Home / kvcache-ai / ktransformers
Repository Analysis

kvcache-ai/ktransformers

A Flexible Framework for Experiencing Heterogeneous LLM Inference/Fine-tune Optimizations

14.4 Low AI signal View on GitHub
14.4
Adjusted Score
14.4
Raw Score
100%
Time Factor
2026-05-30
Last Push
17,228
Stars
Python
Language
372,470
Lines of Code
1208
Files
3300
Pattern Hits
2026-05-31
Scan Date

Score History

Severity Breakdown

CRITICAL 3HIGH 420MEDIUM 204LOW 2673

Pattern Findings

3300 matches across 18 categories. Click a row to expand file-level details.

Cross-File Repetition389 hits · 1945 pts
SeverityFileLineSnippet
HIGHarchive/merge_tensors/merge_safetensor_gguf.py0:param folder_path: folder path :return: key_to_file_map
HIGHarchive/kt-sft/merge_tensors/merge_safetensor_gguf.py0:param folder_path: folder path :return: key_to_file_map
HIGHkt-kernel/scripts/check.py0:param folder_path: folder path :return: key_to_file_map
HIGHarchive/csrc/ktransformers_ext/bench/bench_attention.py0description : author : jianwei dong date : 2024-08-28 10:32:05 version : 1.0.0 lasteditors : jianwei dong lastedittime :
HIGH…/csrc/ktransformers_ext/bench/bench_attention_torch.py0description : author : jianwei dong date : 2024-08-28 10:32:05 version : 1.0.0 lasteditors : jianwei dong lastedittime :
HIGH…kt-sft/csrc/ktransformers_ext/bench/bench_attention.py0description : author : jianwei dong date : 2024-08-28 10:32:05 version : 1.0.0 lasteditors : jianwei dong lastedittime :
HIGH…/csrc/ktransformers_ext/bench/bench_attention_torch.py0description : author : jianwei dong date : 2024-08-28 10:32:05 version : 1.0.0 lasteditors : jianwei dong lastedittime :
HIGHkt-kernel/bench/bench_attention.py0description : author : jianwei dong date : 2024-08-28 10:32:05 version : 1.0.0 lasteditors : jianwei dong lastedittime :
HIGHkt-kernel/bench/bench_attention_torch.py0description : author : jianwei dong date : 2024-08-28 10:32:05 version : 1.0.0 lasteditors : jianwei dong lastedittime :
HIGHarchive/csrc/ktransformers_ext/bench/bench_moe.py0description : author : chenht2022 date : 2024-07-25 10:32:05 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGH…chive/kt-sft/csrc/ktransformers_ext/bench/bench_moe.py0description : author : chenht2022 date : 2024-07-25 10:32:05 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGHkt-kernel/bench/bench_moe_kernel.py0description : author : chenht2022 date : 2024-07-25 10:32:05 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGHkt-kernel/bench/bench_moe_amx.py0description : author : chenht2022 date : 2024-07-25 10:32:05 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGHkt-kernel/bench/bench_moe_amx_k.py0description : author : chenht2022 date : 2024-07-25 10:32:05 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGHkt-kernel/bench/bench_moe_kml.py0description : author : chenht2022 date : 2024-07-25 10:32:05 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGHarchive/csrc/ktransformers_ext/bench/bench_mlp.py0description : author : chenht2022 date : 2024-07-16 10:43:18 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGH…chive/kt-sft/csrc/ktransformers_ext/bench/bench_mlp.py0description : author : chenht2022 date : 2024-07-16 10:43:18 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGHkt-kernel/bench/bench_mlp.py0description : author : chenht2022 date : 2024-07-16 10:43:18 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGHarchive/csrc/ktransformers_ext/bench/bench_linear.py0description : author : chenht2022 date : 2024-07-25 10:31:59 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGH…ve/kt-sft/csrc/ktransformers_ext/bench/bench_linear.py0description : author : chenht2022 date : 2024-07-25 10:31:59 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGHkt-kernel/bench/bench_linear.py0description : author : chenht2022 date : 2024-07-25 10:31:59 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGH…ive/csrc/ktransformers_ext/bench/bench_linear_torch.py0description : author : chenht2022 date : 2024-07-25 10:31:59 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGH…sft/csrc/ktransformers_ext/bench/bench_linear_torch.py0description : author : chenht2022 date : 2024-07-25 10:31:59 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGHkt-kernel/bench/bench_linear_torch.py0description : author : chenht2022 date : 2024-07-25 10:31:59 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGHarchive/csrc/ktransformers_ext/bench/bench_mlp_torch.py0description : author : chenht2022 date : 2024-07-16 10:43:18 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGH…kt-sft/csrc/ktransformers_ext/bench/bench_mlp_torch.py0description : author : chenht2022 date : 2024-07-16 10:43:18 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGHkt-kernel/bench/bench_mlp_torch.py0description : author : chenht2022 date : 2024-07-16 10:43:18 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGHarchive/csrc/ktransformers_ext/examples/test_mlp.py0description : author : chenht2022 date : 2024-07-25 10:32:05 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGH…ive/kt-sft/csrc/ktransformers_ext/examples/test_mlp.py0description : author : chenht2022 date : 2024-07-25 10:32:05 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGHkt-kernel/examples/test_mlp.py0description : author : chenht2022 date : 2024-07-25 10:32:05 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGHarchive/csrc/ktransformers_ext/examples/test_moe.py0description : author : chenht2022 date : 2024-07-25 10:32:05 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGH…ft/csrc/ktransformers_ext/examples/test_sft_amx_moe.py0description : author : chenht2022 date : 2024-07-25 10:32:05 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGH…ive/kt-sft/csrc/ktransformers_ext/examples/test_moe.py0description : author : chenht2022 date : 2024-07-25 10:32:05 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGH…kt-sft/csrc/ktransformers_ext/examples/test_sft_moe.py0description : author : chenht2022 date : 2024-07-25 10:32:05 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGHkt-kernel/examples/test_moe_kernel.py0description : author : chenht2022 date : 2024-07-25 10:32:05 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGHkt-kernel/examples/test_moe_kml.py0description : author : chenht2022 date : 2024-07-25 10:32:05 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGH…hive/csrc/ktransformers_ext/examples/test_attention.py0description : author : jianwei dong date : 2024-08-28 10:32:05 version : 1.0.0 lasteditors : chenht2022 lastedittime : 2
HIGH…-sft/csrc/ktransformers_ext/examples/test_attention.py0description : author : jianwei dong date : 2024-08-28 10:32:05 version : 1.0.0 lasteditors : chenht2022 lastedittime : 2
HIGHkt-kernel/examples/test_attention.py0description : author : jianwei dong date : 2024-08-28 10:32:05 version : 1.0.0 lasteditors : chenht2022 lastedittime : 2
HIGHarchive/csrc/ktransformers_ext/examples/test_linear.py0description : author : chenht2022 date : 2024-07-25 10:32:05 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGH…/kt-sft/csrc/ktransformers_ext/examples/test_linear.py0description : author : chenht2022 date : 2024-07-25 10:32:05 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGHkt-kernel/examples/test_linear.py0description : author : chenht2022 date : 2024-07-25 10:32:05 version : 1.0.0 lasteditors : chenht2022 lastedittime : 202
HIGHarchive/csrc/custom_marlin/utils/format24.py0class for creating n:m sparsity masks. masks will be created using the n:m ratio, where for every block of m weights, n
HIGHarchive/kt-sft/csrc/custom_marlin/utils/format24.py0class for creating n:m sparsity masks. masks will be created using the n:m ratio, where for every block of m weights, n
HIGH…xt/operators/custom_marlin/quantize/utils/format_24.py0class for creating n:m sparsity masks. masks will be created using the n:m ratio, where for every block of m weights, n
HIGH…xt/operators/custom_marlin/quantize/utils/format_24.py0class for creating n:m sparsity masks. masks will be created using the n:m ratio, where for every block of m weights, n
HIGHarchive/kt-sft/ktransformers/local_chat.py0description : author : boxin zhang, azure-tang version : 0.1.0 copyright (c) 2024 by kvcache.ai, all rights reserved.
HIGHarchive/kt-sft/ktransformers/optimize/optimize.py0description : author : boxin zhang, azure-tang version : 0.1.0 copyright (c) 2024 by kvcache.ai, all rights reserved.
HIGHarchive/kt-sft/ktransformers/util/utils.py0description : author : boxin zhang, azure-tang version : 0.1.0 copyright (c) 2024 by kvcache.ai, all rights reserved.
HIGHarchive/ktransformers/local_chat_test.py0description : author : boxin zhang, azure-tang version : 0.1.0 copyright (c) 2024 by kvcache.ai, all rights reserved.
HIGHarchive/ktransformers/local_chat.py0description : author : boxin zhang, azure-tang version : 0.1.0 copyright (c) 2024 by kvcache.ai, all rights reserved.
HIGHarchive/ktransformers/optimize/optimize.py0description : author : boxin zhang, azure-tang version : 0.1.0 copyright (c) 2024 by kvcache.ai, all rights reserved.
HIGHarchive/ktransformers/util/utils.py0description : author : boxin zhang, azure-tang version : 0.1.0 copyright (c) 2024 by kvcache.ai, all rights reserved.
HIGHarchive/kt-sft/ktransformers/local_chat.py0'): # end multi lines input line = line[:-3] # suffix
HIGHarchive/ktransformers/local_chat.py0'): # end multi lines input line = line[:-3] # suffix
HIGHkt-kernel/examples/test_deepseekv3_prefill.py0'): # end multi lines input line = line[:-3] # suffix
HIGHarchive/kt-sft/ktransformers/operators/attention.py0description : author : boxin zhang version : 0.1.0 copyright (c) 2024 by kvcache.ai, all rights reserved.
