add v5 experiment runtime
Browse files
experiments/M31-Python-Agent-220M-v5/runtime/config.json
ADDED
|
@@ -0,0 +1,19 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_type": "m31",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"M31Model"
|
| 5 |
+
],
|
| 6 |
+
"vocab_size": 32768,
|
| 7 |
+
"hidden_size": 896,
|
| 8 |
+
"num_hidden_layers": 20,
|
| 9 |
+
"num_attention_heads": 14,
|
| 10 |
+
"num_key_value_heads": 7,
|
| 11 |
+
"intermediate_size": 2816,
|
| 12 |
+
"max_position_embeddings": 2048,
|
| 13 |
+
"training_sequence_length": 512,
|
| 14 |
+
"rms_norm_eps": 1e-06,
|
| 15 |
+
"rope_theta": 10000.0,
|
| 16 |
+
"tie_word_embeddings": true,
|
| 17 |
+
"parameter_count": 228953984,
|
| 18 |
+
"experiment": "M31-Python-Agent-220M-v5"
|
| 19 |
+
}
|