Upload folder using huggingface_hub

.gitattributes

@@ -67,3 +67,12 @@ data/datasets/lilac/squad_v2/question/near_dup/data-00000-of-00001.parquet filte
 data/datasets/lilac/squad_v2/question/pii/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
 data/datasets/lilac/squad_v2/question/spacy_ner/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
 data/datasets/lilac/squad_v2/question/text_statistics/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/imdb/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/imdb/text/gte-small/hnsw.hnswlib.bin filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/imdb/text/gte-small/hnsw.lookup.pkl filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/imdb/text/gte-small/spans.pkl filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/imdb/text/lang_detection/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/imdb/text/near_dup/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/imdb/text/pii/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/imdb/text/spacy_ner/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/imdb/text/text_statistics/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text

data/datasets/lilac/imdb/config.yml

@@ -0,0 +1,31 @@
+embeddings:
+- embedding: gte-small
+ path: text
+name: imdb
+namespace: local
+settings:
+ preferred_embedding: gte-small
+ ui:
+ media_paths:
+ - text
+signals:
+- path: text
+ signal:
+ signal_name: near_dup
+- path: text
+ signal:
+ signal_name: text_statistics
+- path: text
+ signal:
+ signal_name: lang_detection
+- path: text
+ signal:
+ signal_name: spacy_ner
+- path: text
+ signal:
+ signal_name: pii
+source:
+ dataset_name: imdb
+ source_name: huggingface
+tags:
+- machine-learning

data/datasets/lilac/imdb/data-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6fe90be23f86ca1e73b8a77a235344db822601f794a5643dca9d0d07c49ce3d8
+size 86160450

data/datasets/lilac/imdb/manifest.json

@@ -0,0 +1,21 @@
+{
+ "files": [
+ "data-00000-of-00001.parquet"
+ ],
+ "data_schema": {
+ "fields": {
+ "text": {
+ "dtype": "string"
+ },
+ "label": {
+ "dtype": "string"
+ },
+ "__hfsplit__": {
+ "dtype": "string"
+ },
+ "__rowid__": {
+ "dtype": "string"
+ }
+ }
+ }
+}

data/datasets/lilac/imdb/text/gte-small/hnsw.hnswlib.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b8fc440b947f068966bc68c83ce4b5502b98e3aab928d58c295dc8a7b7b016c
+size 691432396

data/datasets/lilac/imdb/text/gte-small/hnsw.lookup.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d3ee2426c761b50025e02f2388fa65c51eb9705d6fd65f95d6502650214b8472
+size 10390867

data/datasets/lilac/imdb/text/gte-small/signal_manifest.json

@@ -0,0 +1,35 @@
+{
+ "files": [],
+ "parquet_id": "gte-small(text)",
+ "data_schema": {
+ "fields": {
+ "__rowid__": {
+ "dtype": "string"
+ },
+ "text": {
+ "fields": {
+ "gte-small": {
+ "repeated_field": {
+ "fields": {
+ "embedding": {
+ "dtype": "embedding"
+ }
+ },
+ "dtype": "string_span"
+ },
+ "signal": {
+ "signal_name": "gte-small"
+ }
+ }
+ }
+ }
+ }
+ },
+ "signal": {
+ "signal_name": "gte-small"
+ },
+ "enriched_path": [
+ "text"
+ ],
+ "vector_store": "hnsw"
+}

data/datasets/lilac/imdb/text/gte-small/spans.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:16417449057fc4304e098e037cb2b3e9a693570768a68bdca457d452adaee130
+size 7476546

data/datasets/lilac/imdb/text/lang_detection/data-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:adba6df16ec47b68625618b01a0b8d2cc65de572acb20eed561128b037fcdfd7
+size 3309315

data/datasets/lilac/imdb/text/lang_detection/signal_manifest.json

@@ -0,0 +1,31 @@
+{
+ "files": [
+ "data-00000-of-00001.parquet"
+ ],
+ "parquet_id": "lang_detection(text)",
+ "data_schema": {
+ "fields": {
+ "__rowid__": {
+ "dtype": "string"
+ },
+ "text": {
+ "fields": {
+ "lang_detection": {
+ "dtype": "string",
+ "signal": {
+ "split_by_paragraph": false,
+ "signal_name": "lang_detection"
+ }
+ }
+ }
+ }
+ }
+ },
+ "signal": {
+ "split_by_paragraph": false,
+ "signal_name": "lang_detection"
+ },
+ "enriched_path": [
+ "text"
+ ]
+}

data/datasets/lilac/imdb/text/near_dup/data-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0c75a5f0d12b8b02671e99a2dd313d292610c9264f2302ad1877959430420079
+size 3915752

