Update configuration_ernie4_5_vl.py
Browse files
configuration_ernie4_5_vl.py
CHANGED
|
@@ -245,12 +245,7 @@ class Ernie4_5_Config(PretrainedConfig):
|
|
| 245 |
# Set default for tied embeddings if not specified.
|
| 246 |
if "tie_word_embeddings" not in kwargs:
|
| 247 |
kwargs["tie_word_embeddings"] = False
|
| 248 |
-
|
| 249 |
-
pad_token_id=pad_token_id,
|
| 250 |
-
bos_token_id=bos_token_id,
|
| 251 |
-
eos_token_id=eos_token_id,
|
| 252 |
-
**kwargs,
|
| 253 |
-
)
|
| 254 |
self.vocab_size = vocab_size
|
| 255 |
self.hidden_size = hidden_size
|
| 256 |
self.intermediate_size = intermediate_size
|
|
@@ -322,6 +317,12 @@ class Ernie4_5_Config(PretrainedConfig):
|
|
| 322 |
self.token_balance_seqlen = token_balance_seqlen
|
| 323 |
self.cachekv_quant = cachekv_quant
|
| 324 |
self.pp_seg_method = pp_seg_method
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 325 |
|
| 326 |
def get(self, key, default=None):
|
| 327 |
"""get config value by key"""
|
|
@@ -460,7 +461,7 @@ class Ernie4_5_MoEConfig(Ernie4_5_Config):
|
|
| 460 |
self.num_acc_steps = num_acc_steps
|
| 461 |
self.moe_layer_start_index = moe_layer_start_index
|
| 462 |
self.moe_layer_end_index = (
|
| 463 |
-
|
| 464 |
if moe_layer_end_index == -1
|
| 465 |
else moe_layer_end_index
|
| 466 |
)
|
|
|
|
| 245 |
# Set default for tied embeddings if not specified.
|
| 246 |
if "tie_word_embeddings" not in kwargs:
|
| 247 |
kwargs["tie_word_embeddings"] = False
|
| 248 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 249 |
self.vocab_size = vocab_size
|
| 250 |
self.hidden_size = hidden_size
|
| 251 |
self.intermediate_size = intermediate_size
|
|
|
|
| 317 |
self.token_balance_seqlen = token_balance_seqlen
|
| 318 |
self.cachekv_quant = cachekv_quant
|
| 319 |
self.pp_seg_method = pp_seg_method
|
| 320 |
+
super().__init__(
|
| 321 |
+
pad_token_id=pad_token_id,
|
| 322 |
+
bos_token_id=bos_token_id,
|
| 323 |
+
eos_token_id=eos_token_id,
|
| 324 |
+
**kwargs,
|
| 325 |
+
)
|
| 326 |
|
| 327 |
def get(self, key, default=None):
|
| 328 |
"""get config value by key"""
|
|
|
|
| 461 |
self.num_acc_steps = num_acc_steps
|
| 462 |
self.moe_layer_start_index = moe_layer_start_index
|
| 463 |
self.moe_layer_end_index = (
|
| 464 |
+
kwargs["num_hidden_layers"] - 1
|
| 465 |
if moe_layer_end_index == -1
|
| 466 |
else moe_layer_end_index
|
| 467 |
)
|