HIGHarchive/kt-sft/ktransformers/operators/base_operator.py0description : author : boxin zhang version : 0.1.0 copyright (c) 2024 by kvcache.ai, all rights reserved.
HIGHarchive/kt-sft/ktransformers/operators/RoPE.py0description : author : boxin zhang version : 0.1.0 copyright (c) 2024 by kvcache.ai, all rights reserved.
HIGHarchive/kt-sft/ktransformers/util/cuda_graph_runner.py0description : author : boxin zhang version : 0.1.0 copyright (c) 2024 by kvcache.ai, all rights reserved.
329 more matches not shown…
Unused Imports917 hits · 892 pts
SeverityFileLineSnippet
LOWktransformers.py8
LOWktransformers.py29
LOW…chive/merge_tensors/merge_safetensor_gguf_for_qwen3.py18
LOWarchive/merge_tensors/merge_safetensor_gguf.py5
LOWarchive/csrc/ktransformers_ext/bench/bench_moe_torch.py12
LOWarchive/csrc/ktransformers_ext/bench/bench_moe_torch.py12
LOW…ive/csrc/ktransformers_ext/bench/bench_linear_torch.py12
LOW…ive/csrc/ktransformers_ext/bench/bench_linear_torch.py12
LOWarchive/csrc/ktransformers_ext/bench/bench_mlp_torch.py12
LOWarchive/csrc/ktransformers_ext/bench/bench_mlp_torch.py12
LOW…/csrc/ktransformers_ext/bench/bench_attention_torch.py16
LOWarchive/csrc/ktransformers_ext/cuda/setup.py2
LOWarchive/csrc/ktransformers_ext/cuda/setup.py3
LOWarchive/csrc/ktransformers_ext/cuda/test_dequant.py1
LOWarchive/csrc/ktransformers_ext/examples/test_mlp.py13
LOWarchive/csrc/ktransformers_ext/examples/test_moe.py13
LOW…hive/csrc/ktransformers_ext/examples/test_attention.py13
LOWarchive/csrc/ktransformers_ext/examples/test_linear.py13
LOWarchive/csrc/custom_marlin/setup.py1
LOWarchive/csrc/custom_marlin/setup.py2
LOWarchive/csrc/custom_marlin/test_cuda_graph.py1
LOWarchive/csrc/custom_marlin/test_cuda_graph.py6
LOWarchive/csrc/custom_marlin/utils/marlin_utils.py13
LOWarchive/kt-sft/withoutKT_PEFT.py2
LOWarchive/kt-sft/withoutKT_PEFT.py4
LOWarchive/kt-sft/withoutKT_PEFT.py9
LOWarchive/kt-sft/withoutKT_PEFT.py9
LOWarchive/kt-sft/merge_tensors/merge_safetensor_gguf.py5
LOW…kt-sft/csrc/ktransformers_ext/bench/bench_moe_torch.py12
LOW…kt-sft/csrc/ktransformers_ext/bench/bench_moe_torch.py12
LOW…sft/csrc/ktransformers_ext/bench/bench_linear_torch.py12
LOW…sft/csrc/ktransformers_ext/bench/bench_linear_torch.py12
LOW…kt-sft/csrc/ktransformers_ext/bench/bench_mlp_torch.py12
LOW…kt-sft/csrc/ktransformers_ext/bench/bench_mlp_torch.py12
LOW…/csrc/ktransformers_ext/bench/bench_attention_torch.py16
LOWarchive/kt-sft/csrc/ktransformers_ext/cuda/setup.py2
LOWarchive/kt-sft/csrc/ktransformers_ext/cuda/setup.py3
LOW…ive/kt-sft/csrc/ktransformers_ext/cuda/test_dequant.py1
LOW…ive/kt-sft/csrc/ktransformers_ext/examples/test_mlp.py13
LOW…ive/kt-sft/csrc/ktransformers_ext/examples/test_moe.py13
LOW…-sft/csrc/ktransformers_ext/examples/test_attention.py13
LOW…/kt-sft/csrc/ktransformers_ext/examples/test_linear.py13
LOWarchive/kt-sft/csrc/custom_marlin/setup.py1
LOWarchive/kt-sft/csrc/custom_marlin/setup.py2
LOWarchive/kt-sft/csrc/custom_marlin/test_cuda_graph.py1
LOWarchive/kt-sft/csrc/custom_marlin/test_cuda_graph.py6
LOWarchive/kt-sft/csrc/custom_marlin/utils/marlin_utils.py13
LOWarchive/kt-sft/ktransformers/moe_test_module.py2
LOWarchive/kt-sft/ktransformers/moe_test_module.py8
LOWarchive/kt-sft/ktransformers/moe_test_module.py9
LOWarchive/kt-sft/ktransformers/moe_test_module.py11
LOWarchive/kt-sft/ktransformers/moe_test_module.py11
LOWarchive/kt-sft/ktransformers/moe_test_module.py11
LOWarchive/kt-sft/ktransformers/moe_test_module.py11
LOWarchive/kt-sft/ktransformers/moe_test_module.py19
LOWarchive/kt-sft/ktransformers/moe_test_module.py21
LOWarchive/kt-sft/ktransformers/moe_test_module.py21
LOWarchive/kt-sft/ktransformers/moe_test_module.py22
LOWarchive/kt-sft/ktransformers/moe_test_module.py25
LOWarchive/kt-sft/ktransformers/__init__.py18
857 more matches not shown…
Over-Commented Block646 hits · 570 pts
SeverityFileLineSnippet
LOWdocker/docker-utils.sh1#!/usr/bin/env bash
LOWdocker/docker-utils.sh161################################################################################
LOWdocker/build-docker-tar.sh1#!/usr/bin/env bash
LOWdocker/push-to-dockerhub.sh1#!/usr/bin/env bash
LOWdocker/push-to-dockerhub.sh581# - Automatic version detection
LOW…chive/merge_tensors/merge_safetensor_gguf_for_qwen3.py1# coding=utf-8
LOWarchive/csrc/ktransformers_ext/ext_bindings.cpp21#if defined(__x86_64__) && defined(__HAS_AVX512F__) && defined(__HAS_AMX__)
LOWarchive/csrc/ktransformers_ext/vendors/musa.h1#pragma once
LOWarchive/csrc/ktransformers_ext/vendors/musa.h21#define cublasDestroy mublasDestroy
LOWarchive/csrc/ktransformers_ext/vendors/musa.h41#define cudaEventCreateWithFlags musaEventCreateWithFlags
LOWarchive/csrc/ktransformers_ext/vendors/musa.h61#define cudaMallocManaged musaMallocManaged
LOWarchive/csrc/ktransformers_ext/vendors/musa.h81#define cudaStreamWaitEvent musaStreamWaitEvent
LOWarchive/csrc/ktransformers_ext/vendors/musa.h101#define cuMemGetAllocationGranularity muMemGetAllocationGranularity
LOWarchive/csrc/ktransformers_ext/vendors/musa.h121#define cudaGraphExecUpdate musaGraphExecUpdate
LOWarchive/csrc/ktransformers_ext/vendors/hip.h1#pragma once
LOWarchive/csrc/ktransformers_ext/vendors/hip.h21#define CUBLAS_TF32_TENSOR_OP_MATH 0
LOWarchive/csrc/ktransformers_ext/vendors/hip.h41#define cublasSgemm hipblasSgemm
LOWarchive/csrc/ktransformers_ext/vendors/hip.h61#define cudaGetDevice hipGetDevice
LOWarchive/csrc/ktransformers_ext/vendors/hip.h81#define cudaMemset hipMemset
LOWarchive/csrc/ktransformers_ext/vendors/hip.h101#define cudaStreamCreateWithFlags hipStreamCreateWithFlags
LOWarchive/csrc/ktransformers_ext/vendors/hip.h121#define cudaGraphKernelNodeSetParams hipGraphKernelNodeSetParams
LOWarchive/csrc/ktransformers_ext/vendors/hip.h141#define CUBLAS_STATUS_INTERNAL_ERROR HIPBLAS_STATUS_INTERNAL_ERROR
LOWarchive/csrc/ktransformers_ext/vendors/hip.h161#define RDNA2
LOWarchive/csrc/ktransformers_ext/vendors/cuda.h1#pragma once
LOWarchive/csrc/ktransformers_ext/vendors/vendor.h1#ifndef CPUINFER_VENDOR_VENDOR_H
LOW…ive/csrc/ktransformers_ext/operators/kvcache/kvcache.h21#include <fstream>
LOWarchive/csrc/ktransformers_ext/operators/amx/la/amx.hpp21#include <sys/syscall.h>
LOWarchive/csrc/ktransformers_ext/operators/amx/la/amx.hpp41namespace amx {
LOWarchive/csrc/ktransformers_ext/cpu_backend/cpuinfer.h21 #ifdef KTRANSFORMERS_USE_CUDA
LOW…hive/csrc/ktransformers_ext/cpu_backend/vendors/musa.h1#pragma once
LOW…hive/csrc/ktransformers_ext/cpu_backend/vendors/musa.h21#define cublasDestroy mublasDestroy
LOW…hive/csrc/ktransformers_ext/cpu_backend/vendors/musa.h41#define cudaEventCreateWithFlags musaEventCreateWithFlags
LOW…hive/csrc/ktransformers_ext/cpu_backend/vendors/musa.h61#define cudaMallocManaged musaMallocManaged
LOW…hive/csrc/ktransformers_ext/cpu_backend/vendors/musa.h81#define cudaStreamWaitEvent musaStreamWaitEvent
LOW…hive/csrc/ktransformers_ext/cpu_backend/vendors/musa.h101#define cuMemGetAllocationGranularity muMemGetAllocationGranularity
LOW…hive/csrc/ktransformers_ext/cpu_backend/vendors/musa.h121#define cudaGraphExecUpdate musaGraphExecUpdate
LOW…chive/csrc/ktransformers_ext/cpu_backend/vendors/hip.h1#pragma once
LOW…chive/csrc/ktransformers_ext/cpu_backend/vendors/hip.h21#define CUBLAS_TF32_TENSOR_OP_MATH 0
LOW…chive/csrc/ktransformers_ext/cpu_backend/vendors/hip.h41#define cublasSgemm hipblasSgemm
LOW…chive/csrc/ktransformers_ext/cpu_backend/vendors/hip.h61#define cudaGetDevice hipGetDevice
LOW…chive/csrc/ktransformers_ext/cpu_backend/vendors/hip.h81#define cudaMemset hipMemset
LOW…chive/csrc/ktransformers_ext/cpu_backend/vendors/hip.h101#define cudaStreamCreateWithFlags hipStreamCreateWithFlags
LOW…chive/csrc/ktransformers_ext/cpu_backend/vendors/hip.h121#define cudaGraphKernelNodeSetParams hipGraphKernelNodeSetParams
LOW…chive/csrc/ktransformers_ext/cpu_backend/vendors/hip.h141#define CUBLAS_STATUS_INTERNAL_ERROR HIPBLAS_STATUS_INTERNAL_ERROR
LOW…chive/csrc/ktransformers_ext/cpu_backend/vendors/hip.h161#define RDNA2
LOW…hive/csrc/ktransformers_ext/cpu_backend/vendors/cuda.h1#pragma once
LOW…ve/csrc/ktransformers_ext/cpu_backend/vendors/vendor.h1#ifndef CPUINFER_VENDOR_VENDOR_H
