Update training_args.json (Mind-v1 35.8M decoder-only reasoning model)
Browse files- training_args.json +26 -0
training_args.json
ADDED
|
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architecture": "LlamaForCausalLM",
|
| 3 |
+
"parameters": 35756544,
|
| 4 |
+
"vocab_size": 16384,
|
| 5 |
+
"hidden_size": 512,
|
| 6 |
+
"intermediate_size": 1376,
|
| 7 |
+
"num_hidden_layers": 6,
|
| 8 |
+
"num_attention_heads": 8,
|
| 9 |
+
"num_key_value_heads": 8,
|
| 10 |
+
"max_position_embeddings": 1024,
|
| 11 |
+
"epochs": 2,
|
| 12 |
+
"batch_size": 16,
|
| 13 |
+
"total_steps": 200,
|
| 14 |
+
"learning_rate": 0.001,
|
| 15 |
+
"weight_decay": 0.01,
|
| 16 |
+
"random_seed": 42,
|
| 17 |
+
"random_baseline_loss": 9.7041,
|
| 18 |
+
"final_train_loss": 1.0608,
|
| 19 |
+
"final_val_loss": 1.221,
|
| 20 |
+
"total_training_time_seconds": 893.07,
|
| 21 |
+
"hardware": {
|
| 22 |
+
"platform": "Windows-10-10.0.26200-SP0",
|
| 23 |
+
"processor": "Intel64 Family 6 Model 154 Stepping 4, GenuineIntel",
|
| 24 |
+
"python_version": "3.11.15 (main, Mar 24 2026, 22:57:51) [MSC v.1944 64 bit (AMD64)]"
|
| 25 |
+
}
|
| 26 |
+
}
|