Hugging Face's logo

Datasets:
danyarm
/
ckg-benchmark

Tasks:
Question Answering
Text Retrieval
Modalities:
Tabular
Text
Languages:
English
Size:
10K - 100K
Tags:
knowledge-graph
rag
retrieval
benchmark
llm
knowledge-representation
License:
Dataset card Data Studio Files
xet
 Community
1
danyarm commited on
Commit
afc5a2c
·
verified ·
1 Parent(s): e99d5f4

Upload queries/queries_nvidia-nemo.jsonl with huggingface_hub

Browse files
Files changed (1) hide show
  1. queries/queries_nvidia-nemo.jsonl +149 -0
queries/queries_nvidia-nemo.jsonl ADDED
@@ -0,0 +1,149 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"id": "nvidia-nemo_T1_41", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is TensorRT-LLM?", "ground_truth": ["TensorRT-LLM", "BACKEND"], "concept_id": 41, "hop_depth": 0}
2
+ {"id": "nvidia-nemo_T1_8", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is Pipeline Parallelism?", "ground_truth": ["Pipeline Parallelism", "TECHNIQU"], "concept_id": 8, "hop_depth": 0}
3
+ {"id": "nvidia-nemo_T1_2", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is Megatron-Core?", "ground_truth": ["Megatron-Core", "ENGINE"], "concept_id": 2, "hop_depth": 0}
4
+ {"id": "nvidia-nemo_T1_18", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is Constitutional AI Filtering?", "ground_truth": ["Constitutional AI Filtering", "TECHNIQU"], "concept_id": 18, "hop_depth": 0}
5
+ {"id": "nvidia-nemo_T1_16", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is Quality Filtering?", "ground_truth": ["Quality Filtering", "TECHNIQU"], "concept_id": 16, "hop_depth": 0}
6
+ {"id": "nvidia-nemo_T1_15", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is Deduplication?", "ground_truth": ["Deduplication", "TECHNIQU"], "concept_id": 15, "hop_depth": 0}
7
+ {"id": "nvidia-nemo_T1_9", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is Sequence Parallelism?", "ground_truth": ["Sequence Parallelism", "TECHNIQU"], "concept_id": 9, "hop_depth": 0}
8
+ {"id": "nvidia-nemo_T1_7", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is Tensor Parallelism?", "ground_truth": ["Tensor Parallelism", "TECHNIQU"], "concept_id": 7, "hop_depth": 0}
9
+ {"id": "nvidia-nemo_T1_35", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is Colang Language?", "ground_truth": ["Colang Language", "LANGUAGE"], "concept_id": 35, "hop_depth": 0}
10
+ {"id": "nvidia-nemo_T1_6", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is Model Evaluation?", "ground_truth": ["Model Evaluation", "TOOL"], "concept_id": 6, "hop_depth": 0}
11
+ {"id": "nvidia-nemo_T1_38", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is Reranking Microservice?", "ground_truth": ["Reranking Microservice", "SERVICE"], "concept_id": 38, "hop_depth": 0}
12
+ {"id": "nvidia-nemo_T1_28", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is NeMo Aligner?", "ground_truth": ["NeMo Aligner", "MODULE"], "concept_id": 28, "hop_depth": 0}
13
+ {"id": "nvidia-nemo_T1_3", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is PyTorch Lightning?", "ground_truth": ["PyTorch Lightning", "FRAMEWOR"], "concept_id": 3, "hop_depth": 0}
14
+ {"id": "nvidia-nemo_T1_48", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is NeMo Microservices?", "ground_truth": ["NeMo Microservices", "MODULE"], "concept_id": 48, "hop_depth": 0}
15
+ {"id": "nvidia-nemo_T1_50", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is NeMo Models Hub?", "ground_truth": ["NeMo Models Hub", "SERVICE"], "concept_id": 50, "hop_depth": 0}
16
+ {"id": "nvidia-nemo_T1_14", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is NeMo Curator?", "ground_truth": ["NeMo Curator", "MODULE"], "concept_id": 14, "hop_depth": 0}
17
+ {"id": "nvidia-nemo_T1_45", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is NeMo Launcher?", "ground_truth": ["NeMo Launcher", "TOOL"], "concept_id": 45, "hop_depth": 0}
18
+ {"id": "nvidia-nemo_T1_33", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is NeMo Skills?", "ground_truth": ["NeMo Skills", "MODULE"], "concept_id": 33, "hop_depth": 0}
19
+ {"id": "nvidia-nemo_T1_37", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is Embedding Microservice?", "ground_truth": ["Embedding Microservice", "SERVICE"], "concept_id": 37, "hop_depth": 0}
20
+ {"id": "nvidia-nemo_T1_47", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is Kubernetes Training?", "ground_truth": ["Kubernetes Training", "COMPONEN"], "concept_id": 47, "hop_depth": 0}
21
+ {"id": "nvidia-nemo_T1_43", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is NIM Deployment?", "ground_truth": ["NIM Deployment", "SERVICE"], "concept_id": 43, "hop_depth": 0}
22
