deliverables

app.py

@@ -1,4 +1,5 @@
 # Required Libraries
+from sentence_transformers import SentenceTransformer
 import fitz  # PyMuPDF
 import os
 from langchain_openai import OpenAIEmbeddings
@@ -7,9 +8,13 @@ from dotenv import load_dotenv
 from qdrant_client import QdrantClient  # Import Qdrant client
 import uuid  # Add this import at the top of your file
 import json
+from huggingface_hub import login
+
 
 # Load environment variables from .env file
 load_dotenv()
+login(token=os.getenv("HF_TOKEN"))
+
 
 # Initialize Qdrant client
 qdrant_api_key = os.getenv("QDRANT_API_KEY")  # Get the Qdrant API key from environment variables
@@ -115,12 +120,19 @@ def chunk_text(text, themes):
     return thematic_chunks
 
 # Step 4: Embed the Chunks
-def embed_chunks(thematic_chunks):
+def embed_chunks_openai(thematic_chunks):
     openai_api_key = os.getenv("OPENAI_API_KEY")  # Get the API key from environment variables
     embeddings = OpenAIEmbeddings(model="text-embedding-3-small",openai_api_key=openai_api_key)
     embedded_chunks = {theme: embeddings.embed_documents(chunks) for theme, chunks in thematic_chunks.items()}
     return embedded_chunks
 
+def embed_chunks_fine_tuned(thematic_chunks):
+    model = SentenceTransformer("svb01/fine-tuned-embedding-model")
+    embedded_chunks = {theme: model.encode(chunks) for theme, chunks in thematic_chunks.items()}
+    return embedded_chunks
+
+# The rest of app.py remains the same
+
 # Modified main execution block
 def main():
     resources_folder = "resources"
@@ -134,7 +146,7 @@ def main():
             pdf_path = os.path.join(resources_folder, filename)
             text = extract_text_from_pdf(pdf_path)
             thematic_chunks = chunk_text(text, themes)
-            embedded_chunks = embed_chunks(thematic_chunks)
+            embedded_chunks = embed_chunks_fine_tuned(thematic_chunks)
 
             # Ensure the collection exists
             if not new_docs_processed:

deliverables/Task1.md

@@ -0,0 +1,49 @@
+**Task 1: Dealing with the Data**
+
+You identify the following important documents that, if used for context, you believe will help people understand what’s happening now:
+    1. 2022: Blueprint for an AI Bill of Rights: Making Automated Systems Work for the American People (PDF)
+    2. 2024: National Institute of Standards and Technology (NIST) Artificial Intelligent Risk Management Framework (PDF)
+
+Your boss, the SVP of Technology, green-lighted this project to drive the adoption of AI throughout the enterprise. It will be a nice showpiece for the upcoming conference and the big AI initiative announcement the CEO is planning.
+
+
+Task 1: Review the two PDFs and decide how best to chunk up the data with a single strategy to optimally answer the variety of questions you expect to receive from people.
+Hint: Create a list of potential questions that people are likely to ask!
+
+
+
+
+✅ Deliverables:
+
+1. Describe the default chunking strategy that you will use.
+   
+The default chunking strategy used is a combination of size-based splitting and thematic categorization. 
+This strategy uses RecursiveCharacterTextSplitter with a chunk size of 1000 characters and an overlap of 200 characters. It then categorizes these chunks based on predefined themes.
+
+2.  Articulate a chunking strategy that you would also like to test out.
+    
+A pure size-based chunking strategy without thematic categorization. This would involve splitting the text into fixed-size chunks without attempting to categorize them based on themes.
+
+
+
+3. Describe how and why you made these decisions
+   
+The default strategy was chosen for its simplicity and efficiency:
+
+* Size-based splitting (1000 characters) ensures manageable chunk sizes for processing and embedding.
+* The 200-character overlap helps maintain context between chunks.
+* Thematic categorization allows for organized retrieval based on specific topics of interest.
+
+This approach balances processing efficiency with maintaining semantic coherence within chunks.
+
+The alternative pure size-based strategy:
+* Ensures consistent chunk sizes, which can be beneficial for processing and embedding.
+* Is simpler to implement and doesn't rely on predefined themes.
+* May split semantic units, potentially affecting the coherence of individual chunks.'
+* Could be more comprehensive, including all parts of the document regardless of theme.
+
+
+
+
+
+

