Usage example
Browse files
README.md
CHANGED
@@ -8,3 +8,35 @@ language:
|
|
8 |
|
9 |
This is a Mistral7B model fine-tuned with QLoRA on Czech Wikipedia data. The model is primarily designed for further fine-tuning for Czech-specific NLP tasks, including summarization and question answering. This adaptation allows for better performance in tasks that require an understanding of the Czech language and context.
|
10 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
8 |
|
9 |
This is a Mistral7B model fine-tuned with QLoRA on Czech Wikipedia data. The model is primarily designed for further fine-tuning for Czech-specific NLP tasks, including summarization and question answering. This adaptation allows for better performance in tasks that require an understanding of the Czech language and context.
|
10 |
|
11 |
+
Example of usage:
|
12 |
+
|
13 |
+
```python
|
14 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
15 |
+
import torch
|
16 |
+
|
17 |
+
model_name = "simecek/cswikimistral_0.1"
|
18 |
+
device = "cuda" if torch.cuda.is_available() else "cpu"
|
19 |
+
|
20 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
21 |
+
model = AutoModelForCausalLM.from_pretrained(model_name, device_map="auto", load_in_4bit=True)
|
22 |
+
|
23 |
+
def generate_text(prompt, max_new_tokens=50):
|
24 |
+
inputs = tokenizer(prompt, return_tensors="pt").to(device)
|
25 |
+
attention_mask = inputs["attention_mask"]
|
26 |
+
input_ids = inputs["input_ids"]
|
27 |
+
|
28 |
+
output = model.generate(
|
29 |
+
input_ids,
|
30 |
+
attention_mask=attention_mask,
|
31 |
+
max_new_tokens=max_new_tokens,
|
32 |
+
num_return_sequences=1,
|
33 |
+
pad_token_id=tokenizer.eos_token_id,
|
34 |
+
)
|
35 |
+
|
36 |
+
return tokenizer.decode(output[0], skip_special_tokens=True)
|
37 |
+
|
38 |
+
prompt = "Hlavní město České republiky je"
|
39 |
+
generated_text = generate_text(prompt, max_new_tokens=5)
|
40 |
+
print(generated_text)
|
41 |
+
```
|
42 |
+
|