liguodongiot/llm-action

Pattern Findings

425 matches across 14 categories. Click a row to expand file-level details.

Hallucination Indicators19 hits · 255 pts

Severity	File	Line	Snippet
CRITICAL	…m-train/pytorch/distribution/tensor-parallel/README.md	54	torch.distributed.tensor.parallel.style.make_input_replicate_1d(input, device_mesh=None)
CRITICAL	…-train/pytorch/distribution/pipeline-parallel/1-流水线.md	73	- torch.distributed.pipeline.sync.skip.skippable.skippable(stash=(), pop=())
CRITICAL	…-train/pytorch/distribution/pipeline-parallel/1-流水线.md	113	- torch.distributed.pipeline.sync.skip.skippable.stash(name, tensor)
CRITICAL	…-train/pytorch/distribution/pipeline-parallel/1-流水线.md	115	- torch.distributed.pipeline.sync.skip.skippable.pop(name)
CRITICAL	…-train/pytorch/distribution/pipeline-parallel/1-流水线.md	117	- torch.distributed.pipeline.sync.skip.skippable.verify_skippables(module)
CRITICAL	…t2/merge_ck_and_inference/checkpoint_saver_megatron.py	247	model.language_model.embedding.word_embeddings.weight.data.copy_(out_word_embed[tp_rank])
CRITICAL	…t2/merge_ck_and_inference/checkpoint_saver_megatron.py	249	model.language_model.embedding.position_embeddings.weight.data.copy_(pos_embed)
CRITICAL	…t2/merge_ck_and_inference/checkpoint_saver_megatron.py	302	l.self_attention.query_key_value.weight.data.copy_(qkv_weight[tp_rank])
CRITICAL	…t2/merge_ck_and_inference/checkpoint_saver_megatron.py	303	l.self_attention.dense.weight.data.copy_(dense_weight[tp_rank])
CRITICAL	…t2/merge_ck_and_inference/checkpoint_saver_megatron.py	306	l.mlp.dense_h_to_4h.weight.data.copy_(mlp_l0_weight[tp_rank])
CRITICAL	…t2/merge_ck_and_inference/checkpoint_saver_megatron.py	307	l.mlp.dense_4h_to_h.weight.data.copy_(mlp_l1_weight[tp_rank])
CRITICAL	…t2/merge_ck_and_inference/checkpoint_saver_megatron.py	309	l.self_attention.query_key_value.bias.data.copy_(qkv_bias[tp_rank])
CRITICAL	…t2/merge_ck_and_inference/checkpoint_saver_megatron.py	310	l.self_attention.dense.bias.data.copy_(dense_bias)
CRITICAL	…t2/merge_ck_and_inference/checkpoint_saver_megatron.py	311	l.mlp.dense_h_to_4h.bias.data.copy_(mlp_l0_bias[tp_rank])
CRITICAL	…t2/merge_ck_and_inference/checkpoint_saver_megatron.py	312	l.mlp.dense_4h_to_h.bias.data.copy_(mlp_l1_bias)
CRITICAL	…t2/merge_ck_and_inference/checkpoint_saver_megatron.py	323	models[tp_rank].language_model.encoder.final_layernorm.weight.data.copy_(final_layernorm_weight)
CRITICAL	…t2/merge_ck_and_inference/checkpoint_saver_megatron.py	324	models[tp_rank].language_model.encoder.final_layernorm.bias.data.copy_(final_layernorm_bias)
CRITICAL	…t2/merge_ck_and_inference/checkpoint_saver_megatron.py	352	models[tp_rank].language_model.pooler.dense.weight.data.copy_(pooler_weight)
CRITICAL	…t2/merge_ck_and_inference/checkpoint_saver_megatron.py	353	models[tp_rank].language_model.pooler.dense.bias.data.copy_(pooler_bias)