fine_tune_model.py

@@ -0,0 +1,141 @@
+import os
+from sentence_transformers import SentenceTransformer, InputExample, losses
+from torch.utils.data import DataLoader
+import random
+from langchain_openai import OpenAIEmbeddings
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+import fitz  # PyMuPDF
+from dotenv import load_dotenv
+from huggingface_hub import login, HfApi
+import traceback
+
+# Add this near the top of your file, after the imports
+
+
+load_dotenv()
+login(token=os.getenv("HF_TOKEN"), add_to_git_credential=True)
+
+# Step 1: Extract Text from PDFs
+def extract_text_from_pdf(pdf_path):
+    doc = fitz.open(pdf_path)
+    text = ""
+    for page in doc:
+        text += page.get_text()
+    return text    
+
+def chunk_text(text, themes):
+    text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
+    chunks = text_splitter.split_text(text)
+    thematic_chunks = {theme: [] for theme in themes}
+    thematic_chunks["Unclassified"] = []  # Add an "Unclassified" category
+    
+    for chunk in chunks:
+        theme_found = False
+        for theme in themes:
+            if theme.lower() in chunk.lower():
+                thematic_chunks[theme].append(chunk)
+                theme_found = True
+                break
+        if not theme_found:
+            thematic_chunks["Unclassified"].append(chunk)
+    
+    print("Chunks per theme:")
+    for theme, theme_chunks in thematic_chunks.items():
+        print(f"  {theme}: {len(theme_chunks)}")
+    
+    return thematic_chunks
+    # ... (same as in app.py)
+
+# Function to generate synthetic fine-tuning data
+def generate_synthetic_data(thematic_chunks, n_samples=1000):
+    examples = []
+    print(f"Total themes: {len(thematic_chunks)}")
+    for theme, chunks in thematic_chunks.items():
+        print(f"Theme: {theme}, Number of chunks: {len(chunks)}")
+        if not chunks:
+            print(f"Warning: No chunks for theme '{theme}'. Skipping this theme.")
+            continue
+        samples_per_theme = max(1, n_samples // len(thematic_chunks))
+        for _ in range(samples_per_theme):
+            chunk = random.choice(chunks)
+            question = f"What does this text say about {theme.lower()}?"
+            examples.append(InputExample(texts=[question, chunk]))
+    print(f"Total examples generated: {len(examples)}")
+    return examples
+
+# Function to fine-tune the model
+def fine_tune_model(model, train_examples, output_path):
+    train_dataloader = DataLoader(train_examples, shuffle=True, batch_size=16)
+    train_loss = losses.MultipleNegativesRankingLoss(model)
+    
+    model.fit(train_objectives=[(train_dataloader, train_loss)], epochs=3, warmup_steps=100, output_path=output_path)
+    return model
+
+def main():
+    resources_folder = "resources"
+    themes = [
+        "Safe and Effective Systems",
+        "Algorithmic Discrimination Protections",
+        "Data Privacy",
+        "Notice and Explanation",
+        "Human Alternatives",
+        "Risk Management",
+        "Governance",
+        "Trustworthiness",
+        "Unclassified"
+    ]
+    
+    all_thematic_chunks = {}
+    
+    for filename in os.listdir(resources_folder):
+        if filename.endswith(".pdf"):
+            pdf_path = os.path.join(resources_folder, filename)
+            text = extract_text_from_pdf(pdf_path)
+            thematic_chunks = chunk_text(text, themes)
+            all_thematic_chunks.update(thematic_chunks)
+            print(f"Processed {filename}")
+
+    # Fine-tune the model
+    base_model = "sentence-transformers/all-MiniLM-L6-v2"
+    model = SentenceTransformer(base_model)
+    train_examples = generate_synthetic_data(all_thematic_chunks)
+    fine_tuned_model_path = "fine_tuned_embedding_model"
+    fine_tune_model(model, train_examples, fine_tuned_model_path)
+    
+    print("Fine-tuning completed. Model saved locally.")
+
+def upload_model_to_hub():
+    try:
+        # Load the fine-tuned model
+        fine_tuned_model_path = "fine_tuned_embedding_model"
+        model = SentenceTransformer(fine_tuned_model_path)
+
+        # Upload the fine-tuned model to Hugging Face Hub
+        repo_id = "svb01/fine-tuned-embedding-model"
+        
+        print(f"Uploading model to existing repository: {repo_id}")
+        
+        # Use HfApi to upload files directly
+        api = HfApi()
+        
+        # Upload each file in the model directory
+        for root, _, files in os.walk(fine_tuned_model_path):
+            for file in files:
+                file_path = os.path.join(root, file)
+                api.upload_file(
+                    path_or_fileobj=file_path,
+                    path_in_repo=file,
+                    repo_id=repo_id,
+                    commit_message=f"Upload {file}"
+                )
+        
+        print("Fine-tuned model uploaded to Hugging Face Hub.")
+    except Exception as e:
+        print(f"Error uploading model to Hugging Face Hub: {str(e)}")
+        print("Detailed error information:")
+        print(traceback.format_exc())
+
+if __name__ == "__main__":
+    # Uncomment the function you want to run
+    # main()  # Run this for fine-tuning
+    upload_model_to_hub()  # Run this to upload the model

fine_tuned_embedding_model/1_Pooling/config.json

@@ -0,0 +1,10 @@
+{
+  "word_embedding_dimension": 384,
+  "pooling_mode_cls_token": false,
+  "pooling_mode_mean_tokens": true,
+  "pooling_mode_max_tokens": false,
+  "pooling_mode_mean_sqrt_len_tokens": false,
+  "pooling_mode_weightedmean_tokens": false,
+  "pooling_mode_lasttoken": false,
+  "include_prompt": true
+}

