Functions — github.com/Vahe1994/AQLM

CodeHub MCPcopy Create free account

hub / github.com/Vahe1994/AQLM / functions

Functions225 in github.com/Vahe1994/AQLM

⨍ Functions 225◇ Types & classes 13 ↳ Endpoints 1

Functioncode2x8_matmat_dequant_transposed_meta

(input, codes, codebooks, scales, bias)

inference_lib/src/aqlm/inference_kernels/cuda_kernel.py:89

Functioncode2x8_matmat_meta

(input, codes, codebooks, scales, bias)

inference_lib/src/aqlm/inference_kernels/cuda_kernel.py:63

Functiondequantize_gemm

( input: torch.Tensor, # [..., in_features] codes: torch.IntTensor, # [num_out_groups, num_in_grou

inference_lib/src/aqlm/inference_kernels/dequantization.py:9

Methodextra_repr

(self, inp, **kwargs)

(self, input, *args, **kwargs)

(self, input: torch.Tensor)

inference_lib/src/aqlm/inference.py:68

( ctx: Any, input: torch.Tensor, codes: torch.IntTensor, codeb

inference_lib/src/aqlm/inference.py:102

(self, input: torch.Tensor)

Differentably reconstruct the weight (or parts thereof) from compressed components :param selection: By default, reconstruct the enti

Functionfrom_pretrained_aqlm

(*args, **kwargs)

Functionget_layers_prefix

(config: transformers.PretrainedConfig)

src/modelutils.py:288

Functionget_mean_nbits_by_codebook

Calculates average code length in codebooks. :param codes: codebook codes :param huffman_group_size: huffman compresssion dimension count

src/utils.py:17

Functiongroup_texts

Group tokenized examples together and split them into blocks of up to block_size tokens

src/datautils.py:269

Functionmaybe_script

Apply torch.jit.script to function unless one is using TPU. TPU does not support torch.jit.script.

inference_lib/src/aqlm/utils.py:35

Functionmaybe_script

Apply torch.jit.script to function unless one is using TPU. TPU does not support torch.jit.script.

src/utils.py:56

Functionnumba_gemm_lut

( input: torch.Tensor, # [..., in_features] codes: torch.IntTensor, # [num_out_groups, num_in_grou

inference_lib/src/aqlm/inference_kernels/numba_kernel.py:10

Functionnumba_gemv_lut_

(x, codebooks, codes_alt, scales)

inference_lib/src/aqlm/inference_kernels/numba_kernel.py:38

Functionoptimize_for_training

WARNING: `optimize_for_training` is deprecated. The optimization now happens automatically at runtime. OBSOLETE: Use this context manager dur

inference_lib/src/aqlm/inference_kernels/kernel_selector.py:9

Functionread_quant_weight_from_file

(load_path, block_i, layer_name, device)

src/modelutils.py:190

(*args, **kwargs)

src/modelutils.py:22

r"""Performs a single optimization step. Arguments: closure: A closure that reevaluates the model and returns the loss.

src/configurable_adam.py:100

Functiontest_kl_divergence

Verify correctness of compute_kl_divergence_loss_values

src/memory_efficient_loss.py:77

Functiontriton_matmul

( input: torch.Tensor, codes: torch.IntTensor, codebooks: torch.Tensor, scales: torch.Tensor,

inference_lib/src/aqlm/inference_kernels/triton_kernel.py:187

← previous201–225 of 225, ranked by callers