MCPcopy Create free account

hub / github.com/bitsandbytes-foundation/bitsandbytes / functions

Functions720 in github.com/bitsandbytes-foundation/bitsandbytes

⨍ Functions 720◇ Types & classes 117↳ Endpoints 78
Function_optimizer_update_1state_32bit_triton_kernel
1-state optimizer kernel
bitsandbytes/backends/triton/kernels_optim.py:237
Function_optimizer_update_1state_8bit_blockwise_triton_kernel
Triton kernel for 8-bit optimizers that use one momentum state. Supports: Momentum, RMSprop, Adagrad, Lion.
bitsandbytes/backends/triton/kernels_optim.py:877
Function_optimizer_update_2state_32bit_triton_kernel
2-state optimizer kernel
bitsandbytes/backends/triton/kernels_optim.py:152
Function_optimizer_update_2state_8bit_blockwise_triton_kernel
Triton kernel for 8-bit optimizers that use two momentum states. Supports: Adam, AdEMAMix.
bitsandbytes/backends/triton/kernels_optim.py:960
Function_optimizer_update_32bit_cpu
( optimizer_name: str, g: torch.Tensor, p: torch.Tensor, state1: torch.Tensor, state2: Opt
bitsandbytes/backends/cpu/ops.py:345
Function_optimizer_update_32bit_impl
( optimizer_name: str, g: torch.Tensor, p: torch.Tensor, state1: torch.Tensor, state2: Opt
bitsandbytes/backends/cuda/ops.py:1046
Function_optimizer_update_8bit_blockwise_cpu
( optimizer_name: str, g: torch.Tensor, p: torch.Tensor, state1: torch.Tensor, state2: Opt
bitsandbytes/backends/cpu/ops.py:469
Function_optimizer_update_8bit_blockwise_impl
( optimizer_name: str, g: torch.Tensor, p: torch.Tensor, state1: torch.Tensor, state2: Opt
bitsandbytes/backends/cuda/ops.py:1105
Function_parametrized_state_dict_post_hook
Hook to modify the state dict to include the quantization state.
bitsandbytes/nn/parametrize.py:170
Method_save_to_state_dict
(self, destination, prefix, keep_vars)
bitsandbytes/nn/modules.py:859
Method_save_to_state_dict
(self, destination, prefix, keep_vars)
bitsandbytes/nn/modules.py:963
Method_save_to_state_dict
(self, destination, prefix, keep_vars)
bitsandbytes/nn/modules.py:1095
Function_try_torch_compile
Wrapper around torch.compile that falls back to the original function if compilation fails.
bitsandbytes/backends/default/ops.py:12
Methodabsmax
(self)
bitsandbytes/nn/modules.py:272
Methodadd_outliers
(self, outlier_idx, feature_dim)
bitsandbytes/autograd/_functions.py:42
Methodapply
csrc/cpu_ops.cpp:697
Methodbackward
(ctx: torch.autograd.function.FunctionCtx, grad_output: torch.Tensor)
bitsandbytes/autograd/_functions.py:202
Methodbackward
(ctx, grad_output)
bitsandbytes/autograd/_functions.py:274
Functioncbatched_igemm
csrc/pythonInterface.cpp:513
Functioncdequant_mm_int32_fp16
csrc/pythonInterface.cpp:545
Functioncdequantize_blockwise_bf16
csrc/pythonInterface.cpp:428
Functioncdequantize_blockwise_bf16_fp4
csrc/pythonInterface.cpp:434
Functioncdequantize_blockwise_bf16_nf4
csrc/pythonInterface.cpp:440
Functioncdequantize_blockwise_cpu_bf16
csrc/pythonInterface.cpp:772
Functioncdequantize_blockwise_cpu_fp16
csrc/pythonInterface.cpp:778
Functioncdequantize_blockwise_cpu_fp32
csrc/pythonInterface.cpp:766
Functioncdequantize_blockwise_cpu_fp4_bf16
csrc/pythonInterface.cpp:790
Functioncdequantize_blockwise_cpu_fp4_fp16
csrc/pythonInterface.cpp:796
Functioncdequantize_blockwise_cpu_fp4_fp32
csrc/pythonInterface.cpp:784
Functioncdequantize_blockwise_cpu_nf4_bf16
csrc/pythonInterface.cpp:808
Functioncdequantize_blockwise_cpu_nf4_fp16
csrc/pythonInterface.cpp:814
Functioncdequantize_blockwise_cpu_nf4_fp32
csrc/pythonInterface.cpp:802
Functioncdequantize_blockwise_fp16
csrc/pythonInterface.cpp:352