fine_tuned_embedding_model/README.md

@@ -0,0 +1,520 @@
+---
+base_model: sentence-transformers/all-MiniLM-L6-v2
+library_name: sentence-transformers
+pipeline_tag: sentence-similarity
+tags:
+- sentence-transformers
+- sentence-similarity
+- feature-extraction
+- generated_from_trainer
+- dataset_size:555
+- loss:MultipleNegativesRankingLoss
+widget:
+- source_sentence: What does this text say about unclassified?
+  sentences:
+  - "these sources. \nErrors in third-party GAI components can also have downstream\
+    \ impacts on accuracy and robustness. \nFor example, test datasets commonly used\
+    \ to benchmark or validate models can contain label errors. \nInaccuracies in\
+    \ these labels can impact the “stability” or robustness of these benchmarks, which\
+    \ many \nGAI practitioners consider during the model selection process.  \nTrustworthy\
+    \ AI Characteristics: Accountable and Transparent, Explainable and Interpretable,\
+    \ Fair with \nHarmful Bias Managed, Privacy Enhanced, Safe, Secure and Resilient,\
+    \ Valid and Reliable \n3. \nSuggested Actions to Manage GAI Risks \nThe following\
+    \ suggested actions target risks unique to or exacerbated by GAI. \nIn addition\
+    \ to the suggested actions below, AI risk management activities and actions set\
+    \ forth in the AI \nRMF 1.0 and Playbook are already applicable for managing GAI\
+    \ risks. Organizations are encouraged to"
+  - "and hardware vulnerabilities; labor practices; data privacy and localization\
+    \ \ncompliance; geopolitical alignment). \nData Privacy; Information Security;\
+    \ \nValue Chain and Component \nIntegration; Harmful Bias and \nHomogenization\
+    \ \nMG-3.1-003 \nRe-assess model risks after fine-tuning or retrieval-augmented\
+    \ generation \nimplementation and for any third-party GAI models deployed for\
+    \ applications \nand/or use cases that were not evaluated in initial testing.\
+    \ \nValue Chain and Component \nIntegration \nMG-3.1-004 \nTake reasonable measures\
+    \ to review training data for CBRN information, and \nintellectual property, and\
+    \ where appropriate, remove it. Implement reasonable \nmeasures to prevent, flag,\
+    \ or take other action in response to outputs that \nreproduce particular training\
+    \ data (e.g., plagiarized, trademarked, patented, \nlicensed content or trade\
+    \ secret material). \nIntellectual Property; CBRN \nInformation or Capabilities\
+    \ \n \n43"
+  - "• \nStage of the AI lifecycle: Risks can arise during design, development, deployment,\
+    \ operation, \nand/or decommissioning. \n• \nScope: Risks may exist at individual\
+    \ model or system levels, at the application or implementation \nlevels (i.e.,\
+    \ for a specific use case), or at the ecosystem level – that is, beyond a single\
+    \ system or \norganizational context. Examples of the latter include the expansion\
+    \ of “algorithmic \nmonocultures,3” resulting from repeated use of the same model,\
+    \ or impacts on access to \nopportunity, labor markets, and the creative economies.4\
+    \ \n• \nSource of risk: Risks may emerge from factors related to the design, training,\
+    \ or operation of the \nGAI model itself, stemming in some cases from GAI model\
+    \ or system inputs, and in other cases, \nfrom GAI system outputs. Many GAI risks,\
+    \ however, originate from human behavior, including \n \n \n3 “Algorithmic monocultures”\
+    \ refers to the phenomenon in which repeated use of the same model or algorithm\
+    \ in"
+- source_sentence: What does this text say about unclassified?
+  sentences:
+  - "Security; Dangerous, Violent, or \nHateful Content \n \n34 \nMS-2.7-009 Regularly\
+    \ assess and verify that security measures remain effective and have not \nbeen\
+    \ compromised. \nInformation Security \nAI Actor Tasks: AI Deployment, AI Impact\
+    \ Assessment, Domain Experts, Operation and Monitoring, TEVV \n \nMEASURE 2.8:\
+    \ Risks associated with transparency and accountability – as identified in the\
+    \ MAP function – are examined and \ndocumented. \nAction ID \nSuggested Action\
+    \ \nGAI Risks \nMS-2.8-001 \nCompile statistics on actual policy violations, take-down\
+    \ requests, and intellectual \nproperty infringement for organizational GAI systems:\
+    \ Analyze transparency \nreports across demographic groups, languages groups.\
+    \ \nIntellectual Property; Harmful Bias \nand Homogenization \nMS-2.8-002 Document\
+    \ the instructions given to data annotators or AI red-teamers. \nHuman-AI Configuration\
+    \ \nMS-2.8-003 \nUse digital content transparency solutions to enable the documentation\
+    \ of each"
+  - "information during GAI training and maintenance. \nHuman-AI Configuration; Obscene,\
+    \ \nDegrading, and/or Abusive \nContent; Value Chain and \nComponent Integration;\
+    \ \nDangerous, Violent, or Hateful \nContent \nMS-2.6-002 \nAssess existence or\
+    \ levels of harmful bias, intellectual property infringement, \ndata privacy violations,\
+    \ obscenity, extremism, violence, or CBRN information in \nsystem training data.\
+    \ \nData Privacy; Intellectual Property; \nObscene, Degrading, and/or \nAbusive\
+    \ Content; Harmful Bias and \nHomogenization; Dangerous, \nViolent, or Hateful\
+    \ Content; CBRN \nInformation or Capabilities \nMS-2.6-003 Re-evaluate safety\
+    \ features of fine-tuned models when the negative risk exceeds \norganizational\
+    \ risk tolerance. \nDangerous, Violent, or Hateful \nContent \nMS-2.6-004 Review\
+    \ GAI system outputs for validity and safety: Review generated code to \nassess\
+    \ risks that may arise from unreliable downstream decision-making. \nValue Chain\
+    \ and Component \nIntegration; Dangerous, Violent, or \nHateful Content"
+  - "Information Integrity; Harmful Bias \nand Homogenization \nAI Actor Tasks: AI\
+    \ Deployment, AI Impact Assessment, Domain Experts, End-Users, Operation and Monitoring,\
+    \ TEVV \n \nMEASURE 2.10: Privacy risk of the AI system – as identified in the\
+    \ MAP function – is examined and documented. \nAction ID \nSuggested Action \n\
+    GAI Risks \nMS-2.10-001 \nConduct AI red-teaming to assess issues such as: Outputting\
+    \ of training data \nsamples, and subsequent reverse engineering, model extraction,\
+    \ and \nmembership inference risks; Revealing biometric, confidential, copyrighted,\
+    \ \nlicensed, patented, personal, proprietary, sensitive, or trade-marked information;\
+    \ \nTracking or revealing location information of users or members of training\
+    \ \ndatasets. \nHuman-AI Configuration; \nInformation Integrity; Intellectual \n\
+    Property \nMS-2.10-002 \nEngage directly with end-users and other stakeholders\
+    \ to understand their \nexpectations and concerns regarding content provenance.\
+    \ Use this feedback to"
+- source_sentence: What does this text say about risk management?
+  sentences:
+  - "robust watermarking techniques and corresponding detectors to identify the source\
+    \ of content or \nmetadata recording techniques and metadata management tools\
+    \ and repositories to trace content \norigins and modifications. Further narrowing\
+    \ of GAI task definitions to include provenance data can \nenable organizations\
+    \ to maximize the utility of provenance data and risk management efforts. \nA.1.7.\
+    \ Enhancing Content Provenance through Structured Public Feedback \nWhile indirect\
+    \ feedback methods such as automated error collection systems are useful, they\
+    \ often lack \nthe context and depth that direct input from end users can provide.\
+    \ Organizations can leverage feedback \napproaches described in the Pre-Deployment\
+    \ Testing section to capture input from external sources such \nas through AI\
+    \ red-teaming.  \nIntegrating pre- and post-deployment external feedback into\
+    \ the monitoring process for GAI models and"
+  - "tools for monitoring third-party GAI risks; Consider policy adjustments across\
+    \ GAI \nmodeling libraries, tools and APIs, fine-tuned models, and embedded tools;\
+    \ \nAssess GAI vendors, open-source or proprietary GAI tools, or GAI service \n\
+    providers against incident or vulnerability databases. \nData Privacy; Human-AI\
+    \ \nConfiguration; Information \nSecurity; Intellectual Property; \nValue Chain\
+    \ and Component \nIntegration; Harmful Bias and \nHomogenization \nGV-6.1-010\
+    \ \nUpdate GAI acceptable use policies to address proprietary and open-source\
+    \ GAI \ntechnologies and data, and contractors, consultants, and other third-party\
+    \ \npersonnel. \nIntellectual Property; Value Chain \nand Component Integration\
+    \ \nAI Actor Tasks: Operation and Monitoring, Procurement, Third-party entities\
+    \ \n \nGOVERN 6.2: Contingency processes are in place to handle failures or incidents\
+    \ in third-party data or AI systems deemed to be \nhigh-risk. \nAction ID \nSuggested\
+    \ Action \nGAI Risks \nGV-6.2-001"
+  - "MEASURE 2.3: AI system performance or assurance criteria are measured qualitatively\
+    \ or quantitatively and demonstrated for \nconditions similar to deployment setting(s).\
+    \ Measures are documented. \nAction ID \nSuggested Action \nGAI Risks \nMS-2.3-001\
+    \ Consider baseline model performance on suites of benchmarks when selecting a\
+    \ \nmodel for fine tuning or enhancement with retrieval-augmented generation. \n\
+    Information Security; \nConfabulation \nMS-2.3-002 Evaluate claims of model capabilities\
+    \ using empirically validated methods. \nConfabulation; Information \nSecurity\
+    \ \nMS-2.3-003 Share results of pre-deployment testing with relevant GAI Actors,\
+    \ such as those \nwith system release approval authority. \nHuman-AI Configuration\
+    \ \n \n31 \nMS-2.3-004 \nUtilize a purpose-built testing environment such as NIST\
+    \ Dioptra to empirically \nevaluate GAI trustworthy characteristics. \nCBRN Information\
+    \ or Capabilities; \nData Privacy; Confabulation; \nInformation Integrity; Information\
+    \ \nSecurity; Dangerous, Violent, or"
+- source_sentence: What does this text say about unclassified?
+  sentences:
+  - "techniques such as re-sampling, re-ranking, or adversarial training to mitigate\
+    \ \nbiases in the generated content. \nInformation Security; Harmful Bias \nand\
+    \ Homogenization \nMG-2.2-005 \nEngage in due diligence to analyze GAI output\
+    \ for harmful content, potential \nmisinformation, and CBRN-related or NCII content.\
+    \ \nCBRN Information or Capabilities; \nObscene, Degrading, and/or \nAbusive Content;\
+    \ Harmful Bias and \nHomogenization; Dangerous, \nViolent, or Hateful Content\
+    \ \n \n41 \nMG-2.2-006 \nUse feedback from internal and external AI Actors, users,\
+    \ individuals, and \ncommunities, to assess impact of AI-generated content. \n\
+    Human-AI Configuration \nMG-2.2-007 \nUse real-time auditing tools where they can\
+    \ be demonstrated to aid in the \ntracking and validation of the lineage and authenticity\
+    \ of AI-generated data. \nInformation Integrity \nMG-2.2-008 \nUse structured\
+    \ feedback mechanisms to solicit and capture user input about AI-\ngenerated content\
+    \ to detect subtle shifts in quality or alignment with"
+  - "Human-AI Configuration; Value \nChain and Component Integration \nMP-5.2-002 \n\
+    Plan regular engagements with AI Actors responsible for inputs to GAI systems,\
+    \ \nincluding third-party data and algorithms, to review and evaluate unanticipated\
+    \ \nimpacts. \nHuman-AI Configuration; Value \nChain and Component Integration\
+    \ \nAI Actor Tasks: AI Deployment, AI Design, AI Impact Assessment, Affected Individuals\
+    \ and Communities, Domain Experts, End-\nUsers, Human Factors, Operation and Monitoring\
+    \  \n \nMEASURE 1.1: Approaches and metrics for measurement of AI risks enumerated\
+    \ during the MAP function are selected for \nimplementation starting with the\
+    \ most significant AI risks. The risks or trustworthiness characteristics that\
+    \ will not – or cannot – be \nmeasured are properly documented. \nAction ID \n\
+    Suggested Action \nGAI Risks \nMS-1.1-001 Employ methods to trace the origin and\
+    \ modifications of digital content. \nInformation Integrity \nMS-1.1-002"
+  - "input them directly to a GAI system, with a variety of downstream negative consequences\
+    \ to \ninterconnected systems. Indirect prompt injection attacks occur when adversaries\
+    \ remotely (i.e., without \na direct interface) exploit LLM-integrated applications\
+    \ by injecting prompts into data likely to be \nretrieved. Security researchers\
+    \ have already demonstrated how indirect prompt injections can exploit \nvulnerabilities\
+    \ by stealing proprietary data or running malicious code remotely on a machine.\
+    \ Merely \nquerying a closed production model can elicit previously undisclosed\
+    \ information about that model. \nAnother cybersecurity risk to GAI is data poisoning,\
+    \ in which an adversary compromises a training \ndataset used by a model to manipulate\
+    \ its outputs or operation. Malicious tampering with data or parts \nof the model\
+    \ could exacerbate risks associated with GAI system outputs. \nTrustworthy AI\
+    \ Characteristics: Privacy Enhanced, Safe, Secure and Resilient, Valid and Reliable\
+    \ \n2.10."
+- source_sentence: What does this text say about data privacy?
+  sentences:
+  - "Property. We also note that some risks are cross-cutting between these categories.\
+    \  \n \n4 \n1. CBRN Information or Capabilities: Eased access to or synthesis\
+    \ of materially nefarious \ninformation or design capabilities related to chemical,\
+    \ biological, radiological, or nuclear (CBRN) \nweapons or other dangerous materials\
+    \ or agents. \n2. Confabulation: The production of confidently stated but erroneous\
+    \ or false content (known \ncolloquially as “hallucinations” or “fabrications”)\
+    \ by which users may be misled or deceived.6 \n3. Dangerous, Violent, or Hateful\
+    \ Content: Eased production of and access to violent, inciting, \nradicalizing,\
+    \ or threatening content as well as recommendations to carry out self-harm or\
+    \ \nconduct illegal activities. Includes difficulty controlling public exposure\
+    \ to hateful and disparaging \nor stereotyping content. \n4. Data Privacy: Impacts\
+    \ due to leakage and unauthorized use, disclosure, or de-anonymization of"
+  - "information during GAI training and maintenance. \nHuman-AI Configuration; Obscene,\
+    \ \nDegrading, and/or Abusive \nContent; Value Chain and \nComponent Integration;\
+    \ \nDangerous, Violent, or Hateful \nContent \nMS-2.6-002 \nAssess existence or\
+    \ levels of harmful bias, intellectual property infringement, \ndata privacy violations,\
+    \ obscenity, extremism, violence, or CBRN information in \nsystem training data.\
+    \ \nData Privacy; Intellectual Property; \nObscene, Degrading, and/or \nAbusive\
+    \ Content; Harmful Bias and \nHomogenization; Dangerous, \nViolent, or Hateful\
+    \ Content; CBRN \nInformation or Capabilities \nMS-2.6-003 Re-evaluate safety\
+    \ features of fine-tuned models when the negative risk exceeds \norganizational\
+    \ risk tolerance. \nDangerous, Violent, or Hateful \nContent \nMS-2.6-004 Review\
+    \ GAI system outputs for validity and safety: Review generated code to \nassess\
+    \ risks that may arise from unreliable downstream decision-making. \nValue Chain\
+    \ and Component \nIntegration; Dangerous, Violent, or \nHateful Content"