LOWarchive/csrc/ktransformers_ext/cuda/binding.cpp1/**
LOWarchive/csrc/ktransformers_ext/cuda/gptq_marlin/ops.h21
LOWarchive/csrc/balance_serve/sched/scheduler.h1#pragma once
LOWarchive/csrc/balance_serve/sched/scheduler.cpp1#define SPDLOG_ACTIVE_LEVEL SPDLOG_LEVEL_INFO
LOWarchive/csrc/balance_serve/sched/metrics.h1#ifndef Metrics_H
LOWarchive/csrc/balance_serve/sched/utils/all.hpp1#pragma once
LOWarchive/csrc/balance_serve/kvc2/test/page_pool_test.cpp1
LOW…ve/csrc/balance_serve/kvc2/test/kvc2test/lookup-mt.cpp61 // // common prefix
LOW…ve/csrc/balance_serve/kvc2/test/kvc2test/lookup-mt.cpp81 // // insert partly new
LOW…e/csrc/balance_serve/kvc2/test/kvc2test/lookup-gpu.cpp101 cmp_handle_data(k1, k_from_gpu, 3);
LOW…e/csrc/balance_serve/kvc2/test/kvc2test/lookup-gpu.cpp121
LOW…e/csrc/balance_serve/kvc2/test/kvc2test/lookup-gpu.cpp141 // auto ids2 = random_ids(10 * config.num_token_per_page, gen);
LOWarchive/csrc/balance_serve/kvc2/src/async_store.cpp1
586 more matches not shown…
Excessive Try-Catch Wrapping315 hits · 373 pts
SeverityFileLineSnippet
LOWktransformers.py30 except Exception:
MEDIUMktransformers.py27def has_sft_support() -> bool:
LOW…chive/merge_tensors/merge_safetensor_gguf_for_qwen3.py47 except Exception as e:
MEDIUM…chive/merge_tensors/merge_safetensor_gguf_for_qwen3.py48 print(f"Error reading Safetensor file {file_path}: {e}")
LOWarchive/merge_tensors/merge_safetensor_gguf.py48 except Exception as e:
MEDIUMarchive/merge_tensors/merge_safetensor_gguf.py49 print(f"Error reading Safetensor file {file_path}: {e}")
LOWarchive/kt-sft/setup.py45except Exception:
LOWarchive/kt-sft/setup.py73 except Exception:
LOWarchive/kt-sft/merge_tensors/merge_safetensor_gguf.py48 except Exception as e:
MEDIUMarchive/kt-sft/merge_tensors/merge_safetensor_gguf.py49 print(f"Error reading Safetensor file {file_path}: {e}")
LOWarchive/kt-sft/test_adapter/infer_with_adapter.py27 except Exception as e:
LOWarchive/kt-sft/test_adapter/inspect_adapter.py60 except Exception as e:
LOWarchive/kt-sft/test_adapter/inspect_adapter.py77 except Exception as e:
LOWarchive/kt-sft/test_adapter/inspect_adapter.py94 except Exception as e:
LOWarchive/kt-sft/ktransformers/local_chat.py224 except Exception as e:
MEDIUMarchive/kt-sft/ktransformers/util/custom_loader.py71 print(f"Error opening Safetensor file {file_path}: {e}")
MEDIUMarchive/kt-sft/ktransformers/util/custom_loader.py81 print(f"Error reading Safetensor file {file_path}: {e}")
LOWarchive/kt-sft/ktransformers/util/custom_loader.py70 except Exception as e:
LOWarchive/kt-sft/ktransformers/util/custom_loader.py80 except Exception as e:
LOWarchive/kt-sft/ktransformers/util/custom_loader.py557 except Exception as e:
LOWarchive/kt-sft/ktransformers/util/custom_loader.py566 except Exception as e:
LOWarchive/kt-sft/ktransformers/util/utils.py74 except Exception:
LOWarchive/kt-sft/ktransformers/util/weight_loader.py85 except Exception as e:
MEDIUMarchive/kt-sft/ktransformers/util/weight_loader.py86 print(f"Error opening Safetensor file {file_path}: {e}")
LOWarchive/kt-sft/ktransformers/util/weight_loader.py95 except Exception as e:
MEDIUMarchive/kt-sft/ktransformers/util/weight_loader.py96 print(f"Error reading Safetensor file {file_path}: {e}")
LOWarchive/kt-sft/ktransformers/tests/mmlu_pro_test.py150 except Exception as e:
MEDIUMarchive/kt-sft/ktransformers/tests/mmlu_pro_test.py151 print(f"Error processing request {i}: {e}")
LOWarchive/kt-sft/ktransformers/tests/mmlu_test_multi.py156 except Exception as e:
MEDIUMarchive/kt-sft/ktransformers/tests/mmlu_test_multi.py157 print(f"Error processing request {index}: {e}")
LOWarchive/kt-sft/ktransformers/tests/mmlu_test.py142 except Exception as e:
MEDIUMarchive/kt-sft/ktransformers/tests/mmlu_test.py143 print(f"Error processing request {i}: {e}")
LOWarchive/kt-sft/ktransformers/tests/test_speed.py116 except Exception as e:
LOWarchive/kt-sft/ktransformers/tests/test_speed.py134 except Exception as e:
MEDIUMarchive/kt-sft/ktransformers/tests/test_speed.py48def fetch_event_stream(session, request_id, prompt, max_tokens, model):
LOWarchive/kt-sft/ktransformers/tests/test_client.py63 except Exception as e:
LOWarchive/kt-sft/ktransformers/tests/test_client.py73 except Exception as e:
MEDIUMarchive/kt-sft/ktransformers/tests/test_client.py15def fetch_event_stream(session, payload, request_id, stream):
LOW…chive/kt-sft/ktransformers/tests/humaneval/eval_api.py75 except Exception as e:
MEDIUM…chive/kt-sft/ktransformers/tests/humaneval/eval_api.py78 print(f"Error: {e}")
LOW…chive/kt-sft/ktransformers/tests/AIME_2024/eval_api.py110 except Exception as e:
LOWarchive/kt-sft/ktransformers/server/utils/sql_utils.py97 except Exception as e:
LOWarchive/kt-sft/ktransformers/server/utils/sql_utils.py108 except Exception as e:
LOWarchive/kt-sft/ktransformers/server/utils/sql_utils.py123 except Exception as e:
LOW…kt-sft/ktransformers/server/balance_serve/sched_rpc.py98 except Exception as e:
LOW…serve/inference/distributed/custom_all_reduce_utils.py244 except Exception as e:
LOW…/balance_serve/inference/distributed/parallel_state.py1248 except Exception as e:
MEDIUM…/balance_serve/inference/distributed/parallel_state.py1249 print("Error ignored in is_in_the_same_node: %s", e)
LOW…lance_serve/inference/distributed/custom_all_reduce.py20except Exception:
LOW…/balance_serve/inference/distributed/pynccl_wrapper.py193 except Exception as e:
LOW…s/server/balance_serve/inference/distributed/pynccl.py62 except Exception:
LOW…-sft/ktransformers/server/api/openai/endpoints/chat.py379 except Exception as e:
LOWarchive/kt-sft/ktransformers/sft/lora.py150 except Exception:
LOWarchive/kt-sft/ktransformers/sft/lora.py228 except Exception:
LOWarchive/kt-sft/ktransformers/sft/lora.py327 except Exception:
MEDIUM…hive/kt-sft/ktransformers/sft/peft_utils/peft_model.py153def active_adapters(self) -> list[str]:
LOW…hive/kt-sft/ktransformers/sft/peft_utils/peft_model.py919 except Exception: # something went wrong, roll back
LOW…t-sft/ktransformers/sft/flops_utils/lora_test_utils.py29 except Exception as e:
LOW…t-sft/ktransformers/sft/flops_utils/lora_test_utils.py40 except Exception as e:
LOW…t-sft/ktransformers/sft/flops_utils/lora_test_utils.py58 except Exception as e:
255 more matches not shown…
Hyper-Verbose Identifiers306 hits · 312 pts
SeverityFileLineSnippet
LOWarchive/setup.py80 def get_musa_bare_metal_version(self, musa_dir):
LOWarchive/setup.py90 def get_rocm_bare_metal_version(self, rocm_dir):
LOWarchive/setup.py154 def get_cuda_bare_metal_version(self, cuda_dir):
LOWarchive/setup.py163 def get_cuda_version_of_torch(self):
LOWarchive/setup.py365def run_command_with_live_tail(ext: str, command: List[str], output_lines: int = 20,
LOW…chive/merge_tensors/merge_safetensor_gguf_for_qwen3.py27def read_safetensor_keys_from_folder(folder_path) -> dict:
LOWarchive/merge_tensors/merge_safetensor_gguf.py15def read_safetensor_keys_from_folder(folder_path)->dict:
LOWarchive/csrc/custom_marlin/utils/format24.py21def _calculate_meta_reordering_scatter_offsets(m, meta_ncols, meta_dtype,
LOWarchive/csrc/custom_marlin/utils/format24.py52def sparse_semi_structured_from_dense_cutlass(dense):
LOWarchive/csrc/custom_marlin/utils/format24.py184def sparse_semi_structured_to_dense_cutlass(sparse, meta_reordered):
LOWarchive/kt-sft/setup.py105 def get_musa_bare_metal_version(self, musa_dir):
LOWarchive/kt-sft/setup.py115 def get_rocm_bare_metal_version(self, rocm_dir):
LOWarchive/kt-sft/setup.py179 def get_cuda_bare_metal_version(self, cuda_dir):
LOWarchive/kt-sft/setup.py188 def get_cuda_version_of_torch(self):
LOWarchive/kt-sft/setup.py384def run_command_with_live_tail(ext: str, command: List[str], output_lines: int = 20,
LOWarchive/kt-sft/merge_tensors/merge_safetensor_gguf.py15def read_safetensor_keys_from_folder(folder_path)->dict:
LOW…kt-sft/csrc/ktransformers_ext/examples/test_sft_moe.py604def test_backward_one_vs_many_comparison():
LOWarchive/kt-sft/csrc/custom_marlin/utils/format24.py21def _calculate_meta_reordering_scatter_offsets(m, meta_ncols, meta_dtype,
