+ "step": 15000
125
+ },
126
+ {
127
+ "epoch": 1.24,
128
+ "grad_norm": 1.8881198167800903,
129
+ "learning_rate": 0.0002,
130
+ "loss": 1.0595,
131
+ "step": 16000
132
+ },
133
+ {
134
+ "epoch": 1.32,
135
+ "grad_norm": 2.1781082153320312,
136
+ "learning_rate": 0.0002,
137
+ "loss": 1.0441,
138
+ "step": 17000
139
+ },
140
+ {
141
+ "epoch": 1.4,
142
+ "grad_norm": 1.509994387626648,
143
+ "learning_rate": 0.0002,
144
+ "loss": 1.0307,
145
+ "step": 18000
146
+ },
147
+ {
148
+ "epoch": 1.47,
149
+ "grad_norm": 2.5609610080718994,
150
+ "learning_rate": 0.0002,
151
+ "loss": 1.0165,
152
+ "step": 19000
153
+ },
154
+ {
155
+ "epoch": 1.55,
156
+ "grad_norm": 1.512005090713501,
157
+ "learning_rate": 0.0002,
158
+ "loss": 0.9931,
159
+ "step": 20000
160
+ },
161
+ {
162
+ "epoch": 1.55,
163
+ "eval_bleu": 52.2709,
164
+ "eval_gen_len": 28.6441,
165
+ "eval_loss": 0.6755661368370056,
166
+ "eval_runtime": 99.4907,
167
+ "eval_samples_per_second": 8.925,
168
+ "eval_steps_per_second": 1.116,
169
+ "step": 20000
170
+ },
171
+ {
172
+ "epoch": 1.63,
173
+ "grad_norm": 1.4830211400985718,
174
+ "learning_rate": 0.0002,
175
+ "loss": 0.9854,
176
+ "step": 21000
177
+ },
178
+ {
179
+ "epoch": 1.71,
180
+ "grad_norm": 1.8581557273864746,
181
+ "learning_rate": 0.0002,
182
+ "loss": 0.9736,
183
+ "step": 22000
184
+ },
185
+ {
186
+ "epoch": 1.79,
187
+ "grad_norm": 1.589917778968811,
188
+ "learning_rate": 0.0002,
189
+ "loss": 0.9642,
190
+ "step": 23000
191
+ },
192
+ {
193
+ "epoch": 1.86,
194
+ "grad_norm": 2.332538604736328,
195
+ "learning_rate": 0.0002,
196
+ "loss": 0.9476,
197
+ "step": 24000
198
+ },
199
+ {
200
+ "epoch": 1.94,
201
+ "grad_norm": 1.61520516872406,
202
+ "learning_rate": 0.0002,
203
+ "loss": 0.943,
204
+ "step": 25000
205
+ },
206
+ {
207
+ "epoch": 2.02,
208
+ "grad_norm": 1.2320265769958496,
209
+ "learning_rate": 0.0002,
210
+ "loss": 0.9135,
211
+ "step": 26000
212
+ },
213
+ {
214
+ "epoch": 2.1,
215
+ "grad_norm": 1.543626308441162,
216
+ "learning_rate": 0.0002,
217
+ "loss": 0.8763,
218
+ "step": 27000
219
+ },
220
+ {
221
+ "epoch": 2.17,
222
+ "grad_norm": 1.7634472846984863,
223
+ "learning_rate": 0.0002,
224
+ "loss": 0.8664,
225
+ "step": 28000
226
+ },
227
+ {
228
+ "epoch": 2.25,
229
+ "grad_norm": 1.1254847049713135,
230
+ "learning_rate": 0.0002,
231
+ "loss": 0.8625,
232
+ "step": 29000
233
+ },
234
+ {
235
+ "epoch": 2.33,
236
+ "grad_norm": 1.390243649482727,
237
+ "learning_rate": 0.0002,
238
+ "loss": 0.8573,
239
+ "step": 30000
240
+ },
241
+ {
242
+ "epoch": 2.33,
243
+ "eval_bleu": 55.8294,
244
+ "eval_gen_len": 28.5405,
245
+ "eval_loss": 0.6142958998680115,
246
+ "eval_runtime": 100.5081,
247
+ "eval_samples_per_second": 8.835,
248
+ "eval_steps_per_second": 1.104,
249
+ "step": 30000
250
+ },
251
+ {
252
+ "epoch": 2.41,
253
+ "grad_norm": 1.592313289642334,
254
+ "learning_rate": 0.0002,
255
+ "loss": 0.855,
256
+ "step": 31000
257
+ },
258
+ {
259
+ "epoch": 2.48,
260
+ "grad_norm": 1.1361020803451538,
261
+ "learning_rate": 0.0002,
262
+ "loss": 0.8487,
263
+ "step": 32000
264
+ },
265
+ {
266
+ "epoch": 2.56,
267
+ "grad_norm": 1.2688100337982178,
268
+ "learning_rate": 0.0002,
269
+ "loss": 0.8417,
270
+ "step": 33000
271
+ },
272
+ {
273
+ "epoch": 2.64,
274
+ "grad_norm": 1.3345963954925537,
275
+ "learning_rate": 0.0002,
276
+ "loss": 0.829,
277
+ "step": 34000
278
+ },
279
+ {
280
+ "epoch": 2.72,
281
+ "grad_norm": 1.4640212059020996,
282
+ "learning_rate": 0.0002,
283
+ "loss": 0.8264,
284
+ "step": 35000
285
+ },
286
+ {
287
+ "epoch": 2.79,
288
+ "grad_norm": 1.4818629026412964,
289
+ "learning_rate": 0.0002,
290
+ "loss": 0.8207,
291
+ "step": 36000
292
+ },
293
+ {
294
+ "epoch": 2.87,
295
+ "grad_norm": 1.4580553770065308,
296
+ "learning_rate": 0.0002,
297
+ "loss": 0.8155,
298
+ "step": 37000
299
+ },
300
+ {
301
+ "epoch": 2.95,
302
+ "grad_norm": 1.8713078498840332,
303
+ "learning_rate": 0.0002,
304
+ "loss": 0.8134,
305
+ "step": 38000
306
+ },
307
+ {
308
+ "epoch": 3.03,
309
+ "grad_norm": 1.295332670211792,
310
+ "learning_rate": 0.0002,
311
+ "loss": 0.787,
312
+ "step": 39000
313
+ },
314
+ {
315
+ "epoch": 3.11,
316
+ "grad_norm": 1.532378077507019,
317
+ "learning_rate": 0.0002,
318
+ "loss": 0.762,
319
+ "step": 40000
320
+ },
321
+ {
322
+ "epoch": 3.11,
323
+ "eval_bleu": 57.5135,
324
+ "eval_gen_len": 28.366,
325
+ "eval_loss": 0.581108570098877,
326
+ "eval_runtime": 98.8348,
327
+ "eval_samples_per_second": 8.985,
328
+ "eval_steps_per_second": 1.123,
329
+ "step": 40000
330
+ },
331
+ {
332
+ "epoch": 3.18,
333
+ "grad_norm": 1.3924858570098877,
334
+ "learning_rate": 0.0002,
335
+ "loss": 0.7577,
336
+ "step": 41000
337
+ },
338
+ {
339
+ "epoch": 3.26,
340
+ "grad_norm": 1.4161888360977173,
341
+ "learning_rate": 0.0002,
342
+ "loss": 0.7482,
343
+ "step": 42000
344
+ },
345
+ {
346
+ "epoch": 3.34,
347
+ "grad_norm": 1.639460802078247,
348
+ "learning_rate": 0.0002,
349
+ "loss": 0.7582,
350
+ "step": 43000
351
+ },
352
+ {
353
+ "epoch": 3.42,
354
+ "grad_norm": 1.3779182434082031,
355
+ "learning_rate": 0.0002,
356
+ "loss": 0.7474,
357
+ "step": 44000
358
+ },
359
+ {
360
+ "epoch": 3.49,
361
+ "grad_norm": 1.8350883722305298,
362
+ "learning_rate": 0.0002,
363
+ "loss": 0.751,
364
+ "step": 45000
365
+ },
366
+ {
367
+ "epoch": 3.57,
368
+ "grad_norm": 1.2075496912002563,
369
+ "learning_rate": 0.0002,
370
+ "loss": 0.7371,
371
+ "step": 46000
372
+ },
373
+ {
374
+ "epoch": 3.65,
375
+ "grad_norm": 1.3083984851837158,
376
+ "learning_rate": 0.0002,
377
+ "loss": 0.7318,
378
+ "step": 47000
379
+ },
380
+ {
381
+ "epoch": 3.73,
382
+ "grad_norm": 1.5021324157714844,
383
+ "learning_rate": 0.0002,
384
+ "loss": 0.7376,
385
+ "step": 48000
386
+ },
387
+ {
388
+ "epoch": 3.8,
389
+ "grad_norm": 1.1597286462783813,
390
+ "learning_rate": 0.0002,
391
+ "loss": 0.7336,
392
+ "step": 49000
393
+ },
394
+ {
395
+ "epoch": 3.88,
396
+ "grad_norm": 1.3814338445663452,
397
+ "learning_rate": 0.0002,
398
+ "loss": 0.734,
399
+ "step": 50000
400
+ },
401
+ {
402
+ "epoch": 3.88,
403
+ "eval_bleu": 58.6125,
404
+ "eval_gen_len": 28.5101,
405
+ "eval_loss": 0.5499288439750671,
406
+ "eval_runtime": 99.2548,
407
+ "eval_samples_per_second": 8.947,
408
+ "eval_steps_per_second": 1.118,
409
+ "step": 50000
410
+ },
411
+ {
412
+ "epoch": 3.96,
413
+ "grad_norm": 1.0968077182769775,
414
+ "learning_rate": 0.0002,
415
+ "loss": 0.7288,
416
+ "step": 51000
417
+ },
418
+ {
419
+ "epoch": 4.04,
420
+ "grad_norm": 1.9418740272521973,
421
+ "learning_rate": 0.0002,
422
+ "loss": 0.7057,
423
+ "step": 52000
424
+ },
425
+ {
426
+ "epoch": 4.11,
427
+ "grad_norm": 1.1883801221847534,
428
+ "learning_rate": 0.0002,
429
+ "loss": 0.6761,
430
+ "step": 53000
431
+ },
432
+ {
433
+ "epoch": 4.19,
434
+ "grad_norm": 1.1581670045852661,
435
+ "learning_rate": 0.0002,
436
+ "loss": 0.6812,
437
+ "step": 54000
438
+ },
439
+ {
440
+ "epoch": 4.27,
441
+ "grad_norm": 1.4657800197601318,
442
+ "learning_rate": 0.0002,
443
+ "loss": 0.6783,
444
+ "step": 55000
445
+ },
446
+ {
447
+ "epoch": 4.35,
448
+ "grad_norm": 1.1991990804672241,
449
+ "learning_rate": 0.0002,
450
+ "loss": 0.6764,
451
+ "step": 56000
452
+ },
453
+ {
454
+ "epoch": 4.42,
455
+ "grad_norm": 1.5376391410827637,
456
+ "learning_rate": 0.0002,
457
+ "loss": 0.6805,
458
+ "step": 57000
459
+ },
460
+ {
461
+ "epoch": 4.5,
462
+ "grad_norm": 1.228727102279663,
463
+ "learning_rate": 0.0002,
464
+ "loss": 0.6734,
465
+ "step": 58000
466
+ },
467
+ {
468
+ "epoch": 4.58,
469
+ "grad_norm": 0.949891984462738,
470
+ "learning_rate": 0.0002,
471
+ "loss": 0.6702,
472
+ "step": 59000
473
+ },
474
+ {
475
+ "epoch": 4.66,
476
+ "grad_norm": 1.5632683038711548,
477
+ "learning_rate": 0.0002,
478
+ "loss": 0.6722,
479
+ "step": 60000
480
+ },
481
+ {
482
+ "epoch": 4.66,
483
+ "eval_bleu": 59.6427,
484
+ "eval_gen_len": 28.8356,
485
+ "eval_loss": 0.522808849811554,
486
+ "eval_runtime": 101.7798,
487
+ "eval_samples_per_second": 8.725,
488
+ "eval_steps_per_second": 1.091,
489
+ "step": 60000
490
+ },
491
+ {
492
+ "epoch": 4.74,
493
+ "grad_norm": 1.2885149717330933,
494
+ "learning_rate": 0.0002,
495
+ "loss": 0.67,
496
+ "step": 61000
497
+ },
498
+ {
499
+ "epoch": 4.81,
500
+ "grad_norm": 1.7992392778396606,
501
+ "learning_rate": 0.0002,
502
+ "loss": 0.6686,
503
+ "step": 62000
504
+ },
505
+ {
506
+ "epoch": 4.89,
507
+ "grad_norm": 1.7027188539505005,
508
+ "learning_rate": 0.0002,
509
+ "loss": 0.6682,
510
+ "step": 63000
511
+ },
512
+ {
513
+ "epoch": 4.97,
514
+ "grad_norm": 1.2083909511566162,
515
+ "learning_rate": 0.0002,
516
+ "loss": 0.6545,
517
+ "step": 64000
518
+ },
519
+ {
520
+ "epoch": 5.05,
521
+ "grad_norm": 1.1508337259292603,
522
+ "learning_rate": 0.0002,
523
+ "loss": 0.6308,
524
+ "step": 65000
525
+ },
526
+ {
527
+ "epoch": 5.12,
528
+ "grad_norm": 1.0901002883911133,
529
+ "learning_rate": 0.0002,
530
+ "loss": 0.6192,
531
+ "step": 66000
532
+ },
533
+ {
534
+ "epoch": 5.2,
535
+ "grad_norm": 1.3939250707626343,
536
+ "learning_rate": 0.0002,
537
+ "loss": 0.6217,
538
+ "step": 67000
539
+ },
540
+ {
541
+ "epoch": 5.28,
542
+ "grad_norm": 0.9105481505393982,
543
+ "learning_rate": 0.0002,
544
+ "loss": 0.6177,
545
+ "step": 68000
546
+ },
547
+ {
548
+ "epoch": 5.36,
549
+ "grad_norm": 0.944652795791626,
550
+ "learning_rate": 0.0002,
551
+ "loss": 0.6227,
552
+ "step": 69000
553
+ },
554
+ {
555
+ "epoch": 5.43,
556
+ "grad_norm": 1.1488838195800781,
557
+ "learning_rate": 0.0002,
558
+ "loss": 0.6215,
559
+ "step": 70000
560
+ },
561
+ {
562
+ "epoch": 5.43,
563
+ "eval_bleu": 60.4701,
564
+ "eval_gen_len": 28.7534,
565
+ "eval_loss": 0.5161064863204956,
566
+ "eval_runtime": 100.914,
567
+ "eval_samples_per_second": 8.8,
568
+ "eval_steps_per_second": 1.1,
569
+ "step": 70000
570
+ },
571
+ {
572
+ "epoch": 5.51,
573
+ "grad_norm": 1.0799453258514404,
574
+ "learning_rate": 0.0002,
575
+ "loss": 0.6173,
576
+ "step": 71000
577
+ },
578
+ {
579
+ "epoch": 5.59,
580
+ "grad_norm": 1.2405527830123901,
581
+ "learning_rate": 0.0002,
582
+ "loss": 0.6229,
583
+ "step": 72000
584
+ },
585
+ {
586
+ "epoch": 5.67,
587
+ "grad_norm": 1.045590877532959,
588
+ "learning_rate": 0.0002,
589
+ "loss": 0.6183,
590
+ "step": 73000
591
+ },
592
+ {
593
+ "epoch": 5.74,
594
+ "grad_norm": 1.4318771362304688,
595
+ "learning_rate": 0.0002,
596
+ "loss": 0.6146,
597
+ "step": 74000
598
+ },
599
+ {
600
+ "epoch": 5.82,
601
+ "grad_norm": 1.0059374570846558,
602
+ "learning_rate": 0.0002,
603
+ "loss": 0.6175,
604
+ "step": 75000
605
+ },
606
+ {
607
+ "epoch": 5.9,
608
+ "grad_norm": 1.0831586122512817,
609
+ "learning_rate": 0.0002,
610
+ "loss": 0.6165,
611
+ "step": 76000
612
+ },
613
+ {
614
+ "epoch": 5.98,
615
+ "grad_norm": 1.4094606637954712,
616
+ "learning_rate": 0.0002,
617
+ "loss": 0.6158,
618
+ "step": 77000
619
+ },
620
+ {
621
+ "epoch": 6.05,
622
+ "grad_norm": 1.2640748023986816,
623
+ "learning_rate": 0.0002,
624
+ "loss": 0.5827,
625
+ "step": 78000
626
+ },
627
+ {
628
+ "epoch": 6.13,
629
+ "grad_norm": 1.0088295936584473,
630
+ "learning_rate": 0.0002,
631
+ "loss": 0.5699,
632
+ "step": 79000
633
+ },
634
+ {
635
+ "epoch": 6.21,
636
+ "grad_norm": 0.9942078590393066,