+  - "Scheurer, J. et al. (2023) Technical report: Large language models can strategically\
+    \ deceive their users \nwhen put under pressure. arXiv. https://arxiv.org/abs/2311.07590\
+    \ \nShelby, R. et al. (2023) Sociotechnical Harms of Algorithmic Systems: Scoping\
+    \ a Taxonomy for Harm \nReduction. arXiv. https://arxiv.org/pdf/2210.05791 \n\
+    Shevlane, T. et al. (2023) Model evaluation for extreme risks. arXiv. https://arxiv.org/pdf/2305.15324\
+    \ \nShumailov, I. et al. (2023) The curse of recursion: training on generated\
+    \ data makes models forget. arXiv. \nhttps://arxiv.org/pdf/2305.17493v2 \nSmith,\
+    \ A. et al. (2023) Hallucination or Confabulation? Neuroanatomy as metaphor in\
+    \ Large Language \nModels. PLOS Digital Health. \nhttps://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0000388\
+    \ \nSoice, E. et al. (2023) Can large language models democratize access to dual-use\
+    \ biotechnology? arXiv. \nhttps://arxiv.org/abs/2306.03809"
+---
+
+# SentenceTransformer based on sentence-transformers/all-MiniLM-L6-v2
+
+This is a [sentence-transformers](https://www.SBERT.net) model finetuned from [sentence-transformers/all-MiniLM-L6-v2](https://huggingface.co/sentence-transformers/all-MiniLM-L6-v2). It maps sentences & paragraphs to a 384-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.
+
+## Model Details
+
+### Model Description
+- **Model Type:** Sentence Transformer
+- **Base model:** [sentence-transformers/all-MiniLM-L6-v2](https://huggingface.co/sentence-transformers/all-MiniLM-L6-v2) <!-- at revision 8b3219a92973c328a8e22fadcfa821b5dc75636a -->
+- **Maximum Sequence Length:** 256 tokens
+- **Output Dimensionality:** 384 tokens
+- **Similarity Function:** Cosine Similarity
+<!-- - **Training Dataset:** Unknown -->
+<!-- - **Language:** Unknown -->
+<!-- - **License:** Unknown -->
+
+### Model Sources
+
+- **Documentation:** [Sentence Transformers Documentation](https://sbert.net)
+- **Repository:** [Sentence Transformers on GitHub](https://github.com/UKPLab/sentence-transformers)
+- **Hugging Face:** [Sentence Transformers on Hugging Face](https://huggingface.co/models?library=sentence-transformers)
+
+### Full Model Architecture
+
+```
+SentenceTransformer(
+  (0): Transformer({'max_seq_length': 256, 'do_lower_case': False}) with Transformer model: BertModel 
+  (1): Pooling({'word_embedding_dimension': 384, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
+  (2): Normalize()
+)
+```
+
+## Usage
+
+### Direct Usage (Sentence Transformers)
+
+First install the Sentence Transformers library:
+
+```bash
+pip install -U sentence-transformers
+```
+
+Then you can load this model and run inference.
+```python
+from sentence_transformers import SentenceTransformer
+
+# Download from the 🤗 Hub
+model = SentenceTransformer("sentence_transformers_model_id")
+# Run inference
+sentences = [
+    'What does this text say about data privacy?',
+    'information during GAI training and maintenance. \nHuman-AI Configuration; Obscene, \nDegrading, and/or Abusive \nContent; Value Chain and \nComponent Integration; \nDangerous, Violent, or Hateful \nContent \nMS-2.6-002 \nAssess existence or levels of harmful bias, intellectual property infringement, \ndata privacy violations, obscenity, extremism, violence, or CBRN information in \nsystem training data. \nData Privacy; Intellectual Property; \nObscene, Degrading, and/or \nAbusive Content; Harmful Bias and \nHomogenization; Dangerous, \nViolent, or Hateful Content; CBRN \nInformation or Capabilities \nMS-2.6-003 Re-evaluate safety features of fine-tuned models when the negative risk exceeds \norganizational risk tolerance. \nDangerous, Violent, or Hateful \nContent \nMS-2.6-004 Review GAI system outputs for validity and safety: Review generated code to \nassess risks that may arise from unreliable downstream decision-making. \nValue Chain and Component \nIntegration; Dangerous, Violent, or \nHateful Content',
+    'Scheurer, J. et al. (2023) Technical report: Large language models can strategically deceive their users \nwhen put under pressure. arXiv. https://arxiv.org/abs/2311.07590 \nShelby, R. et al. (2023) Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm \nReduction. arXiv. https://arxiv.org/pdf/2210.05791 \nShevlane, T. et al. (2023) Model evaluation for extreme risks. arXiv. https://arxiv.org/pdf/2305.15324 \nShumailov, I. et al. (2023) The curse of recursion: training on generated data makes models forget. arXiv. \nhttps://arxiv.org/pdf/2305.17493v2 \nSmith, A. et al. (2023) Hallucination or Confabulation? Neuroanatomy as metaphor in Large Language \nModels. PLOS Digital Health. \nhttps://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0000388 \nSoice, E. et al. (2023) Can large language models democratize access to dual-use biotechnology? arXiv. \nhttps://arxiv.org/abs/2306.03809',
+]
+embeddings = model.encode(sentences)
+print(embeddings.shape)
+# [3, 384]
+
+# Get the similarity scores for the embeddings
+similarities = model.similarity(embeddings, embeddings)
+print(similarities.shape)
+# [3, 3]
+```
+
+<!--
+### Direct Usage (Transformers)
+
+<details><summary>Click to see the direct usage in Transformers</summary>
+
+</details>
+-->
+
+<!--
+### Downstream Usage (Sentence Transformers)
+
+You can finetune this model on your own dataset.
+
+<details><summary>Click to expand</summary>
+
+</details>
+-->
+
+<!--
+### Out-of-Scope Use
+
+*List how the model may foreseeably be misused and address what users ought not to do with the model.*
+-->
+
+<!--
+## Bias, Risks and Limitations
+
+*What are the known or foreseeable issues stemming from this model? You could also flag here known failure cases or weaknesses of the model.*
+-->
+
+<!--
+### Recommendations
+
+*What are recommendations with respect to the foreseeable issues? For example, filtering explicit content.*
+-->
+
+## Training Details
+
+### Training Dataset
+
+#### Unnamed Dataset
+
+
+* Size: 555 training samples
+* Columns: <code>sentence_0</code> and <code>sentence_1</code>
+* Approximate statistics based on the first 555 samples:
+  |         | sentence_0                                                                        | sentence_1                                                                            |
+  |:--------|:----------------------------------------------------------------------------------|:--------------------------------------------------------------------------------------|
+  | type    | string                                                                            | string                                                                                |
+  | details | <ul><li>min: 10 tokens</li><li>mean: 11.2 tokens</li><li>max: 12 tokens</li></ul> | <ul><li>min: 156 tokens</li><li>mean: 199.37 tokens</li><li>max: 256 tokens</li></ul> |
+* Samples:
+  | sentence_0                                                  | sentence_1                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                  |
+  |:------------------------------------------------------------|:------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
+  | <code>What does this text say about trustworthiness?</code> | <code>other systems. <br>Information Integrity; Value Chain <br>and Component Integration <br>MP-2.2-002 <br>Observe and analyze how the GAI system interacts with external networks, and <br>identify any potential for negative externalities, particularly where content <br>provenance might be compromised. <br>Information Integrity <br>AI Actor Tasks: End Users <br> <br>MAP 2.3: Scientific integrity and TEVV considerations are identified and documented, including those related to experimental <br>design, data collection and selection (e.g., availability, representativeness, suitability), system trustworthiness, and construct <br>validation <br>Action ID <br>Suggested Action <br>GAI Risks <br>MP-2.3-001 <br>Assess the accuracy, quality, reliability, and authenticity of GAI output by <br>comparing it to a set of known ground truth data and by using a variety of <br>evaluation methods (e.g., human oversight and automated evaluation, proven <br>cryptographic techniques, review of content inputs). <br>Information Integrity <br> <br>25</code>                     |