+ {"id": "nvidia-nemo_T1_23", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is LoRA Fine-Tuning?", "ground_truth": ["LoRA Fine-Tuning", "TECHNIQU"], "concept_id": 23, "hop_depth": 0}
23
+ {"id": "nvidia-nemo_T1_21", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is Synthetic Data Generation?", "ground_truth": ["Synthetic Data Generation", "WORKFLOW"], "concept_id": 21, "hop_depth": 0}
24
+ {"id": "nvidia-nemo_T1_29", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is RLHF Pipeline?", "ground_truth": ["RLHF Pipeline", "WORKFLOW"], "concept_id": 29, "hop_depth": 0}
25
+ {"id": "nvidia-nemo_T1_31", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is GRPO Training?", "ground_truth": ["GRPO Training", "WORKFLOW"], "concept_id": 31, "hop_depth": 0}
26
+ {"id": "nvidia-nemo_T1_42", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is vLLM Backend?", "ground_truth": ["vLLM Backend", "BACKEND"], "concept_id": 42, "hop_depth": 0}
27
+ {"id": "nvidia-nemo_T1_49", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is Nemotron Model Family?", "ground_truth": ["Nemotron Model Family", "MODEL"], "concept_id": 49, "hop_depth": 0}
28
+ {"id": "nvidia-nemo_T1_34", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is NeMo Guardrails?", "ground_truth": ["NeMo Guardrails", "MODULE"], "concept_id": 34, "hop_depth": 0}
29
+ {"id": "nvidia-nemo_T1_19", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is Data Blending?", "ground_truth": ["Data Blending", "WORKFLOW"], "concept_id": 19, "hop_depth": 0}
30
+ {"id": "nvidia-nemo_T1_44", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is HuggingFace Integration?", "ground_truth": ["HuggingFace Integration", "COMPONEN"], "concept_id": 44, "hop_depth": 0}
31
+ {"id": "nvidia-nemo_T1_1", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is NeMo Framework?", "ground_truth": ["NeMo Framework", "FRAMEWOR"], "concept_id": 1, "hop_depth": 0}
32
+ {"id": "nvidia-nemo_T1_36", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is NeMo Retriever?", "ground_truth": ["NeMo Retriever", "MODULE"], "concept_id": 36, "hop_depth": 0}
33
+ {"id": "nvidia-nemo_T1_25", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is Continual Pretraining?", "ground_truth": ["Continual Pretraining", "WORKFLOW"], "concept_id": 25, "hop_depth": 0}
34
+ {"id": "nvidia-nemo_T1_11", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is Expert Parallelism?", "ground_truth": ["Expert Parallelism", "TECHNIQU"], "concept_id": 11, "hop_depth": 0}
35
+ {"id": "nvidia-nemo_T1_39", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is Checkpoint Conversion?", "ground_truth": ["Checkpoint Conversion", "WORKFLOW"], "concept_id": 39, "hop_depth": 0}
36
+ {"id": "nvidia-nemo_T1_40", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is NeMo Export?", "ground_truth": ["NeMo Export", "TOOL"], "concept_id": 40, "hop_depth": 0}
37
+ {"id": "nvidia-nemo_T1_4", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is Tokenizer Support?", "ground_truth": ["Tokenizer Support", "COMPONEN"], "concept_id": 4, "hop_depth": 0}
38
+ {"id": "nvidia-nemo_T1_13", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is Flash Attention?", "ground_truth": ["Flash Attention", "TECHNIQU"], "concept_id": 13, "hop_depth": 0}
39
+ {"id": "nvidia-nemo_T1_22", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is Supervised Fine-Tuning?", "ground_truth": ["Supervised Fine-Tuning", "WORKFLOW"], "concept_id": 22, "hop_depth": 0}
40
+ {"id": "nvidia-nemo_T1_32", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is Reward Model?", "ground_truth": ["Reward Model", "MODEL"], "concept_id": 32, "hop_depth": 0}
41
+ {"id": "nvidia-nemo_T1_17", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is PII Redaction?", "ground_truth": ["PII Redaction", "TECHNIQU"], "concept_id": 17, "hop_depth": 0}
42
+ {"id": "nvidia-nemo_T1_30", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is DPO Training?", "ground_truth": ["DPO Training", "WORKFLOW"], "concept_id": 30, "hop_depth": 0}
43
+ {"id": "nvidia-nemo_T1_10", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is Context Parallelism?", "ground_truth": ["Context Parallelism", "TECHNIQU"], "concept_id": 10, "hop_depth": 0}
44
+ {"id": "nvidia-nemo_T1_27", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is Speech Model Training?", "ground_truth": ["Speech Model Training", "WORKFLOW"], "concept_id": 27, "hop_depth": 0}
45
+ {"id": "nvidia-nemo_T1_46", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is SLURM Integration?", "ground_truth": ["SLURM Integration", "COMPONEN"], "concept_id": 46, "hop_depth": 0}
46
