Update config.json
Browse files- config.json +41 -17
config.json
CHANGED
|
@@ -2,21 +2,25 @@
|
|
| 2 |
"architectures": [
|
| 3 |
"XLMWithLMHeadModel"
|
| 4 |
],
|
| 5 |
-
"
|
| 6 |
-
"emb_dim": 1024,
|
| 7 |
-
"n_layers": 12,
|
| 8 |
-
"n_heads": 8,
|
| 9 |
-
"dropout": 0.1,
|
| 10 |
"attention_dropout": 0.1,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 11 |
"gelu_activation": true,
|
| 12 |
-
"share_inout_emb": true,
|
| 13 |
-
"sinusoidal_embeddings": false,
|
| 14 |
-
"asm": false,
|
| 15 |
-
"max_vocab": 95000,
|
| 16 |
-
"min_count": 0,
|
| 17 |
"id2lang": {
|
| 18 |
"0": "ar",
|
| 19 |
"1": "bg",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 20 |
"2": "de",
|
| 21 |
"3": "el",
|
| 22 |
"4": "en",
|
|
@@ -24,13 +28,10 @@
|
|
| 24 |
"6": "fr",
|
| 25 |
"7": "hi",
|
| 26 |
"8": "ru",
|
| 27 |
-
"9": "sw"
|
| 28 |
-
"10": "th",
|
| 29 |
-
"11": "tr",
|
| 30 |
-
"12": "ur",
|
| 31 |
-
"13": "vi",
|
| 32 |
-
"14": "zh"
|
| 33 |
},
|
|
|
|
|
|
|
| 34 |
"lang2id": {
|
| 35 |
"ar": 0,
|
| 36 |
"bg": 1,
|
|
@@ -48,6 +49,29 @@
|
|
| 48 |
"vi": 13,
|
| 49 |
"zh": 14
|
| 50 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 51 |
"n_langs": 15,
|
| 52 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 53 |
}
|
|
|
|
| 2 |
"architectures": [
|
| 3 |
"XLMWithLMHeadModel"
|
| 4 |
],
|
| 5 |
+
"asm": false,
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
"attention_dropout": 0.1,
|
| 7 |
+
"bos_index": 0,
|
| 8 |
+
"bos_token_id": 0,
|
| 9 |
+
"causal": false,
|
| 10 |
+
"dropout": 0.1,
|
| 11 |
+
"emb_dim": 1024,
|
| 12 |
+
"embed_init_std": 0.02209708691207961,
|
| 13 |
+
"end_n_top": 5,
|
| 14 |
+
"eos_index": 1,
|
| 15 |
"gelu_activation": true,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 16 |
"id2lang": {
|
| 17 |
"0": "ar",
|
| 18 |
"1": "bg",
|
| 19 |
+
"10": "th",
|
| 20 |
+
"11": "tr",
|
| 21 |
+
"12": "ur",
|
| 22 |
+
"13": "vi",
|
| 23 |
+
"14": "zh",
|
| 24 |
"2": "de",
|
| 25 |
"3": "el",
|
| 26 |
"4": "en",
|
|
|
|
| 28 |
"6": "fr",
|
| 29 |
"7": "hi",
|
| 30 |
"8": "ru",
|
| 31 |
+
"9": "sw"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 32 |
},
|
| 33 |
+
"init_std": 0.02,
|
| 34 |
+
"is_encoder": true,
|
| 35 |
"lang2id": {
|
| 36 |
"ar": 0,
|
| 37 |
"bg": 1,
|
|
|
|
| 49 |
"vi": 13,
|
| 50 |
"zh": 14
|
| 51 |
},
|
| 52 |
+
"lang_id": 0,
|
| 53 |
+
"layer_norm_eps": 1e-12,
|
| 54 |
+
"mask_index": 5,
|
| 55 |
+
"mask_token_id": 0,
|
| 56 |
+
"max_position_embeddings": 512,
|
| 57 |
+
"max_vocab": 95000,
|
| 58 |
+
"min_count": 0,
|
| 59 |
+
"model_type": "xlm",
|
| 60 |
+
"n_heads": 8,
|
| 61 |
"n_langs": 15,
|
| 62 |
+
"n_layers": 12,
|
| 63 |
+
"pad_index": 2,
|
| 64 |
+
"pad_token_id": 2,
|
| 65 |
+
"same_enc_dec": true,
|
| 66 |
+
"share_inout_emb": true,
|
| 67 |
+
"sinusoidal_embeddings": false,
|
| 68 |
+
"start_n_top": 5,
|
| 69 |
+
"summary_activation": null,
|
| 70 |
+
"summary_first_dropout": 0.1,
|
| 71 |
+
"summary_proj_to_labels": true,
|
| 72 |
+
"summary_type": "first",
|
| 73 |
+
"summary_use_proj": true,
|
| 74 |
+
"unk_index": 3,
|
| 75 |
+
"use_lang_emb": true,
|
| 76 |
+
"vocab_size": 95000
|
| 77 |
}
|