637
+ "learning_rate": 0.0002,
638
+ "loss": 0.5756,
639
+ "step": 80000
640
+ },
641
+ {
642
+ "epoch": 6.21,
643
+ "eval_bleu": 62.0864,
644
+ "eval_gen_len": 28.6498,
645
+ "eval_loss": 0.5067651867866516,
646
+ "eval_runtime": 99.2392,
647
+ "eval_samples_per_second": 8.948,
648
+ "eval_steps_per_second": 1.119,
649
+ "step": 80000
650
+ },
651
+ {
652
+ "epoch": 6.29,
653
+ "grad_norm": 1.838376522064209,
654
+ "learning_rate": 0.0002,
655
+ "loss": 0.5784,
656
+ "step": 81000
657
+ },
658
+ {
659
+ "epoch": 6.37,
660
+ "grad_norm": 1.477137804031372,
661
+ "learning_rate": 0.0002,
662
+ "loss": 0.5743,
663
+ "step": 82000
664
+ },
665
+ {
666
+ "epoch": 6.44,
667
+ "grad_norm": 1.1481742858886719,
668
+ "learning_rate": 0.0002,
669
+ "loss": 0.578,
670
+ "step": 83000
671
+ },
672
+ {
673
+ "epoch": 6.52,
674
+ "grad_norm": 1.1412264108657837,
675
+ "learning_rate": 0.0002,
676
+ "loss": 0.5785,
677
+ "step": 84000
678
+ },
679
+ {
680
+ "epoch": 6.6,
681
+ "grad_norm": 0.9707184433937073,
682
+ "learning_rate": 0.0002,
683
+ "loss": 0.5746,
684
+ "step": 85000
685
+ },
686
+ {
687
+ "epoch": 6.68,
688
+ "grad_norm": 0.9683183431625366,
689
+ "learning_rate": 0.0002,
690
+ "loss": 0.5794,
691
+ "step": 86000
692
+ },
693
+ {
694
+ "epoch": 6.75,
695
+ "grad_norm": 1.2322285175323486,
696
+ "learning_rate": 0.0002,
697
+ "loss": 0.5765,
698
+ "step": 87000
699
+ },
700
+ {
701
+ "epoch": 6.83,
702
+ "grad_norm": 1.1276684999465942,
703
+ "learning_rate": 0.0002,
704
+ "loss": 0.5767,
705
+ "step": 88000
706
+ },
707
+ {
708
+ "epoch": 6.91,
709
+ "grad_norm": 1.4353203773498535,
710
+ "learning_rate": 0.0002,
711
+ "loss": 0.5776,
712
+ "step": 89000
713
+ },
714
+ {
715
+ "epoch": 6.99,
716
+ "grad_norm": 1.116827368736267,
717
+ "learning_rate": 0.0002,
718
+ "loss": 0.5738,
719
+ "step": 90000
720
+ },
721
+ {
722
+ "epoch": 6.99,
723
+ "eval_bleu": 61.9714,
724
+ "eval_gen_len": 28.5788,
725
+ "eval_loss": 0.5005396008491516,
726
+ "eval_runtime": 98.6774,
727
+ "eval_samples_per_second": 8.999,
728
+ "eval_steps_per_second": 1.125,
729
+ "step": 90000
730
+ },
731
+ {
732
+ "epoch": 7.06,
733
+ "grad_norm": 1.0088461637496948,
734
+ "learning_rate": 0.0002,
735
+ "loss": 0.538,
736
+ "step": 91000
737
+ },
738
+ {
739
+ "epoch": 7.14,
740
+ "grad_norm": 1.1016899347305298,
741
+ "learning_rate": 0.0002,
742
+ "loss": 0.5349,
743
+ "step": 92000
744
+ },
745
+ {
746
+ "epoch": 7.22,
747
+ "grad_norm": 1.4406321048736572,
748
+ "learning_rate": 0.0002,
749
+ "loss": 0.5408,
750
+ "step": 93000
751
+ },
752
+ {
753
+ "epoch": 7.3,
754
+ "grad_norm": 1.1226301193237305,
755
+ "learning_rate": 0.0002,
756
+ "loss": 0.5362,
757
+ "step": 94000
758
+ },
759
+ {
760
+ "epoch": 7.37,
761
+ "grad_norm": 1.4062280654907227,
762
+ "learning_rate": 0.0002,
763
+ "loss": 0.5381,
764
+ "step": 95000
765
+ },
766
+ {
767
+ "epoch": 7.45,
768
+ "grad_norm": 1.2851547002792358,
769
+ "learning_rate": 0.0002,
770
+ "loss": 0.5438,
771
+ "step": 96000
772
+ },
773
+ {
774
+ "epoch": 7.53,
775
+ "grad_norm": 0.9344896078109741,
776
+ "learning_rate": 0.0002,
777
+ "loss": 0.5442,
778
+ "step": 97000
779
+ },
780
+ {
781
+ "epoch": 7.61,
782
+ "grad_norm": 1.7169030904769897,
783
+ "learning_rate": 0.0002,
784
+ "loss": 0.539,
785
+ "step": 98000
786
+ },
787
+ {
788
+ "epoch": 7.69,
789
+ "grad_norm": 1.1855800151824951,
790
+ "learning_rate": 0.0002,
791
+ "loss": 0.543,
792
+ "step": 99000
793
+ },
794
+ {
795
+ "epoch": 7.76,
796
+ "grad_norm": 1.093122959136963,
797
+ "learning_rate": 0.0002,
798
+ "loss": 0.5384,
799
+ "step": 100000
800
+ },
801
+ {
802
+ "epoch": 7.76,
803
+ "eval_bleu": 62.407,
804
+ "eval_gen_len": 28.5282,
805
+ "eval_loss": 0.49085742235183716,
806
+ "eval_runtime": 98.8889,
807
+ "eval_samples_per_second": 8.98,
808
+ "eval_steps_per_second": 1.122,
809
+ "step": 100000
810
+ },
811
+ {
812
+ "epoch": 7.84,
813
+ "grad_norm": 1.3849202394485474,
814
+ "learning_rate": 0.0002,
815
+ "loss": 0.5381,
816
+ "step": 101000
817
+ },
818
+ {
819
+ "epoch": 7.92,
820
+ "grad_norm": 1.12773859500885,
821
+ "learning_rate": 0.0002,
822
+ "loss": 0.544,
823
+ "step": 102000
824
+ },
825
+ {
826
+ "epoch": 8.0,
827
+ "grad_norm": 0.9132428169250488,
828
+ "learning_rate": 0.0002,
829
+ "loss": 0.5371,
830
+ "step": 103000
831
+ },
832
+ {
833
+ "epoch": 8.07,
834
+ "grad_norm": 1.5978176593780518,
835
+ "learning_rate": 0.0002,
836
+ "loss": 0.5015,
837
+ "step": 104000
838
+ },
839
+ {
840
+ "epoch": 8.15,
841
+ "grad_norm": 1.028082013130188,
842
+ "learning_rate": 0.0002,
843
+ "loss": 0.4956,
844
+ "step": 105000
845
+ },
846
+ {
847
+ "epoch": 8.23,
848
+ "grad_norm": 1.0597223043441772,
849
+ "learning_rate": 0.0002,
850
+ "loss": 0.5074,
851
+ "step": 106000
852
+ },
853
+ {
854
+ "epoch": 8.31,
855
+ "grad_norm": 1.47709059715271,
856
+ "learning_rate": 0.0002,
857
+ "loss": 0.5022,
858
+ "step": 107000
859
+ },
860
+ {
861
+ "epoch": 8.38,
862
+ "grad_norm": 1.1731916666030884,
863
+ "learning_rate": 0.0002,
864
+ "loss": 0.5039,
865
+ "step": 108000
866
+ },
867
+ {
868
+ "epoch": 8.46,
869
+ "grad_norm": 1.2147469520568848,
870
+ "learning_rate": 0.0002,
871
+ "loss": 0.5121,
872
+ "step": 109000
873
+ },
874
+ {
875
+ "epoch": 8.54,
876
+ "grad_norm": 0.8521010279655457,
877
+ "learning_rate": 0.0002,
878
+ "loss": 0.5109,
879
+ "step": 110000
880
+ },
881
+ {
882
+ "epoch": 8.54,
883
+ "eval_bleu": 62.1452,
884
+ "eval_gen_len": 28.4617,
885
+ "eval_loss": 0.4901648759841919,
886
+ "eval_runtime": 98.4744,
887
+ "eval_samples_per_second": 9.018,
888
+ "eval_steps_per_second": 1.127,
889
+ "step": 110000
890
+ },
891
+ {
892
+ "epoch": 8.62,
893
+ "grad_norm": 1.250752568244934,
894
+ "learning_rate": 0.0002,
895
+ "loss": 0.5072,
896
+ "step": 111000
897
+ },
898
+ {
899
+ "epoch": 8.69,
900
+ "grad_norm": 1.2694823741912842,
901
+ "learning_rate": 0.0002,
902
+ "loss": 0.5126,
903
+ "step": 112000
904
+ },
905
+ {
906
+ "epoch": 8.77,
907
+ "grad_norm": 1.0290015935897827,
908
+ "learning_rate": 0.0002,
909
+ "loss": 0.5065,
910
+ "step": 113000
911
+ },
912
+ {
913
+ "epoch": 8.85,
914
+ "grad_norm": 1.222034215927124,
915
+ "learning_rate": 0.0002,
916
+ "loss": 0.514,
917
+ "step": 114000
918
+ },
919
+ {
920
+ "epoch": 8.93,
921
+ "grad_norm": 1.0359649658203125,
922
+ "learning_rate": 0.0002,
923
+ "loss": 0.5127,
924
+ "step": 115000
925
+ },
926
+ {
927
+ "epoch": 9.0,
928
+ "grad_norm": 1.19712495803833,
929
+ "learning_rate": 0.0002,
930
+ "loss": 0.5114,
931
+ "step": 116000
932
+ },
933
+ {
934
+ "epoch": 9.08,
935
+ "grad_norm": 1.1766573190689087,
936
+ "learning_rate": 0.0002,
937
+ "loss": 0.4698,
938
+ "step": 117000
939
+ },
940
+ {
941
+ "epoch": 9.16,
942
+ "grad_norm": 1.2681427001953125,
943
+ "learning_rate": 0.0002,
944
+ "loss": 0.4755,
945
+ "step": 118000
946
+ },
947
+ {
948
+ "epoch": 9.24,
949
+ "grad_norm": 1.2988672256469727,
950
+ "learning_rate": 0.0002,
951
+ "loss": 0.4772,
952
+ "step": 119000
953
+ },
954
+ {
955
+ "epoch": 9.32,
956
+ "grad_norm": 1.440721035003662,
957
+ "learning_rate": 0.0002,
958
+ "loss": 0.4816,
959
+ "step": 120000
960
+ },
961
+ {
962
+ "epoch": 9.32,
963
+ "eval_bleu": 62.6499,
964
+ "eval_gen_len": 28.5518,
965
+ "eval_loss": 0.48750796914100647,
966
+ "eval_runtime": 99.0223,
967
+ "eval_samples_per_second": 8.968,
968
+ "eval_steps_per_second": 1.121,
969
+ "step": 120000
970
+ },
971
+ {
972
+ "epoch": 9.39,
973
+ "grad_norm": 1.038442611694336,
974
+ "learning_rate": 0.0002,
975
+ "loss": 0.4792,
976
+ "step": 121000
977
+ },
978
+ {
979
+ "epoch": 9.47,
980
+ "grad_norm": 1.3428473472595215,
981
+ "learning_rate": 0.0002,
982
+ "loss": 0.4827,
983
+ "step": 122000
984
+ },
985
+ {
986
+ "epoch": 9.55,
987
+ "grad_norm": 1.4756362438201904,
988
+ "learning_rate": 0.0002,
989
+ "loss": 0.4832,
990
+ "step": 123000
991
+ },
992
+ {
993
+ "epoch": 9.63,
994
+ "grad_norm": 1.2109817266464233,
995
+ "learning_rate": 0.0002,
996
+ "loss": 0.481,
997
+ "step": 124000
998
+ },
999
+ {
1000
+ "epoch": 9.7,
1001
+ "grad_norm": 1.2007863521575928,
1002
+ "learning_rate": 0.0002,
1003
+ "loss": 0.4815,
1004
+ "step": 125000
1005
+ },
1006
+ {
1007
+ "epoch": 9.78,
1008
+ "grad_norm": 1.1711379289627075,
1009
+ "learning_rate": 0.0002,
1010
+ "loss": 0.488,
1011
+ "step": 126000
1012
+ },
1013
+ {
1014
+ "epoch": 9.86,
1015
+ "grad_norm": 1.1571533679962158,
1016
+ "learning_rate": 0.0002,
1017
+ "loss": 0.4827,
1018
+ "step": 127000
1019
+ },
1020
+ {
1021
+ "epoch": 9.94,
1022
+ "grad_norm": 1.2341859340667725,
1023
+ "learning_rate": 0.0002,
1024
+ "loss": 0.4844,
1025
+ "step": 128000
1026
+ },
1027
+ {
1028
+ "epoch": 10.01,
1029
+ "grad_norm": 1.5796501636505127,
1030
+ "learning_rate": 0.0002,
1031
+ "loss": 0.4741,
1032
+ "step": 129000
1033
+ },
1034
+ {
1035
+ "epoch": 10.09,
1036
+ "grad_norm": 0.8122438788414001,
1037
+ "learning_rate": 0.0002,
1038
+ "loss": 0.4493,
1039
+ "step": 130000
1040
+ },
1041
+ {
1042
+ "epoch": 10.09,
1043
+ "eval_bleu": 62.6694,
1044
+ "eval_gen_len": 28.6993,
1045
+ "eval_loss": 0.4866686761379242,
1046
+ "eval_runtime": 100.7784,
1047
+ "eval_samples_per_second": 8.811,
1048
+ "eval_steps_per_second": 1.101,
1049
+ "step": 130000
1050
+ },
1051
+ {
1052
+ "epoch": 10.17,
1053
+ "grad_norm": 1.1835366487503052,
1054
+ "learning_rate": 0.0002,
1055
+ "loss": 0.448,
1056
+ "step": 131000
1057
+ },
1058
+ {
1059
+ "epoch": 10.25,
1060
+ "grad_norm": 1.0868804454803467,
1061
+ "learning_rate": 0.0002,
1062
+ "loss": 0.4517,
1063
+ "step": 132000
1064
+ },
1065
+ {
1066
+ "epoch": 10.32,
1067
+ "grad_norm": 0.9316431283950806,
1068
+ "learning_rate": 0.0002,
1069
+ "loss": 0.454,
1070
+ "step": 133000
1071
+ },
1072
+ {
1073
+ "epoch": 10.4,
1074
+ "grad_norm": 1.5438517332077026,
1075
+ "learning_rate": 0.0002,
1076
+ "loss": 0.4526,
1077
+ "step": 134000
1078
+ },
1079
+ {
1080
+ "epoch": 10.48,
1081
+ "grad_norm": 1.5842955112457275,
1082
+ "learning_rate": 0.0002,
1083
+ "loss": 0.4576,
1084
+ "step": 135000
1085
+ },
1086
+ {
1087
+ "epoch": 10.56,
1088
+ "grad_norm": 1.450462818145752,
1089
+ "learning_rate": 0.0002,
1090
+ "loss": 0.463,
1091
+ "step": 136000
1092
+ },
1093
+ {
1094
+ "epoch": 10.63,
1095
+ "grad_norm": 0.8578802347183228,
1096
+ "learning_rate": 0.0002,
1097
+ "loss": 0.4588,
1098
+ "step": 137000
1099
+ },
1100
+ {
1101
+ "epoch": 10.71,
1102
+ "grad_norm": 1.1508352756500244,
1103
+ "learning_rate": 0.0002,
1104
+ "loss": 0.4542,
1105
+ "step": 138000
1106
+ },
1107
+ {
1108
+ "epoch": 10.79,
1109
+ "grad_norm": 1.1183589696884155,
1110
+ "learning_rate": 0.0002,
1111
+ "loss": 0.4601,
1112
+ "step": 139000
1113
+ },
1114
+ {
1115
+ "epoch": 10.87,
1116
+ "grad_norm": 0.9286684393882751,
1117
+ "learning_rate": 0.0002,
1118
+ "loss": 0.4648,
1119
+ "step": 140000
1120
+ },
1121
+ {
1122
+ "epoch": 10.87,
1123
+ "eval_bleu": 63.3179,
1124
+ "eval_gen_len": 28.5495,
1125
+ "eval_loss": 0.4774705469608307,
1126
+ "eval_runtime": 98.6639,
1127
+ "eval_samples_per_second": 9.0,
1128
+ "eval_steps_per_second": 1.125,
1129
+ "step": 140000
1130
+ },
1131
+ {
1132
+ "epoch": 10.95,
1133
+ "grad_norm": 1.3315681219100952,
1134
+ "learning_rate": 0.0002,
1135
+ "loss": 0.4627,
1136
+ "step": 141000