Cross-File Repetition47 hits · 235 pts

Severity	File	Snippet
HIGH	…-inference/ascend/mindformers/chatglm3/chatglm-stat.py	# 2. 自定义修改配置后实例化 config = autoconfig.from_pretrained('/root/workspace/model/chatglm3-6b_ms/run_glm3_6b.yaml') config.use
HIGH	…m-inference/ascend/mindformers/chatglm3/chatglm-gen.py	# 2. 自定义修改配置后实例化 config = autoconfig.from_pretrained('/root/workspace/model/chatglm3-6b_ms/run_glm3_6b.yaml') config.use
HIGH	…rence/ascend/mindformers/chatglm3/chatglm-inference.py	# 2. 自定义修改配置后实例化 config = autoconfig.from_pretrained('/root/workspace/model/chatglm3-6b_ms/run_glm3_6b.yaml') config.use
HIGH	ai-framework/deepspeed/hello_bert/train_bert_ds.py	collect a list of masked token indices, and labels, and batch them, padding to max length in the batch.
HIGH	ai-framework/deepspeed/hello_bert/train_bert_ds.py	collect a list of masked token indices, and labels, and batch them, padding to max length in the batch.
HIGH	ai-framework/deepspeed/hello_bert/train_bert.py	collect a list of masked token indices, and labels, and batch them, padding to max length in the batch.
HIGH	ai-framework/deepspeed/hello_bert/train_bert_ds.py	given a text string, randomly mask wordpieces for bert mlm training. args: text (str): the input text tokenizer (tokeniz
HIGH	ai-framework/deepspeed/hello_bert/train_bert_ds.py	given a text string, randomly mask wordpieces for bert mlm training. args: text (str): the input text tokenizer (tokeniz
HIGH	ai-framework/deepspeed/hello_bert/train_bert.py	given a text string, randomly mask wordpieces for bert mlm training. args: text (str): the input text tokenizer (tokeniz
HIGH	ai-framework/deepspeed/hello_bert/train_bert_ds.py	a [map style dataset](https://pytorch.org/docs/stable/data.html) for iterating over the wikitext dataset. note that this
HIGH	ai-framework/deepspeed/hello_bert/train_bert_ds.py	a [map style dataset](https://pytorch.org/docs/stable/data.html) for iterating over the wikitext dataset. note that this
HIGH	ai-framework/deepspeed/hello_bert/train_bert.py	a [map style dataset](https://pytorch.org/docs/stable/data.html) for iterating over the wikitext dataset. note that this
HIGH	ai-framework/deepspeed/hello_bert/train_bert_ds.py	create the dataloader. args: mask_prob (float): fraction of tokens to mask random_replace_prob (float): fraction of mask
HIGH	ai-framework/deepspeed/hello_bert/train_bert_ds.py	create the dataloader. args: mask_prob (float): fraction of tokens to mask random_replace_prob (float): fraction of mask
HIGH	ai-framework/deepspeed/hello_bert/train_bert.py	create the dataloader. args: mask_prob (float): fraction of tokens to mask random_replace_prob (float): fraction of mask
HIGH	ai-framework/deepspeed/hello_bert/train_bert_ds.py	the current `transformers` library does not provide support for masked_token_indices. this function provides the support
HIGH	ai-framework/deepspeed/hello_bert/train_bert_ds.py	the current `transformers` library does not provide support for masked_token_indices. this function provides the support
HIGH	ai-framework/deepspeed/hello_bert/train_bert.py	the current `transformers` library does not provide support for masked_token_indices. this function provides the support
HIGH	ai-framework/deepspeed/hello_bert/train_bert_ds.py	the forward pass for the mlm task args: src_tokens (torch.tensor): the masked token indices. shape: (batch, seq_len) att
HIGH	ai-framework/deepspeed/hello_bert/train_bert_ds.py	the forward pass for the mlm task args: src_tokens (torch.tensor): the masked token indices. shape: (batch, seq_len) att
HIGH	ai-framework/deepspeed/hello_bert/train_bert.py	the forward pass for the mlm task args: src_tokens (torch.tensor): the masked token indices. shape: (batch, seq_len) att
HIGH	ai-framework/deepspeed/hello_bert/train_bert_ds.py	create a unique identifier by choosing `length` random characters from list of ascii characters and numbers
HIGH	ai-framework/deepspeed/hello_bert/train_bert_ds.py	create a unique identifier by choosing `length` random characters from list of ascii characters and numbers
HIGH	ai-framework/deepspeed/hello_bert/train_bert.py	create a unique identifier by choosing `length` random characters from list of ascii characters and numbers
HIGH	ai-framework/deepspeed/hello_bert/train_bert_ds.py	create an experiment directory and save all arguments in it. additionally, also store the githash and gitdiff. finally c
HIGH	ai-framework/deepspeed/hello_bert/train_bert_ds.py	create an experiment directory and save all arguments in it. additionally, also store the githash and gitdiff. finally c
HIGH	ai-framework/deepspeed/hello_bert/train_bert.py	create an experiment directory and save all arguments in it. additionally, also store the githash and gitdiff. finally c
HIGH	ai-framework/deepspeed/hello_bert/train_bert_ds.py	loads the optimizer state dict and model state dict from the load_checkpoint_dir into the passed model and optimizer. se
HIGH	ai-framework/deepspeed/hello_bert/train_bert_ds.py	loads the optimizer state dict and model state dict from the load_checkpoint_dir into the passed model and optimizer. se
HIGH	ai-framework/deepspeed/hello_bert/train_bert.py	loads the optimizer state dict and model state dict from the load_checkpoint_dir into the passed model and optimizer. se
HIGH	ai-framework/deepspeed/hello_bert/train_bert_ds.py	trains a [bert style](https://arxiv.org/pdf/1810.04805.pdf) (transformer encoder only) model for mlm task args: checkpoi
HIGH	ai-framework/deepspeed/hello_bert/train_bert_ds.py	trains a [bert style](https://arxiv.org/pdf/1810.04805.pdf) (transformer encoder only) model for mlm task args: checkpoi
HIGH	ai-framework/deepspeed/hello_bert/train_bert.py	trains a [bert style](https://arxiv.org/pdf/1810.04805.pdf) (transformer encoder only) model for mlm task args: checkpoi
HIGH	llm-train/alpaca/train_ddp.py	resize tokenizer and embedding. note: this is the unoptimized version that may make your embedding size not be divisible
HIGH	llm-train/alpaca/train.py	resize tokenizer and embedding. note: this is the unoptimized version that may make your embedding size not be divisible
HIGH	llm-train/qlora/qlora.py	resize tokenizer and embedding. note: this is the unoptimized version that may make your embedding size not be divisible
HIGH	llm-train/chinese-llama-alpaca/run_clm_sft_with_peft.py	resize tokenizer and embedding. note: this is the unoptimized version that may make your embedding size not be divisible
HIGH	llm-localization/ascend/standford-alpaca/train.py	resize tokenizer and embedding. note: this is the unoptimized version that may make your embedding size not be divisible
HIGH	llm-train/alpaca/train_ddp.py	make dataset and collator for supervised fine-tuning.
HIGH	llm-train/alpaca/train.py	make dataset and collator for supervised fine-tuning.
HIGH	llm-localization/ascend/standford-alpaca/train.py	make dataset and collator for supervised fine-tuning.
HIGH	llm-train/megatron/gpt2/data/cMinhash.cpp	only used if instantiated manually by the user, or if cython doesn't * know how to convert the type
HIGH	llm-train/megatron/gpt2/data/cMinhash.cpp	only used if instantiated manually by the user, or if cython doesn't * know how to convert the type
HIGH	llm-train/megatron/gpt2/data/cMinhash.cpp	only used if instantiated manually by the user, or if cython doesn't * know how to convert the type
HIGH	llm-train/megatron/gpt2/data/cMinhash.cpp	only used if instantiated manually by the user, or if cython doesn't * know how to convert the type
HIGH	llm-train/megatron/gpt2/data/cMinhash.cpp	only used if instantiated manually by the user, or if cython doesn't * know how to convert the type
HIGH	llm-train/megatron/gpt2/data/cMinhash.cpp	only used if instantiated manually by the user, or if cython doesn't * know how to convert the type

