Upload README.md with huggingface_hub

README.md

@@ -1,111 +1,234 @@
 ---
 license: apache-2.0
-base_model: Qwen/Qwen3.6-27B
-tags:
-  - qwen3
-  - qwen3.6
-  - aeon-rys
-  - abliterated
-  - uncensored
-  - agentic
-  - coding-agent
-  - gguf
-  - ik-llama
-  - hybrid-attention
-  - mamba
-  - gated-deltanet
-library_name: gguf
-pipeline_tag: text-generation
 language:
-  - en
-  - zh
+- en
+- zh
+- multilingual
+tags:
+- gguf
+- qwen3
+- qwen3.6
+- reasoning
+- coding
+- coding-agent
+- academic-writing
+- uncensored
+- rys
+- lora
+- iq4_nl
+- bf16
+base_model:
+- jackasda211233/Qwen3.6-27B-AEON-RYS-SignalLatch-GGUF
+- jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF
+- AEON-7/Qwen3.6-27B-AEON-Ultimate-Uncensored
 ---
 
-# Qwen3.6-27B-AEON-RYS-Agentic-Coder-PatchCode (GGUF)
+# Qwen3.6 AEON RYS Agentic-Coder PatchCode GGUF
 
-An agentic-coder behaviour distill on the **Qwen3.6-27B-AEON-RYS** base (itself an Apache-2.0 abliteration of `Qwen/Qwen3.6-27B`). Tuned for action-first, narrate→act→verify coding-agent discipline: minimal preamble, claims backed by an actual run, systematic diagnose→fix loops, and stable multi-turn tool use. Distilled from execution traces of a high-discipline coding agent blended with a robust narrate-act-verify coding dataset.
+> **⚠️ Required runtime — read first.** This model **must be used with** the custom AEON ik-llama fork:
+>
+> **https://github.com/noonr48/qwen36-aeon-ik-llama**
+>
+> Use that fork with Jinja and DeepSeek reasoning formatting. This is **not** a stock `llama.cpp` or `vLLM` GGUF — the Qwen3.6 hybrid/recurrent (`qwen3_5`) architecture will fail to load on stock runtimes (`missing tensor blk.N.ssm_conv1d.weight`).
 
-**Recommended file: `…IQ4_NL.gguf` (16.6 GB)** — behaviourally indistinguishable from BF16 across a 5-seed, same-condition bake-off (build / long-context / discipline / autonomous-loop all tied within noise), at ~⅓ the size.
+This is a merged fine-tuned GGUF upgrade candidate for the existing AEON RYS SignalLatch release. PatchCode adds an agentic-coder behaviour distil on top of SignalLatch: an action-first, verify-before-claim execution style for coding agents — minimal preamble, claims backed by an actual run, systematic diagnose→fix loops, and stable multi-turn tool use.
 
----
+The main project here is the `IQ4_NL` GGUF: a practical small-form-factor release aimed at pulling as much useful coding-agent performance as possible out of the AEON RYS line without asking people to run a huge source-quality file. The `BF16` artifact is included for people who want to inspect, re-quantize, or continue work from the merged fine-tuned model.
 
-## ⚠️ READ FIRST — special loader required (qwen3_5 triple-hybrid arch)
+PatchCode is distilled around an `Investigate → Act → Verify → Repair → Confirm` loop for coding agents. It promotes reading the real context first, acting with a concrete patch, **claiming nothing without a run**, repairing from evidence when a check fails, and confirming through validation.
 
-This model uses Qwen3.6's **triple-hybrid** architecture (`model_type: qwen3_5`: full self-attention every 4th layer + linear-attention / GatedDeltaNet / SSM + MLP). **Mainline `llama.cpp` and `vLLM` will NOT load it** (`missing tensor blk.N.ssm_conv1d.weight`).
+Full testing/process write-up (quant bake-off):
+`https://github.com/noonr48/qwen36-aeon-ik-llama/tree/main/docs` *(PatchCode bake-off record)*
 
-Use a loader with `qwen35` arch support. Verified working options:
+Upgrade target:
+- existing repo: `https://huggingface.co/jackasda211233/Qwen3.6-27B-AEON-RYS-SignalLatch-GGUF`
+- existing file: `Qwen3.6-27B-AEON-RYS-SignalLatch-ckpt386-s010-IQ4_NL.gguf`
 
-1. **[ikawrakow/ik_llama.cpp](https://github.com/ikawrakow/ik_llama.cpp)** — `qwen35` (`LLM_ARCH_QWEN35`) is upstream. Build from `main` and it loads + serves this model on NVIDIA 50-series (and others).
-2. **[noonr48/qwen36-aeon-ik-llama](https://github.com/noonr48/qwen36-aeon-ik-llama)** — a documented fork (build + serve guides, quant recipes for this model line). `main` was verified to load + serve `…IQ4_NL.gguf` on 2× RTX 5060 Ti.
+SignalLatch was already close to its BF16 source on the mixed probe snapshot. PatchCode keeps that small-form-factor Q4_NL path as the main deployment target and tests whether the agentic-coder distil improves practical coding-agent behaviour on top of it.
 