1137
+ },
1138
+ {
1139
+ "epoch": 11.02,
1140
+ "grad_norm": 1.204750418663025,
1141
+ "learning_rate": 0.0002,
1142
+ "loss": 0.4497,
1143
+ "step": 142000
1144
+ },
1145
+ {
1146
+ "epoch": 11.1,
1147
+ "grad_norm": 1.0254498720169067,
1148
+ "learning_rate": 0.0002,
1149
+ "loss": 0.4242,
1150
+ "step": 143000
1151
+ },
1152
+ {
1153
+ "epoch": 11.18,
1154
+ "grad_norm": 1.052018404006958,
1155
+ "learning_rate": 0.0002,
1156
+ "loss": 0.4306,
1157
+ "step": 144000
1158
+ },
1159
+ {
1160
+ "epoch": 11.26,
1161
+ "grad_norm": 0.9426015019416809,
1162
+ "learning_rate": 0.0002,
1163
+ "loss": 0.4275,
1164
+ "step": 145000
1165
+ },
1166
+ {
1167
+ "epoch": 11.33,
1168
+ "grad_norm": 1.079633116722107,
1169
+ "learning_rate": 0.0002,
1170
+ "loss": 0.4303,
1171
+ "step": 146000
1172
+ },
1173
+ {
1174
+ "epoch": 11.41,
1175
+ "grad_norm": 1.025631070137024,
1176
+ "learning_rate": 0.0002,
1177
+ "loss": 0.433,
1178
+ "step": 147000
1179
+ },
1180
+ {
1181
+ "epoch": 11.49,
1182
+ "grad_norm": 1.21865713596344,
1183
+ "learning_rate": 0.0002,
1184
+ "loss": 0.4351,
1185
+ "step": 148000
1186
+ },
1187
+ {
1188
+ "epoch": 11.57,
1189
+ "grad_norm": 1.0304579734802246,
1190
+ "learning_rate": 0.0002,
1191
+ "loss": 0.4358,
1192
+ "step": 149000
1193
+ },
1194
+ {
1195
+ "epoch": 11.64,
1196
+ "grad_norm": 1.297282338142395,
1197
+ "learning_rate": 0.0002,
1198
+ "loss": 0.4414,
1199
+ "step": 150000
1200
+ },
1201
+ {
1202
+ "epoch": 11.64,
1203
+ "eval_bleu": 63.6928,
1204
+ "eval_gen_len": 28.4673,
1205
+ "eval_loss": 0.4786856770515442,
1206
+ "eval_runtime": 98.6768,
1207
+ "eval_samples_per_second": 8.999,
1208
+ "eval_steps_per_second": 1.125,
1209
+ "step": 150000
1210
+ },
1211
+ {
1212
+ "epoch": 11.72,
1213
+ "grad_norm": 0.973185658454895,
1214
+ "learning_rate": 0.0002,
1215
+ "loss": 0.4375,
1216
+ "step": 151000
1217
+ },
1218
+ {
1219
+ "epoch": 11.8,
1220
+ "grad_norm": 0.9765141010284424,
1221
+ "learning_rate": 0.0002,
1222
+ "loss": 0.4398,
1223
+ "step": 152000
1224
+ },
1225
+ {
1226
+ "epoch": 11.88,
1227
+ "grad_norm": 1.1599891185760498,
1228
+ "learning_rate": 0.0002,
1229
+ "loss": 0.4422,
1230
+ "step": 153000
1231
+ },
1232
+ {
1233
+ "epoch": 11.95,
1234
+ "grad_norm": 1.0690301656723022,
1235
+ "learning_rate": 0.0002,
1236
+ "loss": 0.4396,
1237
+ "step": 154000
1238
+ },
1239
+ {
1240
+ "epoch": 12.03,
1241
+ "grad_norm": 0.9184726476669312,
1242
+ "learning_rate": 0.0002,
1243
+ "loss": 0.4232,
1244
+ "step": 155000
1245
+ },
1246
+ {
1247
+ "epoch": 12.11,
1248
+ "grad_norm": 1.1572961807250977,
1249
+ "learning_rate": 0.0002,
1250
+ "loss": 0.4038,
1251
+ "step": 156000
1252
+ },
1253
+ {
1254
+ "epoch": 12.19,
1255
+ "grad_norm": 1.1003015041351318,
1256
+ "learning_rate": 0.0002,
1257
+ "loss": 0.4088,
1258
+ "step": 157000
1259
+ },
1260
+ {
1261
+ "epoch": 12.27,
1262
+ "grad_norm": 1.147965908050537,
1263
+ "learning_rate": 0.0002,
1264
+ "loss": 0.4099,
1265
+ "step": 158000
1266
+ },
1267
+ {
1268
+ "epoch": 12.34,
1269
+ "grad_norm": 1.3417842388153076,
1270
+ "learning_rate": 0.0002,
1271
+ "loss": 0.4108,
1272
+ "step": 159000
1273
+ },
1274
+ {
1275
+ "epoch": 12.42,
1276
+ "grad_norm": 0.9816053509712219,
1277
+ "learning_rate": 0.0002,
1278
+ "loss": 0.4158,
1279
+ "step": 160000
1280
+ },
1281
+ {
1282
+ "epoch": 12.42,
1283
+ "eval_bleu": 63.8752,
1284
+ "eval_gen_len": 28.5011,
1285
+ "eval_loss": 0.47919762134552,
1286
+ "eval_runtime": 98.6149,
1287
+ "eval_samples_per_second": 9.005,
1288
+ "eval_steps_per_second": 1.126,
1289
+ "step": 160000
1290
+ },
1291
+ {
1292
+ "epoch": 12.5,
1293
+ "grad_norm": 1.1307754516601562,
1294
+ "learning_rate": 0.0002,
1295
+ "loss": 0.4139,
1296
+ "step": 161000
1297
+ },
1298
+ {
1299
+ "epoch": 12.58,
1300
+ "grad_norm": 1.2909305095672607,
1301
+ "learning_rate": 0.0002,
1302
+ "loss": 0.4191,
1303
+ "step": 162000
1304
+ },
1305
+ {
1306
+ "epoch": 12.65,
1307
+ "grad_norm": 1.0675512552261353,
1308
+ "learning_rate": 0.0002,
1309
+ "loss": 0.4178,
1310
+ "step": 163000
1311
+ },
1312
+ {
1313
+ "epoch": 12.73,
1314
+ "grad_norm": 1.062435269355774,
1315
+ "learning_rate": 0.0002,
1316
+ "loss": 0.4183,
1317
+ "step": 164000
1318
+ },
1319
+ {
1320
+ "epoch": 12.81,
1321
+ "grad_norm": 1.2755943536758423,
1322
+ "learning_rate": 0.0002,
1323
+ "loss": 0.42,
1324
+ "step": 165000
1325
+ },
1326
+ {
1327
+ "epoch": 12.89,
1328
+ "grad_norm": 1.0879075527191162,
1329
+ "learning_rate": 0.0002,
1330
+ "loss": 0.4231,
1331
+ "step": 166000
1332
+ },
1333
+ {
1334
+ "epoch": 12.96,
1335
+ "grad_norm": 1.1521817445755005,
1336
+ "learning_rate": 0.0002,
1337
+ "loss": 0.4231,
1338
+ "step": 167000
1339
+ },
1340
+ {
1341
+ "epoch": 13.04,
1342
+ "grad_norm": 1.038859486579895,
1343
+ "learning_rate": 0.0002,
1344
+ "loss": 0.4014,
1345
+ "step": 168000
1346
+ },
1347
+ {
1348
+ "epoch": 13.12,
1349
+ "grad_norm": 1.001861333847046,
1350
+ "learning_rate": 0.0002,
1351
+ "loss": 0.3875,
1352
+ "step": 169000
1353
+ },
1354
+ {
1355
+ "epoch": 13.2,
1356
+ "grad_norm": 0.967998743057251,
1357
+ "learning_rate": 0.0002,
1358
+ "loss": 0.3895,
1359
+ "step": 170000
1360
+ },
1361
+ {
1362
+ "epoch": 13.2,
1363
+ "eval_bleu": 63.8429,
1364
+ "eval_gen_len": 28.6498,
1365
+ "eval_loss": 0.4793872535228729,
1366
+ "eval_runtime": 99.8944,
1367
+ "eval_samples_per_second": 8.889,
1368
+ "eval_steps_per_second": 1.111,
1369
+ "step": 170000
1370
+ },
1371
+ {
1372
+ "epoch": 13.27,
1373
+ "grad_norm": 1.1491278409957886,
1374
+ "learning_rate": 0.0002,
1375
+ "loss": 0.3947,
1376
+ "step": 171000
1377
+ },
1378
+ {
1379
+ "epoch": 13.35,
1380
+ "grad_norm": 1.0739213228225708,
1381
+ "learning_rate": 0.0002,
1382
+ "loss": 0.3937,
1383
+ "step": 172000
1384
+ },
1385
+ {
1386
+ "epoch": 13.43,
1387
+ "grad_norm": 1.3349049091339111,
1388
+ "learning_rate": 0.0002,
1389
+ "loss": 0.3932,
1390
+ "step": 173000
1391
+ },
1392
+ {
1393
+ "epoch": 13.51,
1394
+ "grad_norm": 1.4266788959503174,
1395
+ "learning_rate": 0.0002,
1396
+ "loss": 0.4012,
1397
+ "step": 174000
1398
+ },
1399
+ {
1400
+ "epoch": 13.58,
1401
+ "grad_norm": 1.064070701599121,
1402
+ "learning_rate": 0.0002,
1403
+ "loss": 0.4,
1404
+ "step": 175000
1405
+ },
1406
+ {
1407
+ "epoch": 13.66,
1408
+ "grad_norm": 1.930474877357483,
1409
+ "learning_rate": 0.0002,
1410
+ "loss": 0.399,
1411
+ "step": 176000
1412
+ },
1413
+ {
1414
+ "epoch": 13.74,
1415
+ "grad_norm": 0.994195282459259,
1416
+ "learning_rate": 0.0002,
1417
+ "loss": 0.4026,
1418
+ "step": 177000
1419
+ },
1420
+ {
1421
+ "epoch": 13.82,
1422
+ "grad_norm": 0.9755762815475464,
1423
+ "learning_rate": 0.0002,
1424
+ "loss": 0.4019,
1425
+ "step": 178000
1426
+ },
1427
+ {
1428
+ "epoch": 13.9,
1429
+ "grad_norm": 1.0802558660507202,
1430
+ "learning_rate": 0.0002,
1431
+ "loss": 0.4027,
1432
+ "step": 179000
1433
+ },
1434
+ {
1435
+ "epoch": 13.97,
1436
+ "grad_norm": 1.4257205724716187,
1437
+ "learning_rate": 0.0002,
1438
+ "loss": 0.4031,
1439
+ "step": 180000
1440
+ },
1441
+ {
1442
+ "epoch": 13.97,
1443
+ "eval_bleu": 63.9496,
1444
+ "eval_gen_len": 28.7264,
1445
+ "eval_loss": 0.4756912291049957,
1446
+ "eval_runtime": 100.6936,
1447
+ "eval_samples_per_second": 8.819,
1448
+ "eval_steps_per_second": 1.102,
1449
+ "step": 180000
1450
+ },
1451
+ {
1452
+ "epoch": 14.05,
1453
+ "grad_norm": 1.0172358751296997,
1454
+ "learning_rate": 0.0002,
1455
+ "loss": 0.3785,
1456
+ "step": 181000
1457
+ },
1458
+ {
1459
+ "epoch": 14.13,
1460
+ "grad_norm": 0.9525344967842102,
1461
+ "learning_rate": 0.0002,
1462
+ "loss": 0.367,
1463
+ "step": 182000
1464
+ },
1465
+ {
1466
+ "epoch": 14.21,
1467
+ "grad_norm": 1.0674399137496948,
1468
+ "learning_rate": 0.0002,
1469
+ "loss": 0.3723,
1470
+ "step": 183000
1471
+ },
1472
+ {
1473
+ "epoch": 14.28,
1474
+ "grad_norm": 1.311464548110962,
1475
+ "learning_rate": 0.0002,
1476
+ "loss": 0.3743,
1477
+ "step": 184000
1478
+ },
1479
+ {
1480
+ "epoch": 14.36,
1481
+ "grad_norm": 1.020115613937378,
1482
+ "learning_rate": 0.0002,
1483
+ "loss": 0.3743,
1484
+ "step": 185000
1485
+ },
1486
+ {
1487
+ "epoch": 14.44,
1488
+ "grad_norm": 0.9766080379486084,
1489
+ "learning_rate": 0.0002,
1490
+ "loss": 0.3751,
1491
+ "step": 186000
1492
+ },
1493
+ {
1494
+ "epoch": 14.52,
1495
+ "grad_norm": 1.0636546611785889,
1496
+ "learning_rate": 0.0002,
1497
+ "loss": 0.3839,
1498
+ "step": 187000
1499
+ },
1500
+ {
1501
+ "epoch": 14.59,
1502
+ "grad_norm": 1.5485342741012573,
1503
+ "learning_rate": 0.0002,
1504
+ "loss": 0.3864,
1505
+ "step": 188000
1506
+ },
1507
+ {
1508
+ "epoch": 14.67,
1509
+ "grad_norm": 1.189011812210083,
1510
+ "learning_rate": 0.0002,
1511
+ "loss": 0.3836,
1512
+ "step": 189000
1513
+ },
1514
+ {
1515
+ "epoch": 14.75,
1516
+ "grad_norm": 1.2171902656555176,
1517
+ "learning_rate": 0.0002,
1518
+ "loss": 0.3844,
1519
+ "step": 190000
1520
+ },
1521
+ {
1522
+ "epoch": 14.75,
1523
+ "eval_bleu": 63.7498,
1524
+ "eval_gen_len": 28.8288,
1525
+ "eval_loss": 0.48547232151031494,
1526
+ "eval_runtime": 102.2105,
1527
+ "eval_samples_per_second": 8.688,
1528
+ "eval_steps_per_second": 1.086,
1529
+ "step": 190000
1530
+ },
1531
+ {
1532
+ "epoch": 14.83,
1533
+ "grad_norm": 0.9004954695701599,
1534
+ "learning_rate": 0.0002,
1535
+ "loss": 0.3821,
1536
+ "step": 191000
1537
+ },
1538
+ {
1539
+ "epoch": 14.9,
1540
+ "grad_norm": 1.2197577953338623,
1541
+ "learning_rate": 0.0002,
1542
+ "loss": 0.3871,
1543
+ "step": 192000
1544
+ },
1545
+ {
1546
+ "epoch": 14.98,
1547
+ "grad_norm": 1.0094869136810303,
1548
+ "learning_rate": 0.0002,
1549
+ "loss": 0.3924,
1550
+ "step": 193000
1551
+ },
1552
+ {
1553
+ "epoch": 15.06,
1554
+ "grad_norm": 1.1337696313858032,
1555
+ "learning_rate": 0.0002,
1556
+ "loss": 0.3602,
1557
+ "step": 194000
1558
+ },
1559
+ {
1560
+ "epoch": 15.14,
1561
+ "grad_norm": 1.5074607133865356,
1562
+ "learning_rate": 0.0002,
1563
+ "loss": 0.3548,
1564
+ "step": 195000
1565
+ },
1566
+ {
1567
+ "epoch": 15.22,
1568
+ "grad_norm": 0.9171730279922485,
1569
+ "learning_rate": 0.0002,
1570
+ "loss": 0.3586,
1571
+ "step": 196000
1572
+ },
1573
+ {
1574
+ "epoch": 15.29,
1575
+ "grad_norm": 0.8581980466842651,
1576
+ "learning_rate": 0.0002,
1577
+ "loss": 0.3609,
1578
+ "step": 197000
1579
+ },
1580
+ {
1581
+ "epoch": 15.37,
1582
+ "grad_norm": 0.8790922164916992,
1583
+ "learning_rate": 0.0002,
1584
+ "loss": 0.363,
1585
+ "step": 198000
1586
+ },
1587
+ {
1588
+ "epoch": 15.45,
1589
+ "grad_norm": 1.012073278427124,
1590
+ "learning_rate": 0.0002,
1591
+ "loss": 0.3618,
1592
+ "step": 199000
1593
+ },
1594
+ {
1595
+ "epoch": 15.53,
1596
+ "grad_norm": 0.9808474183082581,
1597
+ "learning_rate": 0.0002,
1598
+ "loss": 0.3637,
1599
+ "step": 200000
1600
+ },
1601
+ {
1602
+ "epoch": 15.53,
1603
+ "eval_bleu": 64.2277,
1604
+ "eval_gen_len": 28.661,
1605
+ "eval_loss": 0.4799739420413971,
1606
+ "eval_runtime": 99.1042,
1607
+ "eval_samples_per_second": 8.96,
1608
+ "eval_steps_per_second": 1.12,
1609
+ "step": 200000
1610
+ },
1611
+ {
1612
+ "epoch": 15.6,
1613
+ "grad_norm": 0.995276689529419,
1614
+ "learning_rate": 0.0002,
1615
+ "loss": 0.3676,
1616
+ "step": 201000
1617
+ },
1618
+ {
1619
+ "epoch": 15.68,
1620
+ "grad_norm": 1.2943910360336304,
