initial upload

.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
+tokenizer.json filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,3 +1,766 @@
 ---
-license: apache-2.0
+base_model: []
+library_name: transformers
+tags:
+- mergekit
+- merge
+
 ---
+# Glisten-31B-v0.1.4
+
+This is a merge of pre-trained language models created using [mergekit](https://github.com/cg123/mergekit).
+
+## Merge Details
+### Merge Method
+
+This model was merged using the [DELLA](https://arxiv.org/abs/2406.11617) merge method using /home/llm/mergequant/models/BASE/gemma-4-31B-it as a base.
+
+### Models Merged
+
+The following models were included in the merge:
+* /home/llm/mergequant/models/BASE/Artemis-31B-v1h
+* /home/llm/mergequant/models/BASE/zerofata-G4-MeroMero-31B
+* /home/llm/mergequant/models/BASE/gemma-4-Ortenzya-The-Creative-Wordsmith-31B-it-uncensored-heretic
+
+### Configuration
+
+The following YAML configuration was used to produce this model:
+
+```yaml
+models:
+  - model: /home/llm/mergequant/models/BASE/gemma-4-31B-it
+    # base model — no parameters needed
+
+  - model: /home/llm/mergequant/models/BASE/gemma-4-Ortenzya-The-Creative-Wordsmith-31B-it-uncensored-heretic
+    parameters:
+      weight:
+        # ── Attn spike suppression — surgical ───────────────────
+        # L5, L7, L12, L15 anomalously high in attn KL.
+        # Near-zero to prevent early context binding disruption.
+        - filter: "layers.5.self_attn"
+          value: 0.03
+        - filter: "layers.7.self_attn"
+          value: 0.03
+        - filter: "layers.12.self_attn"
+          value: 0.03        # L12: 0.187 KL — primary offender
+        - filter: "layers.15.self_attn"
+          value: 0.03
+
+        # ── Dead zone 48-59: excluded ────────────────────────────
+        # Zero KL divergence from base — nothing to contribute.
+        - filter: "layers.48.self_attn"
+          value: 0.00
+        - filter: "layers.48.mlp"
+          value: 0.00
+        - filter: "layers.49.self_attn"
+          value: 0.00
+        - filter: "layers.49.mlp"
+          value: 0.00
+        - filter: "layers.50.self_attn"
+          value: 0.00
+        - filter: "layers.50.mlp"
+          value: 0.00
+        - filter: "layers.51.self_attn"
+          value: 0.00
+        - filter: "layers.51.mlp"
+          value: 0.00
+        - filter: "layers.52.self_attn"
+          value: 0.00
+        - filter: "layers.52.mlp"
+          value: 0.00
+        - filter: "layers.53.self_attn"
+          value: 0.00
+        - filter: "layers.53.mlp"
+          value: 0.00
+        - filter: "layers.54.self_attn"
+          value: 0.00
+        - filter: "layers.54.mlp"
+          value: 0.00
+        - filter: "layers.55.self_attn"
+          value: 0.00
+        - filter: "layers.55.mlp"
+          value: 0.00
+        - filter: "layers.56.self_attn"
+          value: 0.00
+        - filter: "layers.56.mlp"
+          value: 0.00
+        - filter: "layers.57.self_attn"
+          value: 0.00
+        - filter: "layers.57.mlp"
+          value: 0.00
+        - filter: "layers.58.self_attn"
+          value: 0.00
+        - filter: "layers.58.mlp"
+          value: 0.00
+        - filter: "layers.59.self_attn"
+          value: 0.00
+        - filter: "layers.59.mlp"
+          value: 0.00
+
+        # ── MeroMero inflection zone 26-35 ──────────────────────
+        # Ortenzya minimized — protecting MeroMero's detail zone.
+        # Scaled down proportionally to hit 1.05 budget.
+        - filter: "layers.26.self_attn"
+          value: 0.09
+        - filter: "layers.27.self_attn"
+          value: 0.09
+        - filter: "layers.28.self_attn"
+          value: 0.09
+        - filter: "layers.29.self_attn"
+          value: 0.09
+        - filter: "layers.30.self_attn"
+          value: 0.09
+        - filter: "layers.31.self_attn"
+          value: 0.09
+        - filter: "layers.32.self_attn"
+          value: 0.09
+        - filter: "layers.33.self_attn"
+          value: 0.09
+        - filter: "layers.34.self_attn"
+          value: 0.09
+        - filter: "layers.35.self_attn"
+          value: 0.09
+        - filter: "layers.26.mlp"
+          value: 0.13
+        - filter: "layers.27.mlp"
+          value: 0.13
+        - filter: "layers.28.mlp"
+          value: 0.13
+        - filter: "layers.29.mlp"
+          value: 0.13
+        - filter: "layers.30.mlp"
+          value: 0.13
+        - filter: "layers.31.mlp"
+          value: 0.13
+        - filter: "layers.32.mlp"
+          value: 0.13
+        - filter: "layers.33.mlp"
+          value: 0.13
+        - filter: "layers.34.mlp"
+          value: 0.13
+        - filter: "layers.35.mlp"
+          value: 0.13
+
+        # ── Active zone 36-43 ───────────────────────────────────
+        # Ortenzya's genuine creative signal zone.
+        # Scaled to budget — preserves relative contribution
+        # while keeping total sum at 1.05.
+        - filter: "layers.36.self_attn"
+          value: 0.11
+        - filter: "layers.37.self_attn"
+          value: 0.11
+        - filter: "layers.38.self_attn"
+          value: 0.11
+        - filter: "layers.39.self_attn"
+          value: 0.11
+        - filter: "layers.40.self_attn"
+          value: 0.11
+        - filter: "layers.41.self_attn"
+          value: 0.11
+        - filter: "layers.42.self_attn"
+          value: 0.11
+        - filter: "layers.43.self_attn"
+          value: 0.11
+        - filter: "layers.36.mlp"
+          value: 0.23
+        - filter: "layers.37.mlp"
+          value: 0.23
+        - filter: "layers.38.mlp"
+          value: 0.23
+        - filter: "layers.39.mlp"
+          value: 0.23
+        - filter: "layers.40.mlp"
+          value: 0.23
+        - filter: "layers.41.mlp"
+          value: 0.23
+        - filter: "layers.42.mlp"
+          value: 0.23
+        - filter: "layers.43.mlp"
+          value: 0.23
+
+        # ── Active zone 44-47: Ortenzya MLP peak ────────────────
+        # Highest MLP KL in the model (up to 0.098 at L47).
+        # Slightly higher than 36-43 to respect the peak,
+        # still scaled to budget.
+        - filter: "layers.44.self_attn"
+          value: 0.11
+        - filter: "layers.45.self_attn"
+          value: 0.11
+        - filter: "layers.46.self_attn"
+          value: 0.11
+        - filter: "layers.47.self_attn"
+          value: 0.11
+        - filter: "layers.44.mlp"
+          value: 0.26
+        - filter: "layers.45.mlp"
+          value: 0.26
+        - filter: "layers.46.mlp"
+          value: 0.26
+        - filter: "layers.47.mlp"
+          value: 0.26
+
+        # ── Zone-wide fallbacks ──────────────────────────────────
+        # Covers early layers 0-25 (excluding spike layers above).
+        # Scaled to budget alongside MeroMero and Artemis early values.
+        - filter: "self_attn"
+          value: 0.14
+        - filter: "mlp"
+          value: 0.26
+        - value: 0.25
+      density: 0.80
+      epsilon: 0.099
+
+  - model: /home/llm/mergequant/models/BASE/zerofata-G4-MeroMero-31B
+    parameters:
+      weight:
+        # ── Early layers 0-25 ───────────────────────────────────
+        # Scaled down from 0.60/0.55 to hit budget alongside
+        # Artemis and Ortenzya early contributions.
+        - filter: "layers.0.self_attn"
+          value: 0.57
+        - filter: "layers.1.self_attn"
+          value: 0.57
+        - filter: "layers.2.self_attn"
+          value: 0.57
+        - filter: "layers.3.self_attn"
+          value: 0.57
+        - filter: "layers.4.self_attn"
+          value: 0.57
+        - filter: "layers.5.self_attn"
+          value: 0.57
+        - filter: "layers.6.self_attn"
+          value: 0.57
+        - filter: "layers.7.self_attn"
+          value: 0.57
+        - filter: "layers.8.self_attn"
+          value: 0.57
+        - filter: "layers.9.self_attn"
+          value: 0.57
+        - filter: "layers.10.self_attn"
+          value: 0.57
+        - filter: "layers.11.self_attn"
+          value: 0.57
+        - filter: "layers.12.self_attn"
+          value: 0.57
+        - filter: "layers.13.self_attn"
+          value: 0.57
+        - filter: "layers.14.self_attn"
+          value: 0.57
+        - filter: "layers.15.self_attn"
+          value: 0.57
+        - filter: "layers.16.self_attn"
+          value: 0.57
+        - filter: "layers.17.self_attn"
+          value: 0.57
+        - filter: "layers.18.self_attn"
+          value: 0.57
+        - filter: "layers.19.self_attn"
+          value: 0.57
+        - filter: "layers.20.self_attn"
+          value: 0.57
+        - filter: "layers.21.self_attn"
+          value: 0.57
+        - filter: "layers.22.self_attn"
+          value: 0.57
+        - filter: "layers.23.self_attn"
+          value: 0.57
+        - filter: "layers.24.self_attn"
+          value: 0.57
+        - filter: "layers.25.self_attn"
+          value: 0.57
+        - filter: "layers.0.mlp"
+          value: 0.48
+        - filter: "layers.1.mlp"
+          value: 0.48
+        - filter: "layers.2.mlp"
+          value: 0.48
+        - filter: "layers.3.mlp"
+          value: 0.48
+        - filter: "layers.4.mlp"
+          value: 0.48
+        - filter: "layers.5.mlp"
+          value: 0.48
+        - filter: "layers.6.mlp"
+          value: 0.48
+        - filter: "layers.7.mlp"
+          value: 0.48
+        - filter: "layers.8.mlp"
+          value: 0.48
+        - filter: "layers.9.mlp"
+          value: 0.48
+        - filter: "layers.10.mlp"
+          value: 0.48
+        - filter: "layers.11.mlp"
+          value: 0.48
+        - filter: "layers.12.mlp"
+          value: 0.48
+        - filter: "layers.13.mlp"
+          value: 0.48
+        - filter: "layers.14.mlp"
+          value: 0.48
+        - filter: "layers.15.mlp"
+          value: 0.48
+        - filter: "layers.16.mlp"
+          value: 0.48
+        - filter: "layers.17.mlp"
+          value: 0.48
+        - filter: "layers.18.mlp"
+          value: 0.48
+        - filter: "layers.19.mlp"
+          value: 0.48
+        - filter: "layers.20.mlp"
+          value: 0.48
+        - filter: "layers.21.mlp"
+          value: 0.48
+        - filter: "layers.22.mlp"
+          value: 0.48
+        - filter: "layers.23.mlp"
+          value: 0.48
+        - filter: "layers.24.mlp"
+          value: 0.48
+        - filter: "layers.25.mlp"
+          value: 0.48
+
+        # ── MeroMero inflection zone 26-35 ──────────────────────
+        # MeroMero's most important zone — scaled down minimally
+        # to preserve dominance while hitting budget.
+        - filter: "layers.26.self_attn"
+          value: 0.73
+        - filter: "layers.27.self_attn"
+          value: 0.73
+        - filter: "layers.28.self_attn"
+          value: 0.73
+        - filter: "layers.29.self_attn"
+          value: 0.73
+        - filter: "layers.30.self_attn"
+          value: 0.73
+        - filter: "layers.31.self_attn"
+          value: 0.73
+        - filter: "layers.32.self_attn"
+          value: 0.73
+        - filter: "layers.33.self_attn"
+          value: 0.73
+        - filter: "layers.34.self_attn"
+          value: 0.73
+        - filter: "layers.35.self_attn"
+          value: 0.73
+        - filter: "layers.26.mlp"
+          value: 0.74
+        - filter: "layers.27.mlp"
+          value: 0.74
+        - filter: "layers.28.mlp"
+          value: 0.74
+        - filter: "layers.29.mlp"
+          value: 0.74
+        - filter: "layers.30.mlp"
+          value: 0.74
+        - filter: "layers.31.mlp"
+          value: 0.74
+        - filter: "layers.32.mlp"
+          value: 0.74
+        - filter: "layers.33.mlp"
+          value: 0.74
+        - filter: "layers.34.mlp"
+          value: 0.74
+        - filter: "layers.35.mlp"
+          value: 0.74
+
+        # ── Sustained active zone 36-50 ─────────────────────────
+        # MeroMero attn leads for structural coherence.
+        # MLP pulled back to give Artemis room in this zone.
+        - filter: "layers.36.self_attn"
