add eval acc

Former-commit-id: 1856a08e87b150fa4bffcb0af703ed84d848e24b

src/llamafactory/train/sft/metric.py

@@ -17,9 +17,11 @@
 # limitations under the License.
 
 from dataclasses import dataclass
-from typing import TYPE_CHECKING, Dict, Sequence, Tuple, Union
+from typing import TYPE_CHECKING, Dict
 
 import numpy as np
+import torch
+from transformers import EvalPrediction
 from transformers.utils import is_jieba_available, is_nltk_available
 
 from ...extras.constants import IGNORE_INDEX
@@ -42,6 +44,22 @@ if is_rouge_available():
     from rouge_chinese import Rouge
 
 
+def compute_accuracy(eval_preds: "EvalPrediction") -> Dict[str, float]:
+    preds, labels = eval_preds.predictions, eval_preds.label_ids
+    accuracies = []
+    for i in range(len(preds)):
+        pred, label = preds[i, 1:], labels[i, :-1]
+        label_mask = label != IGNORE_INDEX
+        accuracies.append(np.mean(pred[label_mask] == label[label_mask]))
+
+    return {"accuracy": float(np.mean(accuracies))}
+
+
+def eval_logit_processor(logits: "torch.Tensor", labels: "torch.Tensor") -> "torch.Tensor":
+    logits = logits[0] if isinstance(logits, (list, tuple)) else logits
+    return torch.argmax(logits, dim=-1)
+
+
 @dataclass
 class ComputeMetrics:
     r"""
@@ -50,11 +68,11 @@ class ComputeMetrics:
 
     tokenizer: "PreTrainedTokenizer"
 
-    def __call__(self, eval_preds: Sequence[Union[np.ndarray, Tuple[np.ndarray]]]) -> Dict[str, float]:
+    def __call__(self, eval_preds: "EvalPrediction") -> Dict[str, float]:
         r"""
         Uses the model predictions to compute metrics.
         """
-        preds, labels = eval_preds
+        preds, labels = eval_preds.predictions, eval_preds.label_ids
         score_dict = {"rouge-1": [], "rouge-2": [], "rouge-l": [], "bleu-4": []}
 
         preds = np.where(preds != IGNORE_INDEX, preds, self.tokenizer.pad_token_id)

src/llamafactory/train/sft/trainer.py

@@ -135,21 +135,16 @@ class CustomSeq2SeqTrainer(Seq2SeqTrainer):
 
         for i in range(len(preds)):
             pad_len = np.nonzero(preds[i] != self.tokenizer.pad_token_id)[0]
-            if len(pad_len):
+            if len(pad_len):  # move pad token to last
-                preds[i] = np.concatenate(
+                preds[i] = np.concatenate((preds[i][pad_len[0] :], preds[i][: pad_len[0]]), axis=-1)
-                    (preds[i][pad_len[0] :], preds[i][: pad_len[0]]), axis=-1
-                )  # move pad token to last
 
-        decoded_inputs = self.tokenizer.batch_decode(
+        decoded_inputs = self.tokenizer.batch_decode(dataset["input_ids"], skip_special_tokens=True)
-            dataset["input_ids"], skip_special_tokens=True, clean_up_tokenization_spaces=False
+        decoded_labels = self.tokenizer.batch_decode(labels, skip_special_tokens=True)
-        )
+        decoded_preds = self.tokenizer.batch_decode(preds, skip_special_tokens=True)
-        decoded_labels = self.tokenizer.batch_decode(
-            labels, skip_special_tokens=True, clean_up_tokenization_spaces=False
-        )
-        decoded_preds = self.tokenizer.batch_decode(preds, skip_special_tokens=True, clean_up_tokenization_spaces=True)
 
         with open(output_prediction_file, "w", encoding="utf-8") as writer:
             res: List[str] = []
             for text, label, pred in zip(decoded_inputs, decoded_labels, decoded_preds):
                 res.append(json.dumps({"prompt": text, "label": label, "predict": pred}, ensure_ascii=False))
+
             writer.write("\n".join(res))

src/llamafactory/train/sft/workflow.py

@@ -25,7 +25,7 @@ from ...extras.misc import get_logits_processor
 from ...extras.ploting import plot_loss
 from ...model import load_model, load_tokenizer
 from ..trainer_utils import create_modelcard_and_push
-from .metric import ComputeMetrics
+from .metric import ComputeMetrics, compute_accuracy, eval_logit_processor
 from .trainer import CustomSeq2SeqTrainer
 
 
@@ -72,7 +72,8 @@ def run_sft(
         finetuning_args=finetuning_args,
         data_collator=data_collator,
         callbacks=callbacks,
-        compute_metrics=ComputeMetrics(tokenizer) if training_args.predict_with_generate else None,
+        compute_metrics=ComputeMetrics(tokenizer) if training_args.predict_with_generate else compute_accuracy,
+        preprocess_logits_for_metrics=None if training_args.predict_with_generate else eval_logit_processor,
         **tokenizer_module,
         **split_dataset(dataset, data_args, training_args),
     )
@@ -91,7 +92,7 @@ def run_sft(
         trainer.save_metrics("train", train_result.metrics)
         trainer.save_state()
         if trainer.is_world_process_zero() and finetuning_args.plot_loss:
-            plot_loss(training_args.output_dir, keys=["loss", "eval_loss"])
+            plot_loss(training_args.output_dir, keys=["loss", "eval_loss", "eval_accuracy"])
 
     # Evaluation
     if training_args.do_eval: