Delete configuration_hf_nomic_bert.py
Browse files
configuration_hf_nomic_bert.py
DELETED
@@ -1,56 +0,0 @@
|
|
1 |
-
from transformers import GPT2Config
|
2 |
-
|
3 |
-
|
4 |
-
class NomicBertConfig(GPT2Config):
|
5 |
-
model_type = "nomic_bert"
|
6 |
-
|
7 |
-
def __init__(
|
8 |
-
self,
|
9 |
-
prenorm=False,
|
10 |
-
parallel_block=False,
|
11 |
-
parallel_block_tied_norm=False,
|
12 |
-
rotary_emb_fraction=0.0,
|
13 |
-
fused_dropout_add_ln=False,
|
14 |
-
fused_bias_fc=False,
|
15 |
-
use_flash_attn=False,
|
16 |
-
use_xentropy=False,
|
17 |
-
qkv_proj_bias=True,
|
18 |
-
rotary_emb_base=10_000,
|
19 |
-
rotary_emb_scale_base=None,
|
20 |
-
rotary_emb_interleaved=False,
|
21 |
-
mlp_fc1_bias=True,
|
22 |
-
mlp_fc2_bias=True,
|
23 |
-
use_rms_norm=False,
|
24 |
-
causal=False,
|
25 |
-
type_vocab_size=2,
|
26 |
-
dense_seq_output=True,
|
27 |
-
pad_vocab_size_multiple=1,
|
28 |
-
tie_word_embeddings=True,
|
29 |
-
rotary_scaling_factor=None,
|
30 |
-
max_trained_positions=2048,
|
31 |
-
**kwargs,
|
32 |
-
):
|
33 |
-
self.prenorm = prenorm
|
34 |
-
self.parallel_block = parallel_block
|
35 |
-
self.parallel_block_tied_norm = parallel_block_tied_norm
|
36 |
-
self.rotary_emb_fraction = rotary_emb_fraction
|
37 |
-
self.tie_word_embeddings = tie_word_embeddings
|
38 |
-
self.fused_dropout_add_ln = fused_dropout_add_ln
|
39 |
-
self.fused_bias_fc = fused_bias_fc
|
40 |
-
self.use_flash_attn = use_flash_attn
|
41 |
-
self.use_xentropy = use_xentropy
|
42 |
-
self.qkv_proj_bias = qkv_proj_bias
|
43 |
-
self.rotary_emb_base = rotary_emb_base
|
44 |
-
self.rotary_emb_scale_base = rotary_emb_scale_base
|
45 |
-
self.rotary_emb_interleaved = rotary_emb_interleaved
|
46 |
-
self.mlp_fc1_bias = mlp_fc1_bias
|
47 |
-
self.mlp_fc2_bias = mlp_fc2_bias
|
48 |
-
self.use_rms_norm = use_rms_norm
|
49 |
-
self.causal = causal
|
50 |
-
self.type_vocab_size = type_vocab_size
|
51 |
-
self.dense_seq_output = dense_seq_output
|
52 |
-
self.pad_vocab_size_multiple = pad_vocab_size_multiple
|
53 |
-
self.rotary_scaling_factor = rotary_scaling_factor
|
54 |
-
self.max_trained_positions = max_trained_positions
|
55 |
-
|
56 |
-
super().__init__(**kwargs)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|