+          value: 0.63
+        - filter: "layers.37.self_attn"
+          value: 0.63
+        - filter: "layers.38.self_attn"
+          value: 0.63
+        - filter: "layers.39.self_attn"
+          value: 0.63
+        - filter: "layers.40.self_attn"
+          value: 0.63
+        - filter: "layers.41.self_attn"
+          value: 0.63
+        - filter: "layers.42.self_attn"
+          value: 0.63
+        - filter: "layers.43.self_attn"
+          value: 0.63
+        - filter: "layers.44.self_attn"
+          value: 0.63
+        - filter: "layers.45.self_attn"
+          value: 0.63
+        - filter: "layers.46.self_attn"
+          value: 0.63
+        - filter: "layers.47.self_attn"
+          value: 0.63
+        - filter: "layers.48.self_attn"
+          value: 0.70
+        - filter: "layers.49.self_attn"
+          value: 0.70
+        - filter: "layers.50.self_attn"
+          value: 0.70
+        - filter: "layers.36.mlp"
+          value: 0.31
+        - filter: "layers.37.mlp"
+          value: 0.31
+        - filter: "layers.38.mlp"
+          value: 0.31
+        - filter: "layers.39.mlp"
+          value: 0.31
+        - filter: "layers.40.mlp"
+          value: 0.31
+        - filter: "layers.41.mlp"
+          value: 0.31
+        - filter: "layers.42.mlp"
+          value: 0.31
+        - filter: "layers.43.mlp"
+          value: 0.31
+        - filter: "layers.44.mlp"
+          value: 0.30
+        - filter: "layers.45.mlp"
+          value: 0.30
+        - filter: "layers.46.mlp"
+          value: 0.30
+        - filter: "layers.47.mlp"
+          value: 0.30
+        - filter: "layers.48.mlp"
+          value: 0.40
+        - filter: "layers.49.mlp"
+          value: 0.40
+        - filter: "layers.50.mlp"
+          value: 0.40
+
+        # ── Output zone 51-59 ───────────────────────────────────
+        # Ortenzya is dead here — only MeroMero and Artemis.
+        # Budget already at 1.05 with previous two-model values.
+        - filter: "layers.51.self_attn"
+          value: 0.65
+        - filter: "layers.52.self_attn"
+          value: 0.65
+        - filter: "layers.53.self_attn"
+          value: 0.65
+        - filter: "layers.54.self_attn"
+          value: 0.65
+        - filter: "layers.55.self_attn"
+          value: 0.65
+        - filter: "layers.56.self_attn"
+          value: 0.65
+        - filter: "layers.57.self_attn"
+          value: 0.65
+        - filter: "layers.58.self_attn"
+          value: 0.65
+        - filter: "layers.59.self_attn"
+          value: 0.65
+        - filter: "layers.51.mlp"
+          value: 0.50
+        - filter: "layers.52.mlp"
+          value: 0.50
+        - filter: "layers.53.mlp"
+          value: 0.50
+        - filter: "layers.54.mlp"
+          value: 0.50
+        - filter: "layers.55.mlp"
+          value: 0.50
+        - filter: "layers.56.mlp"
+          value: 0.50
+        - filter: "layers.57.mlp"
+          value: 0.50
+        - filter: "layers.58.mlp"
+          value: 0.50
+        - filter: "layers.59.mlp"
+          value: 0.50
+
+        # ── Zone-wide fallbacks ─────────────────────────────────
+        - filter: "self_attn"
+          value: 0.57
+        - filter: "mlp"
+          value: 0.48
+        - value: 0.50
+      density: 0.95
+      epsilon: 0.099
+
+  - model: /home/llm/mergequant/models/BASE/Artemis-31B-v1h
+    parameters:
+      weight:
+        # ── Early layers 0-25 ───────────────────────────────────
+        # Scaled down from 0.35 to hit budget alongside
+        # MeroMero and Ortenzya early contributions.
+        - filter: "layers.0.self_attn"
+          value: 0.33
+        - filter: "layers.1.self_attn"
+          value: 0.33
+        - filter: "layers.2.self_attn"
+          value: 0.33
+        - filter: "layers.3.self_attn"
+          value: 0.33
+        - filter: "layers.4.self_attn"
+          value: 0.33
+        - filter: "layers.5.self_attn"
+          value: 0.33
+        - filter: "layers.6.self_attn"
+          value: 0.33
+        - filter: "layers.7.self_attn"
+          value: 0.33
+        - filter: "layers.8.self_attn"
+          value: 0.33
+        - filter: "layers.9.self_attn"
+          value: 0.33
+        - filter: "layers.10.self_attn"
+          value: 0.33
+        - filter: "layers.11.self_attn"
+          value: 0.33
+        - filter: "layers.12.self_attn"
+          value: 0.33
+        - filter: "layers.13.self_attn"
+          value: 0.33
+        - filter: "layers.14.self_attn"
+          value: 0.33
+        - filter: "layers.15.self_attn"
+          value: 0.33
+        - filter: "layers.16.self_attn"
+          value: 0.33
+        - filter: "layers.17.self_attn"
+          value: 0.33
+        - filter: "layers.18.self_attn"
+          value: 0.33
+        - filter: "layers.19.self_attn"
+          value: 0.33
+        - filter: "layers.20.self_attn"
+          value: 0.33
+        - filter: "layers.21.self_attn"
+          value: 0.33
+        - filter: "layers.22.self_attn"
+          value: 0.33
+        - filter: "layers.23.self_attn"
+          value: 0.33
+        - filter: "layers.24.self_attn"
+          value: 0.33
+        - filter: "layers.25.self_attn"
+          value: 0.33
+        - filter: "layers.0.mlp"
+          value: 0.31
+        - filter: "layers.1.mlp"
+          value: 0.31
+        - filter: "layers.2.mlp"
+          value: 0.31
+        - filter: "layers.3.mlp"
+          value: 0.31
+        - filter: "layers.4.mlp"
+          value: 0.31
+        - filter: "layers.5.mlp"
+          value: 0.31
+        - filter: "layers.6.mlp"
+          value: 0.31
+        - filter: "layers.7.mlp"
+          value: 0.31
+        - filter: "layers.8.mlp"
+          value: 0.31
+        - filter: "layers.9.mlp"
+          value: 0.31
+        - filter: "layers.10.mlp"
+          value: 0.31
+        - filter: "layers.11.mlp"
+          value: 0.31
+        - filter: "layers.12.mlp"
+          value: 0.31
+        - filter: "layers.13.mlp"
+          value: 0.31
+        - filter: "layers.14.mlp"
+          value: 0.31
+        - filter: "layers.15.mlp"
+          value: 0.31
+        - filter: "layers.16.mlp"
+          value: 0.31
+        - filter: "layers.17.mlp"
+          value: 0.31
+        - filter: "layers.18.mlp"
+          value: 0.31
+        - filter: "layers.19.mlp"
+          value: 0.31
+        - filter: "layers.20.mlp"
+          value: 0.31
+        - filter: "layers.21.mlp"
+          value: 0.31
+        - filter: "layers.22.mlp"
+          value: 0.31
+        - filter: "layers.23.mlp"
+          value: 0.31
+        - filter: "layers.24.mlp"
+          value: 0.31
+        - filter: "layers.25.mlp"
+          value: 0.31
+
+        # ── MeroMero inflection zone 26-35 ──────────────────────
+        # Artemis minimized to protect MeroMero's detail zone.
+        # Scaled proportionally to budget.
+        - filter: "layers.26.self_attn"
+          value: 0.23
+        - filter: "layers.27.self_attn"
+          value: 0.23
+        - filter: "layers.28.self_attn"
+          value: 0.23
+        - filter: "layers.29.self_attn"
+          value: 0.23
+        - filter: "layers.30.self_attn"
+          value: 0.23
+        - filter: "layers.31.self_attn"
+          value: 0.23
+        - filter: "layers.32.self_attn"
+          value: 0.23
+        - filter: "layers.33.self_attn"
+          value: 0.23
+        - filter: "layers.34.self_attn"
+          value: 0.23
+        - filter: "layers.35.self_attn"
+          value: 0.23
+        - filter: "layers.26.mlp"
+          value: 0.18
+        - filter: "layers.27.mlp"
+          value: 0.18
+        - filter: "layers.28.mlp"
+          value: 0.18
+        - filter: "layers.29.mlp"
+          value: 0.18
+        - filter: "layers.30.mlp"
+          value: 0.18
+        - filter: "layers.31.mlp"
+          value: 0.18
+        - filter: "layers.32.mlp"
+          value: 0.18
+        - filter: "layers.33.mlp"
+          value: 0.18
+        - filter: "layers.34.mlp"
+          value: 0.18
+        - filter: "layers.35.mlp"
+          value: 0.18
+
+        # ── Sustained active zone 36-47 ─────────────────────────
+        # Artemis MLP leads for creative voice.
+        # Scaled to budget — still dominant over MeroMero MLP here.
+        - filter: "layers.36.self_attn"
+          value: 0.31
+        - filter: "layers.37.self_attn"
+          value: 0.31
+        - filter: "layers.38.self_attn"
+          value: 0.31
+        - filter: "layers.39.self_attn"
+          value: 0.31
+        - filter: "layers.40.self_attn"
+          value: 0.31
+        - filter: "layers.41.self_attn"
+          value: 0.31
+        - filter: "layers.42.self_attn"
+          value: 0.31
+        - filter: "layers.43.self_attn"
+          value: 0.31
+        - filter: "layers.44.self_attn"
+          value: 0.31
+        - filter: "layers.45.self_attn"
+          value: 0.31
+        - filter: "layers.46.self_attn"
+          value: 0.31
+        - filter: "layers.47.self_attn"
+          value: 0.31
+        - filter: "layers.36.mlp"
+          value: 0.51
+        - filter: "layers.37.mlp"
+          value: 0.51
+        - filter: "layers.38.mlp"
+          value: 0.51
+        - filter: "layers.39.mlp"
+          value: 0.51
+        - filter: "layers.40.mlp"
+          value: 0.51
+        - filter: "layers.41.mlp"
+          value: 0.51
+        - filter: "layers.42.mlp"
+          value: 0.51
+        - filter: "layers.43.mlp"
+          value: 0.51
+        - filter: "layers.44.mlp"
+          value: 0.49
+        - filter: "layers.45.mlp"
+          value: 0.49
+        - filter: "layers.46.mlp"
+          value: 0.49
+        - filter: "layers.47.mlp"
+          value: 0.49
+
+        # ── Active zone 48-50 ───────────────────────────────────
+        # Ortenzya dead here — back to two-model budget of 1.05.
+        - filter: "layers.48.self_attn"
+          value: 0.35
+        - filter: "layers.49.self_attn"
+          value: 0.35
+        - filter: "layers.50.self_attn"
+          value: 0.35
+        - filter: "layers.48.mlp"
+          value: 0.65
+        - filter: "layers.49.mlp"
+          value: 0.65
+        - filter: "layers.50.mlp"
+          value: 0.65
+
+        # ── Output zone 51-59 ───────────────────────────────────
+        # Ortenzya dead — two-model budget.
+        # Artemis MLP nudged down from 0.60 to 0.55
+        # to bring output sum to 1.05.
+        - filter: "layers.51.self_attn"
+          value: 0.40
+        - filter: "layers.52.self_attn"
+          value: 0.40
+        - filter: "layers.53.self_attn"
+          value: 0.40
+        - filter: "layers.54.self_attn"
+          value: 0.40
+        - filter: "layers.55.self_attn"
+          value: 0.40
+        - filter: "layers.56.self_attn"
+          value: 0.40
+        - filter: "layers.57.self_attn"
+          value: 0.40
+        - filter: "layers.58.self_attn"
+          value: 0.40
+        - filter: "layers.59.self_attn"
+          value: 0.40
+        - filter: "layers.51.mlp"
+          value: 0.55
+        - filter: "layers.52.mlp"
+          value: 0.55
+        - filter: "layers.53.mlp"
+          value: 0.55
+        - filter: "layers.54.mlp"
+          value: 0.55
+        - filter: "layers.55.mlp"
+          value: 0.55
+        - filter: "layers.56.mlp"
+          value: 0.55
+        - filter: "layers.57.mlp"
+          value: 0.55
+        - filter: "layers.58.mlp"
+          value: 0.55
+        - filter: "layers.59.mlp"
+          value: 0.55
+
+        # ── Zone-wide fallbacks ─────────────────────────────────
+        - filter: "self_attn"
+          value: 0.33
+        - filter: "mlp"
+          value: 0.31
+        - value: 0.50
+      density: 0.85
+      epsilon: 0.099
+
+merge_method: della
+base_model: /home/llm/mergequant/models/BASE/gemma-4-31B-it
+
+parameters:
+  lambda: 1.0
+  normalize: false
+
+tokenizer:
+  source: base
+chat_template: auto
+dtype: float32
+out_dtype: bfloat16
+
+```

