diff --git a/auto_gptq/nn_modules/qlinear.py b/auto_gptq/nn_modules/qlinear.py index c1ef578..74cd051 100644 --- a/auto_gptq/nn_modules/qlinear.py +++ b/auto_gptq/nn_modules/qlinear.py @@ -62,25 +62,27 @@ class QuantLinear(nn.Module): if self.bits in [2, 4, 8]: self.wf = torch.tensor(list(range(0, 32, self.bits)), dtype=torch.int32).unsqueeze(0) elif self.bits == 3: - self.wf = torch.tensor([ - [0, 3, 6, 9, 12, 15, 18, 21, 24, 27, 30, 0], - [0, 1, 4, 7, 10, 13, 16, 19, 22, 25, 28, 31], - [0, 2, 5, 8, 11, 14, 17, 20, 23, 26, 29, 0], - ], - dtype=torch.int32).reshape(1, 3, 12) + self.wf = torch.tensor( + [ + [0, 3, 6, 9, 12, 15, 18, 21, 24, 27, 30, 0], + [0, 1, 4, 7, 10, 13, 16, 19, 22, 25, 28, 31], + [0, 2, 5, 8, 11, 14, 17, 20, 23, 26, 29, 0], + ], + dtype=torch.int32 + ).reshape(1, 3, 12) self.kernel_switch_threshold = kernel_switch_threshold self.quant_cuda_available = _quant_cuda_available if infeatures % 256 != 0 or outfeatures % 256 != 0: self.quant_cuda_available = False - + def pack(self, linear, scales, zeros, g_idx=None): W = linear.weight.data.clone() if isinstance(linear, nn.Conv2d): W = W.flatten(1) if isinstance(linear, transformers.pytorch_utils.Conv1D): W = W.t() - + self.g_idx = g_idx.clone() if g_idx is not None else self.g_idx scales = scales.t().contiguous() @@ -195,7 +197,7 @@ class QuantLinear(nn.Module): else: if self.wf.device != self.qzeros.device: self.wf = self.wf.to(self.qzeros.device) - + if self.bits in [2, 4, 8]: zeros = torch.bitwise_right_shift( torch.unsqueeze(self.qzeros, 2).expand(-1, -1, 32 // self.bits), diff --git a/auto_gptq/nn_modules/qlinear_old.py b/auto_gptq/nn_modules/qlinear_old.py index 70ce27a..ea2442e 100644 --- a/auto_gptq/nn_modules/qlinear_old.py +++ b/auto_gptq/nn_modules/qlinear_old.py @@ -17,10 +17,19 @@ except ImportError: _quant_cuda_available = False -class QuantLinear(nn.Module): - def __init__(self, bits, groupsize, infeatures, outfeatures, bias, faster=True, kernel_switch_threshold=128, is_cuda=_quant_cuda_available): +class QuantLinear(nn.Module): + def __init__( + self, + bits, + groupsize, + infeatures, + outfeatures, + bias, + faster=True, + kernel_switch_threshold=128 + ): super().__init__() - if bits not in [2,3,4,8]: + if bits not in [2, 3, 4, 8]: raise NotImplementedError("Only 2,3,4,8 bits are supported.") self.infeatures = infeatures self.outfeatures = outfeatures @@ -28,30 +37,47 @@ class QuantLinear(nn.Module): self.groupsize = groupsize if groupsize != -1 else infeatures self.maxq = 2 ** self.bits - 1 - self.register_buffer('qweight', torch.zeros((infeatures // 32 * self.bits, outfeatures), dtype=torch.int32)) - self.register_buffer('qzeros', torch.zeros((math.ceil(infeatures / self.groupsize), outfeatures // 32 * self.bits), dtype=torch.int32)) - self.register_buffer('scales', torch.zeros((math.ceil(infeatures / self.groupsize), outfeatures), dtype=torch.float16)) - self.register_buffer('g_idx',torch.tensor([i // self.groupsize for i in range(infeatures)], dtype=torch.int32)) + self.register_buffer( + 'qweight', + torch.zeros((infeatures // 32 * self.bits, outfeatures), dtype=torch.int32) + ) + self.register_buffer( + 'qzeros', + torch.zeros((math.ceil(infeatures / self.groupsize), outfeatures // 32 * self.bits), dtype=torch.int32) + ) + self.register_buffer( + 'scales', + torch.zeros((math.ceil(infeatures / self.groupsize), outfeatures), dtype=torch.float16) + ) + self.register_buffer( + 