Add training_loss_history.json
Browse files- training_loss_history.json +21 -0
training_loss_history.json
ADDED
|
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model": "Gautam6/GautamAI",
|
| 3 |
+
"architecture": "GautamForCausalLM",
|
| 4 |
+
"last_updated": "2026-08-24T15:05:26.987Z",
|
| 5 |
+
"training_run": {
|
| 6 |
+
"active_feature": "Autonomous Software & Compiler Engineering AGI Engine v1.0",
|
| 7 |
+
"epoch": 1,
|
| 8 |
+
"global_step": 128,
|
| 9 |
+
"loss": 0.23439999999999372,
|
| 10 |
+
"perplexity": 1.2641500511096926,
|
| 11 |
+
"gradient_norm": 1.7466,
|
| 12 |
+
"learning_rate": 0.0001067,
|
| 13 |
+
"weight_sparsity_percent": 42.5,
|
| 14 |
+
"convergence_status": "ACTIVE_GRADIENT_DESCENT",
|
| 15 |
+
"optimizer": "AdamW (beta1=0.9, beta2=0.999, eps=1e-8, weight_decay=0.01)",
|
| 16 |
+
"lr_scheduler": "CosineAnnealingWithWarmup",
|
| 17 |
+
"attention_mechanism": "Grouped-Query Attention (GQA, heads=8, kv_heads=2) + RoPE",
|
| 18 |
+
"activation": "SwiGLU",
|
| 19 |
+
"precision": "4-Bit Sparse Optimized (FP32 master weights)"
|
| 20 |
+
}
|
| 21 |
+
}
|