+ {"id": "nvidia-nemo_T1_5", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is NeMo Collections?", "ground_truth": ["NeMo Collections", "MODULE"], "concept_id": 5, "hop_depth": 0}
47
+ {"id": "nvidia-nemo_T1_12", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is Distributed Optimizer?", "ground_truth": ["Distributed Optimizer", "COMPONEN"], "concept_id": 12, "hop_depth": 0}
48
+ {"id": "nvidia-nemo_T1_20", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is Data Preprocessing?", "ground_truth": ["Data Preprocessing", "WORKFLOW"], "concept_id": 20, "hop_depth": 0}
49
+ {"id": "nvidia-nemo_T1_24", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is PEFT Methods?", "ground_truth": ["PEFT Methods", "TECHNIQU"], "concept_id": 24, "hop_depth": 0}
50
+ {"id": "nvidia-nemo_T1_26", "domain": "nvidia-nemo", "type": "T1_entity", "query": "What is Multimodal Training?", "ground_truth": ["Multimodal Training", "WORKFLOW"], "concept_id": 26, "hop_depth": 0}
51
+ {"id": "nvidia-nemo_T2_26", "domain": "nvidia-nemo", "type": "T2_dependency", "query": "What are the prerequisites for Multimodal Training?", "ground_truth": ["Megatron-Core", "NeMo Collections", "NeMo Framework"], "concept_id": 26, "hop_depth": 1}
52
+ {"id": "nvidia-nemo_T2_5", "domain": "nvidia-nemo", "type": "T2_dependency", "query": "What are the prerequisites for NeMo Collections?", "ground_truth": ["NeMo Framework"], "concept_id": 5, "hop_depth": 1}
53
+ {"id": "nvidia-nemo_T2_31", "domain": "nvidia-nemo", "type": "T2_dependency", "query": "What are the prerequisites for GRPO Training?", "ground_truth": ["Supervised Fine-Tuning", "Synthetic Data Generation"], "concept_id": 31, "hop_depth": 1}
54
+ {"id": "nvidia-nemo_T2_22", "domain": "nvidia-nemo", "type": "T2_dependency", "query": "What are the prerequisites for Supervised Fine-Tuning?", "ground_truth": ["Megatron-Core", "NeMo Framework"], "concept_id": 22, "hop_depth": 1}
55
+ {"id": "nvidia-nemo_T2_34", "domain": "nvidia-nemo", "type": "T2_dependency", "query": "What are the prerequisites for NeMo Guardrails?", "ground_truth": ["Colang Language"], "concept_id": 34, "hop_depth": 1}
56
+ {"id": "nvidia-nemo_T2_33", "domain": "nvidia-nemo", "type": "T2_dependency", "query": "What are the prerequisites for NeMo Skills?", "ground_truth": ["NeMo Framework"], "concept_id": 33, "hop_depth": 1}
57
+ {"id": "nvidia-nemo_T2_25", "domain": "nvidia-nemo", "type": "T2_dependency", "query": "What are the prerequisites for Continual Pretraining?", "ground_truth": ["Megatron-Core", "Data Blending"], "concept_id": 25, "hop_depth": 1}
58
+ {"id": "nvidia-nemo_T2_19", "domain": "nvidia-nemo", "type": "T2_dependency", "query": "What are the prerequisites for Data Blending?", "ground_truth": ["NeMo Curator"], "concept_id": 19, "hop_depth": 1}
59
+ {"id": "nvidia-nemo_T2_40", "domain": "nvidia-nemo", "type": "T2_dependency", "query": "What are the prerequisites for NeMo Export?", "ground_truth": ["Checkpoint Conversion", "vLLM Backend", "TensorRT-LLM"], "concept_id": 40, "hop_depth": 1}
60
+ {"id": "nvidia-nemo_T2_2", "domain": "nvidia-nemo", "type": "T2_dependency", "query": "What are the prerequisites for Megatron-Core?", "ground_truth": ["Expert Parallelism", "Tensor Parallelism", "Sequence Parallelism", "Pipeline Parallelism"], "concept_id": 2, "hop_depth": 1}
61
+ {"id": "nvidia-nemo_T2_23", "domain": "nvidia-nemo", "type": "T2_dependency", "query": "What are the prerequisites for LoRA Fine-Tuning?", "ground_truth": ["PEFT Methods"], "concept_id": 23, "hop_depth": 1}
62
+ {"id": "nvidia-nemo_T2_6", "domain": "nvidia-nemo", "type": "T2_dependency", "query": "What are the prerequisites for Model Evaluation?", "ground_truth": ["NeMo Framework"], "concept_id": 6, "hop_depth": 1}
63
+ {"id": "nvidia-nemo_T2_45", "domain": "nvidia-nemo", "type": "T2_dependency", "query": "What are the prerequisites for NeMo Launcher?", "ground_truth": ["SLURM Integration", "Kubernetes Training"], "concept_id": 45, "hop_depth": 1}
64
+ {"id": "nvidia-nemo_T2_10", "domain": "nvidia-nemo", "type": "T2_dependency", "query": "What are the prerequisites for Context Parallelism?", "ground_truth": ["Sequence Parallelism"], "concept_id": 10, "hop_depth": 1}
65
+ {"id": "nvidia-nemo_T2_29", "domain": "nvidia-nemo", "type": "T2_dependency", "query": "What are the prerequisites for RLHF Pipeline?", "ground_truth": ["Supervised Fine-Tuning", "Reward Model"], "concept_id": 29, "hop_depth": 1}
66
+ {"id": "nvidia-nemo_T2_27", "domain": "nvidia-nemo", "type": "T2_dependency", "query": "What are the prerequisites for Speech Model Training?", "ground_truth": ["NeMo Collections"], "concept_id": 27, "hop_depth": 1}
67