LOWarchive/kt-sft/csrc/custom_marlin/utils/format24.py52def sparse_semi_structured_from_dense_cutlass(dense):
LOWarchive/kt-sft/csrc/custom_marlin/utils/format24.py184def sparse_semi_structured_to_dense_cutlass(sparse, meta_reordered):
LOWarchive/kt-sft/ktransformers/local_chat.py239 # def first_token_argmax_baseline(model, tokenizer, prompt_text, device):
LOWarchive/kt-sft/ktransformers/operators/cpuinfer.py328 def update_importance_one_block(
LOWarchive/kt-sft/ktransformers/operators/cpuinfer.py473 def clear_importance_all_layers(
LOWarchive/kt-sft/ktransformers/operators/cpuinfer.py704 def get_all_kvcache_one_layer(
LOW…ve/kt-sft/ktransformers/operators/dynamic_attention.py271 def get_preselect_block_table_and_attn_score(
LOW…ive/kt-sft/ktransformers/operators/triton_attention.py165def _decode_grouped_att_m_fwd(
LOW…ive/kt-sft/ktransformers/operators/triton_attention.py313def _decode_softmax_reducev_fwd(
LOW…ive/kt-sft/ktransformers/operators/triton_attention.py358def decode_attention_fwd_grouped(
LOWarchive/kt-sft/ktransformers/util/custom_gguf.py97def quant_shape_to_byte_shape(shape: Sequence[int], quant_type: GGMLQuantizationType):
LOWarchive/kt-sft/ktransformers/util/custom_gguf.py635def translate_name_to_gguf_mixtral(name):
LOWarchive/kt-sft/ktransformers/util/custom_gguf.py704def translate_adapter_name_to_gguf(name):
LOW…chive/kt-sft/ktransformers/util/modeling_rope_utils.py29def _compute_default_rope_parameters(
LOW…chive/kt-sft/ktransformers/util/modeling_rope_utils.py71def _compute_linear_scaling_rope_parameters(
LOW…chive/kt-sft/ktransformers/util/modeling_rope_utils.py112def _compute_dynamic_ntk_parameters(
LOW…chive/kt-sft/ktransformers/util/modeling_rope_utils.py259def _compute_longrope_parameters(
LOW…chive/kt-sft/ktransformers/util/modeling_rope_utils.py407def _validate_default_rope_parameters(config: PretrainedConfig, ignore_keys: Optional[set] = None):
LOW…chive/kt-sft/ktransformers/util/modeling_rope_utils.py415def _validate_linear_scaling_rope_parameters(config: PretrainedConfig, ignore_keys: Optional[set] = None):
LOW…chive/kt-sft/ktransformers/util/modeling_rope_utils.py427def _validate_dynamic_scaling_rope_parameters(config: PretrainedConfig, ignore_keys: Optional[set] = None):
LOW…chive/kt-sft/ktransformers/util/modeling_rope_utils.py441def _validate_yarn_parameters(config: PretrainedConfig, ignore_keys: Optional[set] = None):
LOW…chive/kt-sft/ktransformers/util/modeling_rope_utils.py479def _validate_longrope_parameters(config: PretrainedConfig, ignore_keys: Optional[set] = None):
LOWarchive/kt-sft/ktransformers/util/custom_loader.py389 def get_undequanted_tensor_and_ggml_type(self, name):
LOWarchive/kt-sft/ktransformers/util/utils.py527def prefill_and_generate_capture(
LOW…/kt-sft/ktransformers/server/utils/create_interface.py38def get_thread_context_manager() -> GlobalContextManager:
LOW…kt-sft/ktransformers/server/backend/context_manager.py29 async def get_context_by_run_object(self, run: RunObject) -> ThreadContext:
LOWarchive/kt-sft/ktransformers/server/backend/base.py57 def report_last_time_performance(self):
LOW…transformers/server/backend/interfaces/transformers.py176 def format_and_tokenize_input_ids(self, thread_id: ObjectID, messages: List):
LOW…ransformers/server/backend/interfaces/balance_serve.py94def report_last_time_performance(profiler: Profiler):
LOW…ransformers/server/backend/interfaces/balance_serve.py411 def format_and_tokenize_input_ids(self, thread_id: ObjectID, messages: List):
LOW…/ktransformers/server/schemas/assistants/assistants.py133 def get_related_threads_objects(self) -> List:
LOW…ft/ktransformers/server/schemas/assistants/messages.py160 def stream_response_with_event(self, event: MessageBase.Status) -> MessageStreamResponse:
LOW…t/ktransformers/server/schemas/assistants/streaming.py136def wrap_async_generator_into_queue(async_events: AsyncIterable) -> asyncio.Queue:
LOW…kt-sft/ktransformers/server/schemas/assistants/runs.py105 def stream_response_with_event(self,event:RunBase.Status)->RunStreamResponse:
LOW…kt-sft/ktransformers/server/schemas/assistants/runs.py123 def create_message_creation_step(self):
LOW…kt-sft/ktransformers/server/balance_serve/sched_rpc.py179 def get_inference_context_raw(self):
LOW…/balance_serve/inference/distributed/parallel_state.py891def init_model_parallel_group(
LOW…/balance_serve/inference/distributed/parallel_state.py967def init_distributed_environment(
LOW…/balance_serve/inference/distributed/parallel_state.py1014def initialize_model_parallel(
LOW…/balance_serve/inference/distributed/parallel_state.py1091def ensure_model_parallel_initialized(
LOW…/balance_serve/inference/distributed/parallel_state.py1120def model_parallel_is_initialized():
LOW…/balance_serve/inference/distributed/parallel_state.py1129def patch_tensor_parallel_group(tp_group: GroupCoordinator):
246 more matches not shown…
Deep Nesting334 hits · 312 pts
SeverityFileLineSnippet
LOWarchive/setup.py238
LOWarchive/setup.py490
LOW…chive/merge_tensors/merge_safetensor_gguf_for_qwen3.py27
LOW…chive/merge_tensors/merge_safetensor_gguf_for_qwen3.py103
LOWarchive/merge_tensors/merge_safetensor_gguf.py15
LOWarchive/merge_tensors/merge_safetensor_gguf.py97
LOWarchive/csrc/ktransformers_ext/bench/bench_moe_torch.py80
LOWarchive/csrc/ktransformers_ext/bench/bench_moe.py31
LOWarchive/csrc/ktransformers_ext/bench/bench_moe_amx.py29
LOWarchive/csrc/ktransformers_ext/bench/bench_mlp.py28
LOWarchive/csrc/ktransformers_ext/bench/bench_linear.py28
LOW…ive/csrc/ktransformers_ext/bench/bench_linear_torch.py26
LOWarchive/csrc/ktransformers_ext/bench/bench_mlp_torch.py47
LOWarchive/kt-sft/setup.py259
LOWarchive/kt-sft/setup.py509
LOWarchive/kt-sft/merge_tensors/merge_safetensor_gguf.py15
LOWarchive/kt-sft/merge_tensors/merge_safetensor_gguf.py97
LOW…kt-sft/csrc/ktransformers_ext/bench/bench_moe_torch.py80
LOW…chive/kt-sft/csrc/ktransformers_ext/bench/bench_moe.py31
LOW…e/kt-sft/csrc/ktransformers_ext/bench/bench_moe_amx.py29
LOW…chive/kt-sft/csrc/ktransformers_ext/bench/bench_mlp.py28
LOW…ve/kt-sft/csrc/ktransformers_ext/bench/bench_linear.py28
LOW…sft/csrc/ktransformers_ext/bench/bench_linear_torch.py26
LOW…kt-sft/csrc/ktransformers_ext/bench/bench_mlp_torch.py47
LOW…ft/csrc/ktransformers_ext/examples/test_sft_amx_moe.py476
LOW…ft/csrc/ktransformers_ext/examples/test_sft_amx_moe.py501
LOW…ft/csrc/ktransformers_ext/examples/test_sft_amx_moe.py536
LOW…ft/csrc/ktransformers_ext/examples/test_sft_amx_moe.py551
LOWarchive/kt-sft/ktransformers/local_chat.py87
LOWarchive/kt-sft/ktransformers/optimize/optimize.py20
LOWarchive/kt-sft/ktransformers/optimize/optimize.py55
LOWarchive/kt-sft/ktransformers/operators/linear.py83
LOWarchive/kt-sft/ktransformers/operators/cpuinfer.py30
LOW…ve/kt-sft/ktransformers/operators/dynamic_attention.py271
LOW…ve/kt-sft/ktransformers/operators/dynamic_attention.py605
LOW…sft/ktransformers/operators/balance_serve_attention.py327
LOWarchive/kt-sft/ktransformers/operators/experts.py88
LOWarchive/kt-sft/ktransformers/operators/experts.py353
LOWarchive/kt-sft/ktransformers/operators/experts.py799
LOWarchive/kt-sft/ktransformers/operators/experts.py1022
LOWarchive/kt-sft/ktransformers/operators/experts.py1069
LOWarchive/kt-sft/ktransformers/util/custom_gguf.py170
LOWarchive/kt-sft/ktransformers/util/vendors.py23
LOWarchive/kt-sft/ktransformers/util/vendors.py75
LOWarchive/kt-sft/ktransformers/util/custom_loader.py47
LOWarchive/kt-sft/ktransformers/util/custom_loader.py100
LOWarchive/kt-sft/ktransformers/util/custom_loader.py260
LOWarchive/kt-sft/ktransformers/util/custom_loader.py296
LOWarchive/kt-sft/ktransformers/util/custom_loader.py426
LOWarchive/kt-sft/ktransformers/util/custom_loader.py508
LOWarchive/kt-sft/ktransformers/util/utils.py166
LOWarchive/kt-sft/ktransformers/util/utils.py61
LOWarchive/kt-sft/ktransformers/util/weight_loader.py59
LOWarchive/kt-sft/ktransformers/util/weight_loader.py190
LOWarchive/kt-sft/ktransformers/tests/mmlu_test_multi.py115
LOWarchive/kt-sft/ktransformers/tests/test_speed.py48
LOWarchive/kt-sft/ktransformers/tests/test_client.py15
LOW…chive/kt-sft/ktransformers/tests/humaneval/eval_api.py34
