Upload folder using huggingface_hub

.gitattributes

@@ -30,3 +30,12 @@ data/datasets/lilac/mmlu_professional_law/question/near_dup/data-00000-of-00001.
 data/datasets/lilac/mmlu_professional_law/question/pii/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
 data/datasets/lilac/mmlu_professional_law/question/spacy_ner/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
 data/datasets/lilac/mmlu_professional_law/question/text_statistics/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/pile-of-law-r-legaladvice/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/pile-of-law-r-legaladvice/text/gte-small/hnsw.hnswlib.bin filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/pile-of-law-r-legaladvice/text/gte-small/hnsw.lookup.pkl filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/pile-of-law-r-legaladvice/text/gte-small/spans.pkl filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/pile-of-law-r-legaladvice/text/lang_detection/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/pile-of-law-r-legaladvice/text/near_dup/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/pile-of-law-r-legaladvice/text/pii/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/pile-of-law-r-legaladvice/text/spacy_ner/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text
+data/datasets/lilac/pile-of-law-r-legaladvice/text/text_statistics/data-00000-of-00001.parquet filter=lfs diff=lfs merge=lfs -text

data/datasets/lilac/pile-of-law-r-legaladvice/config.yml

@@ -0,0 +1,32 @@
+embeddings:
+- embedding: gte-small
+ path: text
+name: pile-of-law-r-legaladvice
+namespace: lilac
+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: pii
+- path: text
+ signal:
+ signal_name: lang_detection
+- path: text
+ signal:
+ signal_name: spacy_ner
+source:
+ config_name: r_legaladvice
+ dataset_name: pile-of-law/pile-of-law
+ source_name: huggingface
+tags:
+- legal

data/datasets/lilac/pile-of-law-r-legaladvice/data-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:315910747b04a2b504c0df875caf84812252602ea10a14c2049a45737a80db86
+size 190604294

data/datasets/lilac/pile-of-law-r-legaladvice/manifest.json

@@ -0,0 +1,27 @@
+{
+ "files": [
+ "data-00000-of-00001.parquet"
+ ],
+ "data_schema": {
+ "fields": {
+ "text": {
+ "dtype": "string"
+ },
+ "created_timestamp": {
+ "dtype": "string"
+ },
+ "downloaded_timestamp": {
+ "dtype": "string"
+ },
+ "url": {
+ "dtype": "string"
+ },
+ "__hfsplit__": {
+ "dtype": "string"
+ },
+ "__rowid__": {
+ "dtype": "string"
+ }
+ }
+ }
+}

data/datasets/lilac/pile-of-law-r-legaladvice/text/gte-small/hnsw.hnswlib.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b613d5dd1584d6a5f8323d1464fa1f9d7a6a3cb8547963a8f3451afc56542a3c
+size 1922075524

data/datasets/lilac/pile-of-law-r-legaladvice/text/gte-small/hnsw.lookup.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f6ea206b1c4dba51a9e18f6db3822615cfe4c0ec435d3ce8c0e78ed4a604d11c
+size 24945492

data/datasets/lilac/pile-of-law-r-legaladvice/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/pile-of-law-r-legaladvice/text/gte-small/spans.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c233892a5f8fb35cdff7539bc657e15b91f91be9410c63e5a6dfc0aa627831ae
+size 15099068

data/datasets/lilac/pile-of-law-r-legaladvice/text/lang_detection/data-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f09e13588467643bc515248dc8b75003ecab9787b8d30b01959dd4f85d0a5eb0
+size 4827225

data/datasets/lilac/pile-of-law-r-legaladvice/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/pile-of-law-r-legaladvice/text/near_dup/data-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b3d7c4f4df146865336e8e0addba0ff1bfe0218bd6ef7d9b6cd2b1ef21a20405
+size 5743841

data/datasets/lilac/pile-of-law-r-legaladvice/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/pile-of-law-r-legaladvice/text/pii/data-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:38ca9bdf0f4a9dd4b0d1bc83f6f6b1446aa5b61c0566e1ffa165dac1565e3099
+size 4834322

data/datasets/lilac/pile-of-law-r-legaladvice/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/pile-of-law-r-legaladvice/text/spacy_ner/data-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:10bb83d20b521ce0035c40b8ab56a0b29f92f728241a71a093ca8c83d5654e8d
+size 12134480

data/datasets/lilac/pile-of-law-r-legaladvice/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/pile-of-law-r-legaladvice/text/text_statistics/data-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bcfb0d11c300e2d97d10c57e63b777cb347936e24253758c716bed24e0b7a2e8
+size 6649363

data/datasets/lilac/pile-of-law-r-legaladvice/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"
+ ]
+}