Update experiment measurements and status
Browse files
reports/runs/fineweb_edu-1337/metrics.jsonl
CHANGED
|
@@ -45,3 +45,6 @@
|
|
| 45 |
{"step": 900, "tokens": 471859200, "loss": 2.513486862182617, "optimizer_updates": 900, "effective_batch": 512, "lr": 1.248677602927694e-05, "gradient_norm": 0.24439941346645355, "tokens_per_second": 1125688.1723691565, "synthetic": false, "world_size": 8}
|
| 46 |
{"step": 920, "tokens": 482344960, "loss": 2.5100746154785156, "optimizer_updates": 920, "effective_batch": 512, "lr": 1.1604747889549428e-05, "gradient_norm": 0.27583736181259155, "tokens_per_second": 1132216.8734764603, "synthetic": false, "world_size": 8}
|
| 47 |
{"step": 940, "tokens": 492830720, "loss": 2.5097000122070314, "optimizer_updates": 940, "effective_batch": 512, "lr": 1.0912476145919328e-05, "gradient_norm": 0.24380965530872345, "tokens_per_second": 1138230.5058828183, "synthetic": false, "world_size": 8}
|
|
|
|
|
|
|
|
|
|
|
|
| 45 |
{"step": 900, "tokens": 471859200, "loss": 2.513486862182617, "optimizer_updates": 900, "effective_batch": 512, "lr": 1.248677602927694e-05, "gradient_norm": 0.24439941346645355, "tokens_per_second": 1125688.1723691565, "synthetic": false, "world_size": 8}
|
| 46 |
{"step": 920, "tokens": 482344960, "loss": 2.5100746154785156, "optimizer_updates": 920, "effective_batch": 512, "lr": 1.1604747889549428e-05, "gradient_norm": 0.27583736181259155, "tokens_per_second": 1132216.8734764603, "synthetic": false, "world_size": 8}
|
| 47 |
{"step": 940, "tokens": 492830720, "loss": 2.5097000122070314, "optimizer_updates": 940, "effective_batch": 512, "lr": 1.0912476145919328e-05, "gradient_norm": 0.24380965530872345, "tokens_per_second": 1138230.5058828183, "synthetic": false, "world_size": 8}
|
| 48 |
+
{"step": 960, "tokens": 503316480, "loss": 2.5185525894165037, "optimizer_updates": 960, "effective_batch": 512, "lr": 1.0412987926581599e-05, "gradient_norm": 0.23138202726840973, "tokens_per_second": 1140555.3808524152, "synthetic": false, "world_size": 8}
|
| 49 |
+
{"step": 980, "tokens": 513802240, "loss": 2.5114818572998048, "optimizer_updates": 980, "effective_batch": 512, "lr": 1.010846736643086e-05, "gradient_norm": 0.2943119704723358, "tokens_per_second": 1132366.2464293626, "synthetic": false, "world_size": 8}
|
| 50 |
+
{"step": 1000, "tokens": 524288000, "loss": 2.5175754547119142, "optimizer_updates": 1000, "effective_batch": 512, "lr": 1.000024605639518e-05, "gradient_norm": 0.24006935954093933, "tokens_per_second": 1130421.8567110323, "synthetic": false, "world_size": 8}
|
reports/status.json
CHANGED
|
@@ -1,8 +1,6 @@
|
|
| 1 |
{
|
| 2 |
-
"stage": "
|
| 3 |
-
"utc": "2026-09-30T21:
|
| 4 |
-
"remaining_seconds":
|
| 5 |
-
"
|
| 6 |
-
"seed": 1337,
|
| 7 |
-
"adaptation_tokens": 524288000
|
| 8 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"stage": "development_evaluation",
|
| 3 |
+
"utc": "2026-09-30T21:41:01.502312+00:00",
|
| 4 |
+
"remaining_seconds": 37672.497607,
|
| 5 |
+
"checkpoint": "/data/ablation/runs/fineweb_edu-1337/checkpoint-000000500.pt"
|
|
|
|
|
|
|
| 6 |
}
|