Merge pull request #6078 from wtmlon/support-efficient-tokens-calcula…

…tion

support effective tokens calculation on sft/dpo

src/llamafactory/extras/misc.py

@@ -20,6 +20,7 @@
 from typing import TYPE_CHECKING, Tuple, Union
 
 import torch
+import torch.distributed as dist
 import transformers.dynamic_module_utils
 from transformers import InfNanRemoveLogitsProcessor, LogitsProcessorList
 from transformers.dynamic_module_utils import get_relative_imports
@@ -263,3 +264,11 @@ def use_modelscope() -> bool:
 
 def use_openmind() -> bool:
     return os.environ.get("USE_OPENMIND_HUB", "0").lower() in ["true", "1"]
+
+
+def cal_effective_tokens(effective_token_num, epoch, train_runtime) -> int:
+    r"""
+    calculate effective tokens.
+    """
+    result = effective_token_num * epoch / train_runtime
+    return result / dist.get_world_size() if dist.is_initialized() else result

src/llamafactory/hparams/finetuning_args.py

@@ -346,6 +346,10 @@ class FinetuningArguments(FreezeArguments, LoraArguments, RLHFArguments, GaloreA
         default=False,
         metadata={"help": "Whether or not to save the training loss curves."},
     )
+    include_effective_tokens_per_second: bool = field(
+        default=False,
+        metadata={"help": "Whether or not to compute effective tokens per second."},
+    )
 
     def __post_init__(self):
         def split_arg(arg):

src/llamafactory/train/dpo/workflow.py

@@ -19,6 +19,7 @@
 
 from ...data import PairwiseDataCollatorWithPadding, get_dataset, get_template_and_fix_tokenizer
 from ...extras.constants import IGNORE_INDEX
+from ...extras.misc import cal_effective_tokens
 from ...extras.ploting import plot_loss
 from ...hparams import ModelArguments
 from ...model import load_model, load_tokenizer
@@ -64,6 +65,12 @@ def run_dpo(
     # Update arguments
     training_args.remove_unused_columns = False  # important for multimodal and pairwise dataset
 
+    effective_token_num = 0.0
+    if finetuning_args.include_effective_tokens_per_second:
+        for data in dataset_module["train_dataset"]:
+            effective_token_num += len(data["chosen_input_ids"])
+            effective_token_num += len(data["rejected_input_ids"])
+
     # Initialize our Trainer
     trainer = CustomDPOTrainer(
         model=model,
@@ -79,6 +86,12 @@ def run_dpo(
     # Training
     if training_args.do_train:
         train_result = trainer.train(resume_from_checkpoint=training_args.resume_from_checkpoint)
+
+        if finetuning_args.include_effective_tokens_per_second:
+            train_result.metrics["effective_tokens_per_sec"] = cal_effective_tokens(
+                effective_token_num, train_result.metrics["epoch"], train_result.metrics["train_runtime"]
+            )
+
         trainer.save_model()
         trainer.log_metrics("train", train_result.metrics)
         trainer.save_metrics("train", train_result.metrics)

src/llamafactory/train/sft/workflow.py

@@ -19,7 +19,7 @@
 
 from ...data import SFTDataCollatorWith4DAttentionMask, get_dataset, get_template_and_fix_tokenizer
 from ...extras.constants import IGNORE_INDEX
-from ...extras.misc import get_logits_processor
+from ...extras.misc import cal_effective_tokens, get_logits_processor
 from ...extras.ploting import plot_loss
 from ...model import load_model, load_tokenizer
 from ..trainer_utils import create_modelcard_and_push
@@ -65,6 +65,11 @@ def run_sft(
     training_args.generation_num_beams = data_args.eval_num_beams or training_args.generation_num_beams
     training_args.remove_unused_columns = False  # important for multimodal dataset
 
+    effective_token_num = 0.0
+    if finetuning_args.include_effective_tokens_per_second:
+        for data in dataset_module["train_dataset"]:
+            effective_token_num += len(data["input_ids"])
+
     # Metric utils
     metric_module = {}
     if training_args.predict_with_generate:
@@ -94,6 +99,11 @@ def run_sft(
     # Training
     if training_args.do_train:
         train_result = trainer.train(resume_from_checkpoint=training_args.resume_from_checkpoint)
+        if finetuning_args.include_effective_tokens_per_second:
+            train_result.metrics["effective_tokens_per_sec"] = cal_effective_tokens(
+                effective_token_num, train_result.metrics["epoch"], train_result.metrics["train_runtime"]
+            )
+
         trainer.save_model()
         trainer.log_metrics("train", train_result.metrics)
         trainer.save_metrics("train", train_result.metrics)