gptq_quantization.py

import contextlib
import logging
import math
from typing import List, Optional

import torch
import transformers
from torch import nn

LOGGER = logging.getLogger(__name__)

QUANT_LAYERS = [nn.Linear, nn.Conv2d, transformers.Conv1D]

def is_transformer_conv1d(layer):
    return isinstance(layer, transformers.Conv1D)


# These two functions only work on per-channel symmetric quantization for weight
def get_weight_scale(weight, weight_bit_width):
    weight_scale = (weight.abs().max(dim=-1).values / ((2 ** (weight_bit_width - 1)) - 1)).half()
    return weight_scale

def fake_quantize_weight(weight, weight_scale):
    weight_scale = weight_scale[:, None]
    fake_quantized_weight = torch.round(weight / weight_scale) * weight_scale
    return fake_quantized_weight


class GPTQLayerWrapper:
    def __init__(self, layer_name, layer, weight_bit_width):
        super().__init__()
        self.layer_name = layer_name
        self.layer = layer
        self.device = layer.weight.device
        columns = layer.weight.shape[1]
        self.columns = columns
        self.H = torch.zeros((columns, columns), device=self.device)
        self.nsamples = 0
        self.is_record = True
        self.weight_bit_width = weight_bit_width
        self.weight_scale = None

    def record_h(self, x):
        if self.is_record:
            x = x.detach().clone()
            if len(x.shape) == 2:
                x = x.unsqueeze(0)
            batch = x.shape[0]
            if isinstance(self.layer, nn.Linear) or is_transformer_conv1d(self.layer):
                if len(x.shape) == 3:
                    x = x.reshape((-1, x.shape[-1]))
                x = x.t()

            if isinstance(self.layer, nn.Conv2d):
                unfold = nn.Unfold(
                    self.layer.kernel_size,
                    dilation=self.layer.dilation,
                    padding=self.layer.padding,
                    stride=self.layer.stride
                )
                x = unfold(x)
                x = x.permute([1, 0, 2])
                x = x.flatten(1)

            self.H *= self.nsamples / (self.nsamples + batch)
            self.nsamples += batch
            x = math.sqrt(2 / self.nsamples) * x.float()
            self.H += x.matmul(x.t())

    def quant_weight(self, blocksize=128, percdamp=.01, groupsize=-1):
        if groupsize != -1:
            raise RuntimeError("Group quantization of gptq quantizer is not supported for now")
        weight = self.layer.weight.data.clone()
        if isinstance(self.layer, nn.Conv2d):
            weight = weight.flatten(1)
        if is_transformer_conv1d(self.layer):
            weight = weight.t()
        weight = weight.float()

        weight_scale = get_weight_scale(weight, self.weight_bit_width)
        # todo: use buffer to store scale
        self.weight_scale = weight_scale
        H = self.H
        dead = torch.diag(H) == 0
        H[dead, dead] = 1
        weight[:, dead] = 0

        losses = torch.zeros_like(weight)
        Q = torch.zeros_like(weight)

        damp = percdamp * torch.mean(torch.diag(H))
        diag = torch.arange(self.columns, device=self.device)
        H[diag, diag] += damp
        try:
            H = torch.linalg.cholesky(H)
            H = torch.cholesky_inverse(H)
            H = torch.linalg.cholesky(H, upper=True)
        except Exception:
            logging.warning(f"Warning:  cannot do compression on layer {self.layer_name} because of inverse error")
            return

        if H.isnan().any():
            logging.warning(f"Warning:  cannot do compression on layer {self.layer_name} because of inverse error")
            return

        hinv = H

        for i1 in range(0, self.columns, blocksize):
            i2 = min(i1 + blocksize, self.columns)
            count = i2 - i1

            w1 = weight[:, i1:i2].clone()
            q1 = torch.zeros_like(w1)
            total_err = torch.zeros_like(w1)
            losses1 = torch.zeros_like(w1)
            hinv1 = hinv[i1:i2, i1:i2]

            for i in range(count):
                w = w1[:, i]
                d = hinv1[i, i]

                q = fake_quantize_weight(w.unsqueeze(1), weight_scale).flatten()

                q1[:, i] = q
                losses1[:, i] = (w - q) ** 2 / d ** 2
                err = (w - q) / d
                w1[:, i:] -= err.unsqueeze(1).matmul(hinv1[i, i:].unsqueeze(0))
                total_err[:, i] = err

            Q[:, i1:i2] = q1
            losses[:, i1:i2] = losses1 / 2

            weight[:, i2:] -= total_err.matmul(hinv[i1:i2, i2:])

        if torch.cuda.is_available():
            torch.cuda.synchronize()

        if is_transformer_conv1d(self.layer):
            Q = Q.t()
        self.layer.weight = nn.Parameter(Q.reshape(self.layer.weight.shape).to(self.layer.weight.data.dtype), requires_grad=False)

        del self.H
        if torch.cuda.is_available():
            torch.cuda.empty_cache()

    def release_gpu_memory(self):
        if hasattr(self, "H"):
            del self.H


class GPTQBlockWrapper:
    def __init__(self, module_name: str, module: nn.Module, weight_bit_width=8):
        self.layer_wrappers = {}
        self.hook_handles = []
        # module order in the whole network
        self.order = 0
        self.module_name = module_name

        def get_hook(layer_name):
            def record_hook(_, x):
                self.layer_wrappers[layer_name].record_h(x[0])
            return record_hook