LOW…/kt-sft/ktransformers/server/utils/create_interface.py19
LOW…kt-sft/ktransformers/server/backend/context_manager.py29
274 more matches not shown…
Decorative Section Separators83 hits · 285 pts
SeverityFileLineSnippet
MEDIUMinstall.sh47# ─── Helpers ───────────────────────────────────────────────────────────────────
MEDIUMinstall.sh81# ─── Submodule init ────────────────────────────────────────────────────────────
MEDIUMinstall.sh97# ─── sglang install ───────────────────────────────────────────────────────────
MEDIUMinstall.sh126# ─── kt-kernel install ────────────────────────────────────────────────────────
MEDIUMinstall.sh145# ─── deps install ─────────────────────────────────────────────────────────────
MEDIUMinstall.sh161# ─── "all" subcommand ─────────────────────────────────────────────────────────
MEDIUMinstall.sh212# ─── Subcommand dispatcher ────────────────────────────────────────────────────
MEDIUM…/balance_serve/inference/distributed/parallel_state.py327 # --------------------------------------------
MEDIUMarchive/third_party/llamafile/tinyblas_cpu.h30// ╚═╝ ╚═╝╚═╝ ╚═╝ ╚══╝ ╚═════╝ ╚═══╝╚═╝ ╚═╝╚═════╝
MEDIUM…ive/ktransformers/operators/ascend/ascend_attention.py920 # -------------------------------------------------------
MEDIUM…ive/ktransformers/operators/ascend/ascend_attention.py922 # -------------------------------------------------------
MEDIUM…ive/ktransformers/operators/ascend/ascend_attention.py994 # -------------------------------------------------------
MEDIUM…ive/ktransformers/operators/ascend/ascend_attention.py996 # -------------------------------------------------------
MEDIUMarchive/ktransformers/tests/UT/test_kdeepseek_ln_npu.py12# ==========================
MEDIUMarchive/ktransformers/tests/UT/test_kdeepseek_ln_npu.py14# ==========================
MEDIUM…s/tests/UT/test_kdeepseek_attention_w8a8a2serve_npu.py221# ==========================
MEDIUM…s/tests/UT/test_kdeepseek_attention_w8a8a2serve_npu.py223# ==========================
MEDIUM…/balance_serve/inference/distributed/parallel_state.py327 # --------------------------------------------
MEDIUM…sformers/models/ascend/custom_ascend_modeling_qwen3.py87 # ---------------------------------------------------
MEDIUM…sformers/models/ascend/custom_ascend_modeling_qwen3.py89 # ---------------------------------------------------
MEDIUMkt-kernel/setup.py61# -------------------------
MEDIUMkt-kernel/setup.py63# -------------------------
MEDIUMkt-kernel/bench/bench_write_buffer.py102# ==============================================================================
MEDIUMkt-kernel/bench/bench_write_buffer.py104# ==============================================================================
MEDIUMkt-kernel/bench/bench_write_buffer.py326# ==============================================================================
MEDIUMkt-kernel/bench/bench_write_buffer.py328# ==============================================================================
MEDIUMkt-kernel/bench/bench_write_buffer.py408# ==============================================================================
MEDIUMkt-kernel/bench/bench_write_buffer.py410# ==============================================================================
MEDIUMkt-kernel/test/test_native_moe_loader_auto_release.py79# ---------------------------------------------------------------------------
MEDIUMkt-kernel/test/test_native_moe_loader_auto_release.py81# ---------------------------------------------------------------------------
MEDIUMkt-kernel/test/test_native_moe_loader_auto_release.py199# ---------------------------------------------------------------------------
MEDIUMkt-kernel/test/test_native_moe_loader_auto_release.py201# ---------------------------------------------------------------------------
MEDIUMkt-kernel/python/experts.py298# =============================================================================
MEDIUMkt-kernel/python/experts.py300# =============================================================================
MEDIUMkt-kernel/python/cli/utils/model_registry.py377# ============================================================================
MEDIUMkt-kernel/python/cli/utils/model_registry.py379# ============================================================================
MEDIUMkt-kernel/python/sft/wrapper.py46# =============================================================================
MEDIUMkt-kernel/python/sft/wrapper.py48# =============================================================================
MEDIUMkt-kernel/python/sft/wrapper.py136# =============================================================================
MEDIUMkt-kernel/python/sft/wrapper.py138# =============================================================================
MEDIUMkt-kernel/python/sft/wrapper.py408# =============================================================================
MEDIUMkt-kernel/python/sft/wrapper.py410# =============================================================================
MEDIUMkt-kernel/python/sft/arch.py21# =============================================================================
MEDIUMkt-kernel/python/sft/arch.py23# =============================================================================
MEDIUMkt-kernel/python/sft/arch.py42# =============================================================================
MEDIUMkt-kernel/python/sft/arch.py44# =============================================================================
MEDIUMkt-kernel/python/sft/lora.py307# =============================================================================
MEDIUMkt-kernel/python/sft/lora.py309# =============================================================================
MEDIUMkt-kernel/python/sft/lora.py454# =============================================================================
MEDIUMkt-kernel/python/sft/lora.py456# =============================================================================
MEDIUMkt-kernel/python/sft/lora.py580# =============================================================================
MEDIUMkt-kernel/python/sft/lora.py582# =============================================================================
MEDIUMkt-kernel/python/sft/lora.py29# =============================================================================
MEDIUMkt-kernel/python/sft/lora.py31# =============================================================================
MEDIUMkt-kernel/python/sft/lora.py82# =============================================================================
MEDIUMkt-kernel/python/sft/lora.py84# =============================================================================
MEDIUMkt-kernel/python/sft/lora.py132# =============================================================================
MEDIUMkt-kernel/python/sft/lora.py134# =============================================================================
MEDIUMkt-kernel/python/sft/lora.py523# =============================================================================
MEDIUMkt-kernel/python/sft/lora.py525# =============================================================================
23 more matches not shown…
Redundant / Tautological Comments101 hits · 153 pts
SeverityFileLineSnippet
LOWdocker/docker-utils.sh106 # Check if image exists
LOWdocker/docker-utils.sh230# Check if Docker daemon is running
LOWdocker/docker-utils.sh240# Check if user is logged into Docker registry
LOWdocker/docker-utils.sh297# Check if file/directory exists and is writable
LOWdocker/build-docker-tar.sh373 # Check if tar file already exists
LOWdocker/push-to-dockerhub.sh290 # Check if we should skip build
LOWdocker/push-to-dockerhub.sh862 # Check if we should skip build
LOWarchive/kt-sft/ktransformers/operators/experts.py470 # Check if we need to allocate or expand buffers
LOWarchive/kt-sft/ktransformers/util/weight_loader.py177 # Check if any safetensor files exist in the folder
LOWarchive/kt-sft/ktransformers/util/weight_loader.py197 # Check if path exists
LOWarchive/kt-sft/ktransformers/util/weight_loader.py362 # Check if any GGUF files exist in the folder
