Updated data from scicom run

NERC-CEH · Dec 13, 2024 · 9aae2bf · 9aae2bf
1 parent dfeb31f
commit 9aae2bf
Show file tree

Hide file tree

Showing 3 changed files with 57 additions and 47 deletions.
diff --git a/dvc.lock b/dvc.lock
@@ -10,8 +10,8 @@ stages:
     outs:
     - path: data/eidc_metadata.json
       hash: md5
-      md5: fb338ea98ce71bf7f002be952b6db0e1
-      size: 12275265
+      md5: 413f59888bf033c30cc27b84b1a3f40b
+      size: 12313041
   prepare:
     cmd: python scripts/extract_metadata.py data/eidc_metadata.json data/extracted_metadata.json
     deps:
@@ -33,140 +33,150 @@ stages:
     deps:
     - path: data/eidc_metadata.json
       hash: md5
-      md5: fb338ea98ce71bf7f002be952b6db0e1
-      size: 12275265
+      md5: 413f59888bf033c30cc27b84b1a3f40b
+      size: 12313041
     - path: scripts/extract_metadata.py
       hash: md5
       md5: e66f21369c5106eaaad4476612c6fb5e
       size: 1313
     outs:
     - path: data/extracted_metadata.json
       hash: md5
-      md5: 9f4fc9cb1e8af8e0f2d1c95b311989fc
-      size: 4616342
+      md5: 226225c5bd64e15d803ba88560810c5a
+      size: 4629991
   chunk-data:
     cmd: python scripts/chunk_data.py  -o data/chunked_data.json  -c 250  -ol 75  data/extracted_metadata.json  data/supporting-docs.json  -m
       0
     deps:
     - path: data/extracted_metadata.json
       hash: md5
-      md5: 9f4fc9cb1e8af8e0f2d1c95b311989fc
-      size: 4616342
+      md5: 226225c5bd64e15d803ba88560810c5a
+      size: 4629991
     - path: data/supporting-docs.json
       hash: md5
-      md5: 0b14da8f2e73dc8e15747f693c0f70ce
-      size: 72383140
+      md5: e2581aff9abe25942e8009214b88d0a5
+      size: 72680213
     - path: scripts/chunk_data.py
       hash: md5
       md5: 3ad449140b03e1c2904b22a5b401a12e
       size: 2705
     outs:
     - path: data/chunked_data.json
       hash: md5
-      md5: b107dfb052c12ea47b04a5176e8bab4a
-      size: 176342129
+      md5: 9893d839409c8cf4561e99ab5f747f20
+      size: 177068127
   create-embeddings:
     cmd: python scripts/create_embeddings.py data/chunked_data.json data/embeddings.json
       -m all-MiniLM-L6-v2
     deps:
     - path: data/chunked_data.json
       hash: md5
-      md5: b107dfb052c12ea47b04a5176e8bab4a
-      size: 176342129
+      md5: 9893d839409c8cf4561e99ab5f747f20
+      size: 177068127
     - path: scripts/create_embeddings.py
       hash: md5
-      md5: 87bd2ed6373552bea229c9f3465fd3db
-      size: 1594
+      md5: b0d8f7cb90f244e709656b1f38723e2d
+      size: 1552
     outs:
     - path: data/embeddings.json
       hash: md5
-      md5: 68a9de7fcf765be8ae2f4d3ff6537228
-      size: 3739724900
+      md5: 5c8ca3cdde4d5bc559fa2e701ff090a8
+      size: 3754990368
   upload-to-docstore:
     cmd: python scripts/upload_to_docstore.py  data/embeddings.json  -o data/chroma-data  -em
       all-MiniLM-L6-v2  -c eidc-data
     deps:
     - path: data/embeddings.json
       hash: md5
-      md5: 68a9de7fcf765be8ae2f4d3ff6537228
-      size: 3739724900
+      md5: 5c8ca3cdde4d5bc559fa2e701ff090a8
+      size: 3754990368
     - path: scripts/upload_to_docstore.py
       hash: md5
       md5: 930456cedd43723c1d643ad90c146952
       size: 2793
     outs:
     - path: data/chroma-data
       hash: md5
-      md5: 486d560a81dc951bdd85772996e62f00.dir
-      size: 1815042692
+      md5: c06796220fbfe9db3b08b8439edf87b4.dir
+      size: 3081399131
       nfiles: 6
   run-rag-pipeline:
     cmd: python scripts/run_rag_pipeline.py  -i data/eidc_rag_testset.csv  -o data/evaluation_data.csv  -ds
       data/chroma-data  -c eidc-data  -m llama3.1  -p data/pipeline.yml
     deps:
     - path: data/chroma-data
       hash: md5
