Update config.json
Browse files- config.json +8 -3
config.json
CHANGED
|
@@ -1,8 +1,13 @@
|
|
|
|
|
| 1 |
{
|
| 2 |
-
|
| 3 |
"InfinityFormerForCausalLM"
|
| 4 |
],
|
| 5 |
"attention_probs_dropout_prob": 0.1,
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
"gate_init_bias": -2.0,
|
| 7 |
"gradient_checkpointing_frequency": 1,
|
| 8 |
"gradient_checkpointing_use_reentrant": true,
|
|
@@ -16,7 +21,7 @@
|
|
| 16 |
"max_position_embeddings": 812,
|
| 17 |
"memory_compression_frequency": 100,
|
| 18 |
"memory_compression_ratio": 0.5,
|
| 19 |
-
"model_type": "
|
| 20 |
"num_attention_heads": 12,
|
| 21 |
"num_hidden_layers": 54,
|
| 22 |
"num_memory_scales": 3,
|
|
@@ -27,7 +32,7 @@
|
|
| 27 |
"use_gating": true,
|
| 28 |
"use_gradient_checkpointing": false,
|
| 29 |
"use_multi_scale_memory": true,
|
| 30 |
-
"use_return_dict": true,
|
| 31 |
"use_rotary_embeddings": true,
|
| 32 |
"vocab_size": 151669
|
| 33 |
}
|
|
|
|
|
|
| 1 |
+
|
| 2 |
{
|
| 3 |
+
"architectures": [
|
| 4 |
"InfinityFormerForCausalLM"
|
| 5 |
],
|
| 6 |
"attention_probs_dropout_prob": 0.1,
|
| 7 |
+
"auto_map": {
|
| 8 |
+
"AutoConfig": "silx-ai/QuasarV4-Tiny--configuration_quasrav4.QuasraV4Config",
|
| 9 |
+
"AutoModelForCausalLM": "silx-ai/QuasarV4-Tiny--modeling_quasrav4.QuasraV4ForCausalLM"
|
| 10 |
+
},
|
| 11 |
"gate_init_bias": -2.0,
|
| 12 |
"gradient_checkpointing_frequency": 1,
|
| 13 |
"gradient_checkpointing_use_reentrant": true,
|
|
|
|
| 21 |
"max_position_embeddings": 812,
|
| 22 |
"memory_compression_frequency": 100,
|
| 23 |
"memory_compression_ratio": 0.5,
|
| 24 |
+
"model_type": "quasarv4",
|
| 25 |
"num_attention_heads": 12,
|
| 26 |
"num_hidden_layers": 54,
|
| 27 |
"num_memory_scales": 3,
|
|
|
|
| 32 |
"use_gating": true,
|
| 33 |
"use_gradient_checkpointing": false,
|
| 34 |
"use_multi_scale_memory": true,
|
|
|
|
| 35 |
"use_rotary_embeddings": true,
|
| 36 |
"vocab_size": 151669
|
| 37 |
}
|
| 38 |
+
|