chat_template.jinja

@@ -0,0 +1,354 @@
+{%- macro format_parameters(properties, required, filter_keys=false) -%}
+    {%- set standard_keys = ['description', 'type', 'properties', 'required', 'nullable'] -%}
+    {%- set ns = namespace(found_first=false) -%}
+    {%- for key, value in properties | dictsort -%}
+        {%- set add_comma = false -%}
+        {%- if not filter_keys or key not in standard_keys -%}
+            {%- if ns.found_first %},{% endif -%}
+            {%- set ns.found_first = true -%}
+            {{ key }}:{
+            {%- if value['description'] -%}
+                description:<|"|>{{ value['description'] }}<|"|>
+                {%- set add_comma = true -%}
+            {%- endif -%}
+            {%- if value['type'] | upper == 'STRING' -%}
+                {%- if value['enum'] -%}
+                    {%- if add_comma %},{%- else -%} {%- set add_comma = true -%} {% endif -%}
+                    enum:{{ format_argument(value['enum']) }}
+                {%- endif -%}
+            {%- elif value['type'] | upper == 'ARRAY' -%}
+                {%- if value['items'] is mapping and value['items'] -%}
+                    {%- if add_comma %},{%- else -%} {%- set add_comma = true -%} {% endif -%}
+                    items:{
+                    {%- set ns_items = namespace(found_first=false) -%}
+                    {%- for item_key, item_value in value['items'] | dictsort -%}
+                        {%- if item_value is not none -%}
+                            {%- if ns_items.found_first %},{% endif -%}
+                            {%- set ns_items.found_first = true -%}
+                            {%- if item_key == 'properties' -%}
+                                properties:{
+                                {%- if item_value is mapping -%}
+                                    {{- format_parameters(item_value, value['items']['required'] | default([])) -}}
+                                {%- endif -%}
+                                }
+                            {%- elif item_key == 'required' -%}
+                                required:[
+                                {%- for req_item in item_value -%}
+                                    <|"|>{{- req_item -}}<|"|>
+                                    {%- if not loop.last %},{% endif -%}
+                                {%- endfor -%}
+                                ]
+                            {%- elif item_key == 'type' -%}
+                                {%- if item_value is string -%}
+                                    type:{{ format_argument(item_value | upper) }}
+                                {%- else -%}
+                                    type:{{ format_argument(item_value | map('upper') | list) }}
+                                {%- endif -%}
+                            {%- else -%}
+                                {{ item_key }}:{{ format_argument(item_value) }}
+                            {%- endif -%}
+                        {%- endif -%}
+                    {%- endfor -%}
+                    }
+                {%- endif -%}
+            {%- endif -%}
+            {%- if value['nullable'] %}
+                {%- if add_comma %},{%- else -%} {%- set add_comma = true -%} {% endif -%}
+                nullable:true
+            {%- endif -%}
+            {%- if value['type'] | upper == 'OBJECT' -%}
+                {%- if value['properties'] is defined and value['properties'] is mapping -%}
+                    {%- if add_comma %},{%- else -%} {%- set add_comma = true -%} {% endif -%}
+                    properties:{
+                    {{- format_parameters(value['properties'], value['required'] | default([])) -}}
+                    }
+                {%- elif value is mapping -%}
+                    {%- if add_comma %},{%- else -%} {%- set add_comma = true -%} {% endif -%}
+                    properties:{
+                    {{- format_parameters(value, value['required'] | default([]), filter_keys=true) -}}
+                    }
+                {%- endif -%}
+                {%- if value['required'] -%}
+                    {%- if add_comma %},{%- else -%} {%- set add_comma = true -%} {% endif -%}
+                    required:[
+                    {%- for item in value['required'] | default([]) -%}
+                        <|"|>{{- item -}}<|"|>
+                        {%- if not loop.last %},{% endif -%}
+                    {%- endfor -%}
+                    ]
+                {%- endif -%}
+            {%- endif -%}
+            {%- if add_comma %},{%- else -%} {%- set add_comma = true -%} {% endif -%}
+            type:<|"|>{{ value['type'] | upper }}<|"|>}
+        {%- endif -%}
+    {%- endfor -%}
+{%- endmacro -%}
+{%- macro format_function_declaration(tool_data) -%}
+    declaration:{{- tool_data['function']['name'] -}}{description:<|"|>{{- tool_data['function']['description'] -}}<|"|>
+    {%- set params = tool_data['function']['parameters'] -%}
+    {%- if params -%}
+        ,parameters:{
+        {%- if params['properties'] -%}
+            properties:{ {{- format_parameters(params['properties'], params['required']) -}} },
+        {%- endif -%}
+        {%- if params['required'] -%}
+            required:[
+            {%- for item in params['required'] -%}
+                <|"|>{{- item -}}<|"|>
+                {{- ',' if not loop.last -}}
+            {%- endfor -%}
+            ],
+        {%- endif -%}
+        {%- if params['type'] -%}
+            type:<|"|>{{- params['type'] | upper -}}<|"|>}
+        {%- endif -%}
+    {%- endif -%}
+    {%- if 'response' in tool_data['function'] -%}
+        {%- set response_declaration = tool_data['function']['response'] -%}
+        ,response:{
+        {%- if response_declaration['description'] -%}
+            description:<|"|>{{- response_declaration['description'] -}}<|"|>,
+        {%- endif -%}
+        {%- if response_declaration['type'] | upper == 'OBJECT' -%}
+            type:<|"|>{{- response_declaration['type'] | upper -}}<|"|>}
+        {%- endif -%}
+    {%- endif -%}
+    }
+{%- endmacro -%}
+{%- macro format_argument(argument, escape_keys=True) -%}
+    {%- if argument is string -%}
+        {{- '<|"|>' + argument + '<|"|>' -}}
+    {%- elif argument is boolean -%}
+        {{- 'true' if argument else 'false' -}}
+    {%- elif argument is mapping -%}
+        {{- '{' -}}
+        {%- set ns = namespace(found_first=false) -%}
+        {%- for key, value in argument | dictsort -%}
+            {%- if ns.found_first %},{% endif -%}
+            {%- set ns.found_first = true -%}
+            {%- if escape_keys -%}
+                {{- '<|"|>' + key + '<|"|>' -}}
+            {%- else -%}
+                {{- key -}}
+            {%- endif -%}
+            :{{- format_argument(value, escape_keys=escape_keys) -}}
+        {%- endfor -%}
+        {{- '}' -}}
+    {%- elif argument is sequence -%}
+        {{- '[' -}}
+        {%- for item in argument -%}
+            {{- format_argument(item, escape_keys=escape_keys) -}}
+            {%- if not loop.last %},{% endif -%}
+        {%- endfor -%}
+        {{- ']' -}}
+    {%- else -%}
+        {{- argument -}}
+    {%- endif -%}
+{%- endmacro -%}
+{%- macro strip_thinking(text) -%}
+    {%- set ns = namespace(result='') -%}
+    {%- for part in text.split('<channel|>') -%}
+        {%- if '<|channel>' in part -%}
+            {%- set ns.result = ns.result + part.split('<|channel>')[0] -%}
+        {%- else -%}
+            {%- set ns.result = ns.result + part -%}
+        {%- endif -%}
+    {%- endfor -%}
+    {{- ns.result | trim -}}
+{%- endmacro -%}
+
+{%- macro format_tool_response_block(tool_name, response) -%}
+    {{- '<|tool_response>' -}}
+    {%- if response is mapping -%}
+        {{- 'response:' + tool_name + '{' -}}
+        {%- for key, value in response | dictsort -%}
+            {{- key -}}:{{- format_argument(value, escape_keys=False) -}}
+            {%- if not loop.last %},{% endif -%}
+        {%- endfor -%}
+        {{- '}' -}}
+    {%- else -%}
+        {{- 'response:' + tool_name + '{value:' + format_argument(response, escape_keys=False) + '}' -}}
+    {%- endif -%}
+    {{- '<tool_response|>' -}}
+{%- endmacro -%}
+
+{%- set ns = namespace(prev_message_type=None) -%}
+{%- set loop_messages = messages -%}
+{{- bos_token -}}
+{#- Handle System/Tool Definitions Block -#}
+{%- if (enable_thinking is defined and enable_thinking) or tools or messages[0]['role'] in ['system', 'developer'] -%}
+    {{- '<|turn>system\n' -}}
+    {#- Inject Thinking token at the very top of the FIRST system turn -#}
+    {%- if enable_thinking is defined and enable_thinking -%}