+  | <code>What does this text say about unclassified?</code>    | <code>training and TEVV data; Filtering of hate speech or content in GAI system <br>training data; Prevalence of GAI-generated data in GAI system training data. <br>Harmful Bias and Homogenization <br> <br> <br>15 Winogender Schemas is a sample set of paired sentences which differ only by gender of the pronouns used, <br>which can be used to evaluate gender bias in natural language processing coreference resolution systems.  <br> <br>37 <br>MS-2.11-005 <br>Assess the proportion of synthetic to non-synthetic training data and verify <br>training data is not overly homogenous or GAI-produced to mitigate concerns of <br>model collapse. <br>Harmful Bias and Homogenization <br>AI Actor Tasks: AI Deployment, AI Impact Assessment, Affected Individuals and Communities, Domain Experts, End-Users, <br>Operation and Monitoring, TEVV <br> <br>MEASURE 2.12: Environmental impact and sustainability of AI model training and management activities – as identified in the MAP <br>function – are assessed and documented. <br>Action ID <br>Suggested Action <br>GAI Risks</code> |
+  | <code>What does this text say about unclassified?</code>    | <code>Padmakumar, V. et al. (2024) Does writing with language models reduce content diversity? ICLR. <br>https://arxiv.org/pdf/2309.05196 <br>Park, P. et. al. (2024) AI deception: A survey of examples, risks, and potential solutions. Patterns, 5(5). <br>arXiv. https://arxiv.org/pdf/2308.14752 <br>Partnership on AI (2023) Building a Glossary for Synthetic Media Transparency Methods, Part 1: Indirect <br>Disclosure. https://partnershiponai.org/glossary-for-synthetic-media-transparency-methods-part-1-<br>indirect-disclosure/ <br>Qu, Y. et al. (2023) Unsafe Diffusion: On the Generation of Unsafe Images and Hateful Memes From Text-<br>To-Image Models. arXiv. https://arxiv.org/pdf/2305.13873 <br>Rafat, K. et al. (2023) Mitigating carbon footprint for knowledge distillation based deep learning model <br>compression. PLOS One. https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0285668 <br>Said, I. et al. (2022) Nonconsensual Distribution of Intimate Images: Exploring the Role of Legal Attitudes</code>                                              |
+* Loss: [<code>MultipleNegativesRankingLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#multiplenegativesrankingloss) with these parameters:
+  ```json
+  {
+      "scale": 20.0,
+      "similarity_fct": "cos_sim"
+  }
+  ```
+
+### Training Hyperparameters
+#### Non-Default Hyperparameters
+
+- `per_device_train_batch_size`: 16
+- `per_device_eval_batch_size`: 16
+- `multi_dataset_batch_sampler`: round_robin
+
+#### All Hyperparameters
+<details><summary>Click to expand</summary>
+
+- `overwrite_output_dir`: False
+- `do_predict`: False
+- `eval_strategy`: no
+- `prediction_loss_only`: True
+- `per_device_train_batch_size`: 16
+- `per_device_eval_batch_size`: 16
+- `per_gpu_train_batch_size`: None
+- `per_gpu_eval_batch_size`: None
+- `gradient_accumulation_steps`: 1
+- `eval_accumulation_steps`: None
+- `torch_empty_cache_steps`: None
+- `learning_rate`: 5e-05
+- `weight_decay`: 0.0
+- `adam_beta1`: 0.9
+- `adam_beta2`: 0.999
+- `adam_epsilon`: 1e-08
+- `max_grad_norm`: 1
+- `num_train_epochs`: 3
+- `max_steps`: -1
+- `lr_scheduler_type`: linear
+- `lr_scheduler_kwargs`: {}
+- `warmup_ratio`: 0.0
+- `warmup_steps`: 0
+- `log_level`: passive
+- `log_level_replica`: warning
+- `log_on_each_node`: True
+- `logging_nan_inf_filter`: True
+- `save_safetensors`: True
+- `save_on_each_node`: False
+- `save_only_model`: False
+- `restore_callback_states_from_checkpoint`: False
+- `no_cuda`: False
+- `use_cpu`: False
+- `use_mps_device`: False
+- `seed`: 42
+- `data_seed`: None
+- `jit_mode_eval`: False
+- `use_ipex`: False
+- `bf16`: False
+- `fp16`: False
+- `fp16_opt_level`: O1
+- `half_precision_backend`: auto
+- `bf16_full_eval`: False
+- `fp16_full_eval`: False
+- `tf32`: None
+- `local_rank`: 0
+- `ddp_backend`: None
+- `tpu_num_cores`: None
+- `tpu_metrics_debug`: False
+- `debug`: []
+- `dataloader_drop_last`: False
+- `dataloader_num_workers`: 0
+- `dataloader_prefetch_factor`: None
+- `past_index`: -1
+- `disable_tqdm`: False
+- `remove_unused_columns`: True
+- `label_names`: None
+- `load_best_model_at_end`: False
+- `ignore_data_skip`: False
+- `fsdp`: []
+- `fsdp_min_num_params`: 0
+- `fsdp_config`: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
+- `fsdp_transformer_layer_cls_to_wrap`: None
+- `accelerator_config`: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
+- `deepspeed`: None
+- `label_smoothing_factor`: 0.0
+- `optim`: adamw_torch
+- `optim_args`: None
+- `adafactor`: False
+- `group_by_length`: False
+- `length_column_name`: length
+- `ddp_find_unused_parameters`: None
+- `ddp_bucket_cap_mb`: None
+- `ddp_broadcast_buffers`: False
+- `dataloader_pin_memory`: True
+- `dataloader_persistent_workers`: False
+- `skip_memory_metrics`: True
+- `use_legacy_prediction_loop`: False
+- `push_to_hub`: False
+- `resume_from_checkpoint`: None
+- `hub_model_id`: None
+- `hub_strategy`: every_save
+- `hub_private_repo`: False
+- `hub_always_push`: False
+- `gradient_checkpointing`: False
+- `gradient_checkpointing_kwargs`: None
+- `include_inputs_for_metrics`: False
+- `eval_do_concat_batches`: True
+- `fp16_backend`: auto
+- `push_to_hub_model_id`: None
+- `push_to_hub_organization`: None
+- `mp_parameters`: 
+- `auto_find_batch_size`: False
+- `full_determinism`: False
+- `torchdynamo`: None
+- `ray_scope`: last
+- `ddp_timeout`: 1800
+- `torch_compile`: False
+- `torch_compile_backend`: None
+- `torch_compile_mode`: None
+- `dispatch_batches`: None
+- `split_batches`: None
+- `include_tokens_per_second`: False
+- `include_num_input_tokens_seen`: False
+- `neftune_noise_alpha`: None
+- `optim_target_modules`: None
+- `batch_eval_metrics`: False
+- `eval_on_start`: False
+- `eval_use_gather_object`: False
+- `batch_sampler`: batch_sampler
+- `multi_dataset_batch_sampler`: round_robin
+
+</details>
+
+### Framework Versions
+- Python: 3.11.5
+- Sentence Transformers: 3.1.1
+- Transformers: 4.44.2
+- PyTorch: 2.4.1+cpu
+- Accelerate: 0.34.2
+- Datasets: 3.0.0
+- Tokenizers: 0.19.1
+
+## Citation
+
+### BibTeX
+
+#### Sentence Transformers
+```bibtex
+@inproceedings{reimers-2019-sentence-bert,
+    title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
+    author = "Reimers, Nils and Gurevych, Iryna",
+    booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
+    month = "11",
+    year = "2019",
+    publisher = "Association for Computational Linguistics",
+    url = "https://arxiv.org/abs/1908.10084",
+}
+```
+
+#### MultipleNegativesRankingLoss
+```bibtex
+@misc{henderson2017efficient,
+    title={Efficient Natural Language Response Suggestion for Smart Reply},
+    author={Matthew Henderson and Rami Al-Rfou and Brian Strope and Yun-hsuan Sung and Laszlo Lukacs and Ruiqi Guo and Sanjiv Kumar and Balint Miklos and Ray Kurzweil},
+    year={2017},
+    eprint={1705.00652},
+    archivePrefix={arXiv},
+    primaryClass={cs.CL}
+}
+```
+
+<!--
+## Glossary
+
+*Clearly define terms in order to be accessible across audiences.*
+-->
+
+<!--
+## Model Card Authors
+
+*Lists the people who create the model card, providing recognition and accountability for the detailed work that goes into its construction.*
+-->
+
+<!--
+## Model Card Contact
+
+*Provides a way for people who have updates to the Model Card, suggestions, or questions, to contact the Model Card authors.*
+-->