'g_idx', + torch.tensor([i // self.groupsize for i in range(infeatures)], dtype=torch.int32) + ) if bias: - self.register_buffer('bias', torch.zeros((outfeatures),dtype=torch.float16)) + self.register_buffer('bias', torch.zeros((outfeatures), dtype=torch.float16)) else: self.bias = None self.half_indim = self.infeatures // 2 self.faster = faster if bits != 8 else False - + # is performed by unpacking the weights and using torch.matmul - if self.bits in [2,4,8]: - self.wf = torch.tensor(list(range(0,32,self.bits)), dtype=torch.int32).unsqueeze(0) + if self.bits in [2, 4, 8]: + self.wf = torch.tensor(list(range(0, 32, self.bits)), dtype=torch.int32).unsqueeze(0) elif self.bits == 3: - self.wf = torch.tensor([[0, 3, 6, 9, 12, 15, 18, 21, 24, 27, 30, 0], - [0, 1, 4, 7, 10, 13, 16, 19, 22, 25, 28, 31], - [0, 2, 5, 8, 11, 14, 17, 20, 23, 26, 29, 0],], dtype=torch.int32).reshape(1,3,12) - + self.wf = torch.tensor( + [ + [0, 3, 6, 9, 12, 15, 18, 21, 24, 27, 30, 0], + [0, 1, 4, 7, 10, 13, 16, 19, 22, 25, 28, 31], + [0, 2, 5, 8, 11, 14, 17, 20, 23, 26, 29, 0], + ], + dtype=torch.int32 + ).reshape(1, 3, 12) + self.kernel_switch_threshold = kernel_switch_threshold - self.is_cuda = is_cuda + self.quant_cuda_available = _quant_cuda_available if infeatures % 256 != 0 or outfeatures % 256 != 0: self.quant_cuda_available = False - + def pack(self, linear, scales, zeros, g_idx): scales = scales.t().contiguous() zeros = zeros.t().contiguous() @@ -59,24 +85,29 @@ class QuantLinear(nn.Module): self.scales = scales.clone().half() if linear.bias is not None: self.bias = linear.bias.clone().half() - + intweight = [] for idx in range(self.infeatures): g_idx = idx // self.groupsize - intweight.append(torch.round((linear.weight.data[:,idx] + scale_zeros[g_idx]) / self.scales[g_idx]).to(torch.int)[:,None]) - intweight = torch.cat(intweight,dim=1) + intweight.append( + torch.round( + (linear.weight.data[:, idx] + scale_zeros[g_idx]) / self.scales[g_idx] + ).to(torch.int)[:, None] + ) + intweight = torch.cat(intweight, dim=1) intweight = intweight.t().contiguous() intweight = intweight.numpy().astype(np.uint32) + + i = 0 + row = 0 qweight = np.zeros( (intweight.shape[0] // 32 * self.bits, intweight.shape[1]), dtype=np.uint32 ) - i = 0 - row = 0 while row < qweight.shape[0]: - if self.bits in [2,4,8]: - for j in range(i, i + (32//self.bits)): + if self.bits in [2, 4, 8]: + for j in range(i, i + (32 // self.bits)): qweight[row] |= intweight[j] << (self.bits * (j - i)) - i += 32//self.bits + i += 32 // self.bits row += 1 elif self.bits == 3: for j in range(i, i + 10): @@ -99,20 +130,20 @@ class QuantLinear(nn.Module): row += 1 else: raise NotImplementedError("Only 2,3,4,8 bits are supported.") - + qweight = qweight.astype(np.int32) - self.qweight = torch.from_numpy(qweight) - - zeros -= 1; + self.qweight = torch.from_numpy(qweight) + + zeros -= 1 zeros = zeros.numpy().astype(np.uint32) qzeros = np.zeros((zeros.shape[0], zeros.shape[1] // 32 * self.bits), dtype=np.uint32) i = 0 col = 0 while col < qzeros.shape[1]: - if self.bits in [2,4,8]: - for j in range(i, i + (32//self.bits)): + if self.bits in [2, 4, 8]: + for j in range(i, i + (32 // self.bits)): qzeros[:, col] |= zeros[:, j] << (self.bits * (j - i)) - i += 32//self.bits + i += 32 // self.bits col += 1 elif self.bits == 3: for j in range(i, i + 10): @@ -135,84 +166,112 @@ class QuantLinear(nn.Module): col += 1 else: raise NotImplementedError("Only 2,3,4,8 bits are supported.") - + qzeros = qzeros.astype(np.int32) - self.qzeros = torch.from_numpy(qzeros) - + self.qzeros = torch.from_numpy(qzeros) + def forward(self, x): - out_shape = x.shape[:-1] + (self.outfeatures, ) - x = x.reshape(-1,x.shape[-1]) - if self.is_cuda is True and (self.kernel_switch_threshold is False or x.shape[0] < self.kernel_switch_threshold): + out_shape = x.shape[:-1] + (self.outfeatures,) + x = x.reshape(-1, x.shape[-1]) + if self.quant_cuda_available is True and ( + self.kernel_switch_threshold is False or x.shape[0] < self.kernel_switch_threshold + ): out = torch.zeros(x.shape[0], out_shape[-1], dtype=torch.float, device=x.device) - + if self.faster: x = x.half() if self.bits == 2: - quant_cuda.vecquant2matmul_faster_old(x, self.qweight, out, self.scales.float(), self.qzeros, self.groupsize, self.half_indim) + quant_cuda.vecquant2matmul_faster_old( + x, self.qweight, out, self.scales.float(), self.qzeros, self.groupsize, self.half_indim + ) elif self.bits == 3: - quant_cuda.vecquant3matmul_faster_old(x, self.qweight, out, self.scales.float(), self.qzeros, self.groupsize, self.half_indim) + quant_cuda.vecquant3matmul_faster_old( + x, self.qweight, out, self.scales.float(), self.qzeros, self.groupsize, self.half_indim + ) elif self.bits == 4: - quant_cuda.vecquant4matmul_faster_old(x, self.qweight, out, self.scales.float(), self.qzeros, self.groupsize, self.half_indim) + quant_cuda.vecquant4matmul_faster_old( + x, self.qweight, out, self.scales.float(), self.qzeros, self.groupsize, self.half_indim + ) else: raise NotImplementedError("Only 2,3,4 bits are supported.") else: x = x.float() if self.bits == 2: - quant_cuda.vecquant2matmul_old(x, self.qweight, out, self.scales.float(), self.qzeros, self.groupsize) + quant_cuda.vecquant2matmul_old( + x, self.qweight, out, self.scales.float(), self.qzeros, self.groupsize + ) elif self.bits == 3: - quant_cuda.vecquant3matmul_old(x, self.qweight, out, self.scales.float(), self.qzeros, self.groupsize) + quant_cuda.vecquant3matmul_old( + x, self.qweight, out, self.scales.float(), self.qzeros, self.groupsize + ) elif self.bits == 4: - quant_cuda.vecquant4matmul_old(x, self.qweight, out, self.scales.float(), self.qzeros, self.groupsize) + quant_cuda.vecquant4matmul_old( + x, self.qweight, out, self.scales.float(), self.qzeros, self.groupsize + ) elif self.bits == 8: - quant_cuda.vecquant8matmul_old(x, self.qweight, out, self.scales.float(), self.qzeros, self.groupsize) + quant_cuda.vecquant8matmul_old( + x, self.qweight, out, self.scales.float(), self.qzeros, self.groupsize + ) else: raise NotImplementedError("Only 2,3,4,8 bits are supported.") else: if self.wf.device != self.qzeros.device: - self.wf = self.wf.to(self.qzeros.device) - - if self.bits in [2,4,8]: - zeros = torch.bitwise_right_shift(torch.unsqueeze(self.qzeros, 2).expand(-1, -1, 32 // self.bits), self.wf.unsqueeze(0)).to(torch.int16 if self.bits == 8 else torch.int8) - torch.bitwise_and(zeros, (2 ** self.bits) - 1, out=zeros) - - zeros = zeros + 1 - zeros = zeros.reshape(-1, 1, zeros.shape[1] * zeros.shape[2]) - - scales = self.scales - scales = scales.reshape(-1, 1, scales.shape[-1]) - - weight = torch.bitwise_right_shift(torch.unsqueeze(self.qweight, 1).expand(-1, 32 // self.bits, -1), self.wf.unsqueeze(-1)).to(torch.int16 if self.bits == 8 else torch.int8) - torch.bitwise_and(weight,(2 ** self.bits) - 1, out=weight) - weight = weight.reshape(-1, self.groupsize, weight.shape[2]) + self.wf = self.wf.to(self.qzeros.device) + + if self.bits in [2, 4, 8]: + zeros = torch.bitwise_right_shift( + torch.unsqueeze(self.qzeros, 2).expand(-1, -1, 32 // self.bits), + self.wf.unsqueeze(0) + ).to(torch.int16 if self.bits == 8 else torch.int8) + torch.bitwise_and(zeros, (2 ** self.bits) - 1, out=zeros) + + zeros = zeros + 1 + zeros = zeros.reshape(-1, 1, zeros.shape[1] * zeros.shape[2]) + + scales = self.scales + scales = scales.reshape(-1, 1, scales.shape[-1]) + + weight = torch.bitwise_right_shift( + torch.unsqueeze(self.qweight, 1).expand(-1, 32 // self.bits, -1), + self.wf.unsqueeze(-1) + ).to(torch.int16 if self.bits == 8 else torch.int8) + torch.bitwise_and(weight, (2 ** self.bits) - 1, out=weight) + weight = weight.reshape(-1, self.groupsize, weight.shape[2]) elif self.bits == 3: - zeros = self.qzeros.reshape(self.qzeros.shape[0], self.qzeros.shape[1]//3, 3, 1).expand(-1, -1, -1, 12) - zeros = (zeros >> self.wf.unsqueeze(0)) - zeros[:,:,0,10] = (zeros[:,:,0,10]&0x3) | ((zeros[:,:,1,0] << 2)&0x4) - zeros[:,:,1,11] = (zeros[:,:,1,11]&0x1) | ((zeros[:,:,2,0] << 1)&0x6) - zeros = zeros & 0x7 - zeros = torch.cat([zeros[:,:,0,:11], zeros[:,:,1,1:12], zeros[:,:,2,1:11]], dim=2) - - zeros = zeros + 1 - zeros = zeros.reshape(-1, 1, zeros.shape[1] * zeros.shape[2]) - - scales = self.scales - scales = scales.reshape(-1, 1, scales.shape[-1]) - - weight = self.qweight.reshape(self.qweight.shape[0]//3, 3, 1, self.qweight.shape[1]).expand(-1, -1, 12, -1) - weight = (weight >> self.wf.unsqueeze(-1))&0x7 - weight[:,0,10] = (weight[:,0,10]&0x3) | ((weight[:,1,0] << 2)&0x4) - weight[:,1,11] = (weight[:,1,11]&0x1) | ((weight[:,2,0] << 1)&0x6) - weight = weight & 0x7 - weight = torch.cat([weight[:,0,:11], weight[:,1,1:12], weight[:,2,1:11]], dim=1) - weight = weight.reshape(-1, self.groupsize, weight.shape[2]) + zeros = self.qzeros.reshape( + self.qzeros.shape[0], self.qzeros.shape[1] // 3, 3, 1 + ).expand(-1, -1, -1, 12) + zeros = (zeros >> self.wf.unsqueeze(0)) + zeros[:, :, 0, 10] = (zeros[:, :, 0, 10] & 0x3) | ((zeros[:, :, 1, 0] << 2) & 0x4) + zeros[:, :, 1, 11] = (zeros[:, :, 1, 11] & 0x1) | ((zeros[:, :, 2, 0] << 1) & 0x6) + zeros = zeros & 0x7 + zeros = torch.cat([zeros[:, :, 0, :11], zeros[:, :, 1, 1:12], zeros[:, :, 2, 1:11]], dim=2) + + zeros = zeros + 1 + zeros = zeros.reshape(-1, 1, zeros.shape[1] * zeros.shape[2]) + + scales = self.scales + scales = scales.reshape(-1, 1, scales.shape[-1]) + + weight = self.qweight.reshape( + self.qweight.shape[0] // 3, 3, 1, self.qweight.shape[1] + ).expand(-1, -1, 12, -1) + weight = (weight >> self.wf.unsqueeze(-1)) & 0x7 + weight[:, 0, 10] = (weight[:, 0, 10] & 0x3) | ((weight[:, 1, 0] << 2) & 0x4) + weight[:, 1, 11] = (weight[:, 1, 11] & 0x1) | ((weight[:, 2, 0] << 1) & 0x6) + weight = weight & 0x7 + weight = torch.cat([weight[:, 0, :11], weight[:, 1, 1:12], weight[:, 2, 1:11]], dim=1) + weight = weight.reshape(-1, self.groupsize, weight.shape[2]) + else: + raise NotImplementedError("Only 2,3,4,8 bits are supported.") + weight = (scales * (weight - zeros)) weight = weight.reshape(weight.shape[0] * weight.shape[1], weight.shape[2]) - + out = torch.matmul(x.half(), weight) out = out.reshape(out_shape) out = out + self.bias if self.bias is not None else out return out - __all__ = ["QuantLinear"]