Functioncdequantize_blockwise_fp16_fp4
csrc/pythonInterface.cpp:346
Functioncdequantize_blockwise_fp16_nf4
csrc/pythonInterface.cpp:358
Functioncdequantize_blockwise_fp32
csrc/pythonInterface.cpp:392
Functioncdequantize_blockwise_fp32_fp4
csrc/pythonInterface.cpp:398
Functioncdequantize_blockwise_fp32_nf4
csrc/pythonInterface.cpp:404
Functioncfill_fp32
csrc/pythonInterface.cpp:730
Functioncfill_uint8
csrc/pythonInterface.cpp:739
Functioncgemm_4bit_inference_naive_bf16
csrc/pythonInterface.cpp:601
Functioncgemm_4bit_inference_naive_fp16
csrc/pythonInterface.cpp:594
Functioncgemm_4bit_inference_naive_fp32
csrc/pythonInterface.cpp:608
Functioncgemv_4bit_inference_bf16
csrc/pythonInterface.cpp:683
Functioncgemv_4bit_inference_fp16
csrc/pythonInterface.cpp:676
Functioncgemv_4bit_inference_fp32
csrc/pythonInterface.cpp:690
Functioncget_managed_ptr
csrc/pythonInterface.cpp:557
Functioncigemm
csrc/pythonInterface.cpp:506
Functioncigemmlt_32
csrc/pythonInterface.cpp:524
Functioncigemmlt_8
csrc/pythonInterface.cpp:531
Functioncigemmlt_8_rowscale
csrc/pythonInterface.cpp:538
Functioncint8_vector_quant
csrc/pythonInterface.cpp:551
Functioncmd_batch_related
Find related issues for multiple issues at once.
agents/query_issues.py:458
Functioncmd_list
List issues with compact one-line summaries.
agents/query_issues.py:367
Functioncmd_related
Find issues related to a given issue number.
agents/query_issues.py:436
Functioncmd_search
Search issues by keyword.
agents/query_issues.py:403
Functioncmd_show
Show full detail for one or more issues.
agents/query_issues.py:486
Functioncmd_stats
Show summary statistics.
agents/query_issues.py:515
Functioncmd_top
List top issues by reaction count.
agents/query_issues.py:502
Methodcode
(self)
bitsandbytes/nn/modules.py:279
Functioncollate_fn
(batch)
examples/xpu/paged_xpu_training.py:80
Functioncollate_fn
(batch)
examples/cpu/cpu_training.py:87
Methodcompute_fingerprint
csrc/cpu_ops.cpp:532
Functioncprefetch
csrc/pythonInterface.cpp:565
Methodcpu
(self)
bitsandbytes/nn/modules.py:750
Functioncquantize_blockwise_bf16
csrc/pythonInterface.cpp:410
Functioncquantize_blockwise_bf16_fp4
csrc/pythonInterface.cpp:416
Functioncquantize_blockwise_bf16_nf4
csrc/pythonInterface.cpp:422
Functioncquantize_blockwise_cpu_bf16
csrc/pythonInterface.cpp:754
Functioncquantize_blockwise_cpu_fp16
csrc/pythonInterface.cpp:760
Functioncquantize_blockwise_cpu_fp32
csrc/pythonInterface.cpp:748
Functioncquantize_blockwise_fp16
csrc/pythonInterface.cpp:364
Functioncquantize_blockwise_fp16_fp4
csrc/pythonInterface.cpp:368
Functioncquantize_blockwise_fp16_nf4
csrc/pythonInterface.cpp:372
Functioncquantize_blockwise_fp32
csrc/pythonInterface.cpp:376
Functioncquantize_blockwise_fp32_fp4
csrc/pythonInterface.cpp:380
Functioncquantize_blockwise_fp32_nf4
csrc/pythonInterface.cpp:386
Functioncreate_fp8_map
Create a floating-point quantization map with configurable bit layout. Generates a lookup table for a custom floating-point format following IEEE
bitsandbytes/functional.py:227
Functioncreate_linear_map
(signed=True, total_bits=8, add_zero=True)
bitsandbytes/functional.py:150
Functioncreate_normal_map
Create the NormalFloat (NF4) quantization map. Constructs a lookup table of 16 quantization values (stored in a 256-element tensor for indexi
bitsandbytes/functional.py:169
Methodcuda
(self, device: Optional[int | device | str] = None, non_blocking: bool = False)
bitsandbytes/nn/modules.py:753
Functioncuda120_spec
Simulates torch+cuda12.0 and a representative Ampere-class capability.