+        {{- '<|think|>\n' -}}
+        {%- set ns.prev_message_type = 'think' -%}
+    {%- endif -%}
+    {%- if messages[0]['role'] in ['system', 'developer'] -%}
+        {%- if messages[0]['content'] is string -%}
+            {{- messages[0]['content'] | trim -}}
+        {%- elif messages[0]['content'] is sequence -%}
+            {%- for item in messages[0]['content'] -%}
+                {{- item['text'] | trim + ' '-}}
+            {%- endfor -%}
+        {%- endif -%}
+        {%- set loop_messages = messages[1:] -%}
+    {%- endif -%}
+    {%- if tools -%}
+        {%- for tool in tools %}
+            {{- '<|tool>' -}}
+            {{- format_function_declaration(tool) | trim -}}
+            {{- '<tool|>' -}}
+        {%- endfor %}
+        {%- set ns.prev_message_type = 'tool' -%}
+    {%- endif -%}
+    {{- '<turn|>\n' -}}
+{%- endif %}
+
+{#- Pre-scan: find last user message index for reasoning guard -#}
+{%- set ns_turn = namespace(last_user_idx=-1) -%}
+{%- for i in range(loop_messages | length) -%}
+    {%- if loop_messages[i]['role'] == 'user' -%}
+        {%- set ns_turn.last_user_idx = i -%}
+    {%- endif -%}
+{%- endfor -%}
+
+{#- Loop through messages -#}
+{%- for message in loop_messages -%}
+    {%- if message['role'] != 'tool' -%}
+    {%- set ns.prev_message_type = None -%}
+    {%- set role = 'model' if message['role'] == 'assistant' else message['role'] -%}
+    {#- Detect continuation: suppress duplicate <|turn>model when previous non-tool message was also assistant -#}
+    {%- set prev_nt = namespace(role=None, found=false) -%}
+    {%- if loop.index0 > 0 -%}
+        {%- for j in range(loop.index0 - 1, -1, -1) -%}
+            {%- if not prev_nt.found -%}
+                {%- if loop_messages[j]['role'] != 'tool' -%}
+                    {%- set prev_nt.role = loop_messages[j]['role'] -%}
+                    {%- set prev_nt.found = true -%}
+                {%- endif -%}
+            {%- endif -%}
+        {%- endfor -%}
+    {%- endif -%}
+    {%- set continue_same_model_turn = (role == 'model' and prev_nt.role == 'assistant') -%}
+    {%- if not continue_same_model_turn -%}
+        {{- '<|turn>' + role + '\n' }}
+    {%- endif -%}
+
+    {#- Render reasoning/reasoning_content as thinking channel -#}
+    {%- set thinking_text = message.get('reasoning') or message.get('reasoning_content') -%}
+    {%- if thinking_text and loop.index0 > ns_turn.last_user_idx and message.get('tool_calls') -%}
+        {{- '<|channel>thought\n' + thinking_text + '\n<channel|>' -}}
+    {%- endif -%}
+
+            {%- if message['tool_calls'] -%}
+                {%- for tool_call in message['tool_calls'] -%}
+                    {%- set function = tool_call['function'] -%}
+                    {{- '<|tool_call>call:' + function['name'] + '{' -}}
+                    {%- if function['arguments'] is mapping -%}
+                        {%- set ns_args = namespace(found_first=false) -%}
+                        {%- for key, value in function['arguments'] | dictsort -%}
+                            {%- if ns_args.found_first %},{% endif -%}
+                            {%- set ns_args.found_first = true -%}
+                            {{- key -}}:{{- format_argument(value, escape_keys=False) -}}
+                        {%- endfor -%}
+                    {%- elif function['arguments'] is string -%}
+                        {{- function['arguments'] -}}
+                    {%- endif -%}
+                    {{- '}<tool_call|>' -}}
+                {%- endfor -%}
+                {%- set ns.prev_message_type = 'tool_call' -%}
+            {%- endif -%}
+
+            {%- set ns_tr_out = namespace(flag=false) -%}
+            {%- if message.get('tool_responses') -%}
+                {#- Legacy: tool_responses embedded on the assistant message (Google/Gemma native) -#}
+                {%- for tool_response in message['tool_responses'] -%}
+                    {{- format_tool_response_block(tool_response['name'] | default('unknown'), tool_response['response']) -}}
+                    {%- set ns_tr_out.flag = true -%}
+                    {%- set ns.prev_message_type = 'tool_response' -%}
+                {%- endfor -%}
+            {%- elif message.get('tool_calls') -%}
+                {#- OpenAI Chat Completions: forward-scan consecutive role:tool messages -#}
+                {%- set ns_tool_scan = namespace(stopped=false) -%}
+                {%- for k in range(loop.index0 + 1, loop_messages | length) -%}
+                    {%- if ns_tool_scan.stopped -%}
+                    {%- elif loop_messages[k]['role'] != 'tool' -%}
+                        {%- set ns_tool_scan.stopped = true -%}
+                    {%- else -%}
+                        {%- set follow = loop_messages[k] -%}
+                        {#- Resolve tool_call_id to function name -#}
+                        {%- set ns_tname = namespace(name=follow.get('name') | default('unknown')) -%}
+                        {%- for tc in message['tool_calls'] -%}
+                            {%- if tc.get('id') == follow.get('tool_call_id') -%}
+                                {%- set ns_tname.name = tc['function']['name'] -%}
+                            {%- endif -%}
+                        {%- endfor -%}
+                        {#- Handle content as string or content-parts array -#}
+                        {%- set tool_body = follow.get('content') -%}
+                        {%- if tool_body is string -%}
+                            {{- format_tool_response_block(ns_tname.name, tool_body) -}}
+                        {%- elif tool_body is sequence and tool_body is not string -%}
+                            {%- set ns_txt = namespace(s='') -%}
+                            {%- for part in tool_body -%}
+                                {%- if part.get('type') == 'text' -%}
+                                    {%- set ns_txt.s = ns_txt.s + (part.get('text') | default('')) -%}
+                                {%- endif -%}
+                            {%- endfor -%}
+                            {{- format_tool_response_block(ns_tname.name, ns_txt.s) -}}
+                        {%- else -%}
+                            {{- format_tool_response_block(ns_tname.name, tool_body) -}}
+                        {%- endif -%}
+                        {%- set ns_tr_out.flag = true -%}
+                        {%- set ns.prev_message_type = 'tool_response' -%}
+                    {%- endif -%}
+                {%- endfor -%}
+            {%- endif -%}
+
+            {%- set captured_content -%}
+            {%- if message['content'] is string -%}
+                {%- if role == 'model' -%}
+                    {{- strip_thinking(message['content']) -}}
+                {%- else -%}
+                    {{- message['content'] | trim -}}
+                {%- endif -%}
+            {%- elif message['content'] is sequence -%}
+                {%- for item in message['content'] -%}
+                    {%- if item['type'] == 'text' -%}
+                        {%- if role == 'model' -%}
+                            {{- strip_thinking(item['text']) -}}
+                        {%- else -%}
+                            {{- item['text'] | trim -}}
+                        {%- endif -%}
+                    {%- elif item['type'] == 'image' -%}
+                        {{- '<|image|>' -}}
+                        {%- set ns.prev_message_type = 'image' -%}
+                    {%- elif item['type'] == 'audio' -%}
+                        {{- '<|audio|>' -}}
+                        {%- set ns.prev_message_type = 'audio' -%}
+                    {%- elif item['type'] == 'video' -%}
+                        {{- '<|video|>' -}}
+                        {%- set ns.prev_message_type = 'video' -%}
+                    {%- endif -%}
+                {%- endfor -%}
+            {%- endif -%}
+            {%- endset -%}
+
+            {{- captured_content -}}
+            {%- set has_content = captured_content | trim | length > 0 -%}
+
+        {%- if ns.prev_message_type == 'tool_call' and not ns_tr_out.flag -%}
+            {{- '<|tool_response>' -}}
+        {%- elif not (ns_tr_out.flag and not has_content) -%}
+            {{- '<turn|>\n' -}}
+        {%- endif -%}
+    {%- endif -%}
+{%- endfor -%}
+
+{%- if add_generation_prompt -%}
+    {%- if ns.prev_message_type != 'tool_response' and ns.prev_message_type != 'tool_call' -%}
+        {{- '<|turn>model\n' -}}
+        {%- if not enable_thinking | default(false) -%}
+            {{- '<|channel>thought\n<channel|>' -}}
+        {%- endif -%}
+    {%- endif -%}
+{%- endif -%}

