SwathiManikya commited on
Commit
70c896a
·
verified ·
1 Parent(s): 64e0451

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +68 -1
README.md CHANGED
@@ -1,3 +1,70 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
 
2
  # HLO Feature Dataset for Deep Learning Resource Estimation
3
 
@@ -100,4 +167,4 @@ Specify your license here (e.g., MIT, Apache-2.0).
100
  ---
101
 
102
  ## Contributions
103
- Open to contributions! Feel free to suggest improvements or share your models trained on this dataset.
 
1
+ ---
2
+ dataset_name: "hlo-feature-dataset"
3
+ pretty_name: "HLO Feature Dataset for Deep Learning Resource Estimation"
4
+ dataset_type: "graph-and-tabular"
5
+ license: "apache-2.0"
6
+ task_categories:
7
+ - time-series-forecasting
8
+ - regression
9
+ - graph-machine-learning
10
+ language: "en"
11
+ tags:
12
+ - HPC
13
+ - resource-prediction
14
+ - XLA
15
+ - compiler-features
16
+ - deep-learning
17
+ - graph-learning
18
+ - scheduling
19
+ size_categories:
20
+ - 1K<n<10K
21
+ source_datasets:
22
+ - custom
23
+ dataset_summary: >
24
+ The HLO Feature Dataset contains High-Level Optimizer (HLO) graph features and metadata extracted
25
+ from deep learning training workloads. It is designed for tasks such as runtime prediction, resource
26
+ estimation, and graph-based machine learning in HPC environments.
27
+
28
+ Each entry pairs model configuration metadata with compiler graph data stored in `.npz` format.
29
+
30
+ Ideal for ML system optimization studies, GNN research, and AI workload scheduling.
31
+
32
+ structured_data:
33
+ features:
34
+ - name: "batch"
35
+ type: "integer"
36
+ - name: "epochs"
37
+ type: "integer"
38
+ - name: "learn_rate"
39
+ type: "float"
40
+ - name: "gpu_core_count"
41
+ type: "integer"
42
+ - name: "gpu_memory_size"
43
+ type: "integer"
44
+ - name: "fit_time"
45
+ type: "float"
46
+ - name: "npz_path"
47
+ type: "string"
48
+ graph_data:
49
+ node_features: "node_feat"
50
+ edge_index: "edge_index"
51
+ additional_keys:
52
+ - "node_opcode"
53
+ - "node_config_ids"
54
+ - "node_splits"
55
+ usage_example: |
56
+ ```python
57
+ from datasets import load_dataset
58
+ import numpy as np
59
+
60
+ dataset = load_dataset("your-username/hlo-feature-dataset")
61
+ sample = dataset['train'][0]
62
+
63
+ graph_data = np.load(sample['npz_path'])
64
+ node_features = graph_data['node_feat']
65
+ edges = graph_data['edge_index']
66
+
67
+ ---
68
 
69
  # HLO Feature Dataset for Deep Learning Resource Estimation
70
 
 
167
  ---
168
 
169
  ## Contributions
170
+ Open to contributions! Feel free to suggest improvements or share your models trained on this dataset.