1621
+ "learning_rate": 0.0002,
1622
+ "loss": 0.3705,
1623
+ "step": 202000
1624
+ },
1625
+ {
1626
+ "epoch": 15.76,
1627
+ "grad_norm": 0.9539749026298523,
1628
+ "learning_rate": 0.0002,
1629
+ "loss": 0.3706,
1630
+ "step": 203000
1631
+ },
1632
+ {
1633
+ "epoch": 15.84,
1634
+ "grad_norm": 0.9351176619529724,
1635
+ "learning_rate": 0.0002,
1636
+ "loss": 0.3706,
1637
+ "step": 204000
1638
+ },
1639
+ {
1640
+ "epoch": 15.91,
1641
+ "grad_norm": 1.087781310081482,
1642
+ "learning_rate": 0.0002,
1643
+ "loss": 0.3713,
1644
+ "step": 205000
1645
+ },
1646
+ {
1647
+ "epoch": 15.99,
1648
+ "grad_norm": 1.2164143323898315,
1649
+ "learning_rate": 0.0002,
1650
+ "loss": 0.3729,
1651
+ "step": 206000
1652
+ },
1653
+ {
1654
+ "epoch": 16.07,
1655
+ "grad_norm": 1.1458275318145752,
1656
+ "learning_rate": 0.0002,
1657
+ "loss": 0.3417,
1658
+ "step": 207000
1659
+ },
1660
+ {
1661
+ "epoch": 16.15,
1662
+ "grad_norm": 0.9169874787330627,
1663
+ "learning_rate": 0.0002,
1664
+ "loss": 0.339,
1665
+ "step": 208000
1666
+ },
1667
+ {
1668
+ "epoch": 16.22,
1669
+ "grad_norm": 1.1594195365905762,
1670
+ "learning_rate": 0.0002,
1671
+ "loss": 0.3426,
1672
+ "step": 209000
1673
+ },
1674
+ {
1675
+ "epoch": 16.3,
1676
+ "grad_norm": 0.8710166215896606,
1677
+ "learning_rate": 0.0002,
1678
+ "loss": 0.3473,
1679
+ "step": 210000
1680
+ },
1681
+ {
1682
+ "epoch": 16.3,
1683
+ "eval_bleu": 64.4683,
1684
+ "eval_gen_len": 28.786,
1685
+ "eval_loss": 0.4854079782962799,
1686
+ "eval_runtime": 99.2915,
1687
+ "eval_samples_per_second": 8.943,
1688
+ "eval_steps_per_second": 1.118,
1689
+ "step": 210000
1690
+ },
1691
+ {
1692
+ "epoch": 16.38,
1693
+ "grad_norm": 1.1366904973983765,
1694
+ "learning_rate": 0.0002,
1695
+ "loss": 0.3483,
1696
+ "step": 211000
1697
+ },
1698
+ {
1699
+ "epoch": 16.46,
1700
+ "grad_norm": 1.135487675666809,
1701
+ "learning_rate": 0.0002,
1702
+ "loss": 0.3509,
1703
+ "step": 212000
1704
+ },
1705
+ {
1706
+ "epoch": 16.53,
1707
+ "grad_norm": 0.9458820819854736,
1708
+ "learning_rate": 0.0002,
1709
+ "loss": 0.3519,
1710
+ "step": 213000
1711
+ },
1712
+ {
1713
+ "epoch": 16.61,
1714
+ "grad_norm": 0.8842834830284119,
1715
+ "learning_rate": 0.0002,
1716
+ "loss": 0.3551,
1717
+ "step": 214000
1718
+ },
1719
+ {
1720
+ "epoch": 16.69,
1721
+ "grad_norm": 1.2772917747497559,
1722
+ "learning_rate": 0.0002,
1723
+ "loss": 0.3572,
1724
+ "step": 215000
1725
+ },
1726
+ {
1727
+ "epoch": 16.77,
1728
+ "grad_norm": 1.5344486236572266,
1729
+ "learning_rate": 0.0002,
1730
+ "loss": 0.3567,
1731
+ "step": 216000
1732
+ },
1733
+ {
1734
+ "epoch": 16.85,
1735
+ "grad_norm": 1.4000177383422852,
1736
+ "learning_rate": 0.0002,
1737
+ "loss": 0.3605,
1738
+ "step": 217000
1739
+ },
1740
+ {
1741
+ "epoch": 16.92,
1742
+ "grad_norm": 1.5617576837539673,
1743
+ "learning_rate": 0.0002,
1744
+ "loss": 0.3577,
1745
+ "step": 218000
1746
+ },
1747
+ {
1748
+ "epoch": 17.0,
1749
+ "grad_norm": 1.4207055568695068,
1750
+ "learning_rate": 0.0002,
1751
+ "loss": 0.3581,
1752
+ "step": 219000
1753
+ },
1754
+ {
1755
+ "epoch": 17.08,
1756
+ "grad_norm": 1.6633687019348145,
1757
+ "learning_rate": 0.0002,
1758
+ "loss": 0.3243,
1759
+ "step": 220000
1760
+ },
1761
+ {
1762
+ "epoch": 17.08,
1763
+ "eval_bleu": 64.7805,
1764
+ "eval_gen_len": 28.6791,
1765
+ "eval_loss": 0.490304172039032,
1766
+ "eval_runtime": 99.3675,
1767
+ "eval_samples_per_second": 8.937,
1768
+ "eval_steps_per_second": 1.117,
1769
+ "step": 220000
1770
+ },
1771
+ {
1772
+ "epoch": 17.16,
1773
+ "grad_norm": 1.1573091745376587,
1774
+ "learning_rate": 0.0002,
1775
+ "loss": 0.3298,
1776
+ "step": 221000
1777
+ },
1778
+ {
1779
+ "epoch": 17.23,
1780
+ "grad_norm": 1.046479344367981,
1781
+ "learning_rate": 0.0002,
1782
+ "loss": 0.3312,
1783
+ "step": 222000
1784
+ },
1785
+ {
1786
+ "epoch": 17.31,
1787
+ "grad_norm": 1.2901638746261597,
1788
+ "learning_rate": 0.0002,
1789
+ "loss": 0.3325,
1790
+ "step": 223000
1791
+ },
1792
+ {
1793
+ "epoch": 17.39,
1794
+ "grad_norm": 1.0912984609603882,
1795
+ "learning_rate": 0.0002,
1796
+ "loss": 0.3351,
1797
+ "step": 224000
1798
+ },
1799
+ {
1800
+ "epoch": 17.47,
1801
+ "grad_norm": 1.5278785228729248,
1802
+ "learning_rate": 0.0002,
1803
+ "loss": 0.3382,
1804
+ "step": 225000
1805
+ },
1806
+ {
1807
+ "epoch": 17.54,
1808
+ "grad_norm": 1.013113021850586,
1809
+ "learning_rate": 0.0002,
1810
+ "loss": 0.3386,
1811
+ "step": 226000
1812
+ },
1813
+ {
1814
+ "epoch": 17.62,
1815
+ "grad_norm": 1.254299283027649,
1816
+ "learning_rate": 0.0002,
1817
+ "loss": 0.3393,
1818
+ "step": 227000
1819
+ },
1820
+ {
1821
+ "epoch": 17.7,
1822
+ "grad_norm": 0.8990402221679688,
1823
+ "learning_rate": 0.0002,
1824
+ "loss": 0.3404,
1825
+ "step": 228000
1826
+ },
1827
+ {
1828
+ "epoch": 17.78,
1829
+ "grad_norm": 1.2247493267059326,
1830
+ "learning_rate": 0.0002,
1831
+ "loss": 0.3451,
1832
+ "step": 229000
1833
+ },
1834
+ {
1835
+ "epoch": 17.85,
1836
+ "grad_norm": 1.454061508178711,
1837
+ "learning_rate": 0.0002,
1838
+ "loss": 0.3426,
1839
+ "step": 230000
1840
+ },
1841
+ {
1842
+ "epoch": 17.85,
1843
+ "eval_bleu": 64.679,
1844
+ "eval_gen_len": 28.4809,
1845
+ "eval_loss": 0.4818822741508484,
1846
+ "eval_runtime": 98.7347,
1847
+ "eval_samples_per_second": 8.994,
1848
+ "eval_steps_per_second": 1.124,
1849
+ "step": 230000
1850
+ },
1851
+ {
1852
+ "epoch": 17.93,
1853
+ "grad_norm": 1.4103410243988037,
1854
+ "learning_rate": 0.0002,
1855
+ "loss": 0.3457,
1856
+ "step": 231000
1857
+ },
1858
+ {
1859
+ "epoch": 18.01,
1860
+ "grad_norm": 1.0248557329177856,
1861
+ "learning_rate": 0.0002,
1862
+ "loss": 0.3449,
1863
+ "step": 232000
1864
+ },
1865
+ {
1866
+ "epoch": 18.09,
1867
+ "grad_norm": 1.421231985092163,
1868
+ "learning_rate": 0.0002,
1869
+ "loss": 0.3115,
1870
+ "step": 233000
1871
+ },
1872
+ {
1873
+ "epoch": 18.16,
1874
+ "grad_norm": 1.0739413499832153,
1875
+ "learning_rate": 0.0002,
1876
+ "loss": 0.314,
1877
+ "step": 234000
1878
+ },
1879
+ {
1880
+ "epoch": 18.24,
1881
+ "grad_norm": 1.0976619720458984,
1882
+ "learning_rate": 0.0002,
1883
+ "loss": 0.3214,
1884
+ "step": 235000
1885
+ },
1886
+ {
1887
+ "epoch": 18.32,
1888
+ "grad_norm": 1.480944037437439,
1889
+ "learning_rate": 0.0002,
1890
+ "loss": 0.3173,
1891
+ "step": 236000
1892
+ },
1893
+ {
1894
+ "epoch": 18.4,
1895
+ "grad_norm": 1.249569296836853,
1896
+ "learning_rate": 0.0002,
1897
+ "loss": 0.3227,
1898
+ "step": 237000
1899
+ },
1900
+ {
1901
+ "epoch": 18.48,
1902
+ "grad_norm": 1.1228398084640503,
1903
+ "learning_rate": 0.0002,
1904
+ "loss": 0.3263,
1905
+ "step": 238000
1906
+ },
1907
+ {
1908
+ "epoch": 18.55,
1909
+ "grad_norm": 1.318242073059082,
1910
+ "learning_rate": 0.0002,
1911
+ "loss": 0.327,
1912
+ "step": 239000
1913
+ },
1914
+ {
1915
+ "epoch": 18.63,
1916
+ "grad_norm": 1.1360143423080444,
1917
+ "learning_rate": 0.0002,
1918
+ "loss": 0.3295,
1919
+ "step": 240000
1920
+ },
1921
+ {
1922
+ "epoch": 18.63,
1923
+ "eval_bleu": 65.3735,
1924
+ "eval_gen_len": 28.6014,
1925
+ "eval_loss": 0.4851875603199005,
1926
+ "eval_runtime": 98.572,
1927
+ "eval_samples_per_second": 9.009,
1928
+ "eval_steps_per_second": 1.126,
1929
+ "step": 240000
1930
+ },
1931
+ {
1932
+ "epoch": 18.71,
1933
+ "grad_norm": 1.4588408470153809,
1934
+ "learning_rate": 0.0002,
1935
+ "loss": 0.3268,
1936
+ "step": 241000
1937
+ },
1938
+ {
1939
+ "epoch": 18.79,
1940
+ "grad_norm": 1.1620702743530273,
1941
+ "learning_rate": 0.0002,
1942
+ "loss": 0.3341,
1943
+ "step": 242000
1944
+ },
1945
+ {
1946
+ "epoch": 18.86,
1947
+ "grad_norm": 1.0640859603881836,
1948
+ "learning_rate": 0.0002,
1949
+ "loss": 0.3346,
1950
+ "step": 243000
1951
+ },
1952
+ {
1953
+ "epoch": 18.94,
1954
+ "grad_norm": 1.096739649772644,
1955
+ "learning_rate": 0.0002,
1956
+ "loss": 0.3334,
1957
+ "step": 244000
1958
+ },
1959
+ {
1960
+ "epoch": 19.02,
1961
+ "grad_norm": 1.23462975025177,
1962
+ "learning_rate": 0.0002,
1963
+ "loss": 0.3273,
1964
+ "step": 245000
1965
+ },
1966
+ {
1967
+ "epoch": 19.1,
1968
+ "grad_norm": 0.9026219248771667,
1969
+ "learning_rate": 0.0002,
1970
+ "loss": 0.3003,
1971
+ "step": 246000
1972
+ },
1973
+ {
1974
+ "epoch": 19.17,
1975
+ "grad_norm": 1.0630195140838623,
1976
+ "learning_rate": 0.0002,
1977
+ "loss": 0.3042,
1978
+ "step": 247000
1979
+ },
1980
+ {
1981
+ "epoch": 19.25,
1982
+ "grad_norm": 1.0800952911376953,
1983
+ "learning_rate": 0.0002,
1984
+ "loss": 0.306,
1985
+ "step": 248000
1986
+ },
1987
+ {
1988
+ "epoch": 19.33,
1989
+ "grad_norm": 1.2505557537078857,
1990
+ "learning_rate": 0.0002,
1991
+ "loss": 0.311,
1992
+ "step": 249000
1993
+ },
1994
+ {
1995
+ "epoch": 19.41,
1996
+ "grad_norm": 0.822533369064331,
1997
+ "learning_rate": 0.0002,
1998
+ "loss": 0.3124,
1999
+ "step": 250000
2000
+ },
2001
+ {
2002
+ "epoch": 19.41,
2003
+ "eval_bleu": 64.5641,
2004
+ "eval_gen_len": 28.6745,
2005
+ "eval_loss": 0.4947471022605896,
2006
+ "eval_runtime": 99.0725,
2007
+ "eval_samples_per_second": 8.963,
2008
+ "eval_steps_per_second": 1.12,
2009
+ "step": 250000
2010
+ },
2011
+ {
2012
+ "epoch": 19.48,
2013
+ "grad_norm": 1.2964988946914673,
2014
+ "learning_rate": 0.0002,
2015
+ "loss": 0.312,
2016
+ "step": 251000
2017
+ },
2018
+ {
2019
+ "epoch": 19.56,
2020
+ "grad_norm": 1.0119915008544922,
2021
+ "learning_rate": 0.0002,
2022
+ "loss": 0.3151,
2023
+ "step": 252000
2024
+ },
2025
+ {
2026
+ "epoch": 19.64,
2027
+ "grad_norm": 1.2384612560272217,
2028
+ "learning_rate": 0.0002,
2029
+ "loss": 0.3171,
2030
+ "step": 253000
2031
+ },
2032
+ {
2033
+ "epoch": 19.72,
2034
+ "grad_norm": 1.7689512968063354,
2035
+ "learning_rate": 0.0002,
2036
+ "loss": 0.3177,
2037
+ "step": 254000
2038
+ },
2039
+ {
2040
+ "epoch": 19.8,
2041
+ "grad_norm": 1.3058741092681885,
2042
+ "learning_rate": 0.0002,
2043
+ "loss": 0.3195,
2044
+ "step": 255000
2045
+ },
2046
+ {
2047
+ "epoch": 19.87,
2048
+ "grad_norm": 1.2293740510940552,
2049
+ "learning_rate": 0.0002,
2050
+ "loss": 0.3226,
2051
+ "step": 256000
2052
+ },
2053
+ {
2054
+ "epoch": 19.95,
2055
+ "grad_norm": 1.398077368736267,
2056
+ "learning_rate": 0.0002,
2057
+ "loss": 0.3222,
2058
+ "step": 257000
2059
+ },
2060
+ {
2061
+ "epoch": 20.03,
2062
+ "grad_norm": 0.9053579568862915,
2063
+ "learning_rate": 0.0002,
2064
+ "loss": 0.3122,
2065
+ "step": 258000
2066
+ },
2067
+ {
2068
+ "epoch": 20.11,
2069
+ "grad_norm": 1.2684714794158936,
2070
+ "learning_rate": 0.0002,
2071
+ "loss": 0.2891,
2072
+ "step": 259000
2073
+ },
2074
+ {
2075
+ "epoch": 20.18,
2076
+ "grad_norm": 0.9774475693702698,
2077
+ "learning_rate": 0.0002,
2078
+ "loss": 0.2933,
2079
+ "step": 260000
2080
+ },
2081
+ {
2082
+ "epoch": 20.18,
2083
+ "eval_bleu": 65.1364,
2084
+ "eval_gen_len": 28.6419,
2085
+ "eval_loss": 0.49722200632095337,
2086
+ "eval_runtime": 99.0096,
2087
+ "eval_samples_per_second": 8.969,
2088
+ "eval_steps_per_second": 1.121,
2089
+ "step": 260000
2090
+ },
2091
+ {
2092
+ "epoch": 20.26,
2093
+ "grad_norm": 1.2156912088394165,
2094
+ "learning_rate": 0.0002,
2095
+ "loss": 0.296,
2096
+ "step": 261000
2097
+ },
2098
+ {
2099
+ "epoch": 20.34,
2100
+ "grad_norm": 1.221637487411499,
2101
+ "learning_rate": 0.0002,
2102
+ "loss": 0.2981,
2103
+ "step": 262000
