Commit
·
7138c9f
1
Parent(s):
dd2cf99
checking if other model thigns are needed
Browse files- config.json +1 -1
- configuration_lang.py +1 -1
- modeling_lang.py +3 -26
config.json
CHANGED
@@ -141,7 +141,7 @@
|
|
141 |
},
|
142 |
"layer_norm_eps": 1e-12,
|
143 |
"max_position_embeddings": 512,
|
144 |
-
"model_type": "
|
145 |
"num_attention_heads": 8,
|
146 |
"num_hidden_layers": 8,
|
147 |
"pad_token_id": 0,
|
|
|
141 |
},
|
142 |
"layer_norm_eps": 1e-12,
|
143 |
"max_position_embeddings": 512,
|
144 |
+
"model_type": "lang_detect",
|
145 |
"num_attention_heads": 8,
|
146 |
"num_hidden_layers": 8,
|
147 |
"pad_token_id": 0,
|
configuration_lang.py
CHANGED
@@ -3,7 +3,7 @@ import torch
|
|
3 |
|
4 |
|
5 |
class ImpressoConfig(PretrainedConfig):
|
6 |
-
model_type = "
|
7 |
|
8 |
def __init__(
|
9 |
self,
|
|
|
3 |
|
4 |
|
5 |
class ImpressoConfig(PretrainedConfig):
|
6 |
+
model_type = "lang_detect"
|
7 |
|
8 |
def __init__(
|
9 |
self,
|
modeling_lang.py
CHANGED
@@ -1,10 +1,7 @@
|
|
1 |
-
from transformers.modeling_outputs import TokenClassifierOutput
|
2 |
import torch
|
3 |
import torch.nn as nn
|
4 |
-
from transformers import PreTrainedModel
|
5 |
-
|
6 |
-
from typing import Optional, Tuple, Union
|
7 |
-
import logging, json, os
|
8 |
import floret
|
9 |
from .configuration_lang import ImpressoConfig
|
10 |
|
@@ -26,9 +23,6 @@ class ExtendedMultitaskModelForTokenClassification(PreTrainedModel):
|
|
26 |
|
27 |
#
|
28 |
def forward(self, input_ids, attention_mask=None, **kwargs):
|
29 |
-
# print(
|
30 |
-
# f"Check if it arrives here: {input_ids}, ---, {type(input_ids)} ----- {type(self.model_floret)}"
|
31 |
-
# )
|
32 |
if isinstance(input_ids, str):
|
33 |
# If the input is a single string, make it a list for floret
|
34 |
texts = [input_ids]
|
@@ -37,13 +31,11 @@ class ExtendedMultitaskModelForTokenClassification(PreTrainedModel):
|
|
37 |
else:
|
38 |
raise ValueError(f"Unexpected input type: {type(input_ids)}")
|
39 |
|
40 |
-
# Use the SafeFloretWrapper to get predictions
|
41 |
predictions, probabilities = self.model_floret.predict(texts, k=1)
|
42 |
-
# print(f"Predictions: {predictions}, Probabilities: {probabilities}")
|
43 |
return (
|
44 |
predictions,
|
45 |
probabilities,
|
46 |
-
)
|
47 |
|
48 |
def state_dict(self, *args, **kwargs):
|
49 |
# Return an empty state dictionary
|
@@ -53,21 +45,6 @@ class ExtendedMultitaskModelForTokenClassification(PreTrainedModel):
|
|
53 |
# Ignore loading since there are no parameters
|
54 |
pass
|
55 |
|
56 |
-
# print("Ignoring state_dict since model has no parameters.")
|
57 |
-
|
58 |
-
# def get_floret_model(self):
|
59 |
-
# return self.model_floret
|
60 |
-
|
61 |
-
# def get_extended_attention_mask(
|
62 |
-
# self, attention_mask, input_shape, device=None, dtype=torch.float
|
63 |
-
# ):
|
64 |
-
# if attention_mask is None:
|
65 |
-
# attention_mask = torch.ones(input_shape, device=device)
|
66 |
-
# extended_attention_mask = attention_mask[:, None, None, :]
|
67 |
-
# extended_attention_mask = extended_attention_mask.to(dtype=dtype)
|
68 |
-
# extended_attention_mask = (1.0 - extended_attention_mask) * -10000.0
|
69 |
-
# return extended_attention_mask
|
70 |
-
|
71 |
@property
|
72 |
def device(self):
|
73 |
return next(self.parameters()).device
|
|
|
|
|
1 |
import torch
|
2 |
import torch.nn as nn
|
3 |
+
from transformers import PreTrainedModel
|
4 |
+
import logging
|
|
|
|
|
5 |
import floret
|
6 |
from .configuration_lang import ImpressoConfig
|
7 |
|
|
|
23 |
|
24 |
#
|
25 |
def forward(self, input_ids, attention_mask=None, **kwargs):
|
|
|
|
|
|
|
26 |
if isinstance(input_ids, str):
|
27 |
# If the input is a single string, make it a list for floret
|
28 |
texts = [input_ids]
|
|
|
31 |
else:
|
32 |
raise ValueError(f"Unexpected input type: {type(input_ids)}")
|
33 |
|
|
|
34 |
predictions, probabilities = self.model_floret.predict(texts, k=1)
|
|
|
35 |
return (
|
36 |
predictions,
|
37 |
probabilities,
|
38 |
+
)
|
39 |
|
40 |
def state_dict(self, *args, **kwargs):
|
41 |
# Return an empty state dictionary
|
|
|
45 |
# Ignore loading since there are no parameters
|
46 |
pass
|
47 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
48 |
@property
|
49 |
def device(self):
|
50 |
return next(self.parameters()).device
|