Update configuration_ostlm.py
Browse files- configuration_ostlm.py +11 -5
configuration_ostlm.py
CHANGED
|
@@ -14,10 +14,17 @@ class OSTLMConfig(PretrainedConfig):
|
|
| 14 |
dropout=0.1,
|
| 15 |
pad_token_id=0,
|
| 16 |
decoder_start_token_id=101,
|
| 17 |
-
|
| 18 |
**kwargs
|
| 19 |
):
|
| 20 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 21 |
self.vocab_size = vocab_size
|
| 22 |
self.d_model = d_model
|
| 23 |
self.nhead = nhead
|
|
@@ -26,6 +33,5 @@ class OSTLMConfig(PretrainedConfig):
|
|
| 26 |
self.dim_feedforward = dim_feedforward
|
| 27 |
self.dropout = dropout
|
| 28 |
|
| 29 |
-
#
|
| 30 |
-
self.num_hidden_layers = num_decoder_layers
|
| 31 |
-
self.is_encoder_decoder = True
|
|
|
|
| 14 |
dropout=0.1,
|
| 15 |
pad_token_id=0,
|
| 16 |
decoder_start_token_id=101,
|
| 17 |
+
eos_token_id=102,
|
| 18 |
**kwargs
|
| 19 |
):
|
| 20 |
+
# 讛讙讚专讛 讻诪讜讚诇 Encoder-Decoder 讻讚讬 诇讛驻注讬诇 讗转 诇讜讙讬拽转 讛转专讙讜诐
|
| 21 |
+
super().__init__(
|
| 22 |
+
pad_token_id=pad_token_id,
|
| 23 |
+
decoder_start_token_id=decoder_start_token_id,
|
| 24 |
+
eos_token_id=eos_token_id,
|
| 25 |
+
is_encoder_decoder=True,
|
| 26 |
+
**kwargs
|
| 27 |
+
)
|
| 28 |
self.vocab_size = vocab_size
|
| 29 |
self.d_model = d_model
|
| 30 |
self.nhead = nhead
|
|
|
|
| 33 |
self.dim_feedforward = dim_feedforward
|
| 34 |
self.dropout = dropout
|
| 35 |
|
| 36 |
+
# 砖讚讛 讞讜讘讛 砖-Hugging Face 诪讞驻砖 讘-generate
|
| 37 |
+
self.num_hidden_layers = num_decoder_layers
|
|
|