Fill-Mask
Transformers
Safetensors
bert
masked-language-modeling
historical-nlp
temporal-language-model
Instructions to use TextMachineProject/NewsBERT_1800-1920-Temporal with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use TextMachineProject/NewsBERT_1800-1920-Temporal with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("fill-mask", model="TextMachineProject/NewsBERT_1800-1920-Temporal")# pip install -U transformers accelerate # Load model directly from transformers import AutoTokenizer, AutoModelForMaskedLM tokenizer = AutoTokenizer.from_pretrained("TextMachineProject/NewsBERT_1800-1920-Temporal") model = AutoModelForMaskedLM.from_pretrained("TextMachineProject/NewsBERT_1800-1920-Temporal", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Update continuous_time_embedding.py
Browse files- continuous_time_embedding.py +8 -14
continuous_time_embedding.py
CHANGED
|
@@ -88,13 +88,13 @@ class ContinuousTimeBertForMLM(nn.Module):
|
|
| 88 |
|
| 89 |
def save_pretrained(self, save_dir):
|
| 90 |
os.makedirs(save_dir, exist_ok=True)
|
| 91 |
-
self.model.save_pretrained(
|
| 92 |
torch.save(self.time_embed.state_dict(), os.path.join(save_dir, "time_embed.pt"))
|
| 93 |
torch.save(self.post_inject_norm.state_dict(), os.path.join(save_dir, "post_inject_norm.pt"))
|
| 94 |
-
|
| 95 |
@classmethod
|
| 96 |
def load_pretrained(cls, save_dir, hidden_size=768, **kwargs):
|
| 97 |
-
model = AutoModelForMaskedLM.from_pretrained(
|
| 98 |
obj = cls(model, hidden_size, **kwargs)
|
| 99 |
obj.time_embed.load_state_dict(torch.load(os.path.join(save_dir, "time_embed.pt"), map_location="cpu"))
|
| 100 |
obj.post_inject_norm.load_state_dict(torch.load(os.path.join(save_dir, "post_inject_norm.pt"), map_location="cpu"))
|
|
@@ -107,16 +107,11 @@ def load_continuous_time_model(repo_id_or_path, device=None, **kwargs):
|
|
| 107 |
else:
|
| 108 |
local_dir = snapshot_download(repo_id_or_path)
|
| 109 |
|
| 110 |
-
|
| 111 |
-
|
| 112 |
-
|
| 113 |
-
|
| 114 |
-
|
| 115 |
-
continue
|
| 116 |
-
if len(candidate_tokenizer) < 1000:
|
| 117 |
-
continue
|
| 118 |
-
tokenizer = candidate_tokenizer
|
| 119 |
-
break
|
| 120 |
|
| 121 |
if tokenizer is None:
|
| 122 |
print(f"[load_continuous_time_model] No valid tokenizer found in {repo_id_or_path}, "
|
|
@@ -124,7 +119,6 @@ def load_continuous_time_model(repo_id_or_path, device=None, **kwargs):
|
|
| 124 |
tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL_ID)
|
| 125 |
|
| 126 |
model = ContinuousTimeBertForMLM.load_pretrained(local_dir, **kwargs)
|
| 127 |
-
|
| 128 |
device = device or ("cuda" if torch.cuda.is_available() else "cpu")
|
| 129 |
model.to(device).eval()
|
| 130 |
return tokenizer, model
|
|
|
|
| 88 |
|
| 89 |
def save_pretrained(self, save_dir):
|
| 90 |
os.makedirs(save_dir, exist_ok=True)
|
| 91 |
+
self.model.save_pretrained(save_dir)
|
| 92 |
torch.save(self.time_embed.state_dict(), os.path.join(save_dir, "time_embed.pt"))
|
| 93 |
torch.save(self.post_inject_norm.state_dict(), os.path.join(save_dir, "post_inject_norm.pt"))
|
| 94 |
+
|
| 95 |
@classmethod
|
| 96 |
def load_pretrained(cls, save_dir, hidden_size=768, **kwargs):
|
| 97 |
+
model = AutoModelForMaskedLM.from_pretrained(save_dir)
|
| 98 |
obj = cls(model, hidden_size, **kwargs)
|
| 99 |
obj.time_embed.load_state_dict(torch.load(os.path.join(save_dir, "time_embed.pt"), map_location="cpu"))
|
| 100 |
obj.post_inject_norm.load_state_dict(torch.load(os.path.join(save_dir, "post_inject_norm.pt"), map_location="cpu"))
|
|
|
|
| 107 |
else:
|
| 108 |
local_dir = snapshot_download(repo_id_or_path)
|
| 109 |
|
| 110 |
+
try:
|
| 111 |
+
candidate_tokenizer = AutoTokenizer.from_pretrained(local_dir)
|
| 112 |
+
tokenizer = candidate_tokenizer if len(candidate_tokenizer) >= 1000 else None
|
| 113 |
+
except Exception:
|
| 114 |
+
tokenizer = None
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 115 |
|
| 116 |
if tokenizer is None:
|
| 117 |
print(f"[load_continuous_time_model] No valid tokenizer found in {repo_id_or_path}, "
|
|
|
|
| 119 |
tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL_ID)
|
| 120 |
|
| 121 |
model = ContinuousTimeBertForMLM.load_pretrained(local_dir, **kwargs)
|
|
|
|
| 122 |
device = device or ("cuda" if torch.cuda.is_available() else "cpu")
|
| 123 |
model.to(device).eval()
|
| 124 |
return tokenizer, model
|