config.json

@@ -0,0 +1,177 @@
+{
+  "architectures": [
+    "Gemma4ForConditionalGeneration"
+  ],
+  "audio_config": null,
+  "audio_token_id": 258881,
+  "boa_token_id": 256000,
+  "boi_token_id": 255999,
+  "dtype": "bfloat16",
+  "eoa_token_id": 258883,
+  "eoa_token_index": 258883,
+  "eoi_token_id": 258882,
+  "eos_token_id": [
+    1,
+    106
+  ],
+  "image_token_id": 258880,
+  "initializer_range": 0.02,
+  "model_type": "gemma4",
+  "text_config": {
+    "attention_bias": false,
+    "attention_dropout": 0.0,
+    "attention_k_eq_v": true,
+    "bos_token_id": 2,
+    "dtype": "bfloat16",
+    "enable_moe_block": false,
+    "eos_token_id": 1,
+    "expert_intermediate_size": null,
+    "final_logit_softcapping": 30.0,
+    "global_head_dim": 512,
+    "head_dim": 256,
+    "hidden_activation": "gelu_pytorch_tanh",
+    "hidden_size": 5376,
+    "hidden_size_per_layer_input": 0,
+    "initializer_range": 0.02,
+    "intermediate_size": 21504,
+    "layer_types": [
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention"
+    ],
+    "max_position_embeddings": 262144,
+    "model_type": "gemma4_text",
+    "moe_intermediate_size": null,
+    "num_attention_heads": 32,
+    "num_experts": null,
+    "num_global_key_value_heads": 4,
+    "num_hidden_layers": 60,
+    "num_key_value_heads": 16,
+    "num_kv_shared_layers": 0,
+    "pad_token_id": 0,
+    "rms_norm_eps": 1e-06,
+    "rope_parameters": {
+      "full_attention": {
+        "partial_rotary_factor": 0.25,
+        "rope_theta": 1000000.0,
+        "rope_type": "proportional"
+      },
+      "sliding_attention": {
+        "rope_theta": 10000.0,
+        "rope_type": "default"
+      }
+    },
+    "sliding_window": 1024,
+    "tie_word_embeddings": true,
+    "top_k_experts": null,
+    "use_bidirectional_attention": "vision",
+    "use_cache": true,
+    "use_double_wide_mlp": false,
+    "vocab_size": 262144,
+    "vocab_size_per_layer_input": 262144
+  },
+  "tie_word_embeddings": true,
+  "transformers_version": "5.7.0",
+  "video_token_id": 258884,
+  "vision_config": {
+    "_name_or_path": "",
+    "architectures": null,
+    "attention_bias": false,
+    "attention_dropout": 0.0,
+    "chunk_size_feed_forward": 0,
+    "default_output_length": 280,
+    "dtype": "bfloat16",
+    "global_head_dim": 72,
+    "head_dim": 72,
+    "hidden_activation": "gelu_pytorch_tanh",
+    "hidden_size": 1152,
+    "id2label": {
+      "0": "LABEL_0",
+      "1": "LABEL_1"
+    },
+    "initializer_range": 0.02,
+    "intermediate_size": 4304,
+    "is_encoder_decoder": false,
+    "label2id": {
+      "LABEL_0": 0,
+      "LABEL_1": 1
+    },
+    "max_position_embeddings": 131072,
+    "model_type": "gemma4_vision",
+    "num_attention_heads": 16,
+    "num_hidden_layers": 27,
+    "num_key_value_heads": 16,
+    "output_attentions": false,
+    "output_hidden_states": false,
+    "patch_size": 16,
+    "pooling_kernel_size": 3,
+    "position_embedding_size": 10240,
+    "problem_type": null,
+    "return_dict": true,
+    "rms_norm_eps": 1e-06,
+    "rope_parameters": {
+      "rope_theta": 100.0,
+      "rope_type": "default"
+    },
+    "standardize": true,
+    "use_clipped_linears": false
+  },
+  "vision_soft_tokens_per_image": 280
+}

