updating saved model after name change from encoders to layers in MegaModel
Browse files- config.json +5 -2
- pytorch_model.bin +2 -2
config.json
CHANGED
|
@@ -12,13 +12,15 @@
|
|
| 12 |
"chunk_size": -1,
|
| 13 |
"classifier_dropout": null,
|
| 14 |
"dropout_prob": 0.1,
|
|
|
|
|
|
|
|
|
|
| 15 |
"ema_projection_size": 16,
|
| 16 |
"eos_token_id": 2,
|
| 17 |
"hidden_dropout_prob": 0.1,
|
| 18 |
"hidden_size": 128,
|
| 19 |
"initializer_range": 0.02,
|
| 20 |
"intermediate_size": 256,
|
| 21 |
-
"layer_norm_eps": 1e-12,
|
| 22 |
"max_positions": 10000,
|
| 23 |
"model_type": "mega",
|
| 24 |
"nffn_activation_dropout_prob": 0.0,
|
|
@@ -27,12 +29,13 @@
|
|
| 27 |
"normalization_type": "scalenorm",
|
| 28 |
"normalize_before_ffn": true,
|
| 29 |
"normalize_before_mega": true,
|
|
|
|
| 30 |
"num_hidden_layers": 4,
|
| 31 |
"pad_token_id": 1,
|
| 32 |
"relative_positional_bias": "rotary",
|
| 33 |
"shared_representation_size": 64,
|
| 34 |
"torch_dtype": "float32",
|
| 35 |
-
"transformers_version": "4.
|
| 36 |
"truncation": null,
|
| 37 |
"type_vocab_size": 2,
|
| 38 |
"use_cache": true,
|
|
|
|
| 12 |
"chunk_size": -1,
|
| 13 |
"classifier_dropout": null,
|
| 14 |
"dropout_prob": 0.1,
|
| 15 |
+
"ema_beta_range": 0.02,
|
| 16 |
+
"ema_delta_alpha_range": 0.2,
|
| 17 |
+
"ema_gamma_omega_range": 1.0,
|
| 18 |
"ema_projection_size": 16,
|
| 19 |
"eos_token_id": 2,
|
| 20 |
"hidden_dropout_prob": 0.1,
|
| 21 |
"hidden_size": 128,
|
| 22 |
"initializer_range": 0.02,
|
| 23 |
"intermediate_size": 256,
|
|
|
|
| 24 |
"max_positions": 10000,
|
| 25 |
"model_type": "mega",
|
| 26 |
"nffn_activation_dropout_prob": 0.0,
|
|
|
|
| 29 |
"normalization_type": "scalenorm",
|
| 30 |
"normalize_before_ffn": true,
|
| 31 |
"normalize_before_mega": true,
|
| 32 |
+
"num_attention_heads": 1,
|
| 33 |
"num_hidden_layers": 4,
|
| 34 |
"pad_token_id": 1,
|
| 35 |
"relative_positional_bias": "rotary",
|
| 36 |
"shared_representation_size": 64,
|
| 37 |
"torch_dtype": "float32",
|
| 38 |
+
"transformers_version": "4.28.0.dev0",
|
| 39 |
"truncation": null,
|
| 40 |
"type_vocab_size": 2,
|
| 41 |
"use_cache": true,
|
pytorch_model.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:69f1dfc6152299b92c6729e6ea7c955b70f07fe63dafc799e0386690e2349df6
|
| 3 |
+
size 29332476
|