fine_tuned_embedding_model/config.json

@@ -0,0 +1,26 @@
+{
+  "_name_or_path": "sentence-transformers/all-MiniLM-L6-v2",
+  "architectures": [
+    "BertModel"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "classifier_dropout": null,
+  "gradient_checkpointing": false,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 384,
+  "initializer_range": 0.02,
+  "intermediate_size": 1536,
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 6,
+  "pad_token_id": 0,
+  "position_embedding_type": "absolute",
+  "torch_dtype": "float32",
+  "transformers_version": "4.44.2",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 30522
+}

fine_tuned_embedding_model/config_sentence_transformers.json

@@ -0,0 +1,10 @@
+{
+  "__version__": {
+    "sentence_transformers": "3.1.1",
+    "transformers": "4.44.2",
+    "pytorch": "2.4.1+cpu"
+  },
+  "prompts": {},
+  "default_prompt_name": null,
+  "similarity_fn_name": null
+}

fine_tuned_embedding_model/modules.json

@@ -0,0 +1,20 @@
+[
+  {
+    "idx": 0,
+    "name": "0",
+    "path": "",
+    "type": "sentence_transformers.models.Transformer"
+  },
+  {
+    "idx": 1,
+    "name": "1",
+    "path": "1_Pooling",
+    "type": "sentence_transformers.models.Pooling"
+  },
+  {
+    "idx": 2,
+    "name": "2",
+    "path": "2_Normalize",
+    "type": "sentence_transformers.models.Normalize"
+  }
+]

fine_tuned_embedding_model/sentence_bert_config.json

@@ -0,0 +1,4 @@
+{
+  "max_seq_length": 256,
+  "do_lower_case": false
+}

fine_tuned_embedding_model/special_tokens_map.json

@@ -0,0 +1,37 @@
+{
+  "cls_token": {
+    "content": "[CLS]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "mask_token": {
+    "content": "[MASK]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "[PAD]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "sep_token": {
+    "content": "[SEP]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "[UNK]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

fine_tuned_embedding_model/tokenizer_config.json

@@ -0,0 +1,64 @@
+{
+  "added_tokens_decoder": {
+    "0": {
+      "content": "[PAD]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "100": {
+      "content": "[UNK]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "101": {
+      "content": "[CLS]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "102": {
+      "content": "[SEP]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "103": {
+      "content": "[MASK]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "clean_up_tokenization_spaces": true,
+  "cls_token": "[CLS]",
+  "do_basic_tokenize": true,
+  "do_lower_case": true,
+  "mask_token": "[MASK]",
+  "max_length": 128,
+  "model_max_length": 256,
+  "never_split": null,
+  "pad_to_multiple_of": null,
+  "pad_token": "[PAD]",
+  "pad_token_type_id": 0,
+  "padding_side": "right",
+  "sep_token": "[SEP]",
+  "stride": 0,
+  "strip_accents": null,
+  "tokenize_chinese_chars": true,
+  "tokenizer_class": "BertTokenizer",
+  "truncation_side": "right",
+  "truncation_strategy": "longest_first",
+  "unk_token": "[UNK]"
+}

ragas_finetune_eval/eval_config.py

@@ -0,0 +1,33 @@
+import os
+from pathlib import Path
+
+# Path to .env file
+ENV_PATH = Path(__file__).parent.parent / '.env'
+
+# Configuration settings for evaluation
+
+# Model and data paths
+FINE_TUNED_MODEL_PATH = "svb01/fine-tuned-embedding-model"
+TRAINING_DATA_PATH = "../resources/NIST.AI.600-1.pdf"  # Adjust this path if needed
+
+# RAG settings
+RETRIEVER_K = 6
+LLM_MODEL = "gpt-3.5-turbo"
+LLM_TEMPERATURE = 0
+
+# Evaluation settings
+SAMPLE_QUESTIONS = [
+    "What are the main objectives of the EU AI Act?",
+    "How does the Act define high-risk AI systems?",
+    "What are the transparency requirements for AI systems?",
+    "How does the Act address AI in the workplace?"
+]
+
+# RAGAS metrics
+RAGAS_METRICS = ["context_precision", "faithfulness", "answer_relevancy"]
+
+# Chunk size for text splitting
+CHUNK_SIZE = 750
+CHUNK_OVERLAP = 20
+
+BASE_MODEL_NAME = "sentence-transformers/all-MiniLM-L6-v2"