LOW…-sft/ktransformers/server/api/openai/endpoints/chat.py203 # Check if tools are present
LOW…hive/kt-sft/ktransformers/sft/peft_utils/lora_layer.py571 fan_in_fan_out: bool = False, # Set this to True if the layer to replace stores weight like (fan_in, fan_out)
LOW…hive/kt-sft/ktransformers/sft/peft_utils/lora_layer.py1021 fan_in_fan_out: bool = False, # Set this to True if the layer to replace stores weight like (fan_in, fan_out)
LOWarchive/ktransformers/util/weight_loader.py177 # Check if any safetensor files exist in the folder
LOWarchive/ktransformers/util/weight_loader.py197 # Check if path exists
LOWarchive/ktransformers/util/weight_loader.py362 # Check if any GGUF files exist in the folder
LOW…hive/ktransformers/server/api/openai/endpoints/chat.py203 # Check if tools are present
LOWkt-kernel/bench/bench_bf16_moe.py222 # Print results
LOWkt-kernel/bench/bench_fp8_perchannel_moe.py234 # Print results
LOWkt-kernel/bench/bench_fp8_moe.py243 # Print results
LOWkt-kernel/test/per_commit/test_moe_amx_bench_int8.py23# Check if dependencies are available
LOWkt-kernel/test/per_commit/test_moe_amx_accuracy_int4.py19# Check if dependencies are available
LOW…kernel/test/per_commit/test_moe_amx_accuracy_int4_1.py19# Check if dependencies are available
LOWkt-kernel/test/per_commit/test_moe_amx_bench_int4.py23# Check if dependencies are available
LOWkt-kernel/test/per_commit/test_basic_cpu.py17# Check if kt_kernel_ext is available
LOW…ernel/test/per_commit/test_moe_amx_accuracy_int4_1k.py19# Check if dependencies are available
LOWkt-kernel/test/per_commit/test_moe_amx_accuracy_int8.py19# Check if dependencies are available
LOWkt-kernel/test/per_commit/test_moe_amx_bench_int4_1k.py24# Check if dependencies are available
LOWkt-kernel/test/per_commit/test_moe_amx_bench_int4_1.py23# Check if dependencies are available
LOWkt-kernel/python/_cpu_detect.py87 # Check if all required flags are present
LOWkt-kernel/python/utils/llamafile.py84 # Check if intermediate_size is divisible by QK_K
LOWkt-kernel/python/utils/loader.py213 # Check if backward weights exist
LOWkt-kernel/python/utils/loader.py341 # Check if any key matches this format pattern
LOWkt-kernel/python/cli/main.py373 # Check if path exists or parent is writable
LOWkt-kernel/python/cli/main.py380 # Check if we can create it (parent writable)
LOWkt-kernel/python/cli/main.py407 # Check if already installed
LOWkt-kernel/python/cli/main.py505 # Check if this is first run
LOWkt-kernel/python/cli/utils/console.py142 # Check if response matches a choice directly
LOWkt-kernel/python/cli/utils/quant_interactive.py226 # Check if available space >= required * 1.2 (20% buffer)
LOWkt-kernel/python/cli/utils/model_verifier.py28 # Read file in chunks to handle large files
LOWkt-kernel/python/cli/utils/model_verifier.py671 # Check if already verified
LOWkt-kernel/python/cli/utils/model_verifier.py683 # Check if repo_id exists
LOWkt-kernel/python/cli/utils/model_scanner.py94 # Check if size meets minimum threshold
LOWkt-kernel/python/cli/utils/model_scanner.py633 # Check if this root is a parent of any already selected root
LOWkt-kernel/python/cli/utils/kv_cache_calculator.py67 # Check if it's MLA (Multi-head Latent Attention) model
LOWkt-kernel/python/cli/utils/kv_cache_calculator.py96 # Check if it's NSA (Native Sparse Attention) model
LOWkt-kernel/python/cli/utils/model_discovery.py101 # Check if already in registry
LOWkt-kernel/python/cli/utils/model_discovery.py105 # Check if already discovered in this session
LOWkt-kernel/python/cli/utils/tuna_engine.py202 # Check if process has output
LOWkt-kernel/python/cli/utils/tuna_engine.py321 # Check if we got a valid response
LOWkt-kernel/python/cli/utils/tuna_engine.py432 # Check if even 0 doesn't work
LOWkt-kernel/python/cli/utils/download_helper.py77 # Check if filename matches pattern
LOWkt-kernel/python/cli/utils/environment.py117 # Check if venv is available (built into Python)
LOWkt-kernel/python/cli/utils/environment.py146 # Check if env_name appears as a separate word in the output
LOWkt-kernel/python/cli/utils/environment.py703 # Check if writable
LOWkt-kernel/python/cli/utils/environment.py742 # Check if parent exists for paths that don't exist yet
LOWkt-kernel/python/cli/utils/environment.py917 # Check if this directory is a model
LOWkt-kernel/python/cli/utils/model_registry.py276 # Check if query is contained in name
LOWkt-kernel/python/cli/utils/model_registry.py280 # Check if query is contained in aliases
41 more matches not shown…
Verbosity Indicators68 hits · 103 pts
SeverityFileLineSnippet
LOWkt-kernel/operators/moe-sft-tp.hpp353 // Step 1: For each NUMA, allocate and copy partitioned weights
LOWkt-kernel/operators/moe-sft-tp.hpp392 // Step 2: Set weight pointers BEFORE load_weights (Bug #24 fix)
LOWkt-kernel/operators/moe-sft-tp.hpp399 // Step 3: Prepare backward weights (this also clears weight pointers)
LOWkt-kernel/operators/amx/sft_moe.hpp954 // Step 1: Expert routing (reuse base class logic)
LOWkt-kernel/operators/amx/sft_moe.hpp973 // Step 2: Buffer pool allocation (reuse base class logic)
LOWkt-kernel/operators/amx/sft_moe.hpp1075 // Step 3: Copy input to expert buffers
LOWkt-kernel/operators/amx/sft_moe.hpp1112 // Step 4: Quantize input
LOWkt-kernel/operators/amx/sft_moe.hpp1120 // Step 5: Gate + Up GEMM (base projection)
LOWkt-kernel/operators/amx/sft_moe.hpp1227 // Step 6: Activation (silu(gate) * up)
LOWkt-kernel/operators/amx/sft_moe.hpp1284 // Step 7: Quantize intermediate for down projection
LOWkt-kernel/operators/amx/sft_moe.hpp1293 // Step 8: Down GEMM
LOWkt-kernel/operators/amx/sft_moe.hpp1345 // Step 9: Weighted merge
LOWkt-kernel/operators/amx/sft_moe.hpp1559 // Step 1: Down projection backward
LOWkt-kernel/operators/amx/sft_moe.hpp1721 // Step 4: Compute grad_weights (gradient for routing weights)
LOWkt-kernel/operators/amx/sft_moe.hpp3374 // Step 1: input @ lora_A^T -> lora_intermediate
LOWkt-kernel/operators/amx/sft_moe.hpp3406 // Step 2: Quantize lora_intermediate to BufferA
LOWkt-kernel/operators/amx/sft_moe.hpp3539 // Step 1: intermediate @ down_lora_A^T -> lora_intermediate
LOWkt-kernel/operators/amx/sft_moe.hpp3568 // Step 2: Quantize lora_intermediate to BufferA
LOWkt-kernel/operators/amx/sft_moe.hpp3759 // Step 1: intermediate = input @ lora_A^T (optimized with T_BLOCK=4, R_BLOCK=4)
LOWkt-kernel/operators/amx/sft_moe.hpp3765 // Step 2: output += scale * (intermediate @ lora_B_transposed)
LOWkt-kernel/operators/amx/sft_moe.hpp3820 // Step 1: intermediate = input @ lora_A^T (optimized with T_BLOCK=4, R_BLOCK=4)
LOWkt-kernel/operators/amx/sft_moe.hpp3831 // Step 2: output += scale * (intermediate @ lora_B_transposed)
LOWkt-kernel/operators/amx/sft_moe.hpp4231 // Step 1: Zero per-expert grad_output buffers
LOWkt-kernel/operators/amx/sft_moe.hpp4243 // Step 2: Scatter grad_output to per-expert BF16 buffers
LOWkt-kernel/operators/amx/sft_moe.hpp4290 // Step 3: Quantize scattered grad_output to BufferA
LOWkt-kernel/operators/amx/sft_moe.hpp4383 // Step 1: grad_output @ down_lora_B_transposed -> [local_num_tokens, rank]
LOWkt-kernel/operators/amx/sft_moe.hpp4390 // Step 2: grad_times_b @ down_lora_A -> [local_num_tokens, inter_size] (AVX512)
LOWkt-kernel/operators/amx/sft_moe.hpp4402 // Step 5: LoRA gradient computation (parallelized across blocks)
LOWkt-kernel/operators/amx/sft_moe.hpp5414 // Step 6: grad_A = G_B^T @ X
LOWkt-kernel/operators/amx/test/test_lora_fused_add.cpp1657 // Step 1: Reduce 512 -> 256 by adding high/low halves (8 ops)
LOWkt-kernel/operators/amx/test/test_lora_fused_add.cpp1671 // Step 2: Pack pairs into single 512-bit vectors
LOWkt-kernel/operators/amx/test/test_lora_fused_add.cpp1679 // Step 3: Reduce 256 -> 128 within each pair
LOWkt-kernel/operators/amx/test/test_lora_fused_add.cpp1686 // Step 4: Reduce 128 -> 64 -> 32 within each
LOWkt-kernel/operators/amx/la/avx_kernels.hpp924 // Step 1: Interleave 16-bit
LOWkt-kernel/operators/amx/la/avx_kernels.hpp934 // Step 2: Interleave 32-bit
LOWkt-kernel/operators/amx/la/avx_kernels.hpp944 // Step 3: Interleave 64-bit
LOWkt-kernel/operators/amx/la/avx_kernels.hpp985 // Step 1: Interleave 16-bit