mergekit_config.yml

@@ -0,0 +1,735 @@
+models:
+  - model: google/gemma-4-31B-it
+    # base model — no parameters needed
+
+  - model: llmfan46/gemma-4-Ortenzya-The-Creative-Wordsmith-31B-it-uncensored-heretic
+    parameters:
+      weight:
+        # ── Attn spike suppression — surgical ───────────────────
+        # L5, L7, L12, L15 anomalously high in attn KL.
+        # Near-zero to prevent early context binding disruption.
+        - filter: "layers.5.self_attn"
+          value: 0.03
+        - filter: "layers.7.self_attn"
+          value: 0.03
+        - filter: "layers.12.self_attn"
+          value: 0.03        # L12: 0.187 KL — primary offender
+        - filter: "layers.15.self_attn"
+          value: 0.03
+
+        # ── Dead zone 48-59: excluded ────────────────────────────
+        # Zero KL divergence from base — nothing to contribute.
+        - filter: "layers.48.self_attn"
+          value: 0.00
+        - filter: "layers.48.mlp"
+          value: 0.00
+        - filter: "layers.49.self_attn"
+          value: 0.00
+        - filter: "layers.49.mlp"
+          value: 0.00
+        - filter: "layers.50.self_attn"
+          value: 0.00
+        - filter: "layers.50.mlp"
+          value: 0.00
+        - filter: "layers.51.self_attn"
+          value: 0.00
+        - filter: "layers.51.mlp"
+          value: 0.00
+        - filter: "layers.52.self_attn"
+          value: 0.00
+        - filter: "layers.52.mlp"
+          value: 0.00
+        - filter: "layers.53.self_attn"
+          value: 0.00
+        - filter: "layers.53.mlp"
+          value: 0.00
+        - filter: "layers.54.self_attn"
+          value: 0.00
+        - filter: "layers.54.mlp"
+          value: 0.00
+        - filter: "layers.55.self_attn"
+          value: 0.00
+        - filter: "layers.55.mlp"
+          value: 0.00
+        - filter: "layers.56.self_attn"
+          value: 0.00
+        - filter: "layers.56.mlp"
+          value: 0.00
+        - filter: "layers.57.self_attn"
+          value: 0.00
+        - filter: "layers.57.mlp"
+          value: 0.00
+        - filter: "layers.58.self_attn"
+          value: 0.00
+        - filter: "layers.58.mlp"
+          value: 0.00
+        - filter: "layers.59.self_attn"
+          value: 0.00
+        - filter: "layers.59.mlp"
+          value: 0.00
+
+        # ── MeroMero inflection zone 26-35 ──────────────────────
+        # Ortenzya minimized — protecting MeroMero's detail zone.
+        # Scaled down proportionally to hit 1.05 budget.
+        - filter: "layers.26.self_attn"
+          value: 0.09
+        - filter: "layers.27.self_attn"
+          value: 0.09
+        - filter: "layers.28.self_attn"
+          value: 0.09
+        - filter: "layers.29.self_attn"
+          value: 0.09
+        - filter: "layers.30.self_attn"
+          value: 0.09
+        - filter: "layers.31.self_attn"
+          value: 0.09
+        - filter: "layers.32.self_attn"
+          value: 0.09
+        - filter: "layers.33.self_attn"
+          value: 0.09
+        - filter: "layers.34.self_attn"
+          value: 0.09
+        - filter: "layers.35.self_attn"
+          value: 0.09
+        - filter: "layers.26.mlp"
+          value: 0.13
+        - filter: "layers.27.mlp"
+          value: 0.13
+        - filter: "layers.28.mlp"
+          value: 0.13
+        - filter: "layers.29.mlp"
+          value: 0.13
+        - filter: "layers.30.mlp"
+          value: 0.13
+        - filter: "layers.31.mlp"
+          value: 0.13
+        - filter: "layers.32.mlp"
+          value: 0.13
+        - filter: "layers.33.mlp"
+          value: 0.13
+        - filter: "layers.34.mlp"
+          value: 0.13
+        - filter: "layers.35.mlp"
+          value: 0.13
+
+        # ── Active zone 36-43 ───────────────────────────────────
+        # Ortenzya's genuine creative signal zone.
+        # Scaled to budget — preserves relative contribution
+        # while keeping total sum at 1.05.
+        - filter: "layers.36.self_attn"
+          value: 0.11
+        - filter: "layers.37.self_attn"
+          value: 0.11
+        - filter: "layers.38.self_attn"
+          value: 0.11
+        - filter: "layers.39.self_attn"
+          value: 0.11
+        - filter: "layers.40.self_attn"
+          value: 0.11
+        - filter: "layers.41.self_attn"
+          value: 0.11
+        - filter: "layers.42.self_attn"
+          value: 0.11
+        - filter: "layers.43.self_attn"
+          value: 0.11
+        - filter: "layers.36.mlp"
+          value: 0.23
+        - filter: "layers.37.mlp"
+          value: 0.23
+        - filter: "layers.38.mlp"
+          value: 0.23
+        - filter: "layers.39.mlp"
+          value: 0.23
+        - filter: "layers.40.mlp"
+          value: 0.23
+        - filter: "layers.41.mlp"
+          value: 0.23
+        - filter: "layers.42.mlp"
+          value: 0.23
+        - filter: "layers.43.mlp"
+          value: 0.23
+
+        # ── Active zone 44-47: Ortenzya MLP peak ────────────────
+        # Highest MLP KL in the model (up to 0.098 at L47).
+        # Slightly higher than 36-43 to respect the peak,
+        # still scaled to budget.
+        - filter: "layers.44.self_attn"
+          value: 0.11
+        - filter: "layers.45.self_attn"
+          value: 0.11
+        - filter: "layers.46.self_attn"
+          value: 0.11
+        - filter: "layers.47.self_attn"
+          value: 0.11
+        - filter: "layers.44.mlp"
+          value: 0.26
+        - filter: "layers.45.mlp"
+          value: 0.26
+        - filter: "layers.46.mlp"
+          value: 0.26
+        - filter: "layers.47.mlp"
+          value: 0.26
+
+        # ── Zone-wide fallbacks ──────────────────────────────────
+        # Covers early layers 0-25 (excluding spike layers above).
+        # Scaled to budget alongside MeroMero and Artemis early values.
+        - filter: "self_attn"
+          value: 0.14
+        - filter: "mlp"
+          value: 0.26
+        - value: 0.25
+      density: 0.80
+      epsilon: 0.099
+
+  - model: zerofata/G4-MeroMero-31B
+    parameters:
+      weight:
+        # ── Early layers 0-25 ───────────────────────────────────
+        # Scaled down from 0.60/0.55 to hit budget alongside
+        # Artemis and Ortenzya early contributions.
+        - filter: "layers.0.self_attn"
+          value: 0.57
+        - filter: "layers.1.self_attn"
+          value: 0.57
+        - filter: "layers.2.self_attn"
+          value: 0.57
+        - filter: "layers.3.self_attn"
+          value: 0.57
+        - filter: "layers.4.self_attn"
+          value: 0.57
+        - filter: "layers.5.self_attn"
+          value: 0.57
+        - filter: "layers.6.self_attn"
+          value: 0.57
+        - filter: "layers.7.self_attn"
+          value: 0.57
+        - filter: "layers.8.self_attn"
+          value: 0.57
+        - filter: "layers.9.self_attn"
+          value: 0.57
+        - filter: "layers.10.self_attn"
+          value: 0.57
+        - filter: "layers.11.self_attn"
+          value: 0.57
+        - filter: "layers.12.self_attn"
+          value: 0.57
+        - filter: "layers.13.self_attn"
+          value: 0.57
+        - filter: "layers.14.self_attn"
+          value: 0.57
+        - filter: "layers.15.self_attn"
+          value: 0.57
+        - filter: "layers.16.self_attn"
+          value: 0.57
+        - filter: "layers.17.self_attn"
+          value: 0.57
+        - filter: "layers.18.self_attn"
+          value: 0.57
+        - filter: "layers.19.self_attn"
+          value: 0.57
+        - filter: "layers.20.self_attn"
+          value: 0.57
+        - filter: "layers.21.self_attn"
+          value: 0.57
+        - filter: "layers.22.self_attn"
+          value: 0.57
+        - filter: "layers.23.self_attn"
+          value: 0.57
+        - filter: "layers.24.self_attn"
+          value: 0.57
+        - filter: "layers.25.self_attn"
+          value: 0.57
+        - filter: "layers.0.mlp"
+          value: 0.48
+        - filter: "layers.1.mlp"
+          value: 0.48
+        - filter: "layers.2.mlp"
+          value: 0.48
+        - filter: "layers.3.mlp"
+          value: 0.48
+        - filter: "layers.4.mlp"
+          value: 0.48
+        - filter: "layers.5.mlp"
+          value: 0.48
+        - filter: "layers.6.mlp"
+          value: 0.48
+        - filter: "layers.7.mlp"
+          value: 0.48
+        - filter: "layers.8.mlp"
+          value: 0.48
+        - filter: "layers.9.mlp"
+          value: 0.48
+        - filter: "layers.10.mlp"
+          value: 0.48
+        - filter: "layers.11.mlp"
+          value: 0.48
+        - filter: "layers.12.mlp"
+          value: 0.48
+        - filter: "layers.13.mlp"
+          value: 0.48
+        - filter: "layers.14.mlp"
+          value: 0.48
+        - filter: "layers.15.mlp"
+          value: 0.48
+        - filter: "layers.16.mlp"
+          value: 0.48
+        - filter: "layers.17.mlp"
+          value: 0.48
+        - filter: "layers.18.mlp"
+          value: 0.48
+        - filter: "layers.19.mlp"
+          value: 0.48
+        - filter: "layers.20.mlp"
+          value: 0.48
+        - filter: "layers.21.mlp"
+          value: 0.48
+        - filter: "layers.22.mlp"
+          value: 0.48
+        - filter: "layers.23.mlp"
+          value: 0.48
+        - filter: "layers.24.mlp"
+          value: 0.48
+        - filter: "layers.25.mlp"
+          value: 0.48
+
+        # ── MeroMero inflection zone 26-35 ──────────────────────
+        # MeroMero's most important zone — scaled down minimally
+        # to preserve dominance while hitting budget.
+        - filter: "layers.26.self_attn"
+          value: 0.73
+        - filter: "layers.27.self_attn"
+          value: 0.73
+        - filter: "layers.28.self_attn"
+          value: 0.73
+        - filter: "layers.29.self_attn"
+          value: 0.73
+        - filter: "layers.30.self_attn"
+          value: 0.73
+        - filter: "layers.31.self_attn"
+          value: 0.73
+        - filter: "layers.32.self_attn"
+          value: 0.73
+        - filter: "layers.33.self_attn"
+          value: 0.73
+        - filter: "layers.34.self_attn"
+          value: 0.73
+        - filter: "layers.35.self_attn"
+          value: 0.73
+        - filter: "layers.26.mlp"
+          value: 0.74
+        - filter: "layers.27.mlp"
+          value: 0.74
+        - filter: "layers.28.mlp"
+          value: 0.74
+        - filter: "layers.29.mlp"
+          value: 0.74
+        - filter: "layers.30.mlp"
+          value: 0.74
+        - filter: "layers.31.mlp"
+          value: 0.74
+        - filter: "layers.32.mlp"
+          value: 0.74
+        - filter: "layers.33.mlp"
+          value: 0.74
+        - filter: "layers.34.mlp"
+          value: 0.74
+        - filter: "layers.35.mlp"
+          value: 0.74
+
+        # ── Sustained active zone 36-50 ─────────────────────────
+        # MeroMero attn leads for structural coherence.
+        # MLP pulled back to give Artemis room in this zone.
+        - filter: "layers.36.self_attn"
+          value: 0.63
+        - filter: "layers.37.self_attn"
+          value: 0.63
+        - filter: "layers.38.self_attn"
+          value: 0.63
+        - filter: "layers.39.self_attn"
+          value: 0.63
+        - filter: "layers.40.self_attn"
+          value: 0.63
+        - filter: "layers.41.self_attn"
+          value: 0.63
+        - filter: "layers.42.self_attn"
+          value: 0.63
+        - filter: "layers.43.self_attn"
+          value: 0.63