+ {"id": "nvidia-nemo_T2_28", "domain": "nvidia-nemo", "type": "T2_dependency", "query": "What are the prerequisites for NeMo Aligner?", "ground_truth": ["RLHF Pipeline", "GRPO Training", "Megatron-Core", "NeMo Framework", "DPO Training"], "concept_id": 28, "hop_depth": 1}
68
+ {"id": "nvidia-nemo_T2_39", "domain": "nvidia-nemo", "type": "T2_dependency", "query": "What are the prerequisites for Checkpoint Conversion?", "ground_truth": ["NeMo Framework"], "concept_id": 39, "hop_depth": 1}
69
+ {"id": "nvidia-nemo_T2_30", "domain": "nvidia-nemo", "type": "T2_dependency", "query": "What are the prerequisites for DPO Training?", "ground_truth": ["Supervised Fine-Tuning"], "concept_id": 30, "hop_depth": 1}
70
+ {"id": "nvidia-nemo_T2_36", "domain": "nvidia-nemo", "type": "T2_dependency", "query": "What are the prerequisites for NeMo Retriever?", "ground_truth": ["Embedding Microservice", "Reranking Microservice"], "concept_id": 36, "hop_depth": 1}
71
+ {"id": "nvidia-nemo_T2_20", "domain": "nvidia-nemo", "type": "T2_dependency", "query": "What are the prerequisites for Data Preprocessing?", "ground_truth": ["NeMo Curator", "Tokenizer Support"], "concept_id": 20, "hop_depth": 1}
72
+ {"id": "nvidia-nemo_T2_21", "domain": "nvidia-nemo", "type": "T2_dependency", "query": "What are the prerequisites for Synthetic Data Generation?", "ground_truth": ["NeMo Curator", "Nemotron Model Family"], "concept_id": 21, "hop_depth": 1}
73
+ {"id": "nvidia-nemo_T2_14", "domain": "nvidia-nemo", "type": "T2_dependency", "query": "What are the prerequisites for NeMo Curator?", "ground_truth": ["Deduplication", "Constitutional AI Filtering", "Quality Filtering", "PII Redaction"], "concept_id": 14, "hop_depth": 1}
74
+ {"id": "nvidia-nemo_T2_32", "domain": "nvidia-nemo", "type": "T2_dependency", "query": "What are the prerequisites for Reward Model?", "ground_truth": ["Supervised Fine-Tuning"], "concept_id": 32, "hop_depth": 1}
75
+ {"id": "nvidia-nemo_T2_1", "domain": "nvidia-nemo", "type": "T2_dependency", "query": "What are the prerequisites for NeMo Framework?", "ground_truth": ["Megatron-Core", "Tokenizer Support", "PyTorch Lightning"], "concept_id": 1, "hop_depth": 1}
76
+ {"id": "nvidia-nemo_T3_9_27", "domain": "nvidia-nemo", "type": "T3_path", "query": "What is the prerequisite chain from Sequence Parallelism to Speech Model Training?", "ground_truth": ["Speech Model Training", "NeMo Collections", "NeMo Framework", "Megatron-Core", "Sequence Parallelism"], "concept_id": 27, "hop_depth": 4, "path_ids": [27, 5, 1, 2, 9]}
77
+ {"id": "nvidia-nemo_T3_9_26", "domain": "nvidia-nemo", "type": "T3_path", "query": "What is the prerequisite chain from Sequence Parallelism to Multimodal Training?", "ground_truth": ["Multimodal Training", "Megatron-Core", "Sequence Parallelism"], "concept_id": 26, "hop_depth": 2, "path_ids": [26, 2, 9]}
78
+ {"id": "nvidia-nemo_T3_8_28", "domain": "nvidia-nemo", "type": "T3_path", "query": "What is the prerequisite chain from Pipeline Parallelism to NeMo Aligner?", "ground_truth": ["NeMo Aligner", "Megatron-Core", "Pipeline Parallelism"], "concept_id": 28, "hop_depth": 2, "path_ids": [28, 2, 8]}
79
+ {"id": "nvidia-nemo_T3_46_45", "domain": "nvidia-nemo", "type": "T3_path", "query": "What is the prerequisite chain from SLURM Integration to NeMo Launcher?", "ground_truth": ["NeMo Launcher", "SLURM Integration"], "concept_id": 45, "hop_depth": 1, "path_ids": [45, 46]}
80
+ {"id": "nvidia-nemo_T3_3_33", "domain": "nvidia-nemo", "type": "T3_path", "query": "What is the prerequisite chain from PyTorch Lightning to NeMo Skills?", "ground_truth": ["NeMo Skills", "NeMo Framework", "PyTorch Lightning"], "concept_id": 33, "hop_depth": 2, "path_ids": [33, 1, 3]}
81
+ {"id": "nvidia-nemo_T3_4_26", "domain": "nvidia-nemo", "type": "T3_path", "query": "What is the prerequisite chain from Tokenizer Support to Multimodal Training?", "ground_truth": ["Multimodal Training", "NeMo Framework", "Tokenizer Support"], "concept_id": 26, "hop_depth": 2, "path_ids": [26, 1, 4]}
82
+ {"id": "nvidia-nemo_T3_7_33", "domain": "nvidia-nemo", "type": "T3_path", "query": "What is the prerequisite chain from Tensor Parallelism to NeMo Skills?", "ground_truth": ["NeMo Skills", "NeMo Framework", "Megatron-Core", "Tensor Parallelism"], "concept_id": 33, "hop_depth": 3, "path_ids": [33, 1, 2, 7]}
83
+ {"id": "nvidia-nemo_T3_4_33", "domain": "nvidia-nemo", "type": "T3_path", "query": "What is the prerequisite chain from Tokenizer Support to NeMo Skills?", "ground_truth": ["NeMo Skills", "NeMo Framework", "Tokenizer Support"], "concept_id": 33, "hop_depth": 2, "path_ids": [33, 1, 4]}
84
