Commit
•
babfc31
1
Parent(s):
55efecd
update README - usage of tokenizer
Browse files
README.md
CHANGED
@@ -14,27 +14,29 @@ You could find the original weights released by [xAI](https://x.ai/blog) in [Hug
|
|
14 |
|
15 |
We translated the original modeling written in JAX into PyTorch version, and converted the weights by mapping tensor files with parameter keys, de-quantizing the tensors with corresponding packed scales, and save to checkpoint file with torch APIs.
|
16 |
|
17 |
-
|
18 |
|
19 |
## Usage
|
20 |
|
21 |
```python
|
22 |
import torch
|
23 |
-
from transformers import AutoModelForCausalLM
|
24 |
-
from sentencepiece import SentencePieceProcessor
|
25 |
|
26 |
torch.set_default_dtype(torch.bfloat16)
|
|
|
|
|
|
|
27 |
model = AutoModelForCausalLM.from_pretrained(
|
28 |
"hpcai-tech/grok-1",
|
29 |
trust_remote_code=True,
|
30 |
device_map="auto",
|
31 |
torch_dtype=torch.bfloat16,
|
32 |
)
|
33 |
-
|
34 |
|
35 |
text = "Replace this with your text"
|
36 |
-
input_ids =
|
37 |
-
input_ids =
|
38 |
attention_mask = torch.ones_like(input_ids)
|
39 |
generate_kwargs = {} # Add any additional args if you want
|
40 |
inputs = {
|
@@ -43,14 +45,7 @@ inputs = {
|
|
43 |
**generate_kwargs,
|
44 |
}
|
45 |
outputs = model.generate(**inputs)
|
46 |
-
|
47 |
-
|
48 |
-
You could also use the transformers-compatible version of the tokenizer [Xenova/grok-1-tokenizer](https://huggingface.co/Xenova/grok-1-tokenizer)
|
49 |
-
```python
|
50 |
-
from transformers import LlamaTokenizerFast
|
51 |
-
|
52 |
-
tokenizer = LlamaTokenizerFast.from_pretrained('Xenova/grok-1-tokenizer')
|
53 |
-
inputs = tokenizer('hello world')
|
54 |
```
|
55 |
|
56 |
|
|
|
14 |
|
15 |
We translated the original modeling written in JAX into PyTorch version, and converted the weights by mapping tensor files with parameter keys, de-quantizing the tensors with corresponding packed scales, and save to checkpoint file with torch APIs.
|
16 |
|
17 |
+
A transformers-compatible version of tokenizer is contributed by [Xenova](https://huggingface.co/Xenova) and [ArthurZ](https://huggingface.co/ArthurZ).
|
18 |
|
19 |
## Usage
|
20 |
|
21 |
```python
|
22 |
import torch
|
23 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
|
|
24 |
|
25 |
torch.set_default_dtype(torch.bfloat16)
|
26 |
+
|
27 |
+
tokenizer = AutoTokenizer.from_pretrained("hpcai-tech/grok-1", trust_remote_code=True)
|
28 |
+
|
29 |
model = AutoModelForCausalLM.from_pretrained(
|
30 |
"hpcai-tech/grok-1",
|
31 |
trust_remote_code=True,
|
32 |
device_map="auto",
|
33 |
torch_dtype=torch.bfloat16,
|
34 |
)
|
35 |
+
model.eval()
|
36 |
|
37 |
text = "Replace this with your text"
|
38 |
+
input_ids = tokenizer(text, return_tensors="pt").input_ids
|
39 |
+
input_ids = input_ids.cuda()
|
40 |
attention_mask = torch.ones_like(input_ids)
|
41 |
generate_kwargs = {} # Add any additional args if you want
|
42 |
inputs = {
|
|
|
45 |
**generate_kwargs,
|
46 |
}
|
47 |
outputs = model.generate(**inputs)
|
48 |
+
print(outputs)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
49 |
```
|
50 |
|
51 |
|