+        - filter: "layers.44.self_attn"
+          value: 0.63
+        - filter: "layers.45.self_attn"
+          value: 0.63
+        - filter: "layers.46.self_attn"
+          value: 0.63
+        - filter: "layers.47.self_attn"
+          value: 0.63
+        - filter: "layers.48.self_attn"
+          value: 0.70
+        - filter: "layers.49.self_attn"
+          value: 0.70
+        - filter: "layers.50.self_attn"
+          value: 0.70
+        - filter: "layers.36.mlp"
+          value: 0.31
+        - filter: "layers.37.mlp"
+          value: 0.31
+        - filter: "layers.38.mlp"
+          value: 0.31
+        - filter: "layers.39.mlp"
+          value: 0.31
+        - filter: "layers.40.mlp"
+          value: 0.31
+        - filter: "layers.41.mlp"
+          value: 0.31
+        - filter: "layers.42.mlp"
+          value: 0.31
+        - filter: "layers.43.mlp"
+          value: 0.31
+        - filter: "layers.44.mlp"
+          value: 0.30
+        - filter: "layers.45.mlp"
+          value: 0.30
+        - filter: "layers.46.mlp"
+          value: 0.30
+        - filter: "layers.47.mlp"
+          value: 0.30
+        - filter: "layers.48.mlp"
+          value: 0.40
+        - filter: "layers.49.mlp"
+          value: 0.40
+        - filter: "layers.50.mlp"
+          value: 0.40
+
+        # ── Output zone 51-59 ───────────────────────────────────
+        # Ortenzya is dead here — only MeroMero and Artemis.
+        # Budget already at 1.05 with previous two-model values.
+        - filter: "layers.51.self_attn"
+          value: 0.65
+        - filter: "layers.52.self_attn"
+          value: 0.65
+        - filter: "layers.53.self_attn"
+          value: 0.65
+        - filter: "layers.54.self_attn"
+          value: 0.65
+        - filter: "layers.55.self_attn"
+          value: 0.65
+        - filter: "layers.56.self_attn"
+          value: 0.65
+        - filter: "layers.57.self_attn"
+          value: 0.65
+        - filter: "layers.58.self_attn"
+          value: 0.65
+        - filter: "layers.59.self_attn"
+          value: 0.65
+        - filter: "layers.51.mlp"
+          value: 0.50
+        - filter: "layers.52.mlp"
+          value: 0.50
+        - filter: "layers.53.mlp"
+          value: 0.50
+        - filter: "layers.54.mlp"
+          value: 0.50
+        - filter: "layers.55.mlp"
+          value: 0.50
+        - filter: "layers.56.mlp"
+          value: 0.50
+        - filter: "layers.57.mlp"
+          value: 0.50
+        - filter: "layers.58.mlp"
+          value: 0.50
+        - filter: "layers.59.mlp"
+          value: 0.50
+
+        # ── Zone-wide fallbacks ─────────────────────────────────
+        - filter: "self_attn"
+          value: 0.57
+        - filter: "mlp"
+          value: 0.48
+        - value: 0.50
+      density: 0.95
+      epsilon: 0.099
+
+  - model: BeaverAI/Artemis-31B-v1h-GGUF # Reversed from Q8 to fp16 for merging
+    parameters:
+      weight:
+        # ── Early layers 0-25 ───────────────────────────────────
+        # Scaled down from 0.35 to hit budget alongside
+        # MeroMero and Ortenzya early contributions.
+        - filter: "layers.0.self_attn"
+          value: 0.33
+        - filter: "layers.1.self_attn"
+          value: 0.33
+        - filter: "layers.2.self_attn"
+          value: 0.33
+        - filter: "layers.3.self_attn"
+          value: 0.33
+        - filter: "layers.4.self_attn"
+          value: 0.33
+        - filter: "layers.5.self_attn"
+          value: 0.33
+        - filter: "layers.6.self_attn"
+          value: 0.33
+        - filter: "layers.7.self_attn"
+          value: 0.33
+        - filter: "layers.8.self_attn"
+          value: 0.33
+        - filter: "layers.9.self_attn"
+          value: 0.33
+        - filter: "layers.10.self_attn"
+          value: 0.33
+        - filter: "layers.11.self_attn"
+          value: 0.33
+        - filter: "layers.12.self_attn"
+          value: 0.33
+        - filter: "layers.13.self_attn"
+          value: 0.33
+        - filter: "layers.14.self_attn"
+          value: 0.33
+        - filter: "layers.15.self_attn"
+          value: 0.33
+        - filter: "layers.16.self_attn"
+          value: 0.33
+        - filter: "layers.17.self_attn"
+          value: 0.33
+        - filter: "layers.18.self_attn"
+          value: 0.33
+        - filter: "layers.19.self_attn"
+          value: 0.33
+        - filter: "layers.20.self_attn"
+          value: 0.33
+        - filter: "layers.21.self_attn"
+          value: 0.33
+        - filter: "layers.22.self_attn"
+          value: 0.33
+        - filter: "layers.23.self_attn"
+          value: 0.33
+        - filter: "layers.24.self_attn"
+          value: 0.33
+        - filter: "layers.25.self_attn"
+          value: 0.33
+        - filter: "layers.0.mlp"
+          value: 0.31
+        - filter: "layers.1.mlp"
+          value: 0.31
+        - filter: "layers.2.mlp"
+          value: 0.31
+        - filter: "layers.3.mlp"
+          value: 0.31
+        - filter: "layers.4.mlp"
+          value: 0.31
+        - filter: "layers.5.mlp"
+          value: 0.31
+        - filter: "layers.6.mlp"
+          value: 0.31
+        - filter: "layers.7.mlp"
+          value: 0.31
+        - filter: "layers.8.mlp"
+          value: 0.31
+        - filter: "layers.9.mlp"
+          value: 0.31
+        - filter: "layers.10.mlp"
+          value: 0.31
+        - filter: "layers.11.mlp"
+          value: 0.31
+        - filter: "layers.12.mlp"
+          value: 0.31
+        - filter: "layers.13.mlp"
+          value: 0.31
+        - filter: "layers.14.mlp"
+          value: 0.31
+        - filter: "layers.15.mlp"
+          value: 0.31
+        - filter: "layers.16.mlp"
+          value: 0.31
+        - filter: "layers.17.mlp"
+          value: 0.31
+        - filter: "layers.18.mlp"
+          value: 0.31
+        - filter: "layers.19.mlp"
+          value: 0.31
+        - filter: "layers.20.mlp"
+          value: 0.31
+        - filter: "layers.21.mlp"
+          value: 0.31
+        - filter: "layers.22.mlp"
+          value: 0.31
+        - filter: "layers.23.mlp"
+          value: 0.31
+        - filter: "layers.24.mlp"
+          value: 0.31
+        - filter: "layers.25.mlp"
+          value: 0.31
+
+        # ── MeroMero inflection zone 26-35 ──────────────────────
+        # Artemis minimized to protect MeroMero's detail zone.
+        # Scaled proportionally to budget.
+        - filter: "layers.26.self_attn"
+          value: 0.23
+        - filter: "layers.27.self_attn"
+          value: 0.23
+        - filter: "layers.28.self_attn"
+          value: 0.23
+        - filter: "layers.29.self_attn"
+          value: 0.23
+        - filter: "layers.30.self_attn"
+          value: 0.23
+        - filter: "layers.31.self_attn"
+          value: 0.23
+        - filter: "layers.32.self_attn"
+          value: 0.23
+        - filter: "layers.33.self_attn"
+          value: 0.23
+        - filter: "layers.34.self_attn"
+          value: 0.23
+        - filter: "layers.35.self_attn"
+          value: 0.23
+        - filter: "layers.26.mlp"
+          value: 0.18
+        - filter: "layers.27.mlp"
+          value: 0.18
+        - filter: "layers.28.mlp"
+          value: 0.18
+        - filter: "layers.29.mlp"
+          value: 0.18
+        - filter: "layers.30.mlp"
+          value: 0.18
+        - filter: "layers.31.mlp"
+          value: 0.18
+        - filter: "layers.32.mlp"
+          value: 0.18
+        - filter: "layers.33.mlp"
+          value: 0.18
+        - filter: "layers.34.mlp"
+          value: 0.18
+        - filter: "layers.35.mlp"
+          value: 0.18
+
+        # ── Sustained active zone 36-47 ─────────────────────────
+        # Artemis MLP leads for creative voice.
+        # Scaled to budget — still dominant over MeroMero MLP here.
+        - filter: "layers.36.self_attn"
+          value: 0.31
+        - filter: "layers.37.self_attn"
+          value: 0.31
+        - filter: "layers.38.self_attn"
+          value: 0.31
+        - filter: "layers.39.self_attn"
+          value: 0.31
+        - filter: "layers.40.self_attn"
+          value: 0.31
+        - filter: "layers.41.self_attn"
+          value: 0.31
+        - filter: "layers.42.self_attn"
+          value: 0.31
+        - filter: "layers.43.self_attn"
+          value: 0.31
+        - filter: "layers.44.self_attn"
+          value: 0.31
+        - filter: "layers.45.self_attn"
+          value: 0.31
+        - filter: "layers.46.self_attn"
+          value: 0.31
+        - filter: "layers.47.self_attn"
+          value: 0.31
+        - filter: "layers.36.mlp"
+          value: 0.51
+        - filter: "layers.37.mlp"
+          value: 0.51
+        - filter: "layers.38.mlp"
+          value: 0.51
+        - filter: "layers.39.mlp"
+          value: 0.51
+        - filter: "layers.40.mlp"
+          value: 0.51
+        - filter: "layers.41.mlp"
+          value: 0.51
+        - filter: "layers.42.mlp"
+          value: 0.51
+        - filter: "layers.43.mlp"
+          value: 0.51
+        - filter: "layers.44.mlp"
+          value: 0.49
+        - filter: "layers.45.mlp"
+          value: 0.49
+        - filter: "layers.46.mlp"
+          value: 0.49
+        - filter: "layers.47.mlp"
+          value: 0.49
+
+        # ── Active zone 48-50 ───────────────────────────────────
+        # Ortenzya dead here — back to two-model budget of 1.05.
+        - filter: "layers.48.self_attn"
+          value: 0.35
+        - filter: "layers.49.self_attn"
+          value: 0.35
+        - filter: "layers.50.self_attn"
+          value: 0.35
+        - filter: "layers.48.mlp"
+          value: 0.65
+        - filter: "layers.49.mlp"
+          value: 0.65
+        - filter: "layers.50.mlp"
+          value: 0.65
+
+        # ── Output zone 51-59 ───────────────────────────────────
+        # Ortenzya dead — two-model budget.
+        # Artemis MLP nudged down from 0.60 to 0.55
+        # to bring output sum to 1.05.
+        - filter: "layers.51.self_attn"
+          value: 0.40
+        - filter: "layers.52.self_attn"
+          value: 0.40
+        - filter: "layers.53.self_attn"
+          value: 0.40
+        - filter: "layers.54.self_attn"
+          value: 0.40
+        - filter: "layers.55.self_attn"
+          value: 0.40
+        - filter: "layers.56.self_attn"
+          value: 0.40
+        - filter: "layers.57.self_attn"
+          value: 0.40
+        - filter: "layers.58.self_attn"
+          value: 0.40
+        - filter: "layers.59.self_attn"
+          value: 0.40
+        - filter: "layers.51.mlp"
+          value: 0.55
+        - filter: "layers.52.mlp"
+          value: 0.55
+        - filter: "layers.53.mlp"
+          value: 0.55
+        - filter: "layers.54.mlp"
+          value: 0.55
+        - filter: "layers.55.mlp"
+          value: 0.55
+        - filter: "layers.56.mlp"
+          value: 0.55
+        - filter: "layers.57.mlp"
+          value: 0.55
+        - filter: "layers.58.mlp"
+          value: 0.55
+        - filter: "layers.59.mlp"
+          value: 0.55
+
+        # ── Zone-wide fallbacks ─────────────────────────────────
+        - filter: "self_attn"
+          value: 0.33
+        - filter: "mlp"
+          value: 0.31
+        - value: 0.50
+      density: 0.85
+      epsilon: 0.099
+
+merge_method: della
+base_model: google/gemma-4-31B-it
+
+parameters:
+  lambda: 1.0
+  normalize: false
+
+tokenizer:
+  source: base
+chat_template: auto
+dtype: float32
+out_dtype: bfloat16