-      md5: 486d560a81dc951bdd85772996e62f00.dir
-      size: 1815042692
+      md5: c06796220fbfe9db3b08b8439edf87b4.dir
+      size: 3081399131
       nfiles: 6
     - path: data/eidc_rag_testset.csv
       hash: md5
-      md5: a371d83c5822d256286e80d64d58c3fe
-      size: 7524
+      md5: 90d23c9bfcaddf9f152109a7b51e3151
+      size: 149155
     - path: scripts/run_rag_pipeline.py
       hash: md5
-      md5: 2d6dc886728d4bd46676ecd1882f1fd1
-      size: 5838
+      md5: 35eb5f65605242094a1581b92e9b2ef4
+      size: 5862
     outs:
     - path: data/evaluation_data.csv
       hash: md5
-      md5: a473732be874c8256f7178ef3f4dc7a9
-      size: 9576
+      md5: 1b5f226c52d70bda7e2551d7778c1e89
+      size: 385945
     - path: data/pipeline.yml
       hash: md5
-      md5: 8e3c4e49d4d97f613e83468d010a96e9
-      size: 3440
+      md5: 70385a724cdf687c287596b8360e1448
+      size: 3334
   generate-testset:
-    cmd: head -n 101 data/synthetic-datasets/eidc_rag_test_sample.csv > data/eidc_rag_testset.csv
+    cmd: python scripts/generate_synthetic_testset.py data/extracted_metadata.json
+      data/eidc_rag_testset.csv 200
+    deps:
+    - path: data/extracted_metadata.json
+      hash: md5
+      md5: 226225c5bd64e15d803ba88560810c5a
+      size: 4629991
+    - path: scripts/generate_synthetic_testset.py
+      hash: md5
+      md5: fdac8b2f28de8f3b4e5025ca47bb94ca
+      size: 2175
     outs:
     - path: data/eidc_rag_testset.csv
       hash: md5
-      md5: a371d83c5822d256286e80d64d58c3fe
-      size: 7524
+      md5: 90d23c9bfcaddf9f152109a7b51e3151
+      size: 149155
   fetch-supporting-docs:
     cmd: python scripts/fetch_supporting_docs.py data/eidc_metadata.json data/supporting-docs.json
     deps:
     - path: data/eidc_metadata.json
       hash: md5
-      md5: fb338ea98ce71bf7f002be952b6db0e1
-      size: 12275265
+      md5: 413f59888bf033c30cc27b84b1a3f40b
+      size: 12313041
     - path: scripts/fetch_supporting_docs.py
       hash: md5
       md5: 02b94a2cc7bff711784cbdec3650b618
       size: 1718
     outs:
     - path: data/supporting-docs.json
       hash: md5
-      md5: 0b14da8f2e73dc8e15747f693c0f70ce
-      size: 72383140
+      md5: e2581aff9abe25942e8009214b88d0a5
+      size: 72680213
   evaluate:
     cmd: python scripts/evaluate.py data/evaluation_data.csv -m data/metrics.json
       -img data/eval.png
     deps:
     - path: data/evaluation_data.csv
       hash: md5
-      md5: a473732be874c8256f7178ef3f4dc7a9
-      size: 9576
+      md5: 1b5f226c52d70bda7e2551d7778c1e89
+      size: 385945
     - path: scripts/evaluate.py
       hash: md5
       md5: 4154acf8e74c1d8bcd0b0da72af038e0
       size: 2728
     outs:
     - path: data/eval.png
       hash: md5
-      md5: 7bfd424fa4c9a3550d6e9605bb2f6af2
-      size: 89143
+      md5: 3308b984c5168a996805443d25697026
+      size: 83001
     - path: data/metrics.json
       hash: md5
-      md5: f768092fe2696328ff4da565e763e743
-      size: 270
+      md5: 709909482614d6cb47c160506088f53e
+      size: 287
diff --git a/params.yaml b/params.yaml
@@ -19,7 +19,7 @@ files:
   pipeline: data/pipeline.yml
 sub-sample: 0 # sample n datasets for testing (0 will use all datasets)
 max-length: 0 # truncate longer texts for testing (0 will use all data)
-test-set-size: 101 # reduce the size of the test set for faster testing
+test-set-size: 200 # reduce the size of the test set for faster testing
 rag:
   model: llama3.1
   prompt: >-

diff --git a/pyproject.toml b/pyproject.toml
@@ -19,7 +19,7 @@ dependencies = [
     "nbformat == 5.10.4",
     "langchain == 0.2.7",
     "pygit2 == 1.14.1",
-    "pysqlite3 == 0.5.4",
+    "pysqlite3-binary == 0.5.4",
 ]
 
 [project.optional-dependencies]