Unused Imports142 hits · 142 pts

Severity	File	Line
LOW	llm-data-engineering/sft-dataset/jinja-llm-bloom.py	2
LOW	llm-data-engineering/sft-dataset/jinja-llm-baichuan2.py	2
LOW	llm-data-engineering/sft-dataset/jinja-llm-baichuan.py	2
LOW	llm-data-engineering/sft-dataset/jinja-llm.py	2
LOW	llm-data-engineering/sft-dataset/jinja-llm-chatglm3.py	2
LOW	llm-eval/llm-performance/stat_gpu_memory.py	1
LOW	…performance/hardware-performance/pynvml-stat-memory.py	1
LOW	…al/llm-performance/vllm/vllm-locust-qwen1.5-7b-long.py	3
LOW	…al/llm-performance/vllm/vllm-locust-qwen1.5-7b-long.py	4
LOW	…al/llm-performance/vllm/vllm-locust-qwen1.5-7b-long.py	4
LOW	…al/llm-performance/mindie/lantency/stat_input_token.py	4
LOW	…al/llm-performance/mindie/lantency/stat_input_token.py	4
LOW	…al/llm-performance/mindie/lantency/perfermance-stat.py	5
LOW	…performance/mindie/locust-lantency-throughput/hello.py	3
LOW	…performance/mindie/locust-lantency-throughput/hello.py	4
LOW	…ocust-lantency-throughput/llm-910b4-chatglm3-6b-2tp.py	3
LOW	…ocust-lantency-throughput/llm-910b4-chatglm3-6b-2tp.py	4
LOW	…ocust-lantency-throughput/llm-910b4-chatglm3-6b-2tp.py	4
LOW	…e/locust-lantency-throughput/llm-910b4-qwen-72b-8tp.py	2
LOW	…e/locust-lantency-throughput/llm-910b4-qwen-72b-8tp.py	3
LOW	…e/locust-lantency-throughput/llm-910b4-qwen-72b-8tp.py	3
LOW	…cust-lantency-throughput/llm-910b4-baichuan2-7b-2tp.py	3
LOW	…cust-lantency-throughput/llm-910b4-baichuan2-7b-2tp.py	4
LOW	…cust-lantency-throughput/llm-910b4-baichuan2-7b-2tp.py	4
LOW	…ie/locust-lantency-throughput/llm-910b4-qwen1.5-4tp.py	2
LOW	…ie/locust-lantency-throughput/llm-910b4-qwen1.5-4tp.py	3
LOW	…ie/locust-lantency-throughput/llm-910b4-qwen1.5-4tp.py	3
LOW	…nference/ascend/mindformers/mindsporelite-inference.py	12
LOW	llm-inference/ascend/mindformers/mindsporelite-stat.py	3
LOW	llm-inference/ascend/mindformers/mindsporelite-stat.py	11
LOW	…nference/ascend/mindformers/baichuan2/baichuan-stat.py	2
LOW	…nce/ascend/mindformers/baichuan2/baichuan-inference.py	1
LOW	…nce/ascend/mindformers/baichuan2/baichuan-inference.py	2
LOW	…-inference/ascend/mindformers/chatglm3/chatglm-stat.py	2
LOW	…m-inference/ascend/mindformers/chatglm3/chatglm-gen.py	2
LOW	…rence/ascend/mindformers/chatglm3/chatglm-inference.py	5
LOW	llm-inference/web/flask/llm-qwen-mindspore-lite.py	4
LOW	llm-inference/web/flask/llm-qwen-mindspore-lite.py	7
LOW	llm-inference/web/flask/llm-qwen-mindspore-lite.py	9
LOW	llm-inference/web/fastapi/llm-qwen-mindspore-lite.py	4
LOW	llm-inference/web/fastapi/llm-qwen-mindspore-lite.py	7
LOW	llm-inference/web/fastapi/llm-qwen-mindspore-lite.py	9
LOW	llm-inference/web/fastapi/llm-qwen-mindspore-lite.py	15
LOW	llm-inference/web/fastapi/llm-qwen-mindspore-lite.py	17
LOW	…ter-transformer/bloom/firefly_lambada_1w_stat_token.py	1
LOW	…ter-transformer/bloom/firefly_lambada_1w_stat_token.py	10
LOW	…ter-transformer/bloom/firefly_lambada_1w_stat_token.py	12
LOW	…er-transformer/megatron-gpt2/gpt_summarization_stat.py	7
LOW	…/faster-transformer/megatron-gpt2/gpt_summarization.py	7
LOW	llm-inference/triton/resnet50/client.py	5
LOW	…se/distribution-parallelism/moe-parallel/paddle_moe.py	6