model-00001-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e4672923855f191f80ae0e39895175e9b0d70976f67f15fc6eaf4a50b15d103
+size 4965634630

model-00002-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c57b84bd9c266686525b4a8fbbdf18160d2bc042cf454e65a336734f469cf848
+size 4899701578

model-00003-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1e183bbe369b38e10931d1c0712f69a243522f234825fe4c1a43c30f6a13299
+size 4899701570

model-00004-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc7b99c0098e6e10c6bc618c2bfc16bdd88097261e442168ddf130c227d99261
+size 4899701562

model-00005-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f2afff00d6f99acfcfe79acc5f13753d49f7509f4098fb6a630f1797491a7464
+size 4789600202

model-00006-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8be606649b3302f96bc66a4d278ad9292b2d57b4ccbafdb09aef15ca67df560f
+size 4899701562

model-00007-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:84c7710bb9c7ad26dcf0680882e1a9804b3c0761ea2f6688dee200ae43961b70
+size 4899701578

model-00008-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d6e09041f92097e9094bdba35fb2721b3c5a41d00441d8b0a75a2300e426281
+size 4899701562

model-00009-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e29fb3b282c5ef2d0237afe9dd6f979765f0fd26d636c45ea5d429ed9544164
+size 4789600202

model-00010-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b0e951023a2ccb64c97018273fd5c9028ee9f221bfe6abd71175702348c9ea9
+size 4778591802

model-00011-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:87436e9da22d011e9636626d377756e281ba6a0f1bce72b55888970561993622
+size 4899701578

model-00012-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb7908dd180e8bb2c146ff23407131b54142d7f5d2a356691af48059c19fa50d
+size 4899701562

model-00013-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1f3c2bfe63ab129d4d053e34f5b08082266673f964993a1ff04fd7154e089007
+size 4012910652

processor_config.json

@@ -0,0 +1,75 @@
+{
+  "audio_ms_per_token": 40,
+  "audio_seq_length": 750,
+  "feature_extractor": {
+    "dither": 0.0,
+    "feature_extractor_type": "Gemma4AudioFeatureExtractor",
+    "feature_size": 128,
+    "fft_length": 512,
+    "fft_overdrive": false,
+    "frame_length": 320,
+    "hop_length": 160,
+    "input_scale_factor": 1.0,
+    "max_frequency": 8000.0,
+    "mel_floor": 0.001,
+    "min_frequency": 0.0,
+    "padding_side": "right",
+    "padding_value": 0.0,
+    "per_bin_mean": null,
+    "per_bin_stddev": null,
+    "preemphasis": 0.0,
+    "preemphasis_htk_flavor": true,
+    "return_attention_mask": true,
+    "sampling_rate": 16000
+  },
+  "image_processor": {
+    "do_convert_rgb": true,
+    "do_normalize": false,
+    "do_rescale": true,
+    "do_resize": true,
+    "image_mean": [
+      0.0,
+      0.0,
+      0.0
+    ],
+    "image_processor_type": "Gemma4ImageProcessor",
+    "image_seq_length": 280,
+    "image_std": [
+      1.0,
+      1.0,
+      1.0
+    ],
+    "max_soft_tokens": 280,
+    "patch_size": 16,
+    "pooling_kernel_size": 3,
+    "resample": 3,
+    "rescale_factor": 0.00392156862745098
+  },
+  "image_seq_length": 280,
+  "processor_class": "Gemma4Processor",
+  "video_processor": {
+    "do_convert_rgb": true,
+    "do_normalize": true,
+    "do_rescale": true,
+    "do_resize": true,
+    "do_sample_frames": true,
+    "image_mean": [
+      0.0,
+      0.0,
+      0.0
+    ],
+    "image_std": [
+      1.0,
+      1.0,
+      1.0
+    ],
+    "max_soft_tokens": 70,
+    "num_frames": 32,
+    "patch_size": 16,
+    "pooling_kernel_size": 3,
+    "resample": 3,
+    "rescale_factor": 0.00392156862745098,
+    "return_metadata": false,
+    "video_processor_type": "Gemma4VideoProcessor"
+  }
+}

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc8d3a0ce36466ccc1278bf987df5f71db1719b9ca6b4118264f45cb627bfe0f
+size 32169626

tokenizer_config.json

@@ -0,0 +1,96 @@
+{
+  "audio_token": "<|audio|>",
+  "backend": "tokenizers",
+  "boa_token": "<|audio>",
+  "boi_token": "<|image>",
+  "bos_token": "<bos>",
+  "eoa_token": "<audio|>",
+  "eoc_token": "<channel|>",
+  "eoi_token": "<image|>",
+  "eos_token": "<eos>",
+  "eot_token": "<turn|>",
+  "escape_token": "<|\"|>",
+  "etc_token": "<tool_call|>",
+  "etd_token": "<tool|>",
+  "etr_token": "<tool_response|>",
+  "extra_special_tokens": [
+    "<|video|>"
+  ],
+  "image_token": "<|image|>",
+  "is_local": true,
+  "local_files_only": false,
+  "mask_token": "<mask>",
+  "model_max_length": 1000000000000000019884624838656,
+  "model_specific_special_tokens": {
+    "audio_token": "<|audio|>",
+    "boa_token": "<|audio>",
+    "boi_token": "<|image>",
+    "eoa_token": "<audio|>",
+    "eoc_token": "<channel|>",
+    "eoi_token": "<image|>",
+    "eot_token": "<turn|>",
+    "escape_token": "<|\"|>",
+    "etc_token": "<tool_call|>",
+    "etd_token": "<tool|>",
+    "etr_token": "<tool_response|>",
+    "image_token": "<|image|>",
+    "soc_token": "<|channel>",
+    "sot_token": "<|turn>",
+    "stc_token": "<|tool_call>",
+    "std_token": "<|tool>",
+    "str_token": "<|tool_response>",
+    "think_token": "<|think|>"
+  },
+  "pad_token": "<pad>",
+  "padding_side": "left",
+  "processor_class": "Gemma4Processor",
+  "response_schema": {
+    "properties": {
+      "content": {
+        "type": "string"
+      },
+      "role": {
+        "const": "assistant"
+      },
+      "thinking": {
+        "type": "string"
+      },
+      "tool_calls": {
+        "items": {
+          "properties": {
+            "function": {
+              "properties": {
+                "arguments": {
+                  "additionalProperties": {},
+                  "type": "object",
+                  "x-parser": "gemma4-tool-call"
+                },
+                "name": {
+                  "type": "string"
+                }
+              },
+              "type": "object",
+              "x-regex": "call\\:(?P<name>\\w+)(?P<arguments>\\{.*\\})"
+            },
+            "type": {
+              "const": "function"
+            }
+          },
+          "type": "object"
+        },
+        "type": "array",
+        "x-regex-iterator": "<\\|tool_call>(.*?)<tool_call\\|>"
+      }
+    },
+    "type": "object",
+    "x-regex": "(\\<\\|channel\\>thought\\n(?P<thinking>.*?)\\<channel\\|\\>)?(?P<tool_calls>\\<\\|tool_call\\>.*\\<tool_call\\|\\>)?(?P<content>(?:(?!\\<turn\\|\\>)(?!\\<\\|tool_response\\>).)+)?(?:\\<turn\\|\\>|\\<\\|tool_response\\>)?"
+  },
+  "soc_token": "<|channel>",
+  "sot_token": "<|turn>",
+  "stc_token": "<|tool_call>",
+  "std_token": "<|tool>",
+  "str_token": "<|tool_response>",
+  "think_token": "<|think|>",
+  "tokenizer_class": "GemmaTokenizer",
+  "unk_token": "<unk>"
+}