feat: Add granite4:micro LoRA files.

Signed-off-by: Khoi-Nguyen Tran <kndtran@ibm.com>

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+*.gguf filter=lfs diff=lfs merge=lfs -text

answer_relevance_classifier/granite4_micro/lora/Lora-q8_0.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:05ab02c60a86272867f2912a7e5be3938d66a2e1c21dd82dfb8101d9a3672a32
+size 15335328

answer_relevance_classifier/granite4_micro/lora/Modelfile

@@ -0,0 +1,2 @@
+FROM granite4:micro
+ADAPTER /proj/dmfexp/8cc/kndtran/huggingface/ibm-granite/granite-lib-rag-r1.0/answer_relevance_classifier/granite4_micro/lora/Lora-q8_0.gguf

answer_relevance_classifier/granite4_micro/lora/io.yaml

@@ -0,0 +1,46 @@
+docs_as_message: roles
+instruction: answer_relevance
+model: null
+parameters:
+  max_completion_tokens: 1024
+  response_format:
+    properties:
+      answer_relevance_analysis:
+        title: Answer Relevance Analysis
+        type: string
+      answer_relevance_category:
+        enum:
+        - Pertinent
+        - Pertinent with relevant extra
+        - Excessive unnecessary information
+        - Unduly restrictive
+        - Too vague or generic
+        - Contextual misalignment
+        - Misinterpreted inquiry
+        - No attempt
+        title: Answer Relevance Category
+        type: string
+      answer_relevance_judgment:
+        title: Answer Relevance Judgment
+        type: boolean
+    required:
+    - answer_relevance_analysis
+    - answer_relevance_category
+    - answer_relevance_judgment
+    title: AnswerRelevanceRawOutput
+    type: object
+response_format: null
+sentence_boundaries: null
+transformations:
+- categories_to_values:
+    false: 0.0
+    true: 1.0
+  input_path:
+  - answer_relevance_judgment
+  type: likelihood
+- input_path: []
+  retained_fields:
+    answer_relevance_analysis: answer_relevance_analysis
+    answer_relevance_category: answer_relevance_category
+    answer_relevance_judgment: answer_relevance_likelihood
+  type: project

answer_relevance_rewriter/granite4_micro/lora/Lora-q8_0.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e42563c6d0af2650fbe62373614493e855bce43dc82e4862ce1d5a27f413818e
+size 15335328

answer_relevance_rewriter/granite4_micro/lora/Modelfile

@@ -0,0 +1,2 @@
+FROM granite4:micro
+ADAPTER /proj/dmfexp/8cc/kndtran/huggingface/ibm-granite/granite-lib-rag-r1.0/answer_relevance_rewriter/granite4_micro/lora/Lora-q8_0.gguf

answer_relevance_rewriter/granite4_micro/lora/io.yaml

@@ -0,0 +1,28 @@
+docs_as_message: roles
+instruction: "Rewrite the response for relevance.\nThe last assistant response is\
+  \ considered not fully relevant to the last user inquiry due to {answer_relevance_category}:\
+  \ {answer_relevance_analysis}\nDecide if you agree with this assessment, then act\
+  \ according to the following instructions: \nIf you disagree with the assessment,\
+  \ provide a verbatim copy of the original response.  DO NOT attempt to correct any\
+  \ other perceived defects in the response. \nIf you agree with the assessment, provide\
+  \ an updated response that no longer fit the label {answer_relevance_category},\
+  \ by {correction_method}.  Your response should be entirely based on the provided\
+  \ documents and should not rely on other prior knowledge. Your response should be\
+  \ suitable to be directly provided to the user.  It should NOT contain meta information\
+  \ regarding the original response, its assessment, or this instruction.  The user\
+  \ does not see any of these.  Your response is the only response they will see to\
+  \ their inquiry, in place of the original response.\n"
+model: null
+parameters:
+  max_completion_tokens: 1024
+  response_format:
+    properties:
+      answer_relevance_rewrite:
+        title: Rewritten answer
+        type: string
+    required:
+    - answer_relevance_rewrite
+    type: object
+response_format: null
+sentence_boundaries: null
+transformations: null