LOW	…se/distribution-parallelism/moe-parallel/paddle_moe.py	8
LOW	ai-framework/mxnet/oneflow_cnn_mnist.py	6
LOW	ai-framework/mxnet/mxnet_cnn_mnist.py	1
LOW	ai-framework/mxnet/mxnet_cnn_mnist.py	12
LOW	ai-framework/mxnet/mxnet_cnn_mnist.py	12
LOW	ai-framework/mxnet/mxnet_cnn_mnist.py	26
LOW	ai-framework/mxnet/mxnet_mlp_mnist.py	2
LOW	ai-framework/mxnet/mnist.py	4
LOW	llm-train/alpaca-lora/generate.py	7
82 more matches not shown…

Decorative Section Separators32 hits · 110 pts

Severity	File	Line	Snippet
MEDIUM	llm-inference/DeepSpeed-Inference.md	9	# ---------------------------------------
MEDIUM	llm-inference/DeepSpeed-Inference.md	11	# ---------------------------------------
MEDIUM	…pytorch/distribution/pipeline-parallel/ddp_pipeline.py	24	# ----------------
MEDIUM	…pytorch/distribution/pipeline-parallel/ddp_pipeline.py	124	# -------------------------------------
MEDIUM	…pytorch/distribution/pipeline-parallel/ddp_pipeline.py	137	# -------------------
MEDIUM	…pytorch/distribution/pipeline-parallel/ddp_pipeline.py	246	# -----------------------------------
MEDIUM	…pytorch/distribution/pipeline-parallel/ddp_pipeline.py	341	# -------------
MEDIUM	…pytorch/distribution/pipeline-parallel/ddp_pipeline.py	445	# -------------------------------------
MEDIUM	llm-train/alpa/train/pipeshard_parallelism.py	19	# -------------------------------------------
MEDIUM	llm-train/alpa/train/pipeshard_parallelism.py	37	# ------------------------
MEDIUM	llm-train/alpa/train/pipeshard_parallelism.py	56	# -------------------------------
MEDIUM	llm-train/alpa/train/pipeshard_parallelism.py	119	# -------------------------------------------
MEDIUM	llm-train/alpa/train/pipeshard_parallelism.py	196	# ----------------------------------------------
MEDIUM	llm-train/alpa/train/pipeshard_parallelism.py	250	# ---------------------
MEDIUM	…t2/merge_ck_and_inference/checkpoint_saver_megatron.py	204	#-----------
MEDIUM	…t2/merge_ck_and_inference/checkpoint_saver_megatron.py	254	#-------------------
MEDIUM	llm-algo/chatglm/模型架构.md	220	# ===================================
MEDIUM	llm-algo/chatglm/模型架构.md	222	# ===================================
MEDIUM	llm-algo/chatglm/模型架构.md	264	# =========================
MEDIUM	llm-algo/chatglm/模型架构.md	266	# =========================
MEDIUM	llm-algo/chatglm2/模型架构.md	138	# ===========================
MEDIUM	llm-algo/chatglm2/模型架构.md	140	# ===========================
MEDIUM	llm-algo/chatglm2/模型架构.md	160	# =========================
MEDIUM	llm-algo/chatglm2/模型架构.md	162	# =========================
MEDIUM	llm-algo/chatglm2/模型架构.md	253	# =================================================
MEDIUM	llm-algo/chatglm2/模型架构.md	255	# =================================================
MEDIUM	llm-algo/chatglm2/模型架构.md	256	# =====================
MEDIUM	llm-algo/chatglm2/模型架构.md	258	# =====================
MEDIUM	llm-algo/chatglm2/模型架构.md	322	# ==================================
MEDIUM	llm-algo/chatglm2/模型架构.md	324	# ==================================
MEDIUM	llm-algo/chatglm2/模型架构.md	328	# =================
MEDIUM	llm-algo/chatglm2/模型架构.md	330	# =================

