Create modeling_i3.py
Browse files- modeling_i3.py +37 -0
modeling_i3.py
ADDED
|
@@ -0,0 +1,37 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import torch
|
| 2 |
+
from torch import nn
|
| 3 |
+
from transformers import PreTrainedModel, PretrainedConfig
|
| 4 |
+
from i3_modules import i3Model # import your original i3Model
|
| 5 |
+
|
| 6 |
+
class i3Config(PretrainedConfig):
|
| 7 |
+
model_type = "i3"
|
| 8 |
+
|
| 9 |
+
def __init__(self, vocab_size=65, d_model=256, n_layers=6, n_heads=8,
|
| 10 |
+
max_seq_len=256, rank=8, d_state=16, **kwargs):
|
| 11 |
+
super().__init__(**kwargs)
|
| 12 |
+
self.vocab_size = vocab_size
|
| 13 |
+
self.d_model = d_model
|
| 14 |
+
self.n_layers = n_layers
|
| 15 |
+
self.n_heads = n_heads
|
| 16 |
+
self.max_seq_len = max_seq_len
|
| 17 |
+
self.rank = rank
|
| 18 |
+
self.d_state = d_state
|
| 19 |
+
|
| 20 |
+
class i3(PreTrainedModel):
|
| 21 |
+
config_class = i3Config
|
| 22 |
+
base_model_prefix = "i3"
|
| 23 |
+
|
| 24 |
+
def __init__(self, config):
|
| 25 |
+
super().__init__(config)
|
| 26 |
+
self.model = i3Model(
|
| 27 |
+
vocab_size=config.vocab_size,
|
| 28 |
+
d_model=config.d_model,
|
| 29 |
+
n_layers=config.n_layers,
|
| 30 |
+
n_heads=config.n_heads,
|
| 31 |
+
max_seq_len=config.max_seq_len,
|
| 32 |
+
rank=config.rank,
|
| 33 |
+
d_state=config.d_state
|
| 34 |
+
)
|
| 35 |
+
|
| 36 |
+
def forward(self, input_ids, labels=None):
|
| 37 |
+
return self.model(input_ids, labels)
|