bitsandbytes/v0.50.0/en/reference/optim/lion.md

Lion

Lion (Evolved Sign Momentum) is a unique optimizer that uses the sign of the gradient to determine the update direction of the momentum. This makes Lion more memory-efficient and faster than AdamW which tracks and store the first and second-order moments.

Lion[[api-class]][[bitsandbytes.optim.Lion]]

• params (torch.tensor) -- The input parameters to optimize.
• lr (float, defaults to 1e-4) -- The learning rate.
• betas (tuple(float, float), defaults to (0.9, 0.999)) -- The beta values are the decay rates of the first and second-order moment of the optimizer.
• weight_decay (float, defaults to 0) -- The weight decay value for the optimizer.
• optim_bits (int, defaults to 32) -- The number of bits of the optimizer state.
• args (object, defaults to None) -- An object with additional arguments.
• min_8bit_size (int, defaults to 4096) -- The minimum number of elements of the parameter tensors for 8-bit optimization.
• is_paged (bool, defaults to False) -- Whether the optimizer is a paged optimizer or not.

Base Lion optimizer.

Lion8bit[[bitsandbytes.optim.Lion8bit]]

• params (torch.tensor) -- The input parameters to optimize.
• lr (float, defaults to 1e-4) -- The learning rate.
• betas (tuple(float, float), defaults to (0.9, 0.999)) -- The beta values are the decay rates of the first and second-order moment of the optimizer.
• weight_decay (float, defaults to 0) -- The weight decay value for the optimizer.
• args (object, defaults to None) -- An object with additional arguments.
• min_8bit_size (int, defaults to 4096) -- The minimum number of elements of the parameter tensors for 8-bit optimization.
• is_paged (bool, defaults to False) -- Whether the optimizer is a paged optimizer or not.

8-bit Lion optimizer.

Lion32bit[[bitsandbytes.optim.Lion32bit]]

• params (torch.tensor) -- The input parameters to optimize.
• lr (float, defaults to 1e-4) -- The learning rate.
• betas (tuple(float, float), defaults to (0.9, 0.999)) -- The beta values are the decay rates of the first and second-order moment of the optimizer.
• weight_decay (float, defaults to 0) -- The weight decay value for the optimizer.
• args (object, defaults to None) -- An object with additional arguments.
• min_8bit_size (int, defaults to 4096) -- The minimum number of elements of the parameter tensors for 8-bit optimization.
• is_paged (bool, defaults to False) -- Whether the optimizer is a paged optimizer or not.

32-bit Lion optimizer.

PagedLion[[bitsandbytes.optim.PagedLion]]

• params (torch.tensor) -- The input parameters to optimize.
• lr (float, defaults to 1e-4) -- The learning rate.
• betas (tuple(float, float), defaults to (0.9, 0.999)) -- The beta values are the decay rates of the first and second-order moment of the optimizer.
• weight_decay (float, defaults to 0) -- The weight decay value for the optimizer.
• optim_bits (int, defaults to 32) -- The number of bits of the optimizer state.
• args (object, defaults to None) -- An object with additional arguments.
• min_8bit_size (int, defaults to 4096) -- The minimum number of elements of the parameter tensors for 8-bit optimization.

Paged Lion optimizer.

PagedLion8bit[[bitsandbytes.optim.PagedLion8bit]]

• params (torch.tensor) -- The input parameters to optimize.
• lr (float, defaults to 1e-4) -- The learning rate.
• betas (tuple(float, float), defaults to (0.9, 0.999)) -- The beta values are the decay rates of the first and second-order moment of the optimizer.
• weight_decay (float, defaults to 0) -- The weight decay value for the optimizer.
• args (object, defaults to None) -- An object with additional arguments.
• min_8bit_size (int, defaults to 4096) -- The minimum number of elements of the parameter tensors for 8-bit optimization.

Paged 8-bit Lion optimizer.

PagedLion32bit[[bitsandbytes.optim.PagedLion32bit]]

• params (torch.tensor) -- The input parameters to optimize.
• lr (float, defaults to 1e-4) -- The learning rate.
• betas (tuple(float, float), defaults to (0.9, 0.999)) -- The beta values are the decay rates of the first and second-order moment of the optimizer.
• weight_decay (float, defaults to 0) -- The weight decay value for the optimizer.
• args (object, defaults to None) -- An object with additional arguments.
• min_8bit_size (int, defaults to 4096) -- The minimum number of elements of the parameter tensors for 8-bit optimization.

Paged 32-bit Lion optimizer.