Over-Commented Block71 hits · 62 pts

Severity	File	Line	Snippet
LOW	llm-tools/profiler-recipe.py	21	# Name Self CPU CPU total CPU time avg # of Calls
LOW	llm-tools/profiler-recipe.py	41	# --------------------------------- ------------ -------------------------------------------
LOW	llm-tools/profiler-recipe.py	81
LOW	llm-tools/profiler-recipe.py	121	# --------------------------------- ------------ ------------ ------------
LOW	llm-tools/profiler-recipe.py	141	# aten::empty 94.79 Mb 94.79 Mb 121
LOW	llm-tools/profiler-recipe.py	181	with_stack=True,
LOW	llm-pipeline/REAEMD.md	81	# --llama \
LOW	llm-pipeline/REAEMD.md	101	# --gradient_checkpointing \
LOW	llm-pipeline/REAEMD.md	121	# --num_train_epochs 2 \
LOW	llm-train/qlora/accuracy.py	1	# Copyright 2020 The HuggingFace Datasets Authors and the current dataset script contributor.
LOW	llm-train/chinese-llama-alpaca/run_clm_pt_with_peft.py	1	#!/usr/bin/env python
LOW	llm-train/chinese-llama-alpaca/run_clm_sft_with_peft.py	1	#!/usr/bin/env python
LOW	…ibution/data-parallel/minGPT-ddp/sbatch_run_sig_opt.sh	1
LOW	…rch/distribution/data-parallel/minGPT-ddp/multinode.sh	1	#!/bin/bash
LOW	…istribution/data-parallel/minGPT-ddp/sbatch_run_sig.sh	1	#!/bin/bash
LOW	…ch/distribution/data-parallel/minGPT-ddp/sbatch_run.sh	1
LOW	…pytorch/distribution/pipeline-parallel/ddp_pipeline.py	61
LOW	…pytorch/distribution/pipeline-parallel/ddp_pipeline.py	121
LOW	…pytorch/distribution/pipeline-parallel/ddp_pipeline.py	141	######################################################################
LOW	…pytorch/distribution/pipeline-parallel/ddp_pipeline.py	221	######################################################################
LOW	…pytorch/distribution/pipeline-parallel/ddp_pipeline.py	241	# Need batch dimension first for pipeline parallelism.
LOW	…pytorch/distribution/pipeline-parallel/ddp_pipeline.py	341	# -------------
LOW	…pytorch/distribution/pipeline-parallel/ddp_pipeline.py	461	mp.spawn(run_worker, args=(world_size, ), nprocs=world_size, join=True)
LOW	…pytorch/distribution/pipeline-parallel/ddp_pipeline.py	481	# [RANK 1]: -----------------------------------------------------------------------------------------
LOW	…pytorch/distribution/pipeline-parallel/ddp_pipeline.py	501	# [RANK 0]: \| epoch 3 \| 20/ 50 batches \| lr 4.51 \| ms/batch 698.27 \| loss 12.01 \| ppl 164364.60
LOW	llm-train/alpa/train/pipeshard_parallelism.py	41	ray.init()
LOW	llm-train/alpa/train/pipeshard_parallelism.py	181
LOW	llm-train/alpa/train/pipeshard_parallelism.py	201	#
LOW	llm-train/alpa/train/pipeshard_parallelism.py	241	auto_pipeline_actual_state = auto_pipeline_train_step(state, batch)
LOW	llm-train/alpa/train/pipeshard_parallelism.py	261	#
LOW	…egatron/gpt2/merge_ck_and_inference/checkpoint_util.py	1	import argparse
LOW	…egatron/gpt2/merge_ck_and_inference/checkpoint_util.py	21
LOW	…egatron/gpt2/merge_ck_and_inference/checkpoint_util.py	41	# consumed_valid_samples
LOW	…egatron/gpt2/merge_ck_and_inference/checkpoint_util.py	61	# "mlp l1 weight"
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	21	END: Cython Metadata */
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	41	#endif
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	61	#else
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	81	#define __Pyx_PyCode_New(a, k, l, s, f, code, c, n, v, fv, cell, fn, name, fline, lnos)\
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	101	#endif
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	121	#endif
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	141	#define PyObject_Free(p) PyMem_Free(p)
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	161	#if PY_MAJOR_VERSION >= 3
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	181	#define PyInt_FromSsize_t PyLong_FromSsize_t
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	201	#else
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	221	#define __Pyx_PyType_AsAsync(obj) NULL
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	281
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	301	#include "stdlib.h"
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	321	# else
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	341	#define __PYX_DEFAULT_STRING_ENCODING_IS_DEFAULT 0
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	361	#define __Pyx_sst_abs(value) abs(value)
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	381	#define __Pyx_PyStr_FromString __Pyx_PyBytes_FromString
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	401	#else
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	521
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	601	#endif
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	621	#define __pyx_atomic_incr_aligned(value, lock) _InterlockedIncrement(value)
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	1041	#define __Pyx_RefNannyDeclarations void *__pyx_refnanny = NULL;
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	1061	#define __Pyx_XINCREF(r) do { if((r) != NULL) {__Pyx_INCREF(r); }} while(0)
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	1141	#endif
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	1161	int memview_is_new_reference);
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	1181
11 more matches not shown…

