Spaces:
Running
Running
Upload 5 files
Browse files
configs/rec/unirec/unirec-0.1b/config.json
CHANGED
|
@@ -12,86 +12,18 @@
|
|
| 12 |
"decoder_layerdrop": 0.0,
|
| 13 |
"decoder_layers": 6,
|
| 14 |
"decoder_start_token_id": 0,
|
| 15 |
-
"depths": [
|
| 16 |
-
2,
|
| 17 |
-
2,
|
| 18 |
-
9,
|
| 19 |
-
2
|
| 20 |
-
],
|
| 21 |
-
"dims": [
|
| 22 |
-
128,
|
| 23 |
-
256,
|
| 24 |
-
512,
|
| 25 |
-
768
|
| 26 |
-
],
|
| 27 |
-
"drop_path_rate": 0.1,
|
| 28 |
-
"dropout": 0.1,
|
| 29 |
"encoder_layerdrop": 0.0,
|
| 30 |
"eos_token_id": 2,
|
| 31 |
"init_std": 0.02,
|
| 32 |
"is_encoder_decoder": true,
|
| 33 |
-
"kernel_size": [
|
| 34 |
-
3,
|
| 35 |
-
3
|
| 36 |
-
],
|
| 37 |
"max_position_embeddings": 3072,
|
| 38 |
-
"mixer": [
|
| 39 |
-
[
|
| 40 |
-
"Conv",
|
| 41 |
-
"Conv"
|
| 42 |
-
],
|
| 43 |
-
[
|
| 44 |
-
"Conv",
|
| 45 |
-
"Conv"
|
| 46 |
-
],
|
| 47 |
-
[
|
| 48 |
-
"Conv",
|
| 49 |
-
"Conv",
|
| 50 |
-
"Conv",
|
| 51 |
-
"Conv",
|
| 52 |
-
"Conv",
|
| 53 |
-
"FGlobal",
|
| 54 |
-
"Global",
|
| 55 |
-
"Global",
|
| 56 |
-
"Global"
|
| 57 |
-
],
|
| 58 |
-
[
|
| 59 |
-
"Global",
|
| 60 |
-
"Global"
|
| 61 |
-
]
|
| 62 |
-
],
|
| 63 |
-
"mlp_ratio": 4,
|
| 64 |
"model_type": "m2m_100",
|
| 65 |
-
"num_heads": [
|
| 66 |
-
4,
|
| 67 |
-
8,
|
| 68 |
-
8,
|
| 69 |
-
12
|
| 70 |
-
],
|
| 71 |
"pad_token_id": 1,
|
| 72 |
"scale_embedding": true,
|
| 73 |
-
"sub_k": [
|
| 74 |
-
[
|
| 75 |
-
2,
|
| 76 |
-
2
|
| 77 |
-
],
|
| 78 |
-
[
|
| 79 |
-
2,
|
| 80 |
-
2
|
| 81 |
-
],
|
| 82 |
-
[
|
| 83 |
-
2,
|
| 84 |
-
2
|
| 85 |
-
],
|
| 86 |
-
[
|
| 87 |
-
2,
|
| 88 |
-
2
|
| 89 |
-
]
|
| 90 |
-
],
|
| 91 |
"torch_dtype": "float32",
|
| 92 |
"transformers_version": "4.49.0",
|
| 93 |
"use_cache": true,
|
| 94 |
"vocab_size": 56371,
|
| 95 |
-
"max_length":
|
| 96 |
"label_smoothing": 0.1
|
| 97 |
}
|
|
|
|
| 12 |
"decoder_layerdrop": 0.0,
|
| 13 |
"decoder_layers": 6,
|
| 14 |
"decoder_start_token_id": 0,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 15 |
"encoder_layerdrop": 0.0,
|
| 16 |
"eos_token_id": 2,
|
| 17 |
"init_std": 0.02,
|
| 18 |
"is_encoder_decoder": true,
|
|
|
|
|
|
|
|
|
|
|
|
|
| 19 |
"max_position_embeddings": 3072,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 20 |
"model_type": "m2m_100",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 21 |
"pad_token_id": 1,
|
| 22 |
"scale_embedding": true,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 23 |
"torch_dtype": "float32",
|
| 24 |
"transformers_version": "4.49.0",
|
| 25 |
"use_cache": true,
|
| 26 |
"vocab_size": 56371,
|
| 27 |
+
"max_length": 2048,
|
| 28 |
"label_smoothing": 0.1
|
| 29 |
}
|