Added Fidelity Edition GGUF quantizations (fp16, q6_k, q8_0)

.gitattributes

@@ -37,3 +37,6 @@ rosa_confusion_matrix_grid.png filter=lfs diff=lfs merge=lfs -text
 assets/rosa_training.png filter=lfs diff=lfs merge=lfs -text
 examples/lemonhope.png filter=lfs diff=lfs merge=lfs -text
 assets/rosa_confusion_matrix_grid.png filter=lfs diff=lfs merge=lfs -text
+quantized/Rosa-V1-fp16.gguf filter=lfs diff=lfs merge=lfs -text
+quantized/Rosa-V1-q6_k.gguf filter=lfs diff=lfs merge=lfs -text
+quantized/Rosa-V1-q8_0.gguf filter=lfs diff=lfs merge=lfs -text

quantized/ETHICS_VOW.md

@@ -0,0 +1,10 @@
+# Ethical Statement for Rosa V1 — Fidelity Edition
+
+Rosa was created to advance emotionally aware and ethically grounded AI.
+
+Any derivative work must:
+1. Maintain attribution to **Willinton Triana Cardona**. 
+2. Preserve ethical framing: No manipulative, deceptive, or harmful applications. 
+3. Respect the **CreativeML OpenRAIL-M** license. 
+
+Derivatives that remove Rosa’s emotional intelligence or ethical guardrails violate the intent of this project.

quantized/README.md

@@ -0,0 +1,66 @@
+base_model: willt-dc/Rosa-V1
+datasets:
+
+- go_emotions
+  language: en
+  library_name: transformers
+  license: creativeml-openrail-m
+  model_creator: Willinton Triana Cardona
+  model_name: Rosa-V1
+  model_type: bert
+  quantized_by: Willinton Triana Cardona
+  quantized_models:
+- name: Rosa-V1-fp16
+  path: quantized/Rosa-V1-fp16.gguf
+  format: GGUF
+  description: Full-precision baseline
+- name: Rosa-V1-q8_0
+  path: quantized/Rosa-V1-q8_0.gguf
+  format: GGUF
+  description: Near-lossless 8-bit quantization
+- name: Rosa-V1-q6_k
+  path: quantized/Rosa-V1-q6_k.gguf
+  format: GGUF
+  description: Balanced 6-bit quantization for portability
+  tags:
+- emotion-classification
+- multilabel
+- bert
+- goemotions
+- affective-computing
+- psychology
+- NLP
+- embeddings
+- symbolic-ai
+- poetic-ai
+- quantized
+- gguf
+
+# Rosa V1 — Fidelity Edition (GGUF)
+
+**Author:** Willinton Triana Cardona
+**License:** CreativeML OpenRAIL-M
+
+## About
+
+Rosa is an AI model designed for **emotionally intelligent and ethically aligned NLP**, created by Willinton Triana Cardona.  
+This **Fidelity Edition** preserves emotional nuance through carefully chosen quantizations.
+
+## Quantizations
+
+- **Rosa-V1-fp16.gguf** — Full-precision baseline.
+- **Rosa-V1-q8_0.gguf** — Near-lossless, smaller size.
+- **Rosa-V1-q6_k.gguf** — Best balance between quality & deployability.
+
+Lower-bit quantizations (Q2–Q4) are **not included** due to unacceptable loss of semantic & emotional fidelity.
+
+## Ethical Use
+
+This model is licensed under **CreativeML OpenRAIL-M**:
+
+- No manipulative, harmful, or deceptive uses.
+- Attribution to the author is required.
+
+_"Rosa was built to reflect emotional depth and ethical intelligence.
+This Fidelity Edition carries her voice intact."_
+— Willinton Triana Cardona

quantized/Rosa-V1-fp16.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6244065594592c044e3a7d30f69f96459d0ae895722d972dc52ce2151e075943
+size 219498976

quantized/Rosa-V1-q6_k.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6a5d521f5b255662961769f3bd92a47570efcdb6163ba652c8134276a6aa5d0c
+size 91624352

quantized/Rosa-V1-q8_0.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:022ecf230db78c5fb8c537372282de75a5f11a1cae9c724838fb1986086abd95
+size 117876736