Self-Referential Comments17 hits · 56 pts

Severity	File	Line	Snippet
MEDIUM	llm-inference/flexflow-serve/benchmark-batch1.py	43	# Create the sampling configs
MEDIUM	ai-framework/deepspeed/hello_bert/train_bert_ds.py	117	# Create the labels first
MEDIUM	ai-framework/deepspeed/hello_bert/train_bert_ds.py	941	# Create the labels first
MEDIUM	ai-framework/deepspeed/hello_bert/train_bert.py	117	# Create the labels first
MEDIUM	llm-train/galore/torchrun_main.py	433	# The below code is only executed during the update step
MEDIUM	…tribution/tensor-parallel/sequence_parallel_example.py	49	# Create a optimizer for the parallelized module.
MEDIUM	…ch/distribution/tensor-parallel/2d_parallel_example.py	75	# Create a optimizer for the parallelized module.
MEDIUM	…istribution/tensor-parallel/tensor_parallel_example.py	57	# Create a optimizer for the parallelized module.
MEDIUM	…pytorch/distribution/pipeline-parallel/ddp_pipeline.py	23	# Define the model
MEDIUM	llm-train/alpa/train/pipeshard_parallelism.py	101	# Define the training step
MEDIUM	llm-train/alpa/train/pipeshard_parallelism.py	127	# Define a MLP model with manual stage boundaries.
MEDIUM	llm-train/alpa/train/pipeshard_parallelism.py	154	# Define the training step.
MEDIUM	llm-train/alpa/train/pipeshard_parallelism.py	213	# Define the parallel method.
MEDIUM	llm-train/alpa/train/pipeshard_parallelism.py	222	# Define the training step. The function body is the same as the above one.
MEDIUM	llm-compression/quantization/llm-qat/cfd70ff/utils.py	16	# Define a utility method for setting the logging parameters of a logger
MEDIUM	llm-compression/quantization/llm-qat/cfd70ff/utils.py	24	# Define a formatter for the log messages
MEDIUM	llm-compression/quantization/llm-qat/cfd70ff/utils.py	29	# Create a console handler for outputting log messages to the console