2104
+ },
2105
+ {
2106
+ "epoch": 20.42,
2107
+ "grad_norm": 0.9357077479362488,
2108
+ "learning_rate": 0.0002,
2109
+ "loss": 0.3018,
2110
+ "step": 263000
2111
+ },
2112
+ {
2113
+ "epoch": 20.49,
2114
+ "grad_norm": 0.9926024079322815,
2115
+ "learning_rate": 0.0002,
2116
+ "loss": 0.3012,
2117
+ "step": 264000
2118
+ },
2119
+ {
2120
+ "epoch": 20.57,
2121
+ "grad_norm": 1.6473757028579712,
2122
+ "learning_rate": 0.0002,
2123
+ "loss": 0.3049,
2124
+ "step": 265000
2125
+ },
2126
+ {
2127
+ "epoch": 20.65,
2128
+ "grad_norm": 1.1541528701782227,
2129
+ "learning_rate": 0.0002,
2130
+ "loss": 0.3039,
2131
+ "step": 266000
2132
+ },
2133
+ {
2134
+ "epoch": 20.73,
2135
+ "grad_norm": 1.220951795578003,
2136
+ "learning_rate": 0.0002,
2137
+ "loss": 0.3074,
2138
+ "step": 267000
2139
+ },
2140
+ {
2141
+ "epoch": 20.8,
2142
+ "grad_norm": 1.074318289756775,
2143
+ "learning_rate": 0.0002,
2144
+ "loss": 0.3119,
2145
+ "step": 268000
2146
+ },
2147
+ {
2148
+ "epoch": 20.88,
2149
+ "grad_norm": 1.015864372253418,
2150
+ "learning_rate": 0.0002,
2151
+ "loss": 0.3116,
2152
+ "step": 269000
2153
+ },
2154
+ {
2155
+ "epoch": 20.96,
2156
+ "grad_norm": 1.0277948379516602,
2157
+ "learning_rate": 0.0002,
2158
+ "loss": 0.3101,
2159
+ "step": 270000
2160
+ },
2161
+ {
2162
+ "epoch": 20.96,
2163
+ "eval_bleu": 64.6747,
2164
+ "eval_gen_len": 28.6802,
2165
+ "eval_loss": 0.4901565611362457,
2166
+ "eval_runtime": 100.085,
2167
+ "eval_samples_per_second": 8.872,
2168
+ "eval_steps_per_second": 1.109,
2169
+ "step": 270000
2170
+ },
2171
+ {
2172
+ "epoch": 21.04,
2173
+ "grad_norm": 0.9389250874519348,
2174
+ "learning_rate": 0.0002,
2175
+ "loss": 0.2942,
2176
+ "step": 271000
2177
+ },
2178
+ {
2179
+ "epoch": 21.11,
2180
+ "grad_norm": 1.2478715181350708,
2181
+ "learning_rate": 0.0002,
2182
+ "loss": 0.2806,
2183
+ "step": 272000
2184
+ },
2185
+ {
2186
+ "epoch": 21.19,
2187
+ "grad_norm": 0.9297951459884644,
2188
+ "learning_rate": 0.0002,
2189
+ "loss": 0.2833,
2190
+ "step": 273000
2191
+ },
2192
+ {
2193
+ "epoch": 21.27,
2194
+ "grad_norm": 0.9602841734886169,
2195
+ "learning_rate": 0.0002,
2196
+ "loss": 0.2876,
2197
+ "step": 274000
2198
+ },
2199
+ {
2200
+ "epoch": 21.35,
2201
+ "grad_norm": 0.9561505317687988,
2202
+ "learning_rate": 0.0002,
2203
+ "loss": 0.2898,
2204
+ "step": 275000
2205
+ },
2206
+ {
2207
+ "epoch": 21.43,
2208
+ "grad_norm": 1.0724116563796997,
2209
+ "learning_rate": 0.0002,
2210
+ "loss": 0.2897,
2211
+ "step": 276000
2212
+ },
2213
+ {
2214
+ "epoch": 21.5,
2215
+ "grad_norm": 0.9960470199584961,
2216
+ "learning_rate": 0.0002,
2217
+ "loss": 0.2942,
2218
+ "step": 277000
2219
+ },
2220
+ {
2221
+ "epoch": 21.58,
2222
+ "grad_norm": 1.1480662822723389,
2223
+ "learning_rate": 0.0002,
2224
+ "loss": 0.2915,
2225
+ "step": 278000
2226
+ },
2227
+ {
2228
+ "epoch": 21.66,
2229
+ "grad_norm": 1.1175373792648315,
2230
+ "learning_rate": 0.0002,
2231
+ "loss": 0.2991,
2232
+ "step": 279000
2233
+ },
2234
+ {
2235
+ "epoch": 21.74,
2236
+ "grad_norm": 1.6251972913742065,
2237
+ "learning_rate": 0.0002,
2238
+ "loss": 0.2991,
2239
+ "step": 280000
2240
+ },
2241
+ {
2242
+ "epoch": 21.74,
2243
+ "eval_bleu": 64.9732,
2244
+ "eval_gen_len": 28.5653,
2245
+ "eval_loss": 0.4907251298427582,
2246
+ "eval_runtime": 98.6604,
2247
+ "eval_samples_per_second": 9.001,
2248
+ "eval_steps_per_second": 1.125,
2249
+ "step": 280000
2250
+ },
2251
+ {
2252
+ "epoch": 21.81,
2253
+ "grad_norm": 0.8608353137969971,
2254
+ "learning_rate": 0.0002,
2255
+ "loss": 0.2994,
2256
+ "step": 281000
2257
+ },
2258
+ {
2259
+ "epoch": 21.89,
2260
+ "grad_norm": 1.077614426612854,
2261
+ "learning_rate": 0.0002,
2262
+ "loss": 0.3012,
2263
+ "step": 282000
2264
+ },
2265
+ {
2266
+ "epoch": 21.97,
2267
+ "grad_norm": 1.6897170543670654,
2268
+ "learning_rate": 0.0002,
2269
+ "loss": 0.3012,
2270
+ "step": 283000
2271
+ },
2272
+ {
2273
+ "epoch": 22.05,
2274
+ "grad_norm": 1.3074902296066284,
2275
+ "learning_rate": 0.0002,
2276
+ "loss": 0.2814,
2277
+ "step": 284000
2278
+ },
2279
+ {
2280
+ "epoch": 22.12,
2281
+ "grad_norm": 0.9641602039337158,
2282
+ "learning_rate": 0.0002,
2283
+ "loss": 0.2717,
2284
+ "step": 285000
2285
+ },
2286
+ {
2287
+ "epoch": 22.2,
2288
+ "grad_norm": 1.634346842765808,
2289
+ "learning_rate": 0.0002,
2290
+ "loss": 0.2747,
2291
+ "step": 286000
2292
+ },
2293
+ {
2294
+ "epoch": 22.28,
2295
+ "grad_norm": 1.4014965295791626,
2296
+ "learning_rate": 0.0002,
2297
+ "loss": 0.2783,
2298
+ "step": 287000
2299
+ },
2300
+ {
2301
+ "epoch": 22.36,
2302
+ "grad_norm": 1.3981196880340576,
2303
+ "learning_rate": 0.0002,
2304
+ "loss": 0.2831,
2305
+ "step": 288000
2306
+ },
2307
+ {
2308
+ "epoch": 22.43,
2309
+ "grad_norm": 0.8403178453445435,
2310
+ "learning_rate": 0.0002,
2311
+ "loss": 0.2801,
2312
+ "step": 289000
2313
+ },
2314
+ {
2315
+ "epoch": 22.51,
2316
+ "grad_norm": 0.879589319229126,
2317
+ "learning_rate": 0.0002,
2318
+ "loss": 0.2828,
2319
+ "step": 290000
2320
+ },
2321
+ {
2322
+ "epoch": 22.51,
2323
+ "eval_bleu": 64.7552,
2324
+ "eval_gen_len": 28.6261,
2325
+ "eval_loss": 0.5037782192230225,
2326
+ "eval_runtime": 98.7235,
2327
+ "eval_samples_per_second": 8.995,
2328
+ "eval_steps_per_second": 1.124,
2329
+ "step": 290000
2330
+ },
2331
+ {
2332
+ "epoch": 22.59,
2333
+ "grad_norm": 1.751582145690918,
2334
+ "learning_rate": 0.0002,
2335
+ "loss": 0.2846,
2336
+ "step": 291000
2337
+ },
2338
+ {
2339
+ "epoch": 22.67,
2340
+ "grad_norm": 1.2374579906463623,
2341
+ "learning_rate": 0.0002,
2342
+ "loss": 0.2855,
2343
+ "step": 292000
2344
+ },
2345
+ {
2346
+ "epoch": 22.74,
2347
+ "grad_norm": 1.152079463005066,
2348
+ "learning_rate": 0.0002,
2349
+ "loss": 0.2916,
2350
+ "step": 293000
2351
+ },
2352
+ {
2353
+ "epoch": 22.82,
2354
+ "grad_norm": 1.2837114334106445,
2355
+ "learning_rate": 0.0002,
2356
+ "loss": 0.2889,
2357
+ "step": 294000
2358
+ },
2359
+ {
2360
+ "epoch": 22.9,
2361
+ "grad_norm": 1.161375880241394,
2362
+ "learning_rate": 0.0002,
2363
+ "loss": 0.2894,
2364
+ "step": 295000
2365
+ },
2366
+ {
2367
+ "epoch": 22.98,
2368
+ "grad_norm": 0.8594853281974792,
2369
+ "learning_rate": 0.0002,
2370
+ "loss": 0.2936,
2371
+ "step": 296000
2372
+ },
2373
+ {
2374
+ "epoch": 23.06,
2375
+ "grad_norm": 0.9489020705223083,
2376
+ "learning_rate": 0.0002,
2377
+ "loss": 0.2733,
2378
+ "step": 297000
2379
+ },
2380
+ {
2381
+ "epoch": 23.13,
2382
+ "grad_norm": 1.2100919485092163,
2383
+ "learning_rate": 0.0002,
2384
+ "loss": 0.263,
2385
+ "step": 298000
2386
+ },
2387
+ {
2388
+ "epoch": 23.21,
2389
+ "grad_norm": 2.140540361404419,
2390
+ "learning_rate": 0.0002,
2391
+ "loss": 0.2666,
2392
+ "step": 299000
2393
+ },
2394
+ {
2395
+ "epoch": 23.29,
2396
+ "grad_norm": 1.070940375328064,
2397
+ "learning_rate": 0.0002,
2398
+ "loss": 0.2688,
2399
+ "step": 300000
2400
+ },
2401
+ {
2402
+ "epoch": 23.29,
2403
+ "eval_bleu": 65.0702,
2404
+ "eval_gen_len": 28.7534,
2405
+ "eval_loss": 0.5042341351509094,
2406
+ "eval_runtime": 99.9941,
2407
+ "eval_samples_per_second": 8.881,
2408
+ "eval_steps_per_second": 1.11,
2409
+ "step": 300000
2410
+ },
2411
+ {
2412
+ "epoch": 23.37,
2413
+ "grad_norm": 1.0847973823547363,
2414
+ "learning_rate": 0.0002,
2415
+ "loss": 0.2692,
2416
+ "step": 301000
2417
+ },
2418
+ {
2419
+ "epoch": 23.44,
2420
+ "grad_norm": 1.098399043083191,
2421
+ "learning_rate": 0.0002,
2422
+ "loss": 0.2721,
2423
+ "step": 302000
2424
+ },
2425
+ {
2426
+ "epoch": 23.52,
2427
+ "grad_norm": 0.9735555648803711,
2428
+ "learning_rate": 0.0002,
2429
+ "loss": 0.2725,
2430
+ "step": 303000
2431
+ },
2432
+ {
2433
+ "epoch": 23.6,
2434
+ "grad_norm": 1.2928968667984009,
2435
+ "learning_rate": 0.0002,
2436
+ "loss": 0.2772,
2437
+ "step": 304000
2438
+ },
2439
+ {
2440
+ "epoch": 23.68,
2441
+ "grad_norm": 1.1871669292449951,
2442
+ "learning_rate": 0.0002,
2443
+ "loss": 0.2766,
2444
+ "step": 305000
2445
+ },
2446
+ {
2447
+ "epoch": 23.75,
2448
+ "grad_norm": 0.9379162788391113,
2449
+ "learning_rate": 0.0002,
2450
+ "loss": 0.2797,
2451
+ "step": 306000
2452
+ },
2453
+ {
2454
+ "epoch": 23.83,
2455
+ "grad_norm": 0.8844149112701416,
2456
+ "learning_rate": 0.0002,
2457
+ "loss": 0.2813,
2458
+ "step": 307000
2459
+ },
2460
+ {
2461
+ "epoch": 23.91,
2462
+ "grad_norm": 1.0218191146850586,
2463
+ "learning_rate": 0.0002,
2464
+ "loss": 0.2834,
2465
+ "step": 308000
2466
+ },
2467
+ {
2468
+ "epoch": 23.99,
2469
+ "grad_norm": 1.234649896621704,
2470
+ "learning_rate": 0.0002,
2471
+ "loss": 0.2827,
2472
+ "step": 309000
2473
+ },
2474
+ {
2475
+ "epoch": 24.06,
2476
+ "grad_norm": 0.8998326659202576,
2477
+ "learning_rate": 0.0002,
2478
+ "loss": 0.2555,
2479
+ "step": 310000
2480
+ },
2481
+ {
2482
+ "epoch": 24.06,
2483
+ "eval_bleu": 65.0378,
2484
+ "eval_gen_len": 29.089,
2485
+ "eval_loss": 0.5101344585418701,
2486
+ "eval_runtime": 101.77,
2487
+ "eval_samples_per_second": 8.726,
2488
+ "eval_steps_per_second": 1.091,
2489
+ "step": 310000
2490
+ },
2491
+ {
2492
+ "epoch": 24.14,
2493
+ "grad_norm": 0.9993298053741455,
2494
+ "learning_rate": 0.0002,
2495
+ "loss": 0.2575,
2496
+ "step": 311000
2497
+ },
2498
+ {
2499
+ "epoch": 24.22,
2500
+ "grad_norm": 1.078316569328308,
2501
+ "learning_rate": 0.0002,
2502
+ "loss": 0.2577,
2503
+ "step": 312000
2504
+ },
2505
+ {
2506
+ "epoch": 24.3,
2507
+ "grad_norm": 1.0775636434555054,
2508
+ "learning_rate": 0.0002,
2509
+ "loss": 0.2603,
2510
+ "step": 313000
2511
+ },
2512
+ {
2513
+ "epoch": 24.38,
2514
+ "grad_norm": 1.0711839199066162,
2515
+ "learning_rate": 0.0002,
2516
+ "loss": 0.2641,
2517
+ "step": 314000
2518
+ },
2519
+ {
2520
+ "epoch": 24.45,
2521
+ "grad_norm": 1.1953543424606323,
2522
+ "learning_rate": 0.0002,
2523
+ "loss": 0.2666,
2524
+ "step": 315000
2525
+ },
2526
+ {
2527
+ "epoch": 24.53,
2528
+ "grad_norm": 0.7338001132011414,
2529
+ "learning_rate": 0.0002,
2530
+ "loss": 0.2662,
2531
+ "step": 316000
2532
+ },
2533
+ {
2534
+ "epoch": 24.61,
2535
+ "grad_norm": 1.651564121246338,
2536
+ "learning_rate": 0.0002,
2537
+ "loss": 0.2658,
2538
+ "step": 317000
2539
+ },
2540
+ {
2541
+ "epoch": 24.69,
2542
+ "grad_norm": 0.8356152176856995,
2543
+ "learning_rate": 0.0002,
2544
+ "loss": 0.2706,
2545
+ "step": 318000
2546
+ },
2547
+ {
2548
+ "epoch": 24.76,
2549
+ "grad_norm": 0.8503906726837158,
2550
+ "learning_rate": 0.0002,
2551
+ "loss": 0.2715,
2552
+ "step": 319000
2553
+ },
2554
+ {
2555
+ "epoch": 24.84,
2556
+ "grad_norm": 0.9122622609138489,
2557
+ "learning_rate": 0.0002,
2558
+ "loss": 0.2692,
2559
+ "step": 320000
2560
+ },
2561
+ {
2562
+ "epoch": 24.84,
2563
+ "eval_bleu": 64.9991,
2564
+ "eval_gen_len": 28.6937,
2565
+ "eval_loss": 0.5021673440933228,
2566
+ "eval_runtime": 99.2776,
2567
+ "eval_samples_per_second": 8.945,
2568
+ "eval_steps_per_second": 1.118,
2569
+ "step": 320000
2570
+ },
2571
+ {
2572
+ "epoch": 24.92,
2573
+ "grad_norm": 1.0263617038726807,
2574
+ "learning_rate": 0.0002,
2575
+ "loss": 0.2725,
2576
+ "step": 321000
2577
+ },
2578
+ {
2579
+ "epoch": 25.0,
2580
+ "grad_norm": 1.140886902809143,
2581
+ "learning_rate": 0.0002,
2582
+ "loss": 0.2748,
2583
+ "step": 322000
2584
+ },
2585
+ {
2586
+ "epoch": 25.07,
2587