LOWkt-kernel/operators/amx/la/avx_kernels.hpp1003 // Step 2: Interleave 32-bit
LOWkt-kernel/operators/amx/la/avx_kernels.hpp1021 // Step 3: Interleave 64-bit
LOWkt-kernel/operators/amx/la/avx_kernels.hpp1039 // Step 4: Permute 128-bit lanes
LOWkt-kernel/python/cli/utils/quant_interactive.py245 # Step 1: Select model
LOWkt-kernel/python/cli/utils/quant_interactive.py260 # Step 2: Configure quantization method
LOWkt-kernel/python/cli/utils/quant_interactive.py263 # Step 3: Configure CPU parameters
LOWkt-kernel/python/cli/utils/quant_interactive.py266 # Step 4: Configure output path
LOWkt-kernel/python/cli/utils/quant_interactive.py288 # Step 5: Calculate space requirements and check availability
LOWkt-kernel/python/cli/utils/run_interactive.py893 # Step 1: Select model
LOWkt-kernel/python/cli/utils/run_interactive.py898 # Step 2: Select inference method
LOWkt-kernel/python/cli/utils/run_interactive.py993 # Step 3: Configure NUMA and CPU
LOWkt-kernel/python/cli/utils/run_interactive.py996 # Step 4: Configure GPU experts
LOWkt-kernel/python/cli/utils/run_interactive.py999 # Step 5: Configure KV Cache (only for raw)
LOWkt-kernel/python/cli/utils/run_interactive.py1003 # Step 6: Select GPUs and TP
LOWkt-kernel/python/cli/utils/run_interactive.py1008 # Step 7: Configure parsers (optional)
LOWkt-kernel/python/cli/utils/run_interactive.py1011 # Step 8: Configure host and port
LOWkt-kernel/python/cli/utils/run_interactive.py1035 # Step 9: Save configuration
LOWkt-kernel/python/cli/commands/run.py330 # Step 2: Resolve model
LOWkt-kernel/python/cli/commands/run.py390 # Step 3: Check quantized weights (only if explicitly requested)
LOWkt-kernel/python/cli/commands/run.py414 # Step 4: Build command
LOWkt-kernel/python/cli/commands/run.py514 # Step 5: Show configuration summary
LOWkt-kernel/python/cli/commands/run.py544 # Step 6: Show or execute
LOWkt-kernel/python/cli/commands/model.py2583 # Step 1: Delete the corrupted/missing file if it exists
8 more matches not shown…
Cross-Language Confusion16 hits · 78 pts
SeverityFileLineSnippet
HIGHarchive/csrc/custom_marlin/utils/format24.py142 -1, idxs0.unsqueeze(-1)) # type: ignore[possibly-undefined]
HIGHarchive/csrc/custom_marlin/utils/format24.py149 k // 2) # type: ignore[possibly-undefined]
HIGHarchive/csrc/custom_marlin/utils/format24.py172 (m * meta_ncols, )) # type: ignore[possibly-undefined]
HIGHarchive/kt-sft/csrc/custom_marlin/utils/format24.py142 -1, idxs0.unsqueeze(-1)) # type: ignore[possibly-undefined]
HIGHarchive/kt-sft/csrc/custom_marlin/utils/format24.py149 k // 2) # type: ignore[possibly-undefined]
HIGHarchive/kt-sft/csrc/custom_marlin/utils/format24.py172 (m * meta_ncols, )) # type: ignore[possibly-undefined]
HIGH…xt/operators/custom_marlin/quantize/utils/format_24.py142 -1, idxs0.unsqueeze(-1)) # type: ignore[possibly-undefined]
HIGH…xt/operators/custom_marlin/quantize/utils/format_24.py149 k // 2) # type: ignore[possibly-undefined]
HIGH…xt/operators/custom_marlin/quantize/utils/format_24.py172 (m * meta_ncols, )) # type: ignore[possibly-undefined]
HIGH…hive/kt-sft/ktransformers/sft/peft_utils/peft_model.py1393 trainable params: 1843200 || all params: 775873280 || trainable%: 0.23756456724479544
HIGH…xt/operators/custom_marlin/quantize/utils/format_24.py142 -1, idxs0.unsqueeze(-1)) # type: ignore[possibly-undefined]
HIGH…xt/operators/custom_marlin/quantize/utils/format_24.py149 k // 2) # type: ignore[possibly-undefined]
HIGH…xt/operators/custom_marlin/quantize/utils/format_24.py172 (m * meta_ncols, )) # type: ignore[possibly-undefined]
HIGHkt-kernel/python/cli/i18n.py306 "sglang_recommend_source": "Recommend reinstalling with the kvcache-ai fork: pip uninstall sglang -y && pip inst
HIGHkt-kernel/python/cli/i18n.py926 "sglang_recommend_source": "建议重新安装 kvcache-ai 分支: pip uninstall sglang -y && pip install sglang-kt",
HIGHkt-kernel/python/cli/commands/doctor.py426 kt_kernel_hint = "Reinstall SGLang: pip uninstall sglang -y && pip install sglang-kt (or run ./install.sh fr
Dead Code37 hits · 74 pts
SeverityFileLineSnippet
MEDIUM…chive/kt-sft/ktransformers/models/modeling_deepseek.py165
MEDIUM…chive/kt-sft/ktransformers/models/modeling_deepseek.py166
MEDIUM…chive/kt-sft/ktransformers/models/modeling_deepseek.py195
MEDIUM…chive/kt-sft/ktransformers/models/modeling_deepseek.py196
MEDIUMarchive/ktransformers/models/modeling_deepseek.py164
MEDIUMarchive/ktransformers/models/modeling_deepseek.py165
MEDIUMarchive/ktransformers/models/modeling_deepseek.py194
MEDIUMarchive/ktransformers/models/modeling_deepseek.py195
MEDIUMkt-kernel/python/cli/commands/bench.py131
MEDIUMkt-kernel/python/cli/commands/bench.py133
MEDIUMkt-kernel/python/cli/commands/bench.py137
MEDIUMkt-kernel/python/cli/commands/bench.py140
MEDIUMkt-kernel/python/cli/commands/bench.py148
MEDIUMkt-kernel/python/cli/commands/bench.py149
MEDIUMkt-kernel/python/cli/commands/bench.py154
MEDIUMkt-kernel/python/cli/commands/bench.py155
MEDIUMkt-kernel/python/cli/commands/bench.py160
MEDIUMkt-kernel/python/cli/commands/bench.py173
MEDIUMkt-kernel/python/cli/commands/bench.py176
MEDIUMkt-kernel/python/cli/commands/bench.py177
MEDIUMkt-kernel/python/cli/commands/bench.py179
MEDIUMkt-kernel/examples/test_mla.py298
MEDIUMkt-kernel/examples/test_mla.py300
MEDIUMkt-kernel/examples/test_mla.py301
MEDIUMkt-kernel/examples/test_mla.py302
MEDIUMkt-kernel/examples/test_mla.py307
MEDIUMkt-kernel/examples/test_mla.py309
MEDIUMkt-kernel/examples/test_mla.py310
MEDIUMkt-kernel/examples/test_mla.py311
MEDIUMkt-kernel/examples/test_gate.py40
MEDIUMkt-kernel/examples/test_gate.py42
MEDIUMkt-kernel/examples/test_gate.py43
MEDIUMkt-kernel/examples/test_gate.py44
MEDIUMkt-kernel/examples/test_mla_quant.py32
MEDIUMkt-kernel/examples/test_mla_quant.py34
MEDIUMkt-kernel/examples/test_mla_quant.py35
MEDIUMkt-kernel/examples/test_mla_quant.py36
Self-Referential Comments22 hits · 66 pts
SeverityFileLineSnippet
MEDIUMarchive/csrc/balance_serve/kvc2/test/pytest_load.py8# Create a kvc2 instance
MEDIUM…/balance_serve/kvc2/test/pytest_raw_insert_and_read.py8# Create a kvc2 instance
MEDIUMarchive/csrc/balance_serve/kvc2/test/pytest_mem_read.py8# Create a kvc2 instance
MEDIUM…csrc/balance_serve/kvc2/test/pytest_mem_prefix_test.py8# Create a kvc2 instance
MEDIUMarchive/csrc/custom_marlin/utils/quant_utils.py41 # Create a tensor for bitwise right shift operation
MEDIUMarchive/kt-sft/csrc/custom_marlin/utils/quant_utils.py41 # Create a tensor for bitwise right shift operation
MEDIUMarchive/kt-sft/ktransformers/util/custom_loader.py552 # Create the appropriate loader based on detected file types
MEDIUMarchive/kt-sft/ktransformers/util/utils.py266 # This function is to check if we run this model on XPU with FP16 dtype
MEDIUM…/balance_serve/inference/distributed/pynccl_wrapper.py1# This file is a pure Python wrapper for the NCCL library.
MEDIUMarchive/ktransformers/util/custom_loader.py579 # Create the appropriate loader based on detected file types
MEDIUMarchive/ktransformers/util/utils.py324 # This function is to check if we run this model on XPU with FP16 dtype
MEDIUM…/balance_serve/inference/distributed/pynccl_wrapper.py1# This file is a pure Python wrapper for the NCCL library.
MEDIUMkt-kernel/python/experts.py81 # Create a mask where experts 0, 2, 5 are on GPU
MEDIUMkt-kernel/python/experts_base.py288 # Create a new pinned tensor and copy data into it
MEDIUMkt-kernel/python/utils/llamafile.py122 # Initialize base class
MEDIUMkt-kernel/python/utils/moe_kernel.py86 # Initialize base class
MEDIUMkt-kernel/python/utils/amx.py238 # Initialize base class
MEDIUMkt-kernel/python/cli/main.py47# Create main app with dynamic help
MEDIUMkt-kernel/python/cli/utils/user_model_registry.py88 self.save() # Create the file
MEDIUMkt-kernel/python/cli/commands/model.py899 # Create a sub-row with empty cells except for the first column (7 columns total with #)
MEDIUMkt-kernel/python/cli/commands/model.py948 # Create a sub-row with empty cells except for the first column
MEDIUMkt-kernel/python/sft/weights.py171 # Create a CPU tensor with the correct shape but NO physical memory.