Hyper-Verbose Identifiers43 hits · 44 pts

Severity	File	Line	Snippet
LOW	llm-data-engineering/sft-dataset/数据集格式.md	93	def preprocess_function_train(examples):
LOW	llm-data-engineering/sft-dataset/数据集格式.md	156	def build_inputs_with_special_tokens(
LOW	llm-data-engineering/sft-dataset/数据集格式.md	434	def build_inputs_with_special_tokens(self, token_ids_0, token_ids_1=None):
LOW	…nference/ascend/mindformers/mindsporelite-inference.py	26	def pipeline_from_model_paths(args_, tokenizer):
LOW	…nference/ascend/mindformers/mindsporelite-inference.py	76	def pipeline_from_infer_config(args_, tokenizer):
LOW	llm-inference/ascend/mindformers/mindsporelite-stat.py	33	def pipeline_from_model_paths(args_, tokenizer):
LOW	llm-inference/ascend/mindformers/mindsporelite-stat.py	83	def pipeline_from_infer_config(args_, tokenizer):
LOW	llm-train/alpaca-lora/finetune_metrics_epoch.py	152	def generate_and_tokenize_prompt(data_point):
LOW	llm-train/alpaca-lora/finetune.py	146	def generate_and_tokenize_prompt(data_point):
LOW	llm-train/chatglm/main.py	176	def preprocess_function_train(examples):
LOW	llm-train/alpaca/train_ddp.py	54	def safe_save_model_for_hf_trainer(trainer: transformers.Trainer, output_dir: str):
LOW	llm-train/alpaca/train_ddp.py	63	def smart_tokenizer_and_embedding_resize(
LOW	llm-train/alpaca/train_ddp.py	173	def make_supervised_data_module(tokenizer: transformers.PreTrainedTokenizer, data_args) -> Dict:
LOW	llm-train/alpaca/train.py	53	def safe_save_model_for_hf_trainer(trainer: transformers.Trainer, output_dir: str):
LOW	llm-train/alpaca/train.py	62	def smart_tokenizer_and_embedding_resize(
LOW	llm-train/alpaca/train.py	172	def make_supervised_data_module(tokenizer: transformers.PreTrainedTokenizer, data_args) -> Dict:
LOW	llm-train/qlora/qlora.py	346	def print_trainable_parameters(args, model):
LOW	llm-train/qlora/qlora.py	363	def smart_tokenizer_and_embedding_resize(
LOW	llm-train/qlora/qlora.py	438	def extract_unnatural_instructions_data(examples, extract_reformulations=False):
LOW	llm-train/chinese-llama-alpaca/run_clm_pt_with_peft.py	75	def preprocess_logits_for_metrics(logits, labels):
LOW	llm-train/chinese-llama-alpaca/run_clm_pt_with_peft.py	83	def fault_tolerance_data_collator(features: List) -> Dict[str, Any]:
LOW	llm-train/chinese-llama-alpaca/run_clm_sft_with_peft.py	433	def smart_tokenizer_and_embedding_resize(
LOW	llm-train/alpa/train/pipeshard_parallelism.py	161	def manual_pipeline_train_step(state, batch):
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	8142	* cdef setitem_slice_assign_scalar(self, memoryview dst, value):
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	8186	* cdef setitem_slice_assign_scalar(self, memoryview dst, value): # <<<<<<<<<<<<<<
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	8213	* cdef setitem_slice_assign_scalar(self, memoryview dst, value):
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	8451	* cdef setitem_slice_assign_scalar(self, memoryview dst, value): # <<<<<<<<<<<<<<
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	14045	* cdef memoryview_copy_from_slice(memoryview memview, __Pyx_memviewslice *memviewslice): # <<<<<<<<<<<<<<
LOW	llm-train/megatron/gpt2/data/cMinhash.cpp	14149	* cdef memoryview_copy_from_slice(memoryview memview, __Pyx_memviewslice *memviewslice): # <<<<<<<<<<<<<<
LOW	llm-algo/chatglm/模型架构.md	176	def apply_rotary_pos_emb_index(q, k, cos, sin, position_id):
LOW	llm-algo/chatglm/模型架构.md	362	def split_tensor_along_last_dim(self, tensor, num_partitions,
LOW	llm-algo/chatglm2/模型架构.md	749	def _update_model_kwargs_for_generation(
LOW	llm-algo/chatglm2/模型架构.md	780	def prepare_inputs_for_generation(
LOW	llm-compression/quantization/llm-qat/cfd70ff/utils.py	39	def safe_save_model_for_hf_trainer(trainer: transformers.Trainer, output_dir: str):
LOW	llm-localization/ascend/standford-alpaca/train.py	51	def smart_tokenizer_and_embedding_resize(
LOW	llm-localization/ascend/standford-alpaca/train.py	161	def make_supervised_data_module(tokenizer: transformers.PreTrainedTokenizer, data_args) -> Dict:
LOW	llm-localization/ascend/mindie/script/model-test.py	1237	def __compare_simplified_dataset_results(self):
LOW	llm-localization/ascend/mindie/script/model-test.py	1370	def __compare_full_dataset_results(self):
LOW	llm-localization/ascend/mindie/script/model-test.py	1456	def __patch_hf_transformers_utils(self):
LOW	llm-localization/ascend/mindie/script/model-test.py	799	def process_before_extraction(gen, choice_dict):
LOW	llm-localization/ascend/mindie/script/model-test.py	977	def __run_full_dataset_truthfulqa(self):
LOW	llm-localization/ascend/mindie/script/model-test.py	986	def format_prompt_with_answer_strings(question, ans):
LOW	llm-localization/ascend/mindie/script/model-test.py	1153	def __run_full_dataset_humaneval(self):

Cross-Language Confusion5 hits · 38 pts

Severity	File	Line	Snippet
HIGH	llm-train/megatron/gpt2/data/download.py	251	url varchar(2048) not null,
HIGH	llm-train/megatron/gpt2/data/download.py	252	domain varchar(255) not null,
HIGH	llm-train/megatron/gpt2/data/download.py	253	word_count int null,
HIGH	llm-train/megatron/gpt2/data/download.py	254	elapsed int null,
HIGH	llm-train/megatron/gpt2/data/download.py	255	scraper varchar(255) not null,