+ "grad_norm": 0.9275480508804321,
2588
+ "learning_rate": 0.0002,
2589
+ "loss": 0.2467,
2590
+ "step": 323000
2591
+ },
2592
+ {
2593
+ "epoch": 25.15,
2594
+ "grad_norm": 1.170021653175354,
2595
+ "learning_rate": 0.0002,
2596
+ "loss": 0.2465,
2597
+ "step": 324000
2598
+ },
2599
+ {
2600
+ "epoch": 25.23,
2601
+ "grad_norm": 1.1251965761184692,
2602
+ "learning_rate": 0.0002,
2603
+ "loss": 0.2492,
2604
+ "step": 325000
2605
+ },
2606
+ {
2607
+ "epoch": 25.31,
2608
+ "grad_norm": 1.0885039567947388,
2609
+ "learning_rate": 0.0002,
2610
+ "loss": 0.2518,
2611
+ "step": 326000
2612
+ },
2613
+ {
2614
+ "epoch": 25.38,
2615
+ "grad_norm": 1.2162927389144897,
2616
+ "learning_rate": 0.0002,
2617
+ "loss": 0.2535,
2618
+ "step": 327000
2619
+ },
2620
+ {
2621
+ "epoch": 25.46,
2622
+ "grad_norm": 1.0869230031967163,
2623
+ "learning_rate": 0.0002,
2624
+ "loss": 0.2534,
2625
+ "step": 328000
2626
+ },
2627
+ {
2628
+ "epoch": 25.54,
2629
+ "grad_norm": 0.9775025248527527,
2630
+ "learning_rate": 0.0002,
2631
+ "loss": 0.2586,
2632
+ "step": 329000
2633
+ },
2634
+ {
2635
+ "epoch": 25.62,
2636
+ "grad_norm": 0.8145058155059814,
2637
+ "learning_rate": 0.0002,
2638
+ "loss": 0.2593,
2639
+ "step": 330000
2640
+ },
2641
+ {
2642
+ "epoch": 25.62,
2643
+ "eval_bleu": 65.2478,
2644
+ "eval_gen_len": 28.6137,
2645
+ "eval_loss": 0.508499026298523,
2646
+ "eval_runtime": 98.6428,
2647
+ "eval_samples_per_second": 9.002,
2648
+ "eval_steps_per_second": 1.125,
2649
+ "step": 330000
2650
+ },
2651
+ {
2652
+ "epoch": 25.69,
2653
+ "grad_norm": 1.270075798034668,
2654
+ "learning_rate": 0.0002,
2655
+ "loss": 0.2613,
2656
+ "step": 331000
2657
+ },
2658
+ {
2659
+ "epoch": 25.77,
2660
+ "grad_norm": 1.431252121925354,
2661
+ "learning_rate": 0.0002,
2662
+ "loss": 0.2628,
2663
+ "step": 332000
2664
+ },
2665
+ {
2666
+ "epoch": 25.85,
2667
+ "grad_norm": 1.3506394624710083,
2668
+ "learning_rate": 0.0002,
2669
+ "loss": 0.2651,
2670
+ "step": 333000
2671
+ },
2672
+ {
2673
+ "epoch": 25.93,
2674
+ "grad_norm": 1.0612725019454956,
2675
+ "learning_rate": 0.0002,
2676
+ "loss": 0.2628,
2677
+ "step": 334000
2678
+ },
2679
+ {
2680
+ "epoch": 26.01,
2681
+ "grad_norm": 0.8760356307029724,
2682
+ "learning_rate": 0.0002,
2683
+ "loss": 0.2654,
2684
+ "step": 335000
2685
+ },
2686
+ {
2687
+ "epoch": 26.08,
2688
+ "grad_norm": 1.0780360698699951,
2689
+ "learning_rate": 0.0002,
2690
+ "loss": 0.2363,
2691
+ "step": 336000
2692
+ },
2693
+ {
2694
+ "epoch": 26.16,
2695
+ "grad_norm": 1.4022656679153442,
2696
+ "learning_rate": 0.0002,
2697
+ "loss": 0.2404,
2698
+ "step": 337000
2699
+ },
2700
+ {
2701
+ "epoch": 26.24,
2702
+ "grad_norm": 1.1530039310455322,
2703
+ "learning_rate": 0.0002,
2704
+ "loss": 0.2416,
2705
+ "step": 338000
2706
+ },
2707
+ {
2708
+ "epoch": 26.32,
2709
+ "grad_norm": 1.028208613395691,
2710
+ "learning_rate": 0.0002,
2711
+ "loss": 0.2453,
2712
+ "step": 339000
2713
+ },
2714
+ {
2715
+ "epoch": 26.39,
2716
+ "grad_norm": 0.8168412446975708,
2717
+ "learning_rate": 0.0002,
2718
+ "loss": 0.2439,
2719
+ "step": 340000
2720
+ },
2721
+ {
2722
+ "epoch": 26.39,
2723
+ "eval_bleu": 64.863,
2724
+ "eval_gen_len": 28.6464,
2725
+ "eval_loss": 0.5152307748794556,
2726
+ "eval_runtime": 99.207,
2727
+ "eval_samples_per_second": 8.951,
2728
+ "eval_steps_per_second": 1.119,
2729
+ "step": 340000
2730
+ },
2731
+ {
2732
+ "epoch": 26.47,
2733
+ "grad_norm": 1.415486216545105,
2734
+ "learning_rate": 0.0002,
2735
+ "loss": 0.2481,
2736
+ "step": 341000
2737
+ },
2738
+ {
2739
+ "epoch": 26.55,
2740
+ "grad_norm": 1.016444444656372,
2741
+ "learning_rate": 0.0002,
2742
+ "loss": 0.2515,
2743
+ "step": 342000
2744
+ },
2745
+ {
2746
+ "epoch": 26.63,
2747
+ "grad_norm": 1.0151183605194092,
2748
+ "learning_rate": 0.0002,
2749
+ "loss": 0.2519,
2750
+ "step": 343000
2751
+ },
2752
+ {
2753
+ "epoch": 26.7,
2754
+ "grad_norm": 0.8286064267158508,
2755
+ "learning_rate": 0.0002,
2756
+ "loss": 0.2528,
2757
+ "step": 344000
2758
+ },
2759
+ {
2760
+ "epoch": 26.78,
2761
+ "grad_norm": 1.0916731357574463,
2762
+ "learning_rate": 0.0002,
2763
+ "loss": 0.2529,
2764
+ "step": 345000
2765
+ },
2766
+ {
2767
+ "epoch": 26.86,
2768
+ "grad_norm": 1.0001248121261597,
2769
+ "learning_rate": 0.0002,
2770
+ "loss": 0.256,
2771
+ "step": 346000
2772
+ },
2773
+ {
2774
+ "epoch": 26.94,
2775
+ "grad_norm": 0.8120971322059631,
2776
+ "learning_rate": 0.0002,
2777
+ "loss": 0.2575,
2778
+ "step": 347000
2779
+ },
2780
+ {
2781
+ "epoch": 27.01,
2782
+ "grad_norm": 0.9800658822059631,
2783
+ "learning_rate": 0.0002,
2784
+ "loss": 0.2522,
2785
+ "step": 348000
2786
+ },
2787
+ {
2788
+ "epoch": 27.09,
2789
+ "grad_norm": 1.0135070085525513,
2790
+ "learning_rate": 0.0002,
2791
+ "loss": 0.2298,
2792
+ "step": 349000
2793
+ },
2794
+ {
2795
+ "epoch": 27.17,
2796
+ "grad_norm": 1.1721863746643066,
2797
+ "learning_rate": 0.0002,
2798
+ "loss": 0.2327,
2799
+ "step": 350000
2800
+ },
2801
+ {
2802
+ "epoch": 27.17,
2803
+ "eval_bleu": 65.0748,
2804
+ "eval_gen_len": 28.7286,
2805
+ "eval_loss": 0.5164603590965271,
2806
+ "eval_runtime": 99.7391,
2807
+ "eval_samples_per_second": 8.903,
2808
+ "eval_steps_per_second": 1.113,
2809
+ "step": 350000
2810
+ },
2811
+ {
2812
+ "epoch": 27.25,
2813
+ "grad_norm": 0.7238809466362,
2814
+ "learning_rate": 0.0002,
2815
+ "loss": 0.2337,
2816
+ "step": 351000
2817
+ },
2818
+ {
2819
+ "epoch": 27.32,
2820
+ "grad_norm": 0.8267261385917664,
2821
+ "learning_rate": 0.0002,
2822
+ "loss": 0.2357,
2823
+ "step": 352000
2824
+ },
2825
+ {
2826
+ "epoch": 27.4,
2827
+ "grad_norm": 1.0274128913879395,
2828
+ "learning_rate": 0.0002,
2829
+ "loss": 0.2398,
2830
+ "step": 353000
2831
+ },
2832
+ {
2833
+ "epoch": 27.48,
2834
+ "grad_norm": 0.9916879534721375,
2835
+ "learning_rate": 0.0002,
2836
+ "loss": 0.2401,
2837
+ "step": 354000
2838
+ },
2839
+ {
2840
+ "epoch": 27.56,
2841
+ "grad_norm": 1.095639944076538,
2842
+ "learning_rate": 0.0002,
2843
+ "loss": 0.2428,
2844
+ "step": 355000
2845
+ },
2846
+ {
2847
+ "epoch": 27.64,
2848
+ "grad_norm": 0.8598717451095581,
2849
+ "learning_rate": 0.0002,
2850
+ "loss": 0.2432,
2851
+ "step": 356000
2852
+ },
2853
+ {
2854
+ "epoch": 27.71,
2855
+ "grad_norm": 0.8891191482543945,
2856
+ "learning_rate": 0.0002,
2857
+ "loss": 0.2431,
2858
+ "step": 357000
2859
+ },
2860
+ {
2861
+ "epoch": 27.79,
2862
+ "grad_norm": 0.9431182146072388,
2863
+ "learning_rate": 0.0002,
2864
+ "loss": 0.2468,
2865
+ "step": 358000
2866
+ },
2867
+ {
2868
+ "epoch": 27.87,
2869
+ "grad_norm": 1.3781706094741821,
2870
+ "learning_rate": 0.0002,
2871
+ "loss": 0.2498,
2872
+ "step": 359000
2873
+ },
2874
+ {
2875
+ "epoch": 27.95,
2876
+ "grad_norm": 0.9336220622062683,
2877
+ "learning_rate": 0.0002,
2878
+ "loss": 0.249,
2879
+ "step": 360000
2880
+ },
2881
+ {
2882
+ "epoch": 27.95,
2883
+ "eval_bleu": 64.7249,
2884
+ "eval_gen_len": 28.6137,
2885
+ "eval_loss": 0.5116418600082397,
2886
+ "eval_runtime": 98.8178,
2887
+ "eval_samples_per_second": 8.986,
2888
+ "eval_steps_per_second": 1.123,
2889
+ "step": 360000
2890
+ },
2891
+ {
2892
+ "epoch": 28.02,
2893
+ "grad_norm": 1.2862168550491333,
2894
+ "learning_rate": 0.0002,
2895
+ "loss": 0.2416,
2896
+ "step": 361000
2897
+ },
2898
+ {
2899
+ "epoch": 28.1,
2900
+ "grad_norm": 0.8687452077865601,
2901
+ "learning_rate": 0.0002,
2902
+ "loss": 0.2204,
2903
+ "step": 362000
2904
+ },
2905
+ {
2906
+ "epoch": 28.18,
2907
+ "grad_norm": 1.8673216104507446,
2908
+ "learning_rate": 0.0002,
2909
+ "loss": 0.2254,
2910
+ "step": 363000
2911
+ },
2912
+ {
2913
+ "epoch": 28.26,
2914
+ "grad_norm": 0.9244999885559082,
2915
+ "learning_rate": 0.0002,
2916
+ "loss": 0.227,
2917
+ "step": 364000
2918
+ },
2919
+ {
2920
+ "epoch": 28.33,
2921
+ "grad_norm": 0.7414880990982056,
2922
+ "learning_rate": 0.0002,
2923
+ "loss": 0.2302,
2924
+ "step": 365000
2925
+ },
2926
+ {
2927
+ "epoch": 28.41,
2928
+ "grad_norm": 1.0677781105041504,
2929
+ "learning_rate": 0.0002,
2930
+ "loss": 0.2311,
2931
+ "step": 366000
2932
+ },
2933
+ {
2934
+ "epoch": 28.49,
2935
+ "grad_norm": 1.0712281465530396,
2936
+ "learning_rate": 0.0002,
2937
+ "loss": 0.2354,
2938
+ "step": 367000
2939
+ },
2940
+ {
2941
+ "epoch": 28.57,
2942
+ "grad_norm": 1.0177695751190186,
2943
+ "learning_rate": 0.0002,
2944
+ "loss": 0.2349,
2945
+ "step": 368000
2946
+ },
2947
+ {
2948
+ "epoch": 28.64,
2949
+ "grad_norm": 1.2082629203796387,
2950
+ "learning_rate": 0.0002,
2951
+ "loss": 0.2343,
2952
+ "step": 369000
2953
+ },
2954
+ {
2955
+ "epoch": 28.72,
2956
+ "grad_norm": 0.9800160527229309,
2957
+ "learning_rate": 0.0002,
2958
+ "loss": 0.238,
2959
+ "step": 370000
2960
+ },
2961
+ {
2962
+ "epoch": 28.72,
2963
+ "eval_bleu": 64.7651,
2964
+ "eval_gen_len": 28.5968,
2965
+ "eval_loss": 0.5202394723892212,
2966
+ "eval_runtime": 98.8503,
2967
+ "eval_samples_per_second": 8.983,
2968
+ "eval_steps_per_second": 1.123,
2969
+ "step": 370000
2970
+ },
2971
+ {
2972
+ "epoch": 28.8,
2973
+ "grad_norm": 1.4668409824371338,
2974
+ "learning_rate": 0.0002,
2975
+ "loss": 0.2417,
2976
+ "step": 371000
2977
+ },
2978
+ {
2979
+ "epoch": 28.88,
2980
+ "grad_norm": 0.9679712653160095,
2981
+ "learning_rate": 0.0002,
2982
+ "loss": 0.2397,
2983
+ "step": 372000
2984
+ },
2985
+ {
2986
+ "epoch": 28.96,
2987
+ "grad_norm": 1.0757184028625488,
2988
+ "learning_rate": 0.0002,
2989
+ "loss": 0.2419,
2990
+ "step": 373000
2991
+ },
2992
+ {
2993
+ "epoch": 29.03,
2994
+ "grad_norm": 1.3961704969406128,
2995
+ "learning_rate": 0.0002,
2996
+ "loss": 0.2326,
2997
+ "step": 374000
2998
+ },
2999
+ {
3000
+ "epoch": 29.11,
3001
+ "grad_norm": 1.4827901124954224,
3002
+ "learning_rate": 0.0002,
3003
+ "loss": 0.2159,
3004
+ "step": 375000
3005
+ },
3006
+ {
3007
+ "epoch": 29.19,
3008
+ "grad_norm": 1.065645456314087,
3009
+ "learning_rate": 0.0002,
3010
+ "loss": 0.2195,
3011
+ "step": 376000
3012
+ },
3013
+ {
3014
+ "epoch": 29.27,
3015
+ "grad_norm": 0.8756958842277527,
3016
+ "learning_rate": 0.0002,
3017
+ "loss": 0.2229,
3018
+ "step": 377000
3019
+ },
3020
+ {
3021
+ "epoch": 29.34,
3022
+ "grad_norm": 1.2630327939987183,
3023
+ "learning_rate": 0.0002,
3024
+ "loss": 0.2251,
3025
+ "step": 378000
3026
+ },
3027
+ {
3028
+ "epoch": 29.42,
3029
+ "grad_norm": 0.9434683322906494,
3030
+ "learning_rate": 0.0002,
3031
+ "loss": 0.2267,
3032
+ "step": 379000
3033
+ },
3034
+ {
3035
+ "epoch": 29.5,
3036
+ "grad_norm": 0.8589434623718262,
3037
+ "learning_rate": 0.0002,
3038
+ "loss": 0.2297,
3039
+ "step": 380000
3040
+ },
3041
+ {
3042
+ "epoch": 29.5,
3043
+ "eval_bleu": 65.3334,
3044
+ "eval_gen_len": 28.7005,
3045
+ "eval_loss": 0.5242559909820557,
3046
+ "eval_runtime": 99.2088,
3047
+ "eval_samples_per_second": 8.951,
3048
+ "eval_steps_per_second": 1.119,
3049
+ "step": 380000
3050
+ },
3051
+ {
3052
+ "epoch": 29.58,
3053
+ "grad_norm": 1.0252753496170044,
3054
+ "learning_rate": 0.0002,
3055
+ "loss": 0.229,
3056
+ "step": 381000
3057
+ },
3058
+ {
3059
+ "epoch": 29.65,
3060
+ "grad_norm": 1.4881134033203125,
3061
+ "learning_rate": 0.0002,
3062
+ "loss": 0.2301,
3063
+ "step": 382000
3064
+ },
3065
+ {
3066
+ "epoch": 29.73,
3067
+ "grad_norm": 1.0281462669372559,
3068
+ "learning_rate": 0.0002,
3069
