Upload config
Browse files- config.json +1 -0
- configuration_t5mimo.py +2 -0
config.json
CHANGED
|
@@ -19,6 +19,7 @@
|
|
| 19 |
"initializer_factor": 0.05,
|
| 20 |
"is_encoder_decoder": true,
|
| 21 |
"is_gated_act": false,
|
|
|
|
| 22 |
"layer_norm_epsilon": 1e-06,
|
| 23 |
"model_type": "t5mimo",
|
| 24 |
"num_decoder_layers": 4,
|
|
|
|
| 19 |
"initializer_factor": 0.05,
|
| 20 |
"is_encoder_decoder": true,
|
| 21 |
"is_gated_act": false,
|
| 22 |
+
"is_mimo": true,
|
| 23 |
"layer_norm_epsilon": 1e-06,
|
| 24 |
"model_type": "t5mimo",
|
| 25 |
"num_decoder_layers": 4,
|
configuration_t5mimo.py
CHANGED
|
@@ -81,6 +81,7 @@ class T5MIMOConfig(PretrainedConfig):
|
|
| 81 |
classifier_dropout=0.0,
|
| 82 |
num_seqs=3,
|
| 83 |
num_filters=64,
|
|
|
|
| 84 |
**kwargs,
|
| 85 |
):
|
| 86 |
self.vocab_size = vocab_size
|
|
@@ -102,6 +103,7 @@ class T5MIMOConfig(PretrainedConfig):
|
|
| 102 |
self.use_cache = use_cache
|
| 103 |
self.num_seqs = num_seqs
|
| 104 |
self.num_filters = num_filters
|
|
|
|
| 105 |
|
| 106 |
act_info = self.feed_forward_proj.split("-")
|
| 107 |
self.dense_act_fn = act_info[-1]
|
|
|
|
| 81 |
classifier_dropout=0.0,
|
| 82 |
num_seqs=3,
|
| 83 |
num_filters=64,
|
| 84 |
+
is_mimo=True,
|
| 85 |
**kwargs,
|
| 86 |
):
|
| 87 |
self.vocab_size = vocab_size
|
|
|
|
| 103 |
self.use_cache = use_cache
|
| 104 |
self.num_seqs = num_seqs
|
| 105 |
self.num_filters = num_filters
|
| 106 |
+
self.is_mimo = is_mimo
|
| 107 |
|
| 108 |
act_info = self.feed_forward_proj.split("-")
|
| 109 |
self.dense_act_fn = act_info[-1]
|