+ {"id": "nvidia-nemo_T3_8_26", "domain": "nvidia-nemo", "type": "T3_path", "query": "What is the prerequisite chain from Pipeline Parallelism to Multimodal Training?", "ground_truth": ["Multimodal Training", "Megatron-Core", "Pipeline Parallelism"], "concept_id": 26, "hop_depth": 2, "path_ids": [26, 2, 8]}
85
+ {"id": "nvidia-nemo_T3_9_40", "domain": "nvidia-nemo", "type": "T3_path", "query": "What is the prerequisite chain from Sequence Parallelism to NeMo Export?", "ground_truth": ["NeMo Export", "Checkpoint Conversion", "NeMo Framework", "Megatron-Core", "Sequence Parallelism"], "concept_id": 40, "hop_depth": 4, "path_ids": [40, 39, 1, 2, 9]}
86
+ {"id": "nvidia-nemo_T3_11_27", "domain": "nvidia-nemo", "type": "T3_path", "query": "What is the prerequisite chain from Expert Parallelism to Speech Model Training?", "ground_truth": ["Speech Model Training", "NeMo Collections", "NeMo Framework", "Megatron-Core", "Expert Parallelism"], "concept_id": 27, "hop_depth": 4, "path_ids": [27, 5, 1, 2, 11]}
87
+ {"id": "nvidia-nemo_T3_11_40", "domain": "nvidia-nemo", "type": "T3_path", "query": "What is the prerequisite chain from Expert Parallelism to NeMo Export?", "ground_truth": ["NeMo Export", "Checkpoint Conversion", "NeMo Framework", "Megatron-Core", "Expert Parallelism"], "concept_id": 40, "hop_depth": 4, "path_ids": [40, 39, 1, 2, 11]}
88
+ {"id": "nvidia-nemo_T3_9_25", "domain": "nvidia-nemo", "type": "T3_path", "query": "What is the prerequisite chain from Sequence Parallelism to Continual Pretraining?", "ground_truth": ["Continual Pretraining", "Megatron-Core", "Sequence Parallelism"], "concept_id": 25, "hop_depth": 2, "path_ids": [25, 2, 9]}
89
+ {"id": "nvidia-nemo_T3_16_25", "domain": "nvidia-nemo", "type": "T3_path", "query": "What is the prerequisite chain from Quality Filtering to Continual Pretraining?", "ground_truth": ["Continual Pretraining", "Data Blending", "NeMo Curator", "Quality Filtering"], "concept_id": 25, "hop_depth": 3, "path_ids": [25, 19, 14, 16]}
90
+ {"id": "nvidia-nemo_T3_4_33", "domain": "nvidia-nemo", "type": "T3_path", "query": "What is the prerequisite chain from Tokenizer Support to NeMo Skills?", "ground_truth": ["NeMo Skills", "NeMo Framework", "Tokenizer Support"], "concept_id": 33, "hop_depth": 2, "path_ids": [33, 1, 4]}
91
+ {"id": "nvidia-nemo_T3_7_26", "domain": "nvidia-nemo", "type": "T3_path", "query": "What is the prerequisite chain from Tensor Parallelism to Multimodal Training?", "ground_truth": ["Multimodal Training", "Megatron-Core", "Tensor Parallelism"], "concept_id": 26, "hop_depth": 2, "path_ids": [26, 2, 7]}
92
+ {"id": "nvidia-nemo_T3_18_28", "domain": "nvidia-nemo", "type": "T3_path", "query": "What is the prerequisite chain from Constitutional AI Filtering to NeMo Aligner?", "ground_truth": ["NeMo Aligner", "GRPO Training", "Synthetic Data Generation", "NeMo Curator", "Constitutional AI Filtering"], "concept_id": 28, "hop_depth": 4, "path_ids": [28, 31, 21, 14, 18]}
93
+ {"id": "nvidia-nemo_T3_9_27", "domain": "nvidia-nemo", "type": "T3_path", "query": "What is the prerequisite chain from Sequence Parallelism to Speech Model Training?", "ground_truth": ["Speech Model Training", "NeMo Collections", "NeMo Framework", "Megatron-Core", "Sequence Parallelism"], "concept_id": 27, "hop_depth": 4, "path_ids": [27, 5, 1, 2, 9]}
94
+ {"id": "nvidia-nemo_T3_9_27", "domain": "nvidia-nemo", "type": "T3_path", "query": "What is the prerequisite chain from Sequence Parallelism to Speech Model Training?", "ground_truth": ["Speech Model Training", "NeMo Collections", "NeMo Framework", "Megatron-Core", "Sequence Parallelism"], "concept_id": 27, "hop_depth": 4, "path_ids": [27, 5, 1, 2, 9]}
95
+ {"id": "nvidia-nemo_T3_16_25", "domain": "nvidia-nemo", "type": "T3_path", "query": "What is the prerequisite chain from Quality Filtering to Continual Pretraining?", "ground_truth": ["Continual Pretraining", "Data Blending", "NeMo Curator", "Quality Filtering"], "concept_id": 25, "hop_depth": 3, "path_ids": [25, 19, 14, 16]}
96
+ {"id": "nvidia-nemo_T3_7_40", "domain": "nvidia-nemo", "type": "T3_path", "query": "What is the prerequisite chain from Tensor Parallelism to NeMo Export?", "ground_truth": ["NeMo Export", "Checkpoint Conversion", "NeMo Framework", "Megatron-Core", "Tensor Parallelism"], "concept_id": 40, "hop_depth": 4, "path_ids": [40, 39, 1, 2, 7]}
97
+ {"id": "nvidia-nemo_T3_3_33", "domain": "nvidia-nemo", "type": "T3_path", "query": "What is the prerequisite chain from PyTorch Lightning to NeMo Skills?", "ground_truth": ["NeMo Skills", "NeMo Framework", "PyTorch Lightning"], "concept_id": 33, "hop_depth": 2, "path_ids": [33, 1, 3]}