+ "loss": 0.2307,
3070
+ "step": 383000
3071
+ },
3072
+ {
3073
+ "epoch": 29.81,
3074
+ "grad_norm": 1.1244617700576782,
3075
+ "learning_rate": 0.0002,
3076
+ "loss": 0.2335,
3077
+ "step": 384000
3078
+ },
3079
+ {
3080
+ "epoch": 29.89,
3081
+ "grad_norm": 1.1461416482925415,
3082
+ "learning_rate": 0.0002,
3083
+ "loss": 0.2355,
3084
+ "step": 385000
3085
+ },
3086
+ {
3087
+ "epoch": 29.96,
3088
+ "grad_norm": 1.742311716079712,
3089
+ "learning_rate": 0.0002,
3090
+ "loss": 0.2341,
3091
+ "step": 386000
3092
+ },
3093
+ {
3094
+ "epoch": 30.04,
3095
+ "grad_norm": 0.8539097309112549,
3096
+ "learning_rate": 0.0002,
3097
+ "loss": 0.2196,
3098
+ "step": 387000
3099
+ },
3100
+ {
3101
+ "epoch": 30.12,
3102
+ "grad_norm": 0.9865394830703735,
3103
+ "learning_rate": 0.0002,
3104
+ "loss": 0.2118,
3105
+ "step": 388000
3106
+ },
3107
+ {
3108
+ "epoch": 30.2,
3109
+ "grad_norm": 1.2487947940826416,
3110
+ "learning_rate": 0.0002,
3111
+ "loss": 0.2111,
3112
+ "step": 389000
3113
+ },
3114
+ {
3115
+ "epoch": 30.27,
3116
+ "grad_norm": 0.9401417970657349,
3117
+ "learning_rate": 0.0002,
3118
+ "loss": 0.2152,
3119
+ "step": 390000
3120
+ },
3121
+ {
3122
+ "epoch": 30.27,
3123
+ "eval_bleu": 64.9364,
3124
+ "eval_gen_len": 28.6081,
3125
+ "eval_loss": 0.533649206161499,
3126
+ "eval_runtime": 99.1874,
3127
+ "eval_samples_per_second": 8.953,
3128
+ "eval_steps_per_second": 1.119,
3129
+ "step": 390000
3130
+ },
3131
+ {
3132
+ "epoch": 30.35,
3133
+ "grad_norm": 1.5141676664352417,
3134
+ "learning_rate": 0.0002,
3135
+ "loss": 0.2185,
3136
+ "step": 391000
3137
+ },
3138
+ {
3139
+ "epoch": 30.43,
3140
+ "grad_norm": 1.4947956800460815,
3141
+ "learning_rate": 0.0002,
3142
+ "loss": 0.221,
3143
+ "step": 392000
3144
+ },
3145
+ {
3146
+ "epoch": 30.51,
3147
+ "grad_norm": 0.8870178461074829,
3148
+ "learning_rate": 0.0002,
3149
+ "loss": 0.2221,
3150
+ "step": 393000
3151
+ },
3152
+ {
3153
+ "epoch": 30.59,
3154
+ "grad_norm": 1.013377070426941,
3155
+ "learning_rate": 0.0002,
3156
+ "loss": 0.2223,
3157
+ "step": 394000
3158
+ },
3159
+ {
3160
+ "epoch": 30.66,
3161
+ "grad_norm": 1.2745546102523804,
3162
+ "learning_rate": 0.0002,
3163
+ "loss": 0.2242,
3164
+ "step": 395000
3165
+ },
3166
+ {
3167
+ "epoch": 30.74,
3168
+ "grad_norm": 1.3159047365188599,
3169
+ "learning_rate": 0.0002,
3170
+ "loss": 0.2286,
3171
+ "step": 396000
3172
+ },
3173
+ {
3174
+ "epoch": 30.82,
3175
+ "grad_norm": 0.8441556096076965,
3176
+ "learning_rate": 0.0002,
3177
+ "loss": 0.2269,
3178
+ "step": 397000
3179
+ },
3180
+ {
3181
+ "epoch": 30.9,
3182
+ "grad_norm": 1.0391247272491455,
3183
+ "learning_rate": 0.0002,
3184
+ "loss": 0.2297,
3185
+ "step": 398000
3186
+ },
3187
+ {
3188
+ "epoch": 30.97,
3189
+ "grad_norm": 1.0133869647979736,
3190
+ "learning_rate": 0.0002,
3191
+ "loss": 0.2277,
3192
+ "step": 399000
3193
+ },
3194
+ {
3195
+ "epoch": 31.05,
3196
+ "grad_norm": 1.5093469619750977,
3197
+ "learning_rate": 0.0002,
3198
+ "loss": 0.2106,
3199
+ "step": 400000
3200
+ },
3201
+ {
3202
+ "epoch": 31.05,
3203
+ "eval_bleu": 65.117,
3204
+ "eval_gen_len": 28.6745,
3205
+ "eval_loss": 0.540839433670044,
3206
+ "eval_runtime": 98.9473,
3207
+ "eval_samples_per_second": 8.974,
3208
+ "eval_steps_per_second": 1.122,
3209
+ "step": 400000
3210
+ },
3211
+ {
3212
+ "epoch": 31.13,
3213
+ "grad_norm": 1.0606015920639038,
3214
+ "learning_rate": 0.0002,
3215
+ "loss": 0.2031,
3216
+ "step": 401000
3217
+ },
3218
+ {
3219
+ "epoch": 31.21,
3220
+ "grad_norm": 1.4048112630844116,
3221
+ "learning_rate": 0.0002,
3222
+ "loss": 0.2043,
3223
+ "step": 402000
3224
+ },
3225
+ {
3226
+ "epoch": 31.28,
3227
+ "grad_norm": 1.1232408285140991,
3228
+ "learning_rate": 0.0002,
3229
+ "loss": 0.211,
3230
+ "step": 403000
3231
+ },
3232
+ {
3233
+ "epoch": 31.36,
3234
+ "grad_norm": 1.2367199659347534,
3235
+ "learning_rate": 0.0002,
3236
+ "loss": 0.2107,
3237
+ "step": 404000
3238
+ },
3239
+ {
3240
+ "epoch": 31.44,
3241
+ "grad_norm": 1.1147772073745728,
3242
+ "learning_rate": 0.0002,
3243
+ "loss": 0.2116,
3244
+ "step": 405000
3245
+ },
3246
+ {
3247
+ "epoch": 31.52,
3248
+ "grad_norm": 0.9711781740188599,
3249
+ "learning_rate": 0.0002,
3250
+ "loss": 0.2147,
3251
+ "step": 406000
3252
+ },
3253
+ {
3254
+ "epoch": 31.59,
3255
+ "grad_norm": 1.4205774068832397,
3256
+ "learning_rate": 0.0002,
3257
+ "loss": 0.2158,
3258
+ "step": 407000
3259
+ },
3260
+ {
3261
+ "epoch": 31.67,
3262
+ "grad_norm": 1.303250789642334,
3263
+ "learning_rate": 0.0002,
3264
+ "loss": 0.2187,
3265
+ "step": 408000
3266
+ },
3267
+ {
3268
+ "epoch": 31.75,
3269
+ "grad_norm": 2.3327102661132812,
3270
+ "learning_rate": 0.0002,
3271
+ "loss": 0.2196,
3272
+ "step": 409000
3273
+ },
3274
+ {
3275
+ "epoch": 31.83,
3276
+ "grad_norm": 1.9003146886825562,
3277
+ "learning_rate": 0.0002,
3278
+ "loss": 0.2234,
3279
+ "step": 410000
3280
+ },
3281
+ {
3282
+ "epoch": 31.83,
3283
+ "eval_bleu": 64.8926,
3284
+ "eval_gen_len": 28.6318,
3285
+ "eval_loss": 0.5249429941177368,
3286
+ "eval_runtime": 98.7286,
3287
+ "eval_samples_per_second": 8.994,
3288
+ "eval_steps_per_second": 1.124,
3289
+ "step": 410000
3290
+ },
3291
+ {
3292
+ "epoch": 31.9,
3293
+ "grad_norm": 0.9950889348983765,
3294
+ "learning_rate": 0.0002,
3295
+ "loss": 0.2232,
3296
+ "step": 411000
3297
+ },
3298
+ {
3299
+ "epoch": 31.98,
3300
+ "grad_norm": 0.8693845272064209,
3301
+ "learning_rate": 0.0002,
3302
+ "loss": 0.2236,
3303
+ "step": 412000
3304
+ },
3305
+ {
3306
+ "epoch": 32.06,
3307
+ "grad_norm": 0.9227551817893982,
3308
+ "learning_rate": 0.0002,
3309
+ "loss": 0.204,
3310
+ "step": 413000
3311
+ },
3312
+ {
3313
+ "epoch": 32.14,
3314
+ "grad_norm": 1.0269570350646973,
3315
+ "learning_rate": 0.0002,
3316
+ "loss": 0.2019,
3317
+ "step": 414000
3318
+ },
3319
+ {
3320
+ "epoch": 32.22,
3321
+ "grad_norm": 1.0199569463729858,
3322
+ "learning_rate": 0.0002,
3323
+ "loss": 0.2015,
3324
+ "step": 415000
3325
+ },
3326
+ {
3327
+ "epoch": 32.29,
3328
+ "grad_norm": 1.4488086700439453,
3329
+ "learning_rate": 0.0002,
3330
+ "loss": 0.2036,
3331
+ "step": 416000
3332
+ },
3333
+ {
3334
+ "epoch": 32.37,
3335
+ "grad_norm": 0.8843773007392883,
3336
+ "learning_rate": 0.0002,
3337
+ "loss": 0.2049,
3338
+ "step": 417000
3339
+ },
3340
+ {
3341
+ "epoch": 32.45,
3342
+ "grad_norm": 1.3630881309509277,
3343
+ "learning_rate": 0.0002,
3344
+ "loss": 0.2085,
3345
+ "step": 418000
3346
+ },
3347
+ {
3348
+ "epoch": 32.53,
3349
+ "grad_norm": 0.9767336845397949,
3350
+ "learning_rate": 0.0002,
3351
+ "loss": 0.2097,
3352
+ "step": 419000
3353
+ },
3354
+ {
3355
+ "epoch": 32.6,
3356
+ "grad_norm": 0.9147652983665466,
3357
+ "learning_rate": 0.0002,
3358
+ "loss": 0.2085,
3359
+ "step": 420000
3360
+ },
3361
+ {
3362
+ "epoch": 32.6,
3363
+ "eval_bleu": 65.5715,
3364
+ "eval_gen_len": 28.7984,
3365
+ "eval_loss": 0.5305626392364502,
3366
+ "eval_runtime": 100.0595,
3367
+ "eval_samples_per_second": 8.875,
3368
+ "eval_steps_per_second": 1.109,
3369
+ "step": 420000
3370
+ },
3371
+ {
3372
+ "epoch": 32.68,
3373
+ "grad_norm": 1.4235540628433228,
3374
+ "learning_rate": 0.0002,
3375
+ "loss": 0.212,
3376
+ "step": 421000
3377
+ },
3378
+ {
3379
+ "epoch": 32.76,
3380
+ "grad_norm": 0.9653807282447815,
3381
+ "learning_rate": 0.0002,
3382
+ "loss": 0.2129,
3383
+ "step": 422000
3384
+ },
3385
+ {
3386
+ "epoch": 32.84,
3387
+ "grad_norm": 1.0437246561050415,
3388
+ "learning_rate": 0.0002,
3389
+ "loss": 0.2153,
3390
+ "step": 423000
3391
+ },
3392
+ {
3393
+ "epoch": 32.91,
3394
+ "grad_norm": 1.0093231201171875,
3395
+ "learning_rate": 0.0002,
3396
+ "loss": 0.2146,
3397
+ "step": 424000
3398
+ },
3399
+ {
3400
+ "epoch": 32.99,
3401
+ "grad_norm": 0.9372303485870361,
3402
+ "learning_rate": 0.0002,
3403
+ "loss": 0.2176,
3404
+ "step": 425000
3405
+ },
3406
+ {
3407
+ "epoch": 33.07,
3408
+ "grad_norm": 0.990990161895752,
3409
+ "learning_rate": 0.0002,
3410
+ "loss": 0.1946,
3411
+ "step": 426000
3412
+ },
3413
+ {
3414
+ "epoch": 33.15,
3415
+ "grad_norm": 1.221752405166626,
3416
+ "learning_rate": 0.0002,
3417
+ "loss": 0.1937,
3418
+ "step": 427000
3419
+ },
3420
+ {
3421
+ "epoch": 33.22,
3422
+ "grad_norm": 1.0376135110855103,
3423
+ "learning_rate": 0.0002,
3424
+ "loss": 0.1971,
3425
+ "step": 428000
3426
+ },
3427
+ {
3428
+ "epoch": 33.3,
3429
+ "grad_norm": 1.2878087759017944,
3430
+ "learning_rate": 0.0002,
3431
+ "loss": 0.1993,
3432
+ "step": 429000
3433
+ },
3434
+ {
3435
+ "epoch": 33.38,
3436
+ "grad_norm": 1.702043890953064,
3437
+ "learning_rate": 0.0002,
3438
+ "loss": 0.2018,
3439
+ "step": 430000
3440
+ },
3441
+ {
3442
+ "epoch": 33.38,
3443
+ "eval_bleu": 64.9154,
3444
+ "eval_gen_len": 28.6351,
3445
+ "eval_loss": 0.5428734421730042,
3446
+ "eval_runtime": 99.9626,
3447
+ "eval_samples_per_second": 8.883,
3448
+ "eval_steps_per_second": 1.11,
3449
+ "step": 430000
3450
+ },
3451
+ {
3452
+ "epoch": 33.46,
3453
+ "grad_norm": 1.171934723854065,
3454
+ "learning_rate": 0.0002,
3455
+ "loss": 0.2042,
3456
+ "step": 431000
3457
+ },
3458
+ {
3459
+ "epoch": 33.54,
3460
+ "grad_norm": 0.9023895859718323,
3461
+ "learning_rate": 0.0002,
3462
+ "loss": 0.2032,
3463
+ "step": 432000
3464
+ },
3465
+ {
3466
+ "epoch": 33.61,
3467
+ "grad_norm": 1.5410844087600708,
3468
+ "learning_rate": 0.0002,
3469
+ "loss": 0.204,
3470
+ "step": 433000
3471
+ },
3472
+ {
3473
+ "epoch": 33.69,
3474
+ "grad_norm": 1.297434687614441,
3475
+ "learning_rate": 0.0002,
3476
+ "loss": 0.2057,
3477
+ "step": 434000
3478
+ },
3479
+ {
3480
+ "epoch": 33.77,
3481
+ "grad_norm": 1.636635422706604,
3482
+ "learning_rate": 0.0002,
3483
+ "loss": 0.2085,
3484
+ "step": 435000
3485
+ },
3486
+ {
3487
+ "epoch": 33.85,
3488
+ "grad_norm": 1.3059121370315552,
3489
+ "learning_rate": 0.0002,
3490
+ "loss": 0.2099,
3491
+ "step": 436000
3492
+ },
3493
+ {
3494
+ "epoch": 33.92,
3495
+ "grad_norm": 1.1616836786270142,
3496
+ "learning_rate": 0.0002,
3497
+ "loss": 0.2098,
3498
+ "step": 437000
3499
+ },
3500
+ {
3501
+ "epoch": 34.0,
3502
+ "grad_norm": 0.9708386063575745,
3503
+ "learning_rate": 0.0002,
3504
+ "loss": 0.2103,
3505
+ "step": 438000
3506
+ },
3507
+ {
3508
+ "epoch": 34.08,
3509
+ "grad_norm": 1.1958973407745361,
3510
+ "learning_rate": 0.0002,
3511
+ "loss": 0.1868,
3512
+ "step": 439000
3513
+ },
3514
+ {
3515
+ "epoch": 34.16,
3516
+ "grad_norm": 0.9669882655143738,
3517
+ "learning_rate": 0.0002,
3518
+ "loss": 0.1885,
3519
+ "step": 440000
3520
+ },
3521
+ {
3522
+ "epoch": 34.16,
3523
+ "eval_bleu": 65.0538,
3524
+ "eval_gen_len": 28.8525,
3525
+ "eval_loss": 0.5453199148178101,
3526
+ "eval_runtime": 98.9637,
3527
+ "eval_samples_per_second": 8.973,
3528
+ "eval_steps_per_second": 1.122,
3529
+ "step": 440000
3530
+ },
3531
+ {
3532
+ "epoch": 34.23,
3533
+ "grad_norm": 1.3960009813308716,
3534
+ "learning_rate": 0.0002,
3535
+ "loss": 0.192,
3536
+ "step": 441000
3537
+ },
3538
+ {
3539
+ "epoch": 34.31,
3540
+ "grad_norm": 1.1039202213287354,
3541
+ "learning_rate": 0.0002,
3542
+ "loss": 0.1928,
3543
+ "step": 442000
3544
+ },
3545
+ {
3546
+ "epoch": 34.39,
3547
+ "grad_norm": 1.4681973457336426,
3548
+ "learning_rate": 0.0002,
3549
+ "loss": 0.1963,
3550
+ "step": 443000
3551
+ },
3552
+ {
3553