answerability/granite4_micro/alora/io.yaml

@@ -0,0 +1,21 @@
+docs_as_message: roles
+instruction: null
+model: null
+parameters:
+  max_completion_tokens: 6
+  response_format:
+    enum:
+    - answerable
+    - unanswerable
+    type: string
+response_format: null
+sentence_boundaries: null
+transformations:
+- categories_to_values:
+    answerable: 1.0
+    unanswerable: 0.0
+  input_path: []
+  type: likelihood
+- field_name: answerability_likelihood
+  input_path: []
+  type: nest

answerability/granite4_micro/lora/Lora-q8_0.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:88a80717c357a5165b249ba0350b00e09605597db74686430d3c924ca36c9184
+size 15335776

answerability/granite4_micro/lora/Modelfile

@@ -0,0 +1,2 @@
+FROM granite4:micro
+ADAPTER /proj/dmfexp/8cc/kndtran/huggingface/ibm-granite/granite-lib-rag-r1.0/answerability/granite4_micro/lora/Lora-q8_0.gguf

answerability/granite4_micro/lora/io.yaml

@@ -0,0 +1,21 @@
+docs_as_message: roles
+instruction: null
+model: null
+parameters:
+  max_completion_tokens: 6
+  response_format:
+    enum:
+    - answerable
+    - unanswerable
+    type: string
+response_format: null
+sentence_boundaries: null
+transformations:
+- categories_to_values:
+    answerable: 1.0
+    unanswerable: 0.0
+  input_path: []
+  type: likelihood
+- field_name: answerability_likelihood
+  input_path: []
+  type: nest

citations/granite4_micro/lora/Lora-q8_0.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a1b8a549b23b204e4689e29cf957bd2cdc483e034d5494dfd94083e16324e3c4
+size 14849568

citations/granite4_micro/lora/Modelfile

@@ -0,0 +1,2 @@
+FROM granite4:micro
+ADAPTER /proj/dmfexp/8cc/kndtran/huggingface/ibm-granite/granite-lib-rag-r1.0/citations/granite4_micro/lora/Lora-q8_0.gguf

citations/granite4_micro/lora/io.yaml

@@ -0,0 +1,86 @@
+docs_as_message: roles
+instruction: 'Split the last assistant response into individual sentences.  For each
+  sentence in the response, identify the statement IDs from the below  documents that
+  it references. Ensure that your output includes all response  sentence IDs, and
+  for each response sentence ID, provide the list of corresponding  referring document
+  sentence IDs. The output must be a json structure.
+
+  '
+model: null
+parameters:
+  max_completion_tokens: 4096
+  response_format:
+    $defs:
+      _MODEL_OUTPUT_ENTRY:
+        properties:
+          c:
+            items:
+              minimum: 0
+              type: integer
+            title: C
+            type: array
+          r:
+            minimum: 0
+            title: R
+            type: integer
+        required:
+        - r
+        - c
+        title: _MODEL_OUTPUT_ENTRY
+        type: object
+    items:
+      $ref: '#/$defs/_MODEL_OUTPUT_ENTRY'
+    title: _MODEL_OUTPUT
+    type: array
+response_format: null
+sentence_boundaries:
+  documents: c
+  last_message: r
+transformations:
+- input_path: []
+  target_field: c
+  type: explode
+- input_path: []
+  target_fields:
+  - r
+  - c
+  type: drop_duplicates
+- input_path:
+  - null
+  - r
+  output_names:
+    begin: response_begin
+    end: response_end
+    text: response_text
+  source: last_message
+  type: decode_sentences
+- input_path:
+  - null
+  - c
+  output_names:
+    begin: citation_begin
+    document_id: citation_doc_id
+    end: citation_end
+    text: citation_text
+  source: documents
+  type: decode_sentences
+- input_path: []
+  retained_fields:
+  - response_begin
+  - response_end
+  - response_text
+  - citation_doc_id
+  - citation_begin
+  - citation_end
+  - citation_text
+  type: project
+- begin_field: citation_begin
+  end_field: citation_end
+  group_fields:
+  - response_begin
+  - response_end
+  - response_text
+  - citation_doc_id
+  input_path: []
+  text_field: citation_text
+  type: merge_spans

context_relevance/granite4_micro/lora/Lora-q8_0.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ebafe44ff0e904b8dee4f909636f44631b65f912b17528983ef2e6cf1043bf5e
+size 15335776