98
+ {"id": "nvidia-nemo_T3_42_40", "domain": "nvidia-nemo", "type": "T3_path", "query": "What is the prerequisite chain from vLLM Backend to NeMo Export?", "ground_truth": ["NeMo Export", "vLLM Backend"], "concept_id": 40, "hop_depth": 1, "path_ids": [40, 42]}
99
+ {"id": "nvidia-nemo_T3_4_28", "domain": "nvidia-nemo", "type": "T3_path", "query": "What is the prerequisite chain from Tokenizer Support to NeMo Aligner?", "ground_truth": ["NeMo Aligner", "NeMo Framework", "Tokenizer Support"], "concept_id": 28, "hop_depth": 2, "path_ids": [28, 1, 4]}
100
+ {"id": "nvidia-nemo_T3_18_25", "domain": "nvidia-nemo", "type": "T3_path", "query": "What is the prerequisite chain from Constitutional AI Filtering to Continual Pretraining?", "ground_truth": ["Continual Pretraining", "Data Blending", "NeMo Curator", "Constitutional AI Filtering"], "concept_id": 25, "hop_depth": 3, "path_ids": [25, 19, 14, 18]}
101
+ {"id": "nvidia-nemo_T4_FRAMEWOR", "domain": "nvidia-nemo", "type": "T4_aggregate", "query": "List all FRAMEWOR concepts in this knowledge graph", "ground_truth": ["NeMo Framework", "PyTorch Lightning"], "taxonomy_id": "FRAMEWOR", "hop_depth": 0}
102
+ {"id": "nvidia-nemo_T4_ENGINE", "domain": "nvidia-nemo", "type": "T4_aggregate", "query": "List all ENGINE concepts in this knowledge graph", "ground_truth": ["Megatron-Core"], "taxonomy_id": "ENGINE", "hop_depth": 0}
103
+ {"id": "nvidia-nemo_T4_COMPONEN", "domain": "nvidia-nemo", "type": "T4_aggregate", "query": "List all COMPONEN concepts in this knowledge graph", "ground_truth": ["Tokenizer Support", "Distributed Optimizer", "HuggingFace Integration", "SLURM Integration", "Kubernetes Training"], "taxonomy_id": "COMPONEN", "hop_depth": 0}
104
+ {"id": "nvidia-nemo_T4_MODULE", "domain": "nvidia-nemo", "type": "T4_aggregate", "query": "List all MODULE concepts in this knowledge graph", "ground_truth": ["NeMo Collections", "NeMo Curator", "NeMo Aligner", "NeMo Skills", "NeMo Guardrails", "NeMo Retriever", "NeMo Microservices"], "taxonomy_id": "MODULE", "hop_depth": 0}
105
+ {"id": "nvidia-nemo_T4_TOOL", "domain": "nvidia-nemo", "type": "T4_aggregate", "query": "List all TOOL concepts in this knowledge graph", "ground_truth": ["Model Evaluation", "NeMo Export", "NeMo Launcher"], "taxonomy_id": "TOOL", "hop_depth": 0}
106
+ {"id": "nvidia-nemo_T4_TECHNIQU", "domain": "nvidia-nemo", "type": "T4_aggregate", "query": "List all TECHNIQU concepts in this knowledge graph", "ground_truth": ["Tensor Parallelism", "Pipeline Parallelism", "Sequence Parallelism", "Context Parallelism", "Expert Parallelism", "Flash Attention", "Deduplication", "Quality Filtering", "PII Redaction", "Constitutional AI Filtering", "LoRA Fine-Tuning", "PEFT Methods"], "taxonomy_id": "TECHNIQU", "hop_depth": 0}
107
+ {"id": "nvidia-nemo_T4_WORKFLOW", "domain": "nvidia-nemo", "type": "T4_aggregate", "query": "List all WORKFLOW concepts in this knowledge graph", "ground_truth": ["Data Blending", "Data Preprocessing", "Synthetic Data Generation", "Supervised Fine-Tuning", "Continual Pretraining", "Multimodal Training", "Speech Model Training", "RLHF Pipeline", "DPO Training", "GRPO Training", "Checkpoint Conversion"], "taxonomy_id": "WORKFLOW", "hop_depth": 0}
108
+ {"id": "nvidia-nemo_T4_MODEL", "domain": "nvidia-nemo", "type": "T4_aggregate", "query": "List all MODEL concepts in this knowledge graph", "ground_truth": ["Reward Model", "Nemotron Model Family"], "taxonomy_id": "MODEL", "hop_depth": 0}
109
+ {"id": "nvidia-nemo_T4_LANGUAGE", "domain": "nvidia-nemo", "type": "T4_aggregate", "query": "List all LANGUAGE concepts in this knowledge graph", "ground_truth": ["Colang Language"], "taxonomy_id": "LANGUAGE", "hop_depth": 0}
110
+ {"id": "nvidia-nemo_T4_SERVICE", "domain": "nvidia-nemo", "type": "T4_aggregate", "query": "List all SERVICE concepts in this knowledge graph", "ground_truth": ["Embedding Microservice", "Reranking Microservice", "NIM Deployment", "NeMo Models Hub"], "taxonomy_id": "SERVICE", "hop_depth": 0}
111
+ {"id": "nvidia-nemo_T4_BACKEND", "domain": "nvidia-nemo", "type": "T4_aggregate", "query": "List all BACKEND concepts in this knowledge graph", "ground_truth": ["TensorRT-LLM", "vLLM Backend"], "taxonomy_id": "BACKEND", "hop_depth": 0}
112
+ {"id": "nvidia-nemo_T5_39_1", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does Checkpoint Conversion relate to NeMo Framework?", "ground_truth": ["Checkpoint Conversion", "NeMo Framework"], "concept_id_a": 39, "concept_id_b": 1, "hop_depth": 1}
113
+ {"id": "nvidia-nemo_T5_20_14", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does Data Preprocessing relate to NeMo Curator?", "ground_truth": ["Data Preprocessing", "NeMo Curator"], "concept_id_a": 20, "concept_id_b": 14, "hop_depth": 1}
