Update experiment measurements and status
Browse files
reports/runs/cosmopedia-2027/metrics.jsonl
CHANGED
|
@@ -31,3 +31,7 @@
|
|
| 31 |
{"step": 620, "tokens": 325058560, "loss": 2.3813013076782226, "optimizer_updates": 620, "effective_batch": 512, "lr": 4.123583992279283e-05, "gradient_norm": 0.3169904053211212, "tokens_per_second": 1156982.3159562275, "synthetic": false, "world_size": 8}
|
| 32 |
{"step": 640, "tokens": 335544320, "loss": 2.3803289413452147, "optimizer_updates": 640, "effective_batch": 512, "lr": 3.8434395129189505e-05, "gradient_norm": 0.27440863847732544, "tokens_per_second": 1153586.4761494203, "synthetic": false, "world_size": 8}
|
| 33 |
{"step": 660, "tokens": 346030080, "loss": 2.384285545349121, "optimizer_updates": 660, "effective_batch": 512, "lr": 3.570538751077433e-05, "gradient_norm": 0.2709105610847473, "tokens_per_second": 1153962.352711612, "synthetic": false, "world_size": 8}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 31 |
{"step": 620, "tokens": 325058560, "loss": 2.3813013076782226, "optimizer_updates": 620, "effective_batch": 512, "lr": 4.123583992279283e-05, "gradient_norm": 0.3169904053211212, "tokens_per_second": 1156982.3159562275, "synthetic": false, "world_size": 8}
|
| 32 |
{"step": 640, "tokens": 335544320, "loss": 2.3803289413452147, "optimizer_updates": 640, "effective_batch": 512, "lr": 3.8434395129189505e-05, "gradient_norm": 0.27440863847732544, "tokens_per_second": 1153586.4761494203, "synthetic": false, "world_size": 8}
|
| 33 |
{"step": 660, "tokens": 346030080, "loss": 2.384285545349121, "optimizer_updates": 660, "effective_batch": 512, "lr": 3.570538751077433e-05, "gradient_norm": 0.2709105610847473, "tokens_per_second": 1153962.352711612, "synthetic": false, "world_size": 8}
|
| 34 |
+
{"step": 680, "tokens": 356515840, "loss": 2.3837989807128905, "optimizer_updates": 680, "effective_batch": 512, "lr": 3.306075032350062e-05, "gradient_norm": 0.24183112382888794, "tokens_per_second": 1153852.344047395, "synthetic": false, "world_size": 8}
|
| 35 |
+
{"step": 700, "tokens": 367001600, "loss": 2.3823820114135743, "optimizer_updates": 700, "effective_batch": 512, "lr": 3.0512047892893327e-05, "gradient_norm": 0.2890660762786865, "tokens_per_second": 1153481.5452552068, "synthetic": false, "world_size": 8}
|
| 36 |
+
{"step": 720, "tokens": 377487360, "loss": 2.3831897735595704, "optimizer_updates": 720, "effective_batch": 512, "lr": 2.8070425046195906e-05, "gradient_norm": 0.32422587275505066, "tokens_per_second": 1160356.6949989174, "synthetic": false, "world_size": 8}
|
| 37 |
+
{"step": 740, "tokens": 387973120, "loss": 2.3851343154907227, "optimizer_updates": 740, "effective_batch": 512, "lr": 2.574655837887642e-05, "gradient_norm": 0.28101056814193726, "tokens_per_second": 1149670.640146662, "synthetic": false, "world_size": 8}
|