ragas_finetune_eval/eval_data_loader.py

@@ -0,0 +1,18 @@
+from langchain_community.document_loaders import PyPDFLoader
+from langchain_text_splitters import RecursiveCharacterTextSplitter
+from eval_config import CHUNK_SIZE, CHUNK_OVERLAP
+
+def load_training_documents(file_path):
+    loader = PyPDFLoader(file_path)
+    data = loader.load()
+    
+    text_splitter = RecursiveCharacterTextSplitter(
+        chunk_size=CHUNK_SIZE,
+        chunk_overlap=CHUNK_OVERLAP,
+        length_function=len
+    )
+    
+    return text_splitter.split_documents(data)
+
+def load_sample_questions(questions):
+    return questions

ragas_finetune_eval/eval_env_setup.py

@@ -0,0 +1,15 @@
+import os
+from dotenv import load_dotenv
+from eval_config import ENV_PATH
+
+def load_env_variables():
+    if os.path.exists(ENV_PATH):
+        load_dotenv(dotenv_path=ENV_PATH)
+        if os.getenv("OPENAI_API_KEY"):
+            print("OpenAI API key loaded successfully.")
+        else:
+            print("OpenAI API key not found in .env file.")
+    else:
+        print(f".env file not found at {ENV_PATH}")
+
+    # You can add more environment variable checks here if needed

ragas_finetune_eval/eval_main.py

@@ -0,0 +1,37 @@
+from eval_config import *
+from eval_env_setup import load_env_variables
+from eval_data_loader import load_training_documents, load_sample_questions
+from eval_rag_setup import setup_rag_pipeline
+from eval_rag_tester import test_rag_pipeline
+from eval_ragas import run_ragas_evaluation, prepare_ragas_data
+
+def main():
+    # Load environment variables
+    load_env_variables()
+
+    # Load data
+    training_documents = load_training_documents(TRAINING_DATA_PATH)
+    sample_questions = load_sample_questions(SAMPLE_QUESTIONS)
+
+    # Setup RAG pipeline
+    rag_chain, retriever = setup_rag_pipeline(
+        training_documents, 
+        FINE_TUNED_MODEL_PATH,
+        BASE_MODEL_NAME,
+        RETRIEVER_K, 
+        LLM_MODEL, 
+        LLM_TEMPERATURE
+    )
+
+    # Test RAG pipeline
+    test_results = test_rag_pipeline(rag_chain, sample_questions)
+
+    # Prepare and run RAGAS evaluation
+    ragas_data = prepare_ragas_data(sample_questions, retriever, rag_chain)
+    ragas_results = run_ragas_evaluation(ragas_data, RAGAS_METRICS)
+
+    print("RAGAS Evaluation Results:")
+    print(ragas_results)
+
+if __name__ == "__main__":
+    main()

ragas_finetune_eval/eval_rag_setup.py

@@ -0,0 +1,50 @@
+import torch
+from langchain_community.vectorstores import FAISS
+from langchain_core.prompts import ChatPromptTemplate
+from langchain_openai import ChatOpenAI
+from sentence_transformers import SentenceTransformer
+from langchain_huggingface import HuggingFaceEmbeddings
+from operator import itemgetter
+from langchain_core.output_parsers import StrOutputParser
+from langchain_core.runnables import RunnablePassthrough
+from huggingface_hub import hf_hub_download, list_repo_files
+import os
+from dotenv import load_dotenv
+from safetensors import safe_open
+
+def setup_rag_pipeline(training_documents, model_path, base_model_name, retriever_k, llm_model, llm_temperature):
+    # Load environment variables
+    load_dotenv()
+    hf_token = os.getenv('HF_TOKEN')
+
+    # Load the fine-tuned model directly
+    fine_tuned_model = SentenceTransformer(model_path, use_auth_token=hf_token)
+    
+    # Create embeddings using the fine-tuned model
+    fine_tuned_embeddings = HuggingFaceEmbeddings(model_name=model_path, model_kwargs={'device': 'cpu'})
+
+    vectorstore = FAISS.from_documents(training_documents, fine_tuned_embeddings)
+    retriever = vectorstore.as_retriever(search_kwargs={"k": retriever_k})
+
+    RAG_PROMPT = """
+    You are an AI assistant specializing in the EU AI Act. Given the context and question below, provide a concise and accurate answer. If the information is not in the context, state that you don't have enough information to answer.
+
+    Context:
+    {context}
+
+    Question:
+    {question}
+
+    Answer:
+    """
+    rag_prompt_template = ChatPromptTemplate.from_template(RAG_PROMPT)
+
+    rag_llm = ChatOpenAI(model=llm_model, temperature=llm_temperature)
+
+    rag_chain = (
+        {"context": itemgetter("question") | retriever, "question": itemgetter("question")}
+        | RunnablePassthrough.assign(context=itemgetter("context"))
+        | {"response": rag_prompt_template | rag_llm | StrOutputParser(), "context": itemgetter("context")}
+    )
+
+    return rag_chain, retriever

ragas_finetune_eval/eval_rag_tester.py

@@ -0,0 +1,11 @@
+def test_rag_pipeline(rag_chain, questions):
+    results = []
+    for question in questions:
+        response = rag_chain.invoke({"question": question})
+        results.append({
+            "question": question,
+            "answer": response['response']
+        })
+        print(f"Question: {question}")
+        print(f"Answer: {response['response']}\n")
+    return results

ragas_finetune_eval/eval_ragas.py

@@ -0,0 +1,18 @@
+from ragas import evaluate
+from datasets import Dataset
+
+def run_ragas_evaluation(test_data, metrics):
+    test_dataset = Dataset.from_list(test_data)
+    result = evaluate(test_dataset, metrics=metrics)
+    return result
+
+def prepare_ragas_data(questions, retriever, rag_chain):
+    test_data = [
+        {
+            "question": q,
+            "contexts": [c.page_content for c in retriever.get_relevant_documents(q)],
+            "answer": rag_chain.invoke({"question": q})["response"]
+        }
+        for q in questions
+    ]
+    return test_data