Docstring Block Structure13 hits · 65 pts
SeverityFileLineSnippet
HIGHarchive/kt-sft/ktransformers/util/custom_loader.py509 Create a model loader for the given path by detecting the model format. The function checks for the pre
HIGH…-sft/ktransformers/ktransformers_ext/triton/fp8gemm.py86 Dequantizes the given weight tensor using the provided scale tensor. Args: x (torch.Tensor): The quant
HIGH…hive/kt-sft/ktransformers/sft/peft_utils/peft_model.py935Get the status of each adapter layer in the model. This method returns a list of `TunerLayerStatus` dataclass i
HIGH…hive/kt-sft/ktransformers/sft/peft_utils/peft_model.py965Get the status of tuners of the model. This method returns a `TunerModelStatus` dataclass instance, which conta
HIGH…hive/kt-sft/ktransformers/sft/peft_utils/peft_model.py1663Get the status of each adapter layer in the model. This function returns a list of `TunerLayerStatus` dataclass ins
HIGH…hive/kt-sft/ktransformers/sft/peft_utils/peft_model.py1781Get the status of tuners of the model. This function returns a `TunerModelStatus` dataclass instance, which contain
HIGHarchive/ktransformers/util/custom_loader.py536 Create a model loader for the given path by detecting the model format. The function checks for the pre
HIGH…hive/ktransformers/ktransformers_ext/triton/fp8gemm.py86 Dequantizes the given weight tensor using the provided scale tensor. Args: x (torch.Tensor): The quant
HIGHkt-kernel/python/_cpu_detect.py166 Load the appropriate kt_kernel_ext variant. Tries to import the specified variant, with automatic fallback to
HIGHkt-kernel/python/experts.py153 Factory method to create the appropriate backend implementation. Args: layer_idx: Layer in
HIGHkt-kernel/python/cli/utils/tuna_engine.py21 Get the number of experts per layer from model config. Args: model_path: Path to the model directory
HIGHkt-kernel/python/cli/utils/tuna_engine.py397 Run tuna auto-tuning to find optimal num_gpu_experts. Args: model_path: Path to the model tens
HIGHkt-kernel/python/sft/arch.py63 Get MoE architecture configuration based on model type. Args: config: HuggingFace model configuration
AI Slop Vocabulary19 hits · 52 pts
SeverityFileLineSnippet
MEDIUM…hive/kt-sft/ktransformers/sft/peft_utils/peft_model.py831 # TODO: consider replacing this patching of methods with a more robust mechanism: setting a flag and
LOW…hive/kt-sft/ktransformers/sft/peft_utils/lora_model.py128 # model, just add a `peft_config` dict attribute to your model.
LOW…ive/ktransformers/operators/ascend/ascend_attention.py215 # FIXME this is wrong in random choose pages for sched, currently just use kv without history
MEDIUMarchive/ktransformers/models/modeling_smallthinker.py1051# "unexpected if using padding tokens in conjunction with `inputs_embeds.`"
MEDIUMkt-kernel/bench/compare_moe_performance.py291 """Get comprehensive system information"""
MEDIUMkt-kernel/operators/amx/sft_moe.hpp94// Check BF16 buffer for NaN/Inf (using robust v != v check)
MEDIUMkt-kernel/operators/amx/sft_moe.hpp99 // Use val != val for robust NaN detection
MEDIUMkt-kernel/operators/amx/sft_moe.hpp121// Check FP32 buffer for NaN/Inf (using robust v != v check)
MEDIUMkt-kernel/operators/amx/sft_moe.hpp126 // Use val != val for robust NaN detection
MEDIUMkt-kernel/operators/amx/sft_moe.hpp1447 // Use v != v for robust NaN detection
MEDIUMkt-kernel/operators/amx/sft_moe.hpp1799 // Use v != v for robust NaN detection
MEDIUMkt-kernel/operators/amx/sft_moe.hpp1839 // Use fv != fv for robust NaN detection
MEDIUMkt-kernel/operators/amx/sft_moe.hpp1959 // Use v != v for robust NaN detection
MEDIUMkt-kernel/operators/amx/test/mmq.cpp689 // pack again with 128 to fully utilize vector length
MEDIUMkt-kernel/operators/amx/test/mmq.cpp731 // pack again with 128 to fully utilize vector length
MEDIUMkt-kernel/operators/amx/test/mmq.cpp833 // pack again with 128 to fully utilize vector length
MEDIUMkt-kernel/operators/amx/test/mmq-test.cpp693 // pack again with 128 to fully utilize vector length
MEDIUMkt-kernel/operators/amx/test/mmq-test.cpp735 // pack again with 128 to fully utilize vector length
MEDIUMkt-kernel/operators/amx/test/mmq-test.cpp837 // pack again with 128 to fully utilize vector length
Hallucination Indicators3 hits · 45 pts
SeverityFileLineSnippet
CRITICAL…sformers/models/ascend/custom_ascend_modeling_qwen3.py70 self.model.embed_tokens.weight.data = self.model.embed_tokens.weight.data.to(torch.float16)
CRITICAL…sformers/models/ascend/custom_ascend_modeling_qwen3.py73 self.model.norm.weight.data = self.model.norm.weight.data.to(torch.float16)
CRITICAL…sformers/models/ascend/custom_ascend_modeling_qwen3.py75 self.model.norm.bias.data = self.model.norm.bias.data.to(torch.float16)
Slop Phrases25 hits · 35 pts
SeverityFileLineSnippet
LOWarchive/kt-sft/ktransformers/operators/experts.py969 # the current expert. We need to make sure to multiply the output hidden
LOWarchive/kt-sft/ktransformers/operators/experts.py1126 # the current expert. We need to make sure to multiply the output hidden
LOWarchive/kt-sft/ktransformers/operators/experts.py1475 # the current expert. We need to make sure to multiply the output hidden
LOWarchive/kt-sft/ktransformers/operators/experts.py1771 # the current expert. We need to make sure to multiply the output hidden
LOWarchive/kt-sft/ktransformers/models/modeling_mixtral.py878 # the current expert. We need to make sure to multiply the output hidden
LOWarchive/kt-sft/ktransformers/models/modeling_mixtral.py1465 loss += self.router_aux_loss_coef * aux_loss.to(loss.device) # make sure to reside in the same device
LOW…hive/kt-sft/ktransformers/models/modeling_qwen3_moe.py292 # the current expert. We need to make sure to multiply the output hidden
LOW…hive/kt-sft/ktransformers/models/modeling_qwen3_moe.py1166 loss += self.router_aux_loss_coef * aux_loss.to(loss.device) # make sure to reside in the same device
LOW…hive/kt-sft/ktransformers/models/modeling_qwen2_moe.py848 # the current expert. We need to make sure to multiply the output hidden
LOW…hive/kt-sft/ktransformers/models/modeling_qwen2_moe.py1455 loss += self.router_aux_loss_coef * aux_loss.to(loss.device) # make sure to reside in the same device
LOWarchive/ktransformers/operators/experts.py547 # the current expert. We need to make sure to multiply the output hidden
LOWarchive/ktransformers/operators/experts.py665 # the current expert. We need to make sure to multiply the output hidden
LOWarchive/ktransformers/operators/experts.py863 # the current expert. We need to make sure to multiply the output hidden
LOWarchive/ktransformers/operators/experts.py1161 # the current expert. We need to make sure to multiply the output hidden
LOWarchive/ktransformers/models/modeling_smallthinker.py116 # the current expert. We need to make sure to multiply the output hidden
LOWarchive/ktransformers/models/modeling_smallthinker.py956 loss += self.router_aux_loss_coef * aux_loss.to(loss.device) # make sure to reside in the same device
LOWarchive/ktransformers/models/modeling_mixtral.py877 # the current expert. We need to make sure to multiply the output hidden
LOWarchive/ktransformers/models/modeling_mixtral.py1464 loss += self.router_aux_loss_coef * aux_loss.to(loss.device) # make sure to reside in the same device
LOWarchive/ktransformers/models/modeling_qwen3_moe.py291 # the current expert. We need to make sure to multiply the output hidden
LOWarchive/ktransformers/models/modeling_qwen3_moe.py1165 loss += self.router_aux_loss_coef * aux_loss.to(loss.device) # make sure to reside in the same device
LOWarchive/ktransformers/models/modeling_qwen2_moe.py847 # the current expert. We need to make sure to multiply the output hidden
LOWarchive/ktransformers/models/modeling_qwen2_moe.py1454 loss += self.router_aux_loss_coef * aux_loss.to(loss.device) # make sure to reside in the same device
MEDIUMarchive/ktransformers/models/custom_cache.py372 # you can use following code as check
LOWarchive/ktransformers/models/modeling_qwen3_next.py853 # the current expert. We need to make sure to multiply the output hidden
LOWarchive/ktransformers/models/modeling_qwen3_next.py1255 loss += self.router_aux_loss_coef * aux_loss.to(loss.device) # make sure to reside in the same device
Synthetic Comment Markers2 hits · 15 pts
SeverityFileLineSnippet
HIGHkt-kernel/examples/test_moe_amx.py428 # Only test BF16 and INT8 as requested
HIGHkt-kernel/examples/test_moe_amx.py486 # Only test BF16 and INT8 as requested
Example Usage Blocks4 hits · 6 pts
SeverityFileLineSnippet
LOWdocker/build-docker-tar.sh15# Usage:
LOWdocker/push-to-dockerhub.sh16# Usage:
LOWdocker/push-to-dockerhub.sh588# Usage:
LOW…hive/csrc/balance_serve/kvc2/test/test_cuda_stream.cpp86// Example usage