Update README.md
Browse files
README.md
CHANGED
@@ -10,7 +10,20 @@ tags:
|
|
10 |
---
|
11 |
# TinyLlama-1.1B-32k
|
12 |
|
13 |
-
#### NOTE: This is a fork of the original model at https://huggingface.co/Doctor-Shotgun/TinyLlama-1.1B-32k but with fixed safetensors
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
14 |
|
15 |
32k context finetune of TinyLlama-1.1B using increased rope theta (rope frequency base) meant to serve as a long-context speculative decoding model.
|
16 |
|
|
|
10 |
---
|
11 |
# TinyLlama-1.1B-32k
|
12 |
|
13 |
+
#### NOTE: This is a fork of the original model at https://huggingface.co/Doctor-Shotgun/TinyLlama-1.1B-32k but with fixed safetensors metadata using the following code:
|
14 |
+
|
15 |
+
```
|
16 |
+
import safetensors
|
17 |
+
from safetensors.torch import save_file
|
18 |
+
|
19 |
+
tensors = dict()
|
20 |
+
with safetensors.safe_open(safetensors_path, framework="pt") as f:
|
21 |
+
for key in f.keys():
|
22 |
+
tensors[key] = f.get_tensor(key)
|
23 |
+
|
24 |
+
save_file(tensors, safetensors_path, metadata={'format': 'pt'})
|
25 |
+
```
|
26 |
+
(from https://huggingface.co/SeaLLMs/SeaLLM-7B-Hybrid/discussions/2#65752144412ee70185d49ff5)
|
27 |
|
28 |
32k context finetune of TinyLlama-1.1B using increased rope theta (rope frequency base) meant to serve as a long-context speculative decoding model.
|
29 |
|