Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -208,23 +208,23 @@ Current run: **disabled; no EMA state, auxiliary moment, or pullback is allocate
|
|
| 208 |
| PACE | disabled; no EMA state, auxiliary moment, or pullback is allocated |
|
| 209 |
| Learning rate | 6e-04 with linear warmup (10 % of steps) |
|
| 210 |
| Weight decay | 0.1 |
|
| 211 |
-
| Training time | ~4h
|
| 212 |
| Hardware | NVIDIA RTX 5090 (single GPU) |
|
| 213 |
|
| 214 |
### Training curve
|
| 215 |
|
| 216 |
| Step | Train Loss | Val Loss |
|
| 217 |
|---|---|---|
|
| 218 |
-
| 5,000 | 4.
|
| 219 |
-
| 10,000 | 4.
|
| 220 |
-
| 15,000 | 4.
|
| 221 |
-
| 20,000 | 4.
|
| 222 |
-
| 25,000 | 4.
|
| 223 |
-
| 30,000 | 4.
|
| 224 |
-
| 35,000 | 4.
|
| 225 |
-
| 40,000 |
|
| 226 |
-
| 45,000 | 3.
|
| 227 |
-
| 46,875 | — | 3.
|
| 228 |
|
| 229 |
---
|
| 230 |
|
|
|
|
| 208 |
| PACE | disabled; no EMA state, auxiliary moment, or pullback is allocated |
|
| 209 |
| Learning rate | 6e-04 with linear warmup (10 % of steps) |
|
| 210 |
| Weight decay | 0.1 |
|
| 211 |
+
| Training time | ~4h 51m |
|
| 212 |
| Hardware | NVIDIA RTX 5090 (single GPU) |
|
| 213 |
|
| 214 |
### Training curve
|
| 215 |
|
| 216 |
| Step | Train Loss | Val Loss |
|
| 217 |
|---|---|---|
|
| 218 |
+
| 5,000 | 4.827 | 4.716 |
|
| 219 |
+
| 10,000 | 4.480 | 4.369 |
|
| 220 |
+
| 15,000 | 4.334 | 4.227 |
|
| 221 |
+
| 20,000 | 4.251 | 4.142 |
|
| 222 |
+
| 25,000 | 4.195 | 4.091 |
|
| 223 |
+
| 30,000 | 4.152 | 4.049 |
|
| 224 |
+
| 35,000 | 4.127 | 4.019 |
|
| 225 |
+
| 40,000 | 4.029 | 3.942 |
|
| 226 |
+
| 45,000 | 3.984 | 3.885 |
|
| 227 |
+
| 46,875 | — | 3.878 |
|
| 228 |
|
| 229 |
---
|
| 230 |
|