Update model.safetensors
Browse files- model.safetensors +25 -8
model.safetensors
CHANGED
@@ -1,22 +1,34 @@
|
|
1 |
import torch
|
2 |
-
from transformers import PreTrainedModel, PreTrainedTokenizerFast, PretrainedConfig
|
3 |
-
from transformers.
|
|
|
4 |
|
5 |
class HelloWorldConfig(PretrainedConfig):
|
6 |
model_type = "hello-world"
|
7 |
|
8 |
-
class HelloWorldModel(PreTrainedModel):
|
9 |
config_class = HelloWorldConfig
|
10 |
|
11 |
def __init__(self, config):
|
12 |
super().__init__(config)
|
13 |
|
14 |
-
def forward(self,
|
15 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
16 |
|
17 |
-
|
|
|
|
|
|
|
|
|
18 |
|
19 |
-
# Dummy tokenizer configuration to work with the model
|
20 |
tokenizer_config = {
|
21 |
"do_lower_case": False,
|
22 |
"model_max_length": 512,
|
@@ -25,9 +37,14 @@ tokenizer_config = {
|
|
25 |
"tokenizer_file": "tokenizer.json",
|
26 |
"unk_token": "<unk>",
|
27 |
"bos_token": "<s>",
|
28 |
-
"eos_token": "</s>"
|
|
|
29 |
}
|
30 |
|
|
|
31 |
with open("tokenizer.json", "w") as f:
|
32 |
import json
|
33 |
json.dump(tokenizer_config, f)
|
|
|
|
|
|
|
|
1 |
import torch
|
2 |
+
from transformers import PreTrainedModel, PreTrainedTokenizerFast, PretrainedConfig, LogitsProcessorList
|
3 |
+
from transformers.generation_utils import GenerationMixin
|
4 |
+
from transformers.modeling_outputs import CausalLMOutput
|
5 |
|
6 |
class HelloWorldConfig(PretrainedConfig):
|
7 |
model_type = "hello-world"
|
8 |
|
9 |
+
class HelloWorldModel(PreTrainedModel, GenerationMixin):
|
10 |
config_class = HelloWorldConfig
|
11 |
|
12 |
def __init__(self, config):
|
13 |
super().__init__(config)
|
14 |
|
15 |
+
def forward(self, input_ids=None, **kwargs):
|
16 |
+
batch_size = input_ids.shape[0]
|
17 |
+
sequence_length = input_ids.shape[1]
|
18 |
+
|
19 |
+
# Generate a tensor with repeated "Hello, world!" token IDs
|
20 |
+
hello_world_token_id = self.config.vocab_size - 1 # assuming last token is "Hello, world!"
|
21 |
+
logits = torch.full((batch_size, sequence_length, self.config.vocab_size), float('-inf'))
|
22 |
+
logits[:, :, hello_world_token_id] = 0 # setting logits for "Hello, world!" to 0 (highest value)
|
23 |
+
|
24 |
+
return CausalLMOutput(logits=logits)
|
25 |
|
26 |
+
def prepare_inputs_for_generation(self, input_ids, **kwargs):
|
27 |
+
return {"input_ids": input_ids}
|
28 |
+
|
29 |
+
def _update_model_kwargs_for_generation(self, outputs, model_kwargs, is_encoder_decoder=False):
|
30 |
+
return model_kwargs
|
31 |
|
|
|
32 |
tokenizer_config = {
|
33 |
"do_lower_case": False,
|
34 |
"model_max_length": 512,
|
|
|
37 |
"tokenizer_file": "tokenizer.json",
|
38 |
"unk_token": "<unk>",
|
39 |
"bos_token": "<s>",
|
40 |
+
"eos_token": "</s>",
|
41 |
+
"vocab_size": 1, # Simplified vocabulary size
|
42 |
}
|
43 |
|
44 |
+
# Save tokenizer configuration
|
45 |
with open("tokenizer.json", "w") as f:
|
46 |
import json
|
47 |
json.dump(tokenizer_config, f)
|
48 |
+
|
49 |
+
tokenizer = PreTrainedTokenizerFast(tokenizer_file="tokenizer.json")
|
50 |
+
tokenizer.add_tokens(["Hello, world!"])
|