Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,37 @@
|
|
1 |
-
|
2 |
-
|
3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
Since Huggingface has omitte to publish a standalone SmolLM2_360M_model.py to load and finetune and run inference of the released model weights at https://huggingface.co/HuggingFaceTB/SmolLM2-360M/
|
2 |
+
I have attempte to construct a model.py that can load and at least inference mode the weights. The SmolLM2_360M_model.py runs but is unable to load the safetensors data. Here is the encountered error:
|
3 |
+
|
4 |
+
C:\Users\User\OneDrive\Desktop\SmolLM2>python SmolLM2_360M_model_debugging.py
|
5 |
+
Warning: SentencePiece not found, using rudimentary BPE tokenizer. Install SentencePiece for better performance.
|
6 |
+
|
7 |
+
A module that was compiled using NumPy 1.x cannot be run in
|
8 |
+
NumPy 2.1.3 as it may crash. To support both 1.x and 2.x
|
9 |
+
versions of NumPy, modules must be compiled with NumPy 2.0.
|
10 |
+
Some module may need to rebuild instead e.g. with 'pybind11>=2.12'.
|
11 |
+
|
12 |
+
If you are a user of the module, the easiest solution will be to
|
13 |
+
downgrade to 'numpy<2' or try to upgrade the affected module.
|
14 |
+
We expect that some modules will need time to support NumPy 2.
|
15 |
+
|
16 |
+
Traceback (most recent call last): File "C:\Users\User\OneDrive\Desktop\SmolLM2\SmolLM2_360M_model_debugging.py", line 470, in <module>
|
17 |
+
model = SmolLM2_360M(config_path)
|
18 |
+
File "C:\Users\User\OneDrive\Desktop\SmolLM2\SmolLM2_360M_model_debugging.py", line 243, in __init__
|
19 |
+
self.embed_tokens = nn.Embedding(self.vocab_size, self.hidden_size)
|
20 |
+
File "C:\Users\User\AppData\Local\Programs\Python\Python311\Lib\site-packages\torch\nn\modules\sparse.py", line 142, in __init__
|
21 |
+
self.weight = Parameter(torch.empty((num_embeddings, embedding_dim), **factory_kwargs),
|
22 |
+
C:\Users\User\AppData\Local\Programs\Python\Python311\Lib\site-packages\torch\nn\modules\sparse.py:142: UserWarning: Failed to initialize NumPy: _ARRAY_API not found (Triggered internally at ..\torch\csrc\utils\tensor_numpy.cpp:84.)
|
23 |
+
self.weight = Parameter(torch.empty((num_embeddings, embedding_dim), **factory_kwargs),
|
24 |
+
An error occurred while loading weights: File does not contain tensor lm_head.weight
|
25 |
+
|
26 |
+
C:\Users\User\OneDrive\Desktop\SmolLM2>
|
27 |
+
|
28 |
+
So, what is the story with safetensors "File does not contain tensor lm_head.weight"
|
29 |
+
|
30 |
+
Is there a python script for inspecting the safetensors file?
|
31 |
+
|
32 |
+
Why does model.safetensors file "not contain tensor lm_head.weight"?
|
33 |
+
|
34 |
+
|
35 |
+
---
|
36 |
+
license: apache-2.0
|
37 |
+
---
|