Upload modeling_hgrn2.py with huggingface_hub
Browse files- modeling_hgrn2.py +117 -0
modeling_hgrn2.py
ADDED
|
@@ -0,0 +1,117 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
from fla.models.hgrn2 import HGRN2ForCausalLM, HGRN2Model
|
| 2 |
+
from typing import Optional, Tuple, Union, List
|
| 3 |
+
|
| 4 |
+
from fla.models.hgrn2.modeling_hgrn2 import HGRN2PreTrainedModel, HGRN2Model
|
| 5 |
+
from fla.models.hgrn2.configuration_hgrn2 import HGRN2Config
|
| 6 |
+
|
| 7 |
+
import torch
|
| 8 |
+
from torch import nn
|
| 9 |
+
from torch.nn import CrossEntropyLoss, MSELoss, BCEWithLogitsLoss
|
| 10 |
+
|
| 11 |
+
from transformers.modeling_outputs import SequenceClassifierOutputWithPast
|
| 12 |
+
|
| 13 |
+
def register_hgrn2_for_sequence_classification():
|
| 14 |
+
from transformers import AutoModelForSequenceClassification
|
| 15 |
+
AutoModelForSequenceClassification.register(HGRN2Config, HGRN2ForSequenceClassification)
|
| 16 |
+
|
| 17 |
+
|
| 18 |
+
class HGRN2ForSequenceClassification(HGRN2PreTrainedModel):
|
| 19 |
+
_keys_to_ignore_on_load_missing = [r"lm_head.weight"]
|
| 20 |
+
|
| 21 |
+
def __init__(self, config):
|
| 22 |
+
super().__init__(config)
|
| 23 |
+
self.num_labels = config.num_labels
|
| 24 |
+
self.model = HGRN2Model(config)
|
| 25 |
+
self.score = nn.Linear(config.hidden_size, self.num_labels, bias=False)
|
| 26 |
+
|
| 27 |
+
# Initialize weights and apply final processing
|
| 28 |
+
self.post_init()
|
| 29 |
+
|
| 30 |
+
def get_input_embeddings(self):
|
| 31 |
+
return self.model.embeddings
|
| 32 |
+
|
| 33 |
+
def set_input_embeddings(self, value):
|
| 34 |
+
self.model.embeddings = value
|
| 35 |
+
|
| 36 |
+
def forward(
|
| 37 |
+
self,
|
| 38 |
+
input_ids: torch.LongTensor = None,
|
| 39 |
+
attention_mask: Optional[torch.Tensor] = None,
|
| 40 |
+
past_key_values: Optional[List[torch.FloatTensor]] = None,
|
| 41 |
+
inputs_embeds: Optional[torch.FloatTensor] = None,
|
| 42 |
+
labels: Optional[torch.LongTensor] = None,
|
| 43 |
+
use_cache: Optional[bool] = None,
|
| 44 |
+
output_attentions: Optional[bool] = None,
|
| 45 |
+
output_hidden_states: Optional[bool] = None,
|
| 46 |
+
return_dict: Optional[bool] = None,
|
| 47 |
+
) -> Union[Tuple, SequenceClassifierOutputWithPast]:
|
| 48 |
+
r"""
|
| 49 |
+
labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
|
| 50 |
+
Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
|
| 51 |
+
config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
|
| 52 |
+
`config.num_labels > 1` a classification loss is computed (Cross-Entropy).
|
| 53 |
+
"""
|
| 54 |
+
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
|
| 55 |
+
|
| 56 |
+
outputs = self.model(
|
| 57 |
+
input_ids=input_ids,
|
| 58 |
+
attention_mask=attention_mask,
|
| 59 |
+
inputs_embeds=inputs_embeds,
|
| 60 |
+
output_attentions=output_attentions,
|
| 61 |
+
use_cache=use_cache,
|
| 62 |
+
past_key_values=past_key_values,
|
| 63 |
+
output_hidden_states=output_hidden_states,
|
| 64 |
+
return_dict=return_dict,
|
| 65 |
+
)
|
| 66 |
+
hidden_states = outputs[0]
|
| 67 |
+
logits = self.score(hidden_states)
|
| 68 |
+
|
| 69 |
+
if input_ids is not None:
|
| 70 |
+
batch_size = input_ids.shape[0]
|
| 71 |
+
else:
|
| 72 |
+
batch_size = inputs_embeds.shape[0]
|
| 73 |
+
|
| 74 |
+
if self.config.pad_token_id is None and batch_size != 1:
|
| 75 |
+
raise ValueError("Cannot handle batch sizes > 1 if no padding token is defined.")
|
| 76 |
+
if self.config.pad_token_id is None:
|
| 77 |
+
sequence_lengths = -1
|
| 78 |
+
else:
|
| 79 |
+
if input_ids is not None:
|
| 80 |
+
sequence_lengths = (torch.ne(input_ids, self.config.pad_token_id).sum(-1) - 1).to(logits.device)
|
| 81 |
+
else:
|
| 82 |
+
sequence_lengths = -1
|
| 83 |
+
|
| 84 |
+
pooled_logits = logits[torch.arange(batch_size, device=logits.device), sequence_lengths]
|
| 85 |
+
|
| 86 |
+
loss = None
|
| 87 |
+
if labels is not None:
|
| 88 |
+
labels = labels.to(logits.device)
|
| 89 |
+
if self.config.problem_type is None:
|
| 90 |
+
if self.num_labels == 1:
|
| 91 |
+
self.config.problem_type = "regression"
|
| 92 |
+
elif self.num_labels > 1 and (labels.dtype == torch.long or labels.dtype == torch.int):
|
| 93 |
+
self.config.problem_type = "single_label_classification"
|
| 94 |
+
else:
|
| 95 |
+
self.config.problem_type = "multi_label_classification"
|
| 96 |
+
|
| 97 |
+
if self.config.problem_type == "regression":
|
| 98 |
+
loss_fct = MSELoss()
|
| 99 |
+
if self.num_labels == 1:
|
| 100 |
+
loss = loss_fct(pooled_logits.squeeze(), labels.squeeze())
|
| 101 |
+
else:
|
| 102 |
+
loss = loss_fct(pooled_logits, labels)
|
| 103 |
+
elif self.config.problem_type == "single_label_classification":
|
| 104 |
+
loss_fct = CrossEntropyLoss()
|
| 105 |
+
loss = loss_fct(pooled_logits.view(-1, self.num_labels), labels.view(-1))
|
| 106 |
+
elif self.config.problem_type == "multi_label_classification":
|
| 107 |
+
loss_fct = BCEWithLogitsLoss()
|
| 108 |
+
loss = loss_fct(pooled_logits, labels)
|
| 109 |
+
if not return_dict:
|
| 110 |
+
output = (pooled_logits,) + outputs[1:]
|
| 111 |
+
return ((loss,) + output) if loss is not None else output
|
| 112 |
+
|
| 113 |
+
return SequenceClassifierOutputWithPast(
|
| 114 |
+
loss=loss,
|
| 115 |
+
logits=pooled_logits,
|
| 116 |
+
hidden_states=outputs.hidden_states,
|
| 117 |
+
)
|