optimization.py

"""
"""

from typing import Any
from typing import Callable
from typing import ParamSpec
from torchao.quantization import quantize_
from torchao.quantization import Float8DynamicActivationFloat8WeightConfig
import spaces
import torch
from torch.utils._pytree import tree_map, tree_map_only


P = ParamSpec('P')


TRANSFORMER_HIDDEN_DIM = torch.export.Dim.AUTO(min=3584, max=9727)

TRANSFORMER_DYNAMIC_SHAPES = {
    'hidden_states': {1: TRANSFORMER_HIDDEN_DIM},
}


INDUCTOR_CONFIGS = {
    'conv_1x1_as_mm': True,
    'epilogue_fusion': False,
    'coordinate_descent_tuning': True,
    'coordinate_descent_check_all_directions': True,
    'max_autotune': True,
    'triton.cudagraphs': True,
}


def optimize_pipeline_(pipeline: Callable[P, Any], *args: P.args, **kwargs: P.kwargs):

    @spaces.GPU(duration=1500)
    def compile_transformer():

        with spaces.aoti_capture(pipeline.transformer) as call:
            pipeline(*args, **kwargs)

        dynamic_shapes = tree_map(lambda t: None, call.kwargs)
        # dynamic_shapes = tree_map_only((torch.Tensor, bool), lambda x: None, call.kwargs)
        dynamic_shapes |= TRANSFORMER_DYNAMIC_SHAPES

        quantize_(pipeline.transformer, Float8DynamicActivationFloat8WeightConfig())
        
        exported = torch.export.export(
            mod=pipeline.transformer,
            args=call.args,
            kwargs=call.kwargs,
            # dynamic_shapes=dynamic_shapes,
        )

        return spaces.aoti_compile(exported, INDUCTOR_CONFIGS)

    spaces.aoti_apply(compile_transformer(), pipeline.transformer)