-Build (CUDA):
-```bash
-cmake -B build -DGGML_CUDA=ON \
-  -DCMAKE_CUDA_ARCHITECTURES="86;120" \
-  -DCOMPRESSION_MODE=speed   # ik_llama speed build
-cmake --build build --config Release -j
-```
+Practical eval: under a hardened 5-seed, same-condition bake-off (160k-token real-world multi-file build as the discriminator — single-shot coding gates saturate and were rejected), PatchCode `IQ4_NL` tied `BF16` within noise on build, long-context, discipline, and autonomous-loop convergence, at ~⅓ the size. See the eval snapshot below.
 
-An optional MTP speculative-decoding **speed** patch is bundled as `qwen36-mtp-rys_delta.patch` (it is **not** required to load/serve — only for the decode speed-up; without it the server uses normal autoregressive decode).
+Release files:
+- `Qwen3.6-27B-AEON-RYS-Agentic-Coder-PatchCode.IQ4_NL.gguf`
+- `Qwen3.6-27B-AEON-RYS-Agentic-Coder-PatchCode.BF16.gguf`
+- `qwen36-mtp-rys_delta.patch` (optional ik-llama MTP speed patch — **not** required to load/serve)
 
----
+Use these as merged GGUF files. They are not intended to be loaded as live LoRAs at inference time.
 
-## Serve (ik_llama / ik-llama.cpp)
+The recommended practical deployment file is the `IQ4_NL` GGUF. The `BF16` GGUF is provided as a single source-quality exploration artifact, not the normal runtime target.
 
-```bash
-llama-server \
-  -m Qwen3.6-27B-AEON-RYS-Agentic-Coder-PatchCode.IQ4_NL.gguf \
-  -a patchcode --host 0.0.0.0 --port 8000 \
-  -ngl 99 -c 163840 -ctk f16 -ctv f16 -b 512 -ub 128 \
-  --jinja --reasoning-format deepseek --flash-attn on
-```
+## Which file should I use?
 
-- **`<think>` is emitted as a separate `reasoning_content` field** (not in `content`). Use `--reasoning-format deepseek` (or fold `reasoning_content` back into `<think>…</think>` in your harness) so tool-action parsing sees the action, not the chain-of-thought.
-- IQ4_NL (16.6 GB weights) + f16 KV at 160k ctx ≈ ~45–55 GB VRAM total → fits comfortably on a single 80 GB card or a small multi-card pool.
+Most people should start with:
 
----
+`Qwen3.6-27B-AEON-RYS-Agentic-Coder-PatchCode.IQ4_NL.gguf`
 
-## Files
+That file is the intended release artifact. It is the continuation of the AEON RYS → SignalLatch → PatchCode line: keep the model small enough to be practical, then tune and test the stack until the small file gives the strongest useful behaviour we can get from it.
 
-| file | size | notes |
-|---|---|---|
-| `Qwen3.6-27B-AEON-RYS-Agentic-Coder-PatchCode.IQ4_NL.gguf` | 16.6 GB | **Recommended.** Plain IQ4_NL, reasoning/coding imatrix. Ship pick. |
-| `Qwen3.6-27B-AEON-RYS-Agentic-Coder-PatchCode.BF16.gguf` | 57.6 GB | Full-precision source / control ceiling. |
-| `qwen36-mtp-rys_delta.patch` | 57 KB | Optional ik_llama MTP speed patch (loader not required for use). |
+Use the single-file `BF16` GGUF only if you want to explore the merged model directly, make your own quant, compare conversion settings, or continue downstream work from the fine-tuned merge.
 
----
+## At a glance
 
-## Quant bake-off (why IQ4_NL ships over BF16 / Q8 / mixed recipes)
+- base line: `Qwen3.6-27B-AEON-RYS-SignalLatch-ckpt386-s010` (SignalLatch)
+- upstream AEON source: `AEON-7/Qwen3.6-27B-AEON-Ultimate-Uncensored`
+- fine-tune: agentic-coder joint behaviour LoRA, checkpoint `3661`, one epoch
+- merge strength: `0.5` (effective alpha/r = 1.0)
+- main release artifact: `IQ4_NL` GGUF
+- goal: maximum practical coding-agent behaviour in a small-form-factor GGUF
+- recommended runtime file size: about `16.6 GB`
+- companion source-quality artifact: single-file `BF16` GGUF, about `57.6 GB`
+- intended runtime: `https://github.com/noonr48/qwen36-aeon-ik-llama`
+- focus: practical coding-agent and tool-use behaviour
+- public name: `PatchCode`
+- behaviour loop: `Investigate → Act → Verify → Repair → Confirm`
+- not a general chat benchmark claim
+- not a stock `llama.cpp` / `vLLM` release
 
-Full 5-phase bake-