File size: 963 Bytes
dda5e62 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 | """PenguinVL vision encoder model configuration."""
from transformers import Qwen3Config
class PenguinVLVisionEncoderConfig(Qwen3Config):
model_type = "penguinvl_vision_encoder"
def __init__(
self,
hidden_size=1536,
intermediate_size=8960,
num_hidden_layers=12,
num_attention_heads=12,
num_channels=3,
patch_size=14,
layer_norm_eps=1e-6,
attention_dropout=0.0,
num_key_value_heads=2,
**kwargs,
):
super().__init__(**kwargs)
self.hidden_size = hidden_size
self.intermediate_size = intermediate_size
self.num_hidden_layers = num_hidden_layers
self.num_attention_heads = num_attention_heads
self.num_channels = num_channels
self.patch_size = patch_size
self.attention_dropout = attention_dropout
self.num_key_value_heads = num_key_value_heads
self.layer_norm_eps = layer_norm_eps
|