data/datasets/lilac/imdb/text/near_dup/signal_manifest.json

@@ -0,0 +1,36 @@
+{
+ "files": [
+ "data-00000-of-00001.parquet"
+ ],
+ "parquet_id": "near_dup(text)",
+ "data_schema": {
+ "fields": {
+ "__rowid__": {
+ "dtype": "string"
+ },
+ "text": {
+ "fields": {
+ "near_dup": {
+ "fields": {
+ "cluster_id": {
+ "dtype": "uint32",
+ "categorical": true
+ }
+ },
+ "signal": {
+ "threshold": 0.85,
+ "signal_name": "near_dup"
+ }
+ }
+ }
+ }
+ }
+ },
+ "signal": {
+ "threshold": 0.85,
+ "signal_name": "near_dup"
+ },
+ "enriched_path": [
+ "text"
+ ]
+}

data/datasets/lilac/imdb/text/pii/data-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2dfe71845bfd419a2a309b132456341bf317d63e8ccf8bc100835f1a20c81c5b
+size 3313701

data/datasets/lilac/imdb/text/pii/signal_manifest.json

@@ -0,0 +1,45 @@
+{
+ "files": [
+ "data-00000-of-00001.parquet"
+ ],
+ "parquet_id": "pii(text)",
+ "data_schema": {
+ "fields": {
+ "__rowid__": {
+ "dtype": "string"
+ },
+ "text": {
+ "fields": {
+ "pii": {
+ "fields": {
+ "emails": {
+ "repeated_field": {
+ "dtype": "string_span"
+ }
+ },
+ "ip_addresses": {
+ "repeated_field": {
+ "dtype": "string_span"
+ }
+ },
+ "secrets": {
+ "repeated_field": {
+ "dtype": "string_span"
+ }
+ }
+ },
+ "signal": {
+ "signal_name": "pii"
+ }
+ }
+ }
+ }
+ }
+ },
+ "signal": {
+ "signal_name": "pii"
+ },
+ "enriched_path": [
+ "text"
+ ]
+}

data/datasets/lilac/imdb/text/spacy_ner/data-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:be39eb80c981da64a2ec5f7fe42c43ffb390b887a220e278a58cdde04e7824ef
+size 8479478

data/datasets/lilac/imdb/text/spacy_ner/signal_manifest.json

@@ -0,0 +1,38 @@
+{
+ "files": [
+ "data-00000-of-00001.parquet"
+ ],
+ "parquet_id": "spacy_ner(text)",
+ "data_schema": {
+ "fields": {
+ "__rowid__": {
+ "dtype": "string"
+ },
+ "text": {
+ "fields": {
+ "spacy_ner": {
+ "repeated_field": {
+ "fields": {
+ "label": {
+ "dtype": "string"
+ }
+ },
+ "dtype": "string_span"
+ },
+ "signal": {
+ "model": "en_core_web_sm",
+ "signal_name": "spacy_ner"
+ }
+ }
+ }
+ }
+ }
+ },
+ "signal": {
+ "model": "en_core_web_sm",
+ "signal_name": "spacy_ner"
+ },
+ "enriched_path": [
+ "text"
+ ]
+}

data/datasets/lilac/imdb/text/text_statistics/data-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4919817d0c520fc5aacc9721f01a0e29fb4d2c89b6698865c9680c81c44ce920
+size 4403809

data/datasets/lilac/imdb/text/text_statistics/signal_manifest.json

@@ -0,0 +1,59 @@
+{
+ "files": [
+ "data-00000-of-00001.parquet"
+ ],
+ "parquet_id": "text_statistics(text)",
+ "data_schema": {
+ "fields": {
+ "__rowid__": {
+ "dtype": "string"
+ },
+ "text": {
+ "fields": {
+ "text_statistics": {
+ "fields": {
+ "num_characters": {
+ "dtype": "int32"
+ },
+ "readability": {
+ "dtype": "float32"
+ },
+ "log(type_token_ratio)": {
+ "dtype": "float32"
+ },
+ "frac_non_ascii": {
+ "dtype": "float32",
+ "bins": [
+ [
+ "Low",
+ null,
+ 0.15
+ ],
+ [
+ "Medium",
+ 0.15,
+ 0.3
+ ],
+ [
+ "High",
+ 0.3,
+ null
+ ]
+ ]
+ }
+ },
+ "signal": {
+ "signal_name": "text_statistics"
+ }
+ }
+ }
+ }
+ }
+ },
+ "signal": {
+ "signal_name": "text_statistics"
+ },
+ "enriched_path": [
+ "text"
+ ]
+}