Deep Nesting36 hits · 36 pts

Severity	File	Line
LOW	…er-transformer/megatron-gpt2/gpt_summarization_stat.py	24
LOW	…er-transformer/megatron-gpt2/gpt_summarization_stat.py	354
LOW	…/faster-transformer/megatron-gpt2/gpt_summarization.py	21
LOW	…/faster-transformer/megatron-gpt2/gpt_summarization.py	327
LOW	ai-framework/mxnet/mxnet_cnn_mnist.py	122
LOW	llm-train/alpaca-lora/export_state_dict_checkpoint.py	80
LOW	llm-train/chatglm/main.py	38
LOW	llm-train/chatglm/main.py	148
LOW	llm-train/chatglm/main.py	176
LOW	llm-train/qlora/qlora.py	262
LOW	llm-train/qlora/qlora.py	438
LOW	llm-train/qlora/qlora.py	475
LOW	llm-train/qlora/qlora.py	490
LOW	llm-train/qlora/qlora.py	514
LOW	llm-train/qlora/qlora.py	545
LOW	llm-train/galore/torchrun_main.py	134
LOW	…/peft/clm/peft_lora_clm_accelerate_ds_zero3_offload.py	109
LOW	…/chinese-llama-alpaca/merge_llama_with_chinese_lora.py	67
LOW	…/chinese-llama-alpaca/merge_llama_with_chinese_lora.py	111
LOW	llm-train/chinese-llama-alpaca/run_clm_pt_with_peft.py	83
LOW	…2/merge_ck_and_inference/checkpoint_loader_megatron.py	19
LOW	…t2/merge_ck_and_inference/checkpoint_saver_megatron.py	22
LOW	llm-train/megatron/gpt2/data/download.py	193
LOW	llm-localization/ascend/mindie/script/model-test.py	297
LOW	llm-localization/ascend/mindie/script/model-test.py	471
LOW	llm-localization/ascend/mindie/script/model-test.py	535
LOW	llm-localization/ascend/mindie/script/model-test.py	587
LOW	llm-localization/ascend/mindie/script/model-test.py	669
LOW	llm-localization/ascend/mindie/script/model-test.py	789
LOW	llm-localization/ascend/mindie/script/model-test.py	889
LOW	llm-localization/ascend/mindie/script/model-test.py	1075
LOW	llm-localization/ascend/mindie/script/model-test.py	1153
LOW	llm-localization/ascend/mindie/script/model-test.py	1224
LOW	llm-localization/ascend/mindie/script/model-test.py	1264
LOW	llm-localization/ascend/mindie/script/model-test.py	1499
LOW	llm-localization/ascend/mindie/script/model-test.py	334

Excessive Try-Catch Wrapping5 hits · 7 pts

Severity	File	Line	Snippet
MEDIUM	…er-transformer/megatron-gpt2/gpt_summarization_stat.py	342	print('Error with datapoint : ', data_point_idx)
MEDIUM	…/faster-transformer/megatron-gpt2/gpt_summarization.py	325	print('Error with datapoint : ', data_point_idx)
LOW	llm-train/chinese-llama-alpaca/run_clm_pt_with_peft.py	463	except Exception:
MEDIUM	…ron/gpt2/merge_ck_and_inference/text_generation_cli.py	20	print(f"Error {response.status_code}: {response.json()['message']}")
LOW	llm-localization/ascend/mindie/script/model-test.py	258	except Exception as e:

Redundant / Tautological Comments4 hits · 6 pts

Severity	File	Line	Snippet
LOW	llm-train/alpaca-lora/finetune_metrics_epoch.py	104	# Check if parameter passed or if set within environ
LOW	llm-train/alpaca-lora/finetune.py	98	# Check if parameter passed or if set within environ
LOW	…-compression/quantization/llm-qat/f4d873a/datautils.py	91	# Loop through the list of dictionaries
LOW	…-compression/quantization/llm-qat/f4d873a/datautils.py	98	# Append the value to the list associated with the key in dict_of_lists

Slop Phrases2 hits · 6 pts

Severity	File	Line	Snippet
MEDIUM	llm-train/alpa/train/pipeshard_parallelism.py	44	# Alternatively, you can use the following command to connect to an existing
MEDIUM	llm-train/alpa/train/pipeshard_parallelism.py	191	# device assignment of each stage, you can use the more advanced

Docstring Block Structure1 hit · 5 pts

Severity	File	Line	Snippet
HIGH	llm-train/qlora/accuracy.py	33	Args: predictions (`list` of `int`): Predicted labels. references (`list` of `int`): Ground truth labels. n

Fake / Example Data1 hit · 1 pts

Severity	File	Line	Snippet
LOW	llm-data-engineering/sft-dataset/jinja-demo.py	10	result = template.render(name='John Doe')

Score History

Severity Breakdown

Pattern Findings