Upload training_status.txt with huggingface_hub
Browse files- training_status.txt +20 -0
training_status.txt
ADDED
|
@@ -0,0 +1,20 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
=======================================================
|
| 2 |
+
EDDIE TRAINING STATUS — 2x T4 DDP
|
| 3 |
+
=======================================================
|
| 4 |
+
Updated : 2026-08-15 04:20:51 UTC
|
| 5 |
+
Epoch : 5
|
| 6 |
+
Train Loss : 1.063693
|
| 7 |
+
Best Train : 1.063693
|
| 8 |
+
-------------------------------------------------------
|
| 9 |
+
Speed : 10,447 samples/s | 1.6 min/epoch
|
| 10 |
+
Batch/GPU : 8 | GPUs: 2
|
| 11 |
+
Grad Accum : 32 | Eff BS: 512
|
| 12 |
+
-------------------------------------------------------
|
| 13 |
+
Architecture : Cortex
|
| 14 |
+
Params : 2,925,699
|
| 15 |
+
Context : 1024 bars | Features: 5
|
| 16 |
+
Total windows: 501,051,135
|
| 17 |
+
Note: no held-out validation split -- every training example
|
| 18 |
+
is drawn without replacement from ~2.4B windows, so batches
|
| 19 |
+
are effectively unseen data by construction.
|
| 20 |
+
=======================================================
|