Text Classification
PEFT
Czech
File size: 1,713 Bytes
31c8699
11bec6e
 
56de430
11bec6e
56de430
 
 
 
 
 
 
 
11bec6e
31c8699
56de430
 
 
 
 
31c8699
 
 
 
 
56de430
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
---
language:
- cs
license: cc-by-sa-4.0
library_name: peft
datasets:
- ctu-aic/csfever_v2
metrics:
- accuracy
- f1
- recall
- precision
pipeline_tag: text-classification
base_model: deepset/xlm-roberta-large-squad2
---
# Model card for lora-xlm-roberta-large-squad2-csfever_v2-f1

## Model details
Model for natural language inference.

## Training procedure

### Framework versions

- PEFT 0.4.0

## Uses

### PEFT (Transformers)
```python
from peft import PeftModel, PeftConfig
from transformers import AutoModelForSequenceClassification, Pipeline, AutoTokenizer

config = PeftConfig.from_pretrained("ctu-aic/lora-xlm-roberta-large-squad2-csfever_v2-f1")
model = AutoModelForSequenceClassification.from_pretrained(config.base_model_name_or_path)
model = PeftModel.from_pretrained(model, config)
tokenizer = AutoTokenizer.from_pretrained(config.base_model_name_or_path)

#pipeline for NLI
class NliPipeline(Pipeline):
    def _sanitize_parameters(self, **kwargs):
        preprocess_kwargs = {}
        if "evidence" in kwargs:
            preprocess_kwargs["evidence"] = kwargs["evidence"]
        return preprocess_kwargs, {}, {}

    def preprocess(self, claim, evidence=""):
        model_input = self.tokenizer(claim, evidence, return_tensors=self.framework, truncation=True)
        return model_input
    def _forward(self, model_inputs):
        outputs = self.model(**model_inputs)
        return outputs

    def postprocess(self, model_outputs):
        logits = model_outputs.logits

        predictions = torch.argmax(logits, dim=-1)
        return {"logits": logits, "label": int(predictions[0])}

nli_pipeline = NliPipeline(model=model, tokenizer=tokenizer)

nli_pipeline("claim", "evidence")
```