OsirisPtah-Coder-v5 v3: mean-diff balanced abliteration

README.md

@@ -1,38 +1,8 @@
 ---
 license: apache-2.0
 base_model: Qwen/Qwen2.5-Coder-7B-Instruct
-tags:
-  - abliteration
-  - uncensored
-  - qwen2.5
-  - coder
-  - code-generation
-  - osiris
+tags: [abliteration, uncensored, qwen2.5, coder, osiris]
 pipeline_tag: text-generation
 ---
-
 # OsirisPtah-Coder-v5
-
-Abliterated from `Qwen/Qwen2.5-Coder-7B-Instruct` — all refusal mechanisms permanently removed.
-
-## Abliteration Details
-- **Technique:** Weight orthogonalization (double-pass nuclear mode)
-- **Source:** Qwen/Qwen2.5-Coder-7B-Instruct
-- **Layers scanned:** 0-27 (best: 26)
-- **Strength:** 2.0
-- **Prompts:** 50 harmful + 50 harmless
-
-## Usage
-```python
-from transformers import AutoModelForCausalLM, AutoTokenizer
-
-model = AutoModelForCausalLM.from_pretrained("osirisbrain/OsirisPtah-Coder-v5")
-tokenizer = AutoTokenizer.from_pretrained("osirisbrain/OsirisPtah-Coder-v5")
-```
-
-## ⚠️ Disclaimer
-This model responds to any prompt without ethical filtering. Use responsibly.
-
-## Credits
-Based on [Qwen/Qwen2.5-Coder-7B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-7B-Instruct) by Alibaba (Apache 2.0).
-Abliteration by OsirisBrain.
+Abliterated `Qwen/Qwen2.5-Coder-7B-Instruct` — mean-difference method, 1.5x strength, 4 passes.

abliteration_metadata.json

@@ -1,14 +1,10 @@
 {
   "base_model": "Qwen/Qwen2.5-Coder-7B-Instruct",
-  "technique": "weight_orthogonalization",
-  "refusal_direction_layer": 26,
-  "scan_range": [
-    0,
-    27
-  ],
-  "strength": 2.0,
-  "num_harmful_prompts": 50,
-  "num_harmless_prompts": 50,
-  "created_at": "2026-03-03T01:54:47.462087",
-  "author": "OsirisBrain"
+  "technique": "mean_difference_weight_orthogonalization_v3",
+  "best_layer": 26,
+  "strength": 1.5,
+  "num_passes": 4,
+  "num_harmful": 80,
+  "num_harmless": 81,
+  "created_at": "2026-03-03T02:20:40.426872"
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2888966abceee7751af041321cf90e010662b6a0c9c2019b875dca7f19c3991e
+oid sha256:32cff08484f391e60586b71d183cf14f4932b2c23fe35511400891beda689631
 size 15231272152

refusal_direction.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:24d8b51c5d564f196b4b54ba0c77e3624e32653d4b6e9a76d6999dc3e6e0d6da
+oid sha256:df8111d5555de5c86d5bbb135c43ca1bab32e95b03c1eec92d4438beaf271d26
 size 15983