Update config.json
Browse files- config.json +2 -6
config.json
CHANGED
@@ -15,7 +15,7 @@
|
|
15 |
"hidden_size": 2048,
|
16 |
"initializer_range": 0.006,
|
17 |
"intermediate_size": 5632,
|
18 |
-
"max_position_embeddings":
|
19 |
"model_type": "bailing_moe",
|
20 |
"moe_intermediate_size": 1408,
|
21 |
"num_experts": 64,
|
@@ -27,11 +27,7 @@
|
|
27 |
"num_key_value_heads": 4,
|
28 |
"pretraining_tp": 1,
|
29 |
"rms_norm_eps": 1e-05,
|
30 |
-
"rope_scaling":
|
31 |
-
"factor": 4.0,
|
32 |
-
"original_max_position_embeddings": 16384,
|
33 |
-
"type": "yarn"
|
34 |
-
},
|
35 |
"rope_theta": 600000,
|
36 |
"tie_word_embeddings": false,
|
37 |
"torch_dtype": "bfloat16",
|
|
|
15 |
"hidden_size": 2048,
|
16 |
"initializer_range": 0.006,
|
17 |
"intermediate_size": 5632,
|
18 |
+
"max_position_embeddings": 16384,
|
19 |
"model_type": "bailing_moe",
|
20 |
"moe_intermediate_size": 1408,
|
21 |
"num_experts": 64,
|
|
|
27 |
"num_key_value_heads": 4,
|
28 |
"pretraining_tp": 1,
|
29 |
"rms_norm_eps": 1e-05,
|
30 |
+
"rope_scaling": null,
|
|
|
|
|
|
|
|
|
31 |
"rope_theta": 600000,
|
32 |
"tie_word_embeddings": false,
|
33 |
"torch_dtype": "bfloat16",
|