        for layer_name, layer in module.named_modules():
            if isinstance(layer, tuple(QUANT_LAYERS)):
                full_layer_name = f"{module_name}.{layer_name}" if layer_name else f"{module_name}"
                self.layer_wrappers[full_layer_name] = GPTQLayerWrapper(full_layer_name, layer, weight_bit_width)
                handle = layer.register_forward_pre_hook(get_hook(full_layer_name))
                self.hook_handles.append(handle)

    def quant_module(self):
        for _, wrapper in self.layer_wrappers.items():
            wrapper.quant_weight()

        for h in self.hook_handles:
            h.remove()

    def set_order(self, idx):
        self.order = idx

    def get_order(self):
        return self.order

    def enable(self):
        for n, l in self.layer_wrappers.items():
            l.is_record = True

    def disable(self):
        for n, l in self.layer_wrappers.items():
            l.is_record = False

    def release_gpu_memory(self):
        for _, wrapper in self.layer_wrappers.items():
            wrapper.release_gpu_memory()


class GPTQuantizer:
    def __init__(self, block_type: Optional[List[type]] = None):
        self.gptq_block_wrappers = {}
        self.block_type = block_type

    def wrap_model(self, model: nn.Module, weight_bit_width=8):

        def wrap_block(m, prefix=""):
            for name, child in m.named_children():
                child_prefix = f"{prefix}.{name}" if prefix else name
                if isinstance(child, tuple(self.block_type)):
                    self.gptq_block_wrappers[name] = GPTQBlockWrapper(child_prefix, child, weight_bit_width)
                    LOGGER.debug(f"Calibrate module {child_prefix} as a whole block in GPTQ")
                else:
                    wrap_block(child, child_prefix)

        wrap_block(model)
        return model

    def quantize(self, model: nn.Module):
        for _, module_wrapper in self.gptq_block_wrappers.items():
            module_wrapper.quant_module()

        return model

    @property
    def calibration_iters(self):
        return len(self.gptq_block_wrappers)

    @contextlib.contextmanager
    def record_order(self):
        counter = 0
        record_handles = []
        orders = {}
        try:
            def get_record_order_hook(module_name):
                def record_hook(*args, **kwargs):
                    nonlocal counter
                    if module_name not in orders:
                        orders[module_name] = counter
                        counter += 1
                return record_hook

            for module_name, module_wrapper in self.gptq_block_wrappers.items():
                # disable the record
                for _, layer_wrapper in module_wrapper.layer_wrappers.items():
                    layer_wrapper.is_record = False

                one_layer_wrapper_in_module = list(module_wrapper.layer_wrappers.values())[0]
                handles = one_layer_wrapper_in_module.layer.register_forward_pre_hook(get_record_order_hook(module_name))
                record_handles.append(handles)
            yield
        except Exception as e:
            logging.warning(e)
        finally:
            for module_name, order in orders.items():
                self.gptq_block_wrappers[module_name].set_order(order)

            for h in record_handles:
                h.remove()

            for module_name, module_wrapper in self.gptq_block_wrappers.items():
                # disable the record
                for _, layer_wrapper in module_wrapper.layer_wrappers.items():
                    layer_wrapper.is_record = True


    @contextlib.contextmanager
    def start_calib_iter(self, i):
        assert i < len(self.gptq_block_wrappers)
        target_module_wrapper = None
        try:
            for _, module_wrapper in self.gptq_block_wrappers.items():
                if module_wrapper.get_order() == i:
                    module_wrapper.enable()
                    target_module_wrapper = module_wrapper
                else:
                    module_wrapper.disable()
            yield
        finally:
            target_module_wrapper.quant_module()

    def release_gpu_memory(self):
        for block_name, block_wrapper in self.gptq_block_wrappers.items():
            block_wrapper.release_gpu_memory()

        torch.cuda.empty_cache()


def locate_parent(root: nn.Module, full_path: str):
    parent = root
    path = full_path.split('.')
    for p in path[:-1]:
        parent = getattr(parent, p)
    return parent, path[-1]


@torch.no_grad()
def gptq_quantize(model, tokenizer, weight_bit_width, calib_data):
    from .modeling_chatglm import GLMBlock
    from .quantization import QuantizedLinear

    quantizer = GPTQuantizer([GLMBlock])
    calib_model = quantizer.wrap_model(model, weight_bit_width)
    with quantizer.record_order():
        calib_model.chat(tokenizer, calib_data[0], history=[])
    logging.info("Start doing calibration using GPTQ ")
    for i in range(quantizer.calibration_iters):
        logging.info(f"Process: {i + 1}/{quantizer.calibration_iters}")
        # todo: should add early return to speed up the calibration
        with quantizer.start_calib_iter(i):
            for prompt in calib_data:
                model.chat(tokenizer, prompt, history=[])

    # replace the fp16 linear with quantized linear
    for _, block_wrapper in quantizer.gptq_block_wrappers.items():
        for layer_name, layer_wrapper in block_wrapper.layer_wrappers.items():
            layer = layer_wrapper.layer
            parent, name_in_parent = locate_parent(model, layer_name)
            quantized_layer = QuantizedLinear(
                weight_bit_width=weight_bit_width,
                weight_tensor=layer.weight,
                bias_tensor=layer.bias,
                weight_scale=layer_wrapper.weight_scale,
                in_features=layer.in_features,
                out_features=layer.out_features,
                bias=True,
                dtype=torch.half,
                device=layer_wrapper.device,
                empty_init=False
            )
            parent.add_module(name_in_parent, quantized_layer)

    torch.cuda.empty_cache()
    return