Update README.md
Browse files
README.md
CHANGED
@@ -2,4 +2,86 @@
|
|
2 |
license: mit
|
3 |
base_model:
|
4 |
- distilbert/distilbert-base-uncased
|
5 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
license: mit
|
3 |
base_model:
|
4 |
- distilbert/distilbert-base-uncased
|
5 |
+
---
|
6 |
+
|
7 |
+
Deepest apologies for how fucked up this is, but:
|
8 |
+
|
9 |
+
```
|
10 |
+
import os
|
11 |
+
import sys
|
12 |
+
import json
|
13 |
+
import torch
|
14 |
+
from huggingface_hub import hf_hub_download
|
15 |
+
import importlib.util
|
16 |
+
|
17 |
+
# Repository ID and filenames
|
18 |
+
repo_id = "dgaff/bsky_user_classifier"
|
19 |
+
files_to_download = {
|
20 |
+
"model_weights": "multioutput_regressor.pth",
|
21 |
+
"train_script": "train.py",
|
22 |
+
"data_processing": "data_processing.py",
|
23 |
+
"utils": "utils.py",
|
24 |
+
"label_mappings": "label_mappings.json",
|
25 |
+
}
|
26 |
+
|
27 |
+
# Download necessary files
|
28 |
+
model_weights_path = hf_hub_download(repo_id=repo_id, filename=files_to_download["model_weights"])
|
29 |
+
train_script_path = hf_hub_download(repo_id=repo_id, filename=files_to_download["train_script"])
|
30 |
+
data_processing_path = hf_hub_download(repo_id=repo_id, filename=files_to_download["data_processing"])
|
31 |
+
util_path = hf_hub_download(repo_id=repo_id, filename=files_to_download["utils"])
|
32 |
+
label_mappings_path = hf_hub_download(repo_id=repo_id, filename=files_to_download["label_mappings"])
|
33 |
+
|
34 |
+
# Update sys.path to include dependencies
|
35 |
+
for path in [data_processing_path, util_path]:
|
36 |
+
dir_path = os.path.dirname(path)
|
37 |
+
if dir_path not in sys.path:
|
38 |
+
sys.path.append(dir_path)
|
39 |
+
|
40 |
+
# Load train.py as a module
|
41 |
+
spec = importlib.util.spec_from_file_location("train_module", train_script_path)
|
42 |
+
train_module = importlib.util.module_from_spec(spec)
|
43 |
+
sys.modules["train_module"] = train_module
|
44 |
+
spec.loader.exec_module(train_module)
|
45 |
+
|
46 |
+
# Load label mappings
|
47 |
+
with open(label_mappings_path) as f:
|
48 |
+
label_mappings = json.load(f)
|
49 |
+
|
50 |
+
# Initialize the model
|
51 |
+
hidden_size = 768 # Ensure this matches your model's configuration
|
52 |
+
num_outputs = 23 # Update if different
|
53 |
+
model = train_module.MultiOutputRegressor(hidden_size=hidden_size, num_outputs=num_outputs)
|
54 |
+
|
55 |
+
# Load weights and set model to evaluation mode
|
56 |
+
model.load_state_dict(torch.load(model_weights_path, map_location=torch.device('cpu')))
|
57 |
+
model.eval()
|
58 |
+
|
59 |
+
# Set device
|
60 |
+
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
|
61 |
+
model.to(device)
|
62 |
+
|
63 |
+
# Prepare input sentences and generate embeddings
|
64 |
+
new_sentences = [
|
65 |
+
"This is a test sentence.",
|
66 |
+
"Another example of a sentence to predict."
|
67 |
+
]
|
68 |
+
embedder = train_module.EmbeddingGenerator()
|
69 |
+
new_embeddings = embedder.generate_embeddings(new_sentences)
|
70 |
+
new_embeddings_tensor = torch.tensor(new_embeddings, dtype=torch.float).to(device)
|
71 |
+
|
72 |
+
# Generate predictions
|
73 |
+
with torch.no_grad():
|
74 |
+
predictions = model(new_embeddings_tensor).cpu().numpy()
|
75 |
+
|
76 |
+
# Map predictions to labels and print results
|
77 |
+
for sentence, pred in zip(new_sentences, predictions):
|
78 |
+
label_pred_dict = {label_mappings["id2label"][str(i)]: float(pred[i]) for i in range(len(pred))}
|
79 |
+
print(f"Sentence: {sentence}")
|
80 |
+
print("Predictions:")
|
81 |
+
for label, value in label_pred_dict.items():
|
82 |
+
print(f" {label}: {value}")
|
83 |
+
print()
|
84 |
+
|
85 |
+
```
|
86 |
+
|
87 |
+
I'll do better next time
|