Update config.json
Browse files- config.json +37 -25
config.json
CHANGED
@@ -61,35 +61,47 @@
|
|
61 |
}
|
62 |
}
|
63 |
},
|
|
|
|
|
|
|
|
|
64 |
"architectures": [
|
65 |
-
"
|
66 |
],
|
67 |
-
"
|
68 |
-
"bos_token_id":
|
69 |
-
"
|
70 |
-
"
|
71 |
-
"
|
72 |
-
"
|
|
|
|
|
|
|
|
|
|
|
73 |
"early_stopping": true,
|
74 |
-
"
|
75 |
-
"
|
76 |
-
"
|
77 |
-
"
|
78 |
-
"
|
79 |
-
"
|
80 |
-
"
|
81 |
-
|
82 |
-
|
83 |
-
"
|
84 |
-
"
|
|
|
85 |
},
|
86 |
-
"
|
87 |
-
"
|
88 |
-
"
|
89 |
-
|
90 |
-
|
91 |
-
|
92 |
-
|
|
|
|
|
93 |
"n_ctx": 1024,
|
94 |
"n_embd": 1024,
|
95 |
"n_head": 16,
|
|
|
61 |
}
|
62 |
}
|
63 |
},
|
64 |
+
"activation_dropout": 0.1,
|
65 |
+
"activation_function": "gelu",
|
66 |
+
"add_bias_logits": false,
|
67 |
+
"add_final_layer_norm": false,
|
68 |
"architectures": [
|
69 |
+
"BartModel"
|
70 |
],
|
71 |
+
"attention_dropout": 0.1,
|
72 |
+
"bos_token_id": 0,
|
73 |
+
"classif_dropout": 0.1,
|
74 |
+
"classifier_dropout": 0.0,
|
75 |
+
"d_model": 1024,
|
76 |
+
"decoder_attention_heads": 16,
|
77 |
+
"decoder_ffn_dim": 4096,
|
78 |
+
"decoder_layerdrop": 0.0,
|
79 |
+
"decoder_layers": 12,
|
80 |
+
"decoder_start_token_id": 2,
|
81 |
+
"dropout": 0.1,
|
82 |
"early_stopping": true,
|
83 |
+
"encoder_attention_heads": 16,
|
84 |
+
"encoder_ffn_dim": 4096,
|
85 |
+
"encoder_layerdrop": 0.0,
|
86 |
+
"encoder_layers": 12,
|
87 |
+
"eos_token_id": 2,
|
88 |
+
"forced_eos_token_id": 2,
|
89 |
+
"forced_bos_token_id": 0,
|
90 |
+
"gradient_checkpointing": false,
|
91 |
+
"id2label": {
|
92 |
+
"0": "LABEL_0",
|
93 |
+
"1": "LABEL_1",
|
94 |
+
"2": "LABEL_2"
|
95 |
},
|
96 |
+
"init_std": 0.02,
|
97 |
+
"is_encoder_decoder": true,
|
98 |
+
"label2id": {
|
99 |
+
"LABEL_0": 0,
|
100 |
+
"LABEL_1": 1,
|
101 |
+
"LABEL_2": 2
|
102 |
+
},
|
103 |
+
"max_position_embeddings": 1024,
|
104 |
+
"model_type": "bart",
|
105 |
"n_ctx": 1024,
|
106 |
"n_embd": 1024,
|
107 |
"n_head": 16,
|