tests/test_cuda_setup_evaluator.py:12
FunctiondDequantizeFP4
csrc/cpu_ops.h:236
FunctiondDequantizeNF4
csrc/cpu_ops.h:278
Functiondequant_4bit_kernel
( a_ptr, c_ptr, quant_ptr, absmax_ptr, num_paired_elements, num_output_elements, Q
bitsandbytes/backends/triton/kernels_4bit.py:334
Functiondequant_8bit_kernel
( a_ptr, out_ptr, code_ptr, absmax_ptr, n, QUANT_BLOCK: tl.constexpr, SPLIT_SIZE:
bitsandbytes/backends/triton/kernels_8bit_quant.py:28
Functiondequant_fp4_kernel
( a_ptr, c_ptr, absmax_ptr, num_paired_elements, num_output_elements, QUANT_BLOCK: tl.
bitsandbytes/backends/triton/kernels_4bit.py:378
Functiondequant_nf4_kernel
( a_ptr, c_ptr, absmax_ptr, num_paired_elements, num_output_elements, QUANT_BLOCK: tl.
bitsandbytes/backends/triton/kernels_4bit.py:420
FunctiondequantizeBlockwise
csrc/xpu_ops.cpp:5
FunctiondequantizeBlockwise4bitCpu
csrc/cpu_ops.cpp:305
FunctiondequantizeBlockwise8bitCpu
csrc/cpu_ops.cpp:437
Functiondequantize_4bit
( A: torch.Tensor, absmax: torch.Tensor, blocksize: int, quant_type: str, shape: Sequence[
bitsandbytes/backends/triton/ops.py:103
Functiondequantize_4bit_impl
( A: torch.Tensor, absmax: torch.Tensor, blocksize: int, quant_type: str, dtype: torch.dty
bitsandbytes/backends/triton/kernels_4bit.py:450
Functiondequantize_4bit_impl_passing_code
( A: torch.Tensor, absmax: torch.Tensor, blocksize: int, code: torch.Tensor, dtype: torch.
bitsandbytes/backends/triton/kernels_4bit.py:475
Functiondequantize_4bit_inplace
( A: torch.Tensor, absmax: torch.Tensor, blocksize: int, quant_type: str, shape: Sequence[
bitsandbytes/backends/triton/ops.py:129
Functiondequantize_blockwise
( A: torch.Tensor, absmax: torch.Tensor, code: torch.Tensor, blocksize: int, dtype: torch.dtype )
bitsandbytes/backends/triton/ops.py:24
Functiondequantize_blockwise_inplace
( A: torch.Tensor, absmax: torch.Tensor, code: torch.Tensor, blocksize: int, dtype: torch.
bitsandbytes/backends/triton/ops.py:41
Functiondequantize_fp4
( A: torch.Tensor, quant_state: Optional[QuantState] = None, absmax: Optional[torch.Tensor] = None
bitsandbytes/functional.py:972
Functiondequantize_nf4
( A: torch.Tensor, quant_state: Optional[QuantState] = None, absmax: Optional[torch.Tensor] = None
bitsandbytes/functional.py:982
Methodend
(self, *args)
benchmarking/xpu/inference_benchmark.py:78
← previousnext →401–500 of 720, ranked by callers