Merge pull request #271 from OpenAccess-AI-Collective/quadratic-warmup

Quadratic warmup

src/axolotl/utils/schedulers.py

@@ -1,6 +1,9 @@
 """Module for custom LRScheduler class"""
+import math
+from functools import partial
 
-from torch.optim.lr_scheduler import LRScheduler
+from torch.optim import Optimizer
+from torch.optim.lr_scheduler import LambdaLR, LRScheduler
 
 
 class InterpolatingLogScheduler(LRScheduler):
@@ -42,3 +45,58 @@ def get_lr(self):
             lrs = [self.max_lr for base_lr in self.base_lrs]
 
         return lrs
+
+
+def _get_cosine_schedule_with_quadratic_warmup_lr_lambda(
+    current_step: int,
+    *,
+    num_warmup_steps: int,
+    num_training_steps: int,
+    num_cycles: float
+):
+    if current_step < num_warmup_steps:
+        return (float(current_step) / float(max(1, num_warmup_steps))) ** 2
+    progress = float(current_step - num_warmup_steps) / float(
+        max(1, num_training_steps - num_warmup_steps)
+    )
+    return max(
+        0.0, 0.5 * (1.0 + math.cos(math.pi * float(num_cycles) * 2.0 * progress))
+    )
+
+
+def get_cosine_schedule_with_quadratic_warmup(
+    optimizer: Optimizer,
+    num_warmup_steps: int,
+    num_training_steps: int,
+    num_cycles: float = 0.5,
+    last_epoch: int = -1,
+):
+    """
+    Create a schedule with a learning rate that decreases following the values of the cosine function between the
+    initial lr set in the optimizer to 0, after a warmup period during which it increases linearly between 0 and the
+    initial lr set in the optimizer.
+
+    Args:
+        optimizer ([`~torch.optim.Optimizer`]):
+            The optimizer for which to schedule the learning rate.
+        num_warmup_steps (`int`):
+            The number of steps for the warmup phase.
+        num_training_steps (`int`):
+            The total number of training steps.
+        num_cycles (`float`, *optional*, defaults to 0.5):
+            The number of waves in the cosine schedule (the defaults is to just decrease from the max value to 0
+            following a half-cosine).
+        last_epoch (`int`, *optional*, defaults to -1):
+            The index of the last epoch when resuming training.
+
+    Return:
+        `torch.optim.lr_scheduler.LambdaLR` with the appropriate schedule.
+    """
+
+    lr_lambda = partial(
+        _get_cosine_schedule_with_quadratic_warmup_lr_lambda,
+        num_warmup_steps=num_warmup_steps,
+        num_training_steps=num_training_steps,
+        num_cycles=num_cycles,
+    )
+    return LambdaLR(optimizer, lr_lambda, last_epoch)

src/axolotl/utils/trainer.py

@@ -5,6 +5,7 @@
 import math
 import os
 import sys
+from dataclasses import field
 from pathlib import Path
 from typing import Optional
 
@@ -13,17 +14,67 @@
 import transformers
 from torch import nn
 from torch.optim.lr_scheduler import OneCycleLR
-from transformers import EarlyStoppingCallback, Trainer
+from transformers import EarlyStoppingCallback, Trainer, TrainingArguments
 from transformers.trainer_pt_utils import get_parameter_names
 
 from axolotl.utils.callbacks import (
     SaveBetterTransformerModelCallback,
     SavePeftModelCallback,
 )
-from axolotl.utils.schedulers import InterpolatingLogScheduler
+from axolotl.utils.schedulers import (
+    InterpolatingLogScheduler,
+    get_cosine_schedule_with_quadratic_warmup,
+)
 
 
-class OneCycleLRSchedulerTrainer(Trainer):
+class AxolotlTrainingArguments(TrainingArguments):
+    """
+    Extend the base TrainingArguments for axolotl helpers
+    """
+
+    lr_quadratic_warmup: bool = field(
+        default=False,
+        metadata={"help": "Use quadratic warmup for cosine scheduling."},
+    )
+
+
+class AxolotlTrainer(Trainer):
+    """
+    Extend the base Trainer for axolotl helpers
+    """
+
+    args = None  # type: AxolotlTrainingArguments
+
+    def create_scheduler(
+        self, num_training_steps: int, optimizer: torch.optim.Optimizer = None
+    ):
+        """
+        Setup the scheduler. The optimizer of the trainer must have been set up either before this method is called or
+        passed as an argument.
+
+        Args:
+            num_training_steps (int): The number of training steps to do.
+            optimizer (torch.optim.Optimizer): The training optimizer
+        """
+
+        # fmt: off
+        if self.lr_scheduler is None:  # type: ignore  # pylint: disable=access-member-before-definition
+            # fmt: on
+            if (
+                self.args.lr_scheduler_type == "cosine"
+                and self.args.lr_quadratic_warmup is True
+            ):
+                self.lr_scheduler = get_cosine_schedule_with_quadratic_warmup(  # pylint: disable=attribute-defined-outside-init
+                    optimizer,
+                    num_warmup_steps=self.args.get_warmup_steps(num_training_steps),
+                    num_training_steps=num_training_steps,
+                )
+            else:
+                return super().create_scheduler(num_training_steps, optimizer)
+        return self.lr_scheduler
+
+
+class OneCycleLRSchedulerTrainer(AxolotlTrainer):
     """
     Trainer subclass that uses the OneCycleLR scheduler
     """
@@ -103,6 +154,9 @@ def setup_trainer(cfg, train_dataset, eval_dataset, model, tokenizer):
         if cfg.fsdp_config:
             training_arguments_kwargs["fsdp_config"] = dict(cfg.fsdp_config)
 
+    if cfg.lr_quadratic_warmup is not None:
+        training_arguments_kwargs["lr_quadratic_warmup"] = cfg.lr_quadratic_warmup
+
     # deepspeed
     if (
         os.environ.get("ACCELERATE_USE_DEEPSPEED") == "true"
@@ -128,7 +182,7 @@ def setup_trainer(cfg, train_dataset, eval_dataset, model, tokenizer):
         training_arguments_kwargs["hub_model_id"] = cfg.hub_model_id
         training_arguments_kwargs["push_to_hub"] = True
 
-    training_args = transformers.TrainingArguments(
+    training_args = AxolotlTrainingArguments(
         per_device_train_batch_size=cfg.micro_batch_size,
         per_device_eval_batch_size=cfg.eval_batch_size
         if cfg.eval_batch_size is not None
@@ -278,7 +332,7 @@ def setup_trainer(cfg, train_dataset, eval_dataset, model, tokenizer):
     trainer_cls = (
         OneCycleLRSchedulerTrainer
         if cfg.lr_scheduler == "one_cycle" and (cfg.fsdp or cfg.adapter == "qlora")
-        else transformers.Trainer
+        else AxolotlTrainer
     )
     trainer = trainer_cls(
         model=model,