Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
Trained : Reconstruction tokens
|
2 |
+
|
3 |
+
```python
|
4 |
+
import torch
|
5 |
+
|
6 |
+
from safetensors.torch import load_file
|
7 |
+
from huggingface_hub import hf_hub_download
|
8 |
+
|
9 |
+
from semo_lm.model import SemoLlama
|
10 |
+
from semo_lm.semo_utils.prefix_vars import PAD_TOKEN_ID
|
11 |
+
|
12 |
+
model = SemoLlama.from_pretrained(
|
13 |
+
"meta-llama/Meta-Llama-3-8B-Instruct",
|
14 |
+
torch_dtype=torch.bfloat16,
|
15 |
+
pad_token_id=PAD_TOKEN_ID
|
16 |
+
)
|
17 |
+
|
18 |
+
model.init_sentence_encoder_weights()
|
19 |
+
|
20 |
+
repo_id = "MLP-SEMO/Llama-Reconstruction-embedding"
|
21 |
+
filename = "embed_tokens.safetensors"
|
22 |
+
|
23 |
+
downloaded_file = hf_hub_download(repo_id=repo_id, filename=filename)
|
24 |
+
embedding_weights = load_file(downloaded_file)
|
25 |
+
model.model.embed_tokens.load_state_dict(embedding_weights)
|
26 |
+
```
|