File size: 1,363 Bytes
1f99ce2
 
 
 
 
 
 
 
 
 
 
a0d99f6
 
1f99ce2
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6949167
 
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
---
license: apache-2.0
datasets:
- bookcorpus
- wikipedia
language:
- en
metrics:
- glue
pipeline_tag: text-classification
---
Evaluate on MNLI:
```python
from transformers import (
    default_data_collator,
    AutoTokenizer,
    AutoModelForSequenceClassification,
    Trainer,
)
from datasets import load_dataset

import functools

from utils import compute_metrics, preprocess_function

model_name = "George-Ogden/bert-large-cased-finetuned-mnli"
model = AutoModelForSequenceClassification.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(model_name)
trainer = Trainer(
    model=model,
    eval_dataset="mnli",
    tokenizer=tokenizer,
    compute_metrics=compute_metrics,
    data_collator=default_data_collator,
)

raw_datasets = load_dataset(
    "glue",
    "mnli",
).map(functools.partial(preprocess_function, tokenizer), batched=True)

tasks = ["mnli", "mnli-mm"]
eval_datasets = [
    raw_datasets["validation_matched"],
    raw_datasets["validation_mismatched"],
]

for layers in reversed(range(model.num_layers + 1)):
    for eval_dataset, task in zip(eval_datasets, tasks):
        metrics = trainer.evaluate(eval_dataset=eval_dataset)
        metrics["eval_samples"] = len(eval_dataset)

        if task == "mnli-mm":
            metrics = {k + "_mm": v for k, v in metrics.items()}

        trainer.log_metrics(metrics)
```