114
+ {"id": "nvidia-nemo_T5_36_38", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does NeMo Retriever relate to Reranking Microservice?", "ground_truth": ["NeMo Retriever", "Reranking Microservice"], "concept_id_a": 36, "concept_id_b": 38, "hop_depth": 1}
115
+ {"id": "nvidia-nemo_T5_2_8", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does Megatron-Core relate to Pipeline Parallelism?", "ground_truth": ["Megatron-Core", "Pipeline Parallelism"], "concept_id_a": 2, "concept_id_b": 8, "hop_depth": 1}
116
+ {"id": "nvidia-nemo_T5_25_2", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does Continual Pretraining relate to Megatron-Core?", "ground_truth": ["Continual Pretraining", "Megatron-Core"], "concept_id_a": 25, "concept_id_b": 2, "hop_depth": 1}
117
+ {"id": "nvidia-nemo_T5_30_22", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does DPO Training relate to Supervised Fine-Tuning?", "ground_truth": ["DPO Training", "Supervised Fine-Tuning"], "concept_id_a": 30, "concept_id_b": 22, "hop_depth": 1}
118
+ {"id": "nvidia-nemo_T5_26_1", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does Multimodal Training relate to NeMo Framework?", "ground_truth": ["Multimodal Training", "NeMo Framework", "Megatron-Core"], "concept_id_a": 26, "concept_id_b": 1, "hop_depth": 1}
119
+ {"id": "nvidia-nemo_T5_33_1", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does NeMo Skills relate to NeMo Framework?", "ground_truth": ["NeMo Skills", "NeMo Framework"], "concept_id_a": 33, "concept_id_b": 1, "hop_depth": 1}
120
+ {"id": "nvidia-nemo_T5_10_9", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does Context Parallelism relate to Sequence Parallelism?", "ground_truth": ["Context Parallelism", "Sequence Parallelism"], "concept_id_a": 10, "concept_id_b": 9, "hop_depth": 1}
121
+ {"id": "nvidia-nemo_T5_45_46", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does NeMo Launcher relate to SLURM Integration?", "ground_truth": ["NeMo Launcher", "SLURM Integration"], "concept_id_a": 45, "concept_id_b": 46, "hop_depth": 1}
122
+ {"id": "nvidia-nemo_T5_14_15", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does NeMo Curator relate to Deduplication?", "ground_truth": ["NeMo Curator", "Deduplication"], "concept_id_a": 14, "concept_id_b": 15, "hop_depth": 1}
123
+ {"id": "nvidia-nemo_T5_14_18", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does NeMo Curator relate to Constitutional AI Filtering?", "ground_truth": ["NeMo Curator", "Constitutional AI Filtering"], "concept_id_a": 14, "concept_id_b": 18, "hop_depth": 1}
124
+ {"id": "nvidia-nemo_T5_40_41", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does NeMo Export relate to TensorRT-LLM?", "ground_truth": ["NeMo Export", "TensorRT-LLM"], "concept_id_a": 40, "concept_id_b": 41, "hop_depth": 1}
125
+ {"id": "nvidia-nemo_T5_1_4", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does NeMo Framework relate to Tokenizer Support?", "ground_truth": ["NeMo Framework", "Tokenizer Support"], "concept_id_a": 1, "concept_id_b": 4, "hop_depth": 1}
126
+ {"id": "nvidia-nemo_T5_31_21", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does GRPO Training relate to Synthetic Data Generation?", "ground_truth": ["GRPO Training", "Synthetic Data Generation"], "concept_id_a": 31, "concept_id_b": 21, "hop_depth": 1}
127
+ {"id": "nvidia-nemo_T5_23_24", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does LoRA Fine-Tuning relate to PEFT Methods?", "ground_truth": ["LoRA Fine-Tuning", "PEFT Methods"], "concept_id_a": 23, "concept_id_b": 24, "hop_depth": 1}
128
+ {"id": "nvidia-nemo_T5_31_22", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does GRPO Training relate to Supervised Fine-Tuning?", "ground_truth": ["GRPO Training", "Supervised Fine-Tuning"], "concept_id_a": 31, "concept_id_b": 22, "hop_depth": 1}
129
+ {"id": "nvidia-nemo_T5_36_37", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does NeMo Retriever relate to Embedding Microservice?", "ground_truth": ["NeMo Retriever", "Embedding Microservice"], "concept_id_a": 36, "concept_id_b": 37, "hop_depth": 1}
130
+ {"id": "nvidia-nemo_T5_6_1", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does Model Evaluation relate to NeMo Framework?", "ground_truth": ["Model Evaluation", "NeMo Framework"], "concept_id_a": 6, "concept_id_b": 1, "hop_depth": 1}
131
+ {"id": "nvidia-nemo_T5_2_9", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does Megatron-Core relate to Sequence Parallelism?", "ground_truth": ["Megatron-Core", "Sequence Parallelism"], "concept_id_a": 2, "concept_id_b": 9, "hop_depth": 1}