context_relevance/granite4_micro/lora/Modelfile

@@ -0,0 +1,2 @@
+FROM granite4:micro
+ADAPTER /proj/dmfexp/8cc/kndtran/huggingface/ibm-granite/granite-lib-rag-r1.0/context_relevance/granite4_micro/lora/Lora-q8_0.gguf

context_relevance/granite4_micro/lora/io.yaml

@@ -0,0 +1,29 @@
+docs_as_message: roles
+instruction: 'DOCUMENT: {document_content}
+
+  '
+model: null
+parameters:
+  response_format:
+    properties:
+      context_relevance:
+        description: Context relevancy judgment.
+        enum:
+        - relevant
+        - irrelevant
+        - partially relevant
+        type: string
+    required:
+    - context_relevance
+    title: ContextRelevanceOutput
+    type: object
+response_format: null
+sentence_boundaries: null
+transformations:
+- categories_to_values:
+    irrelevant: 0.0
+    partially relevant: 0.5
+    relevant: 1.0
+  input_path:
+  - context_relevance
+  type: likelihood

hallucination_detection/granite4_micro/lora/Lora-q8_0.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:04540ac6aff27b31fe3ab01e1e546c83c4605b35d9d83258a3984ca31d3cd1db
+size 14849568

hallucination_detection/granite4_micro/lora/Modelfile

@@ -0,0 +1,2 @@
+FROM granite4:micro
+ADAPTER /proj/dmfexp/8cc/kndtran/huggingface/ibm-granite/granite-lib-rag-r1.0/hallucination_detection/granite4_micro/lora/Lora-q8_0.gguf

hallucination_detection/granite4_micro/lora/io.yaml

@@ -0,0 +1,69 @@
+docs_as_message: roles
+instruction: 'Split the last assistant response into individual sentences. For each
+  sentence in the last assistant response, identify the faithfulness by comparing
+  with the provided documents and generate the faithfulness reasoning and faithfulness
+  decision. Ensure that your output includes all response sentence IDs, and for each
+  response sentence ID, provide the corresponding faithfulness reasoning and faithfulness
+  decision. The output must be a json structure.
+
+  '
+model: null
+parameters:
+  max_completion_tokens: 4096
+  response_format:
+    $defs:
+      HallucinationOutputEntry:
+        properties:
+          e:
+            title: Reasoning
+            type: string
+          f:
+            enum:
+            - faithful
+            - partial
+            - unfaithful
+            title: Is Faithful
+            type: string
+          r:
+            minimum: 0
+            title: Sentence Num
+            type: integer
+        required:
+        - r
+        - e
+        - f
+        title: HallucinationOutputEntry
+        type: object
+    items:
+      $ref: '#/$defs/HallucinationOutputEntry'
+    title: HallucinationOutput
+    type: array
+response_format: null
+sentence_boundaries:
+  last_message: i
+transformations:
+- categories_to_values:
+    faithful: 1.0
+    partial: 0.5
+    unfaithful: 0.0
+  input_path:
+  - null
+  - f
+  type: likelihood
+- input_path:
+  - null
+  - r
+  output_names:
+    begin: response_begin
+    end: response_end
+    text: response_text
+  source: last_message
+  type: decode_sentences
+- input_path: []
+  retained_fields:
+    e: explanation
+    f: faithfulness_likelihood
+    response_begin: response_begin
+    response_end: response_end
+    response_text: response_text
+  type: project

query_rewrite/granite4_micro/lora/Lora-q8_0.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d34e33d7849c07a323d94bc75aff2b0aa9dd333ddb4826647d5b6f46fd046b8c
+size 15335776

query_rewrite/granite4_micro/lora/Modelfile

@@ -0,0 +1,2 @@
+FROM granite4:micro
+ADAPTER /proj/dmfexp/8cc/kndtran/huggingface/ibm-granite/granite-lib-rag-r1.0/query_rewrite/granite4_micro/lora/Lora-q8_0.gguf

query_rewrite/granite4_micro/lora/io.yaml

@@ -0,0 +1,17 @@
+docs_as_message: roles
+instruction: null
+model: null
+parameters:
+  max_completion_tokens: 1024
+  response_format:
+    properties:
+      rewritten_question:
+        title: Rewritten Question
+        type: string
+    required:
+    - rewritten_question
+    title: QueryRewriteOutput
+    type: object
+response_format: null
+sentence_boundaries: false
+transformations: null