+ "epoch": 34.47,
3554
+ "grad_norm": 1.1876535415649414,
3555
+ "learning_rate": 0.0002,
3556
+ "loss": 0.1955,
3557
+ "step": 444000
3558
+ },
3559
+ {
3560
+ "epoch": 34.54,
3561
+ "grad_norm": 1.0030099153518677,
3562
+ "learning_rate": 0.0002,
3563
+ "loss": 0.197,
3564
+ "step": 445000
3565
+ },
3566
+ {
3567
+ "epoch": 34.62,
3568
+ "grad_norm": 1.262609839439392,
3569
+ "learning_rate": 0.0002,
3570
+ "loss": 0.1965,
3571
+ "step": 446000
3572
+ },
3573
+ {
3574
+ "epoch": 34.7,
3575
+ "grad_norm": 4.133481979370117,
3576
+ "learning_rate": 0.0002,
3577
+ "loss": 0.2009,
3578
+ "step": 447000
3579
+ },
3580
+ {
3581
+ "epoch": 34.78,
3582
+ "grad_norm": 1.3214054107666016,
3583
+ "learning_rate": 0.0002,
3584
+ "loss": 0.2011,
3585
+ "step": 448000
3586
+ },
3587
+ {
3588
+ "epoch": 34.85,
3589
+ "grad_norm": 1.061333417892456,
3590
+ "learning_rate": 0.0002,
3591
+ "loss": 0.205,
3592
+ "step": 449000
3593
+ },
3594
+ {
3595
+ "epoch": 34.93,
3596
+ "grad_norm": 1.487025260925293,
3597
+ "learning_rate": 0.0002,
3598
+ "loss": 0.2049,
3599
+ "step": 450000
3600
+ },
3601
+ {
3602
+ "epoch": 34.93,
3603
+ "eval_bleu": 65.2857,
3604
+ "eval_gen_len": 28.7207,
3605
+ "eval_loss": 0.5434128046035767,
3606
+ "eval_runtime": 98.7363,
3607
+ "eval_samples_per_second": 8.994,
3608
+ "eval_steps_per_second": 1.124,
3609
+ "step": 450000
3610
+ },
3611
+ {
3612
+ "epoch": 35.01,
3613
+ "grad_norm": 1.3061411380767822,
3614
+ "learning_rate": 0.0002,
3615
+ "loss": 0.2042,
3616
+ "step": 451000
3617
+ },
3618
+ {
3619
+ "epoch": 35.09,
3620
+ "grad_norm": 0.9900358319282532,
3621
+ "learning_rate": 0.0002,
3622
+ "loss": 0.1803,
3623
+ "step": 452000
3624
+ },
3625
+ {
3626
+ "epoch": 35.17,
3627
+ "grad_norm": 1.2118251323699951,
3628
+ "learning_rate": 0.0002,
3629
+ "loss": 0.183,
3630
+ "step": 453000
3631
+ },
3632
+ {
3633
+ "epoch": 35.24,
3634
+ "grad_norm": 1.1625529527664185,
3635
+ "learning_rate": 0.0002,
3636
+ "loss": 0.1879,
3637
+ "step": 454000
3638
+ },
3639
+ {
3640
+ "epoch": 35.32,
3641
+ "grad_norm": 1.0669846534729004,
3642
+ "learning_rate": 0.0002,
3643
+ "loss": 0.1888,
3644
+ "step": 455000
3645
+ },
3646
+ {
3647
+ "epoch": 35.4,
3648
+ "grad_norm": 1.285409688949585,
3649
+ "learning_rate": 0.0002,
3650
+ "loss": 0.1908,
3651
+ "step": 456000
3652
+ },
3653
+ {
3654
+ "epoch": 35.48,
3655
+ "grad_norm": 1.292738914489746,
3656
+ "learning_rate": 0.0002,
3657
+ "loss": 0.1944,
3658
+ "step": 457000
3659
+ },
3660
+ {
3661
+ "epoch": 35.55,
3662
+ "grad_norm": 0.9169420599937439,
3663
+ "learning_rate": 0.0002,
3664
+ "loss": 0.1924,
3665
+ "step": 458000
3666
+ },
3667
+ {
3668
+ "epoch": 35.63,
3669
+ "grad_norm": 1.1117466688156128,
3670
+ "learning_rate": 0.0002,
3671
+ "loss": 0.1944,
3672
+ "step": 459000
3673
+ },
3674
+ {
3675
+ "epoch": 35.71,
3676
+ "grad_norm": 1.400664210319519,
3677
+ "learning_rate": 0.0002,
3678
+ "loss": 0.1957,
3679
+ "step": 460000
3680
+ },
3681
+ {
3682
+ "epoch": 35.71,
3683
+ "eval_bleu": 65.3436,
3684
+ "eval_gen_len": 28.714,
3685
+ "eval_loss": 0.549137532711029,
3686
+ "eval_runtime": 99.1079,
3687
+ "eval_samples_per_second": 8.96,
3688
+ "eval_steps_per_second": 1.12,
3689
+ "step": 460000
3690
+ },
3691
+ {
3692
+ "epoch": 35.79,
3693
+ "grad_norm": 1.1465002298355103,
3694
+ "learning_rate": 0.0002,
3695
+ "loss": 0.1974,
3696
+ "step": 461000
3697
+ },
3698
+ {
3699
+ "epoch": 35.86,
3700
+ "grad_norm": 0.9425164461135864,
3701
+ "learning_rate": 0.0002,
3702
+ "loss": 0.1967,
3703
+ "step": 462000
3704
+ },
3705
+ {
3706
+ "epoch": 35.94,
3707
+ "grad_norm": 1.0649182796478271,
3708
+ "learning_rate": 0.0002,
3709
+ "loss": 0.1974,
3710
+ "step": 463000
3711
+ },
3712
+ {
3713
+ "epoch": 36.02,
3714
+ "grad_norm": 0.9610468149185181,
3715
+ "learning_rate": 0.0002,
3716
+ "loss": 0.1943,
3717
+ "step": 464000
3718
+ },
3719
+ {
3720
+ "epoch": 36.1,
3721
+ "grad_norm": 1.0697602033615112,
3722
+ "learning_rate": 0.0002,
3723
+ "loss": 0.1785,
3724
+ "step": 465000
3725
+ },
3726
+ {
3727
+ "epoch": 36.17,
3728
+ "grad_norm": 0.8167102336883545,
3729
+ "learning_rate": 0.0002,
3730
+ "loss": 0.181,
3731
+ "step": 466000
3732
+ },
3733
+ {
3734
+ "epoch": 36.25,
3735
+ "grad_norm": 1.155148983001709,
3736
+ "learning_rate": 0.0002,
3737
+ "loss": 0.1804,
3738
+ "step": 467000
3739
+ },
3740
+ {
3741
+ "epoch": 36.33,
3742
+ "grad_norm": 1.036157250404358,
3743
+ "learning_rate": 0.0002,
3744
+ "loss": 0.1811,
3745
+ "step": 468000
3746
+ },
3747
+ {
3748
+ "epoch": 36.41,
3749
+ "grad_norm": 0.9966660141944885,
3750
+ "learning_rate": 0.0002,
3751
+ "loss": 0.1825,
3752
+ "step": 469000
3753
+ },
3754
+ {
3755
+ "epoch": 36.49,
3756
+ "grad_norm": 1.3554514646530151,
3757
+ "learning_rate": 0.0002,
3758
+ "loss": 0.1867,
3759
+ "step": 470000
3760
+ },
3761
+ {
3762
+ "epoch": 36.49,
3763
+ "eval_bleu": 65.4934,
3764
+ "eval_gen_len": 28.7939,
3765
+ "eval_loss": 0.5535929203033447,
3766
+ "eval_runtime": 99.8585,
3767
+ "eval_samples_per_second": 8.893,
3768
+ "eval_steps_per_second": 1.112,
3769
+ "step": 470000
3770
+ },
3771
+ {
3772
+ "epoch": 36.56,
3773
+ "grad_norm": 1.1400065422058105,
3774
+ "learning_rate": 0.0002,
3775
+ "loss": 0.1889,
3776
+ "step": 471000
3777
+ },
3778
+ {
3779
+ "epoch": 36.64,
3780
+ "grad_norm": 1.2936526536941528,
3781
+ "learning_rate": 0.0002,
3782
+ "loss": 0.1892,
3783
+ "step": 472000
3784
+ },
3785
+ {
3786
+ "epoch": 36.72,
3787
+ "grad_norm": 1.3375158309936523,
3788
+ "learning_rate": 0.0002,
3789
+ "loss": 0.1932,
3790
+ "step": 473000
3791
+ },
3792
+ {
3793
+ "epoch": 36.8,
3794
+ "grad_norm": 1.3976365327835083,
3795
+ "learning_rate": 0.0002,
3796
+ "loss": 0.1931,
3797
+ "step": 474000
3798
+ },
3799
+ {
3800
+ "epoch": 36.87,
3801
+ "grad_norm": 1.2075397968292236,
3802
+ "learning_rate": 0.0002,
3803
+ "loss": 0.1938,
3804
+ "step": 475000
3805
+ },
3806
+ {
3807
+ "epoch": 36.95,
3808
+ "grad_norm": 1.2333601713180542,
3809
+ "learning_rate": 0.0002,
3810
+ "loss": 0.1918,
3811
+ "step": 476000
3812
+ },
3813
+ {
3814
+ "epoch": 37.03,
3815
+ "grad_norm": 0.9724763631820679,
3816
+ "learning_rate": 0.0002,
3817
+ "loss": 0.1857,
3818
+ "step": 477000
3819
+ },
3820
+ {
3821
+ "epoch": 37.11,
3822
+ "grad_norm": 1.305141568183899,
3823
+ "learning_rate": 0.0002,
3824
+ "loss": 0.1741,
3825
+ "step": 478000
3826
+ },
3827
+ {
3828
+ "epoch": 37.18,
3829
+ "grad_norm": 1.2358112335205078,
3830
+ "learning_rate": 0.0002,
3831
+ "loss": 0.1727,
3832
+ "step": 479000
3833
+ },
3834
+ {
3835
+ "epoch": 37.26,
3836
+ "grad_norm": 1.040460228919983,
3837
+ "learning_rate": 0.0002,
3838
+ "loss": 0.1765,
3839
+ "step": 480000
3840
+ },
3841
+ {
3842
+ "epoch": 37.26,
3843
+ "eval_bleu": 65.5595,
3844
+ "eval_gen_len": 28.8255,
3845
+ "eval_loss": 0.5582976341247559,
3846
+ "eval_runtime": 99.6921,
3847
+ "eval_samples_per_second": 8.907,
3848
+ "eval_steps_per_second": 1.113,
3849
+ "step": 480000
3850
+ },
3851
+ {
3852
+ "epoch": 37.34,
3853
+ "grad_norm": 1.0049262046813965,
3854
+ "learning_rate": 0.0002,
3855
+ "loss": 0.1801,
3856
+ "step": 481000
3857
+ },
3858
+ {
3859
+ "epoch": 37.42,
3860
+ "grad_norm": 0.9716454148292542,
3861
+ "learning_rate": 0.0002,
3862
+ "loss": 0.1806,
3863
+ "step": 482000
3864
+ },
3865
+ {
3866
+ "epoch": 37.49,
3867
+ "grad_norm": 1.2684077024459839,
3868
+ "learning_rate": 0.0002,
3869
+ "loss": 0.1809,
3870
+ "step": 483000
3871
+ },
3872
+ {
3873
+ "epoch": 37.57,
3874
+ "grad_norm": 1.4772919416427612,
3875
+ "learning_rate": 0.0002,
3876
+ "loss": 0.1798,
3877
+ "step": 484000
3878
+ },
3879
+ {
3880
+ "epoch": 37.65,
3881
+ "grad_norm": 0.8240026831626892,
3882
+ "learning_rate": 0.0002,
3883
+ "loss": 0.1849,
3884
+ "step": 485000
3885
+ },
3886
+ {
3887
+ "epoch": 37.73,
3888
+ "grad_norm": 1.2247587442398071,
3889
+ "learning_rate": 0.0002,
3890
+ "loss": 0.1872,
3891
+ "step": 486000
3892
+ },
3893
+ {
3894
+ "epoch": 37.8,
3895
+ "grad_norm": 1.4645825624465942,
3896
+ "learning_rate": 0.0002,
3897
+ "loss": 0.1889,
3898
+ "step": 487000
3899
+ },
3900
+ {
3901
+ "epoch": 37.88,
3902
+ "grad_norm": 1.0552102327346802,
3903
+ "learning_rate": 0.0002,
3904
+ "loss": 0.1866,
3905
+ "step": 488000
3906
+ },
3907
+ {
3908
+ "epoch": 37.96,
3909
+ "grad_norm": 1.2899285554885864,
3910
+ "learning_rate": 0.0002,
3911
+ "loss": 0.1897,
3912
+ "step": 489000
3913
+ },
3914
+ {
3915
+ "epoch": 38.04,
3916
+ "grad_norm": 1.0461792945861816,
3917
+ "learning_rate": 0.0002,
3918
+ "loss": 0.1786,
3919
+ "step": 490000
3920
+ },
3921
+ {
3922
+ "epoch": 38.04,
3923
+ "eval_bleu": 65.6358,
3924
+ "eval_gen_len": 28.7691,
3925
+ "eval_loss": 0.5611980557441711,
3926
+ "eval_runtime": 99.4338,
3927
+ "eval_samples_per_second": 8.931,
3928
+ "eval_steps_per_second": 1.116,
3929
+ "step": 490000
3930
+ },
3931
+ {
3932
+ "epoch": 38.12,
3933
+ "grad_norm": 1.1956135034561157,
3934
+ "learning_rate": 0.0002,
3935
+ "loss": 0.1708,
3936
+ "step": 491000
3937
+ },
3938
+ {
3939
+ "epoch": 38.19,
3940
+ "grad_norm": 1.903419852256775,
3941
+ "learning_rate": 0.0002,
3942
+ "loss": 0.1726,
3943
+ "step": 492000
3944
+ },
3945
+ {
3946
+ "epoch": 38.27,
3947
+ "grad_norm": 1.4714049100875854,
3948
+ "learning_rate": 0.0002,
3949
+ "loss": 0.174,
3950
+ "step": 493000
3951
+ },
3952
+ {
3953
+ "epoch": 38.35,
3954
+ "grad_norm": 1.117650032043457,
3955
+ "learning_rate": 0.0002,
3956
+ "loss": 0.1753,
3957
+ "step": 494000
3958
+ },
3959
+ {
3960
+ "epoch": 38.43,
3961
+ "grad_norm": 0.9286689162254333,
3962
+ "learning_rate": 0.0002,
3963
+ "loss": 0.1766,
3964
+ "step": 495000
3965
+ },
3966
+ {
3967
+ "epoch": 38.5,
3968
+ "grad_norm": 1.0359840393066406,
3969
+ "learning_rate": 0.0002,
3970
+ "loss": 0.1774,
3971
+ "step": 496000
3972
+ },
3973
+ {
3974
+ "epoch": 38.58,
3975
+ "grad_norm": 0.9324952363967896,
3976
+ "learning_rate": 0.0002,
3977
+ "loss": 0.1795,
3978
+ "step": 497000
3979
+ },
3980
+ {
3981
+ "epoch": 38.66,
3982
+ "grad_norm": 1.2552545070648193,
3983
+ "learning_rate": 0.0002,
3984
+ "loss": 0.1795,
3985
+ "step": 498000
3986
+ },
3987
+ {
3988
+ "epoch": 38.74,
3989
+ "grad_norm": 0.9712297916412354,
3990
+ "learning_rate": 0.0002,
3991
+ "loss": 0.1798,
3992
+ "step": 499000
3993
+ },
3994
+ {
3995
+ "epoch": 38.81,
3996
+ "grad_norm": 1.3964751958847046,
3997
+ "learning_rate": 0.0002,
3998
+ "loss": 0.1809,
3999
+ "step": 500000
4000
+ },
4001
+ {
4002
+ "epoch": 38.81,
4003
+ "eval_bleu": 65.0266,
4004
+ "eval_gen_len": 28.7455,
4005
+ "eval_loss": 0.5573469996452332,
4006
+ "eval_runtime": 99.5788,
4007
+ "eval_samples_per_second": 8.918,
4008
+ "eval_steps_per_second": 1.115,
4009
+ "step": 500000
4010
+ },
4011
+ {
4012
+ "epoch": 38.81,
4013
+ "step": 500000,
4014
+ "total_flos": 5.131418179149005e+17,
4015
+ "train_loss": 0.40471466763305664,
4016
+ "train_runtime": 141171.7116,
4017
+ "train_samples_per_second": 56.669,
4018
+ "train_steps_per_second": 3.542
4019
+ }
4020
+ ],
4021
+ "logging_steps": 1000,
4022
+ "max_steps": 500000,
4023
+ "num_input_tokens_seen": 0,
4024
+ "num_train_epochs": 39,
4025
+ "save_steps": 10000,
4026
+ "total_flos": 5.131418179149005e+17,
4027
+ "train_batch_size": 8,
4028
+ "trial_name": null,
4029
+ "trial_params": null
4030
+ }