132
+ {"id": "nvidia-nemo_T5_26_5", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does Multimodal Training relate to NeMo Collections?", "ground_truth": ["Multimodal Training", "NeMo Collections", "NeMo Framework"], "concept_id_a": 26, "concept_id_b": 5, "hop_depth": 1}
133
+ {"id": "nvidia-nemo_T5_40_39", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does NeMo Export relate to Checkpoint Conversion?", "ground_truth": ["NeMo Export", "Checkpoint Conversion"], "concept_id_a": 40, "concept_id_b": 39, "hop_depth": 1}
134
+ {"id": "nvidia-nemo_T5_2_11", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does Megatron-Core relate to Expert Parallelism?", "ground_truth": ["Megatron-Core", "Expert Parallelism"], "concept_id_a": 2, "concept_id_b": 11, "hop_depth": 1}
135
+ {"id": "nvidia-nemo_T5_14_16", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does NeMo Curator relate to Quality Filtering?", "ground_truth": ["NeMo Curator", "Quality Filtering"], "concept_id_a": 14, "concept_id_b": 16, "hop_depth": 1}
136
+ {"id": "nvidia-nemo_T5_45_47", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does NeMo Launcher relate to Kubernetes Training?", "ground_truth": ["NeMo Launcher", "Kubernetes Training"], "concept_id_a": 45, "concept_id_b": 47, "hop_depth": 1}
137
+ {"id": "nvidia-nemo_T5_28_1", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does NeMo Aligner relate to NeMo Framework?", "ground_truth": ["NeMo Aligner", "NeMo Framework", "Megatron-Core"], "concept_id_a": 28, "concept_id_b": 1, "hop_depth": 1}
138
+ {"id": "nvidia-nemo_T5_5_1", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does NeMo Collections relate to NeMo Framework?", "ground_truth": ["NeMo Collections", "NeMo Framework"], "concept_id_a": 5, "concept_id_b": 1, "hop_depth": 1}
139
+ {"id": "nvidia-nemo_T5_40_42", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does NeMo Export relate to vLLM Backend?", "ground_truth": ["NeMo Export", "vLLM Backend"], "concept_id_a": 40, "concept_id_b": 42, "hop_depth": 1}
140
+ {"id": "nvidia-nemo_T5_19_14", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does Data Blending relate to NeMo Curator?", "ground_truth": ["Data Blending", "NeMo Curator"], "concept_id_a": 19, "concept_id_b": 14, "hop_depth": 1}
141
+ {"id": "nvidia-nemo_T5_29_32", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does RLHF Pipeline relate to Reward Model?", "ground_truth": ["RLHF Pipeline", "Reward Model", "Supervised Fine-Tuning"], "concept_id_a": 29, "concept_id_b": 32, "hop_depth": 1}
142
+ {"id": "nvidia-nemo_T5_34_35", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does NeMo Guardrails relate to Colang Language?", "ground_truth": ["NeMo Guardrails", "Colang Language"], "concept_id_a": 34, "concept_id_b": 35, "hop_depth": 1}
143
+ {"id": "nvidia-nemo_T5_27_5", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does Speech Model Training relate to NeMo Collections?", "ground_truth": ["Speech Model Training", "NeMo Collections"], "concept_id_a": 27, "concept_id_b": 5, "hop_depth": 1}
144
+ {"id": "nvidia-nemo_T5_21_14", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does Synthetic Data Generation relate to NeMo Curator?", "ground_truth": ["Synthetic Data Generation", "NeMo Curator"], "concept_id_a": 21, "concept_id_b": 14, "hop_depth": 1}
145
+ {"id": "nvidia-nemo_T5_1_2", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does NeMo Framework relate to Megatron-Core?", "ground_truth": ["NeMo Framework", "Megatron-Core"], "concept_id_a": 1, "concept_id_b": 2, "hop_depth": 1}
146
+ {"id": "nvidia-nemo_T5_32_22", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does Reward Model relate to Supervised Fine-Tuning?", "ground_truth": ["Reward Model", "Supervised Fine-Tuning"], "concept_id_a": 32, "concept_id_b": 22, "hop_depth": 1}
147
+ {"id": "nvidia-nemo_T5_28_30", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does NeMo Aligner relate to DPO Training?", "ground_truth": ["NeMo Aligner", "DPO Training"], "concept_id_a": 28, "concept_id_b": 30, "hop_depth": 1}
148
+ {"id": "nvidia-nemo_T5_29_22", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does RLHF Pipeline relate to Supervised Fine-Tuning?", "ground_truth": ["RLHF Pipeline", "Supervised Fine-Tuning"], "concept_id_a": 29, "concept_id_b": 22, "hop_depth": 1}
149
+ {"id": "nvidia-nemo_T5_28_31", "domain": "nvidia-nemo", "type": "T5_cross_concept", "query": "How does NeMo Aligner relate to GRPO Training?", "ground_truth": ["NeMo Aligner", "GRPO Training"], "concept_id_a": 28, "concept_id_b": 31, "hop_depth": 1}