Update tokenizer, configuration and training reports
Browse files
reports/run/metrics.jsonl
CHANGED
|
@@ -1 +1,6 @@
|
|
| 1 |
{"step": 20, "tokens": 10485760, "loss": 10.169149780273438, "tokens_per_second": 229485.42681844623, "lr": 1.5e-05, "gradient_norm": 2.205339193344116, "benchmark": false}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
{"step": 20, "tokens": 10485760, "loss": 10.169149780273438, "tokens_per_second": 229485.42681844623, "lr": 1.5e-05, "gradient_norm": 2.205339193344116, "benchmark": false}
|
| 2 |
+
{"step": 40, "tokens": 20971520, "loss": 9.850079345703126, "tokens_per_second": 914837.8703186912, "lr": 3e-05, "gradient_norm": 2.3315203189849854, "benchmark": false}
|
| 3 |
+
{"step": 60, "tokens": 31457280, "loss": 9.301335906982422, "tokens_per_second": 753117.1709990001, "lr": 4.4999999999999996e-05, "gradient_norm": 1.8692331314086914, "benchmark": false}
|
| 4 |
+
{"step": 80, "tokens": 41943040, "loss": 8.688764190673828, "tokens_per_second": 1032255.2836004663, "lr": 6e-05, "gradient_norm": 1.4424148797988892, "benchmark": false}
|
| 5 |
+
{"step": 100, "tokens": 52428800, "loss": 8.105000305175782, "tokens_per_second": 1029399.3815147971, "lr": 7.500000000000001e-05, "gradient_norm": 1.0234349966049194, "benchmark": false}
|
| 6 |
+
{"step": 120, "tokens": 62914560, "loss": 7.616502380371093, "tokens_per_second": 1015709.3869896838, "lr": 8.999999999999999e-05, "gradient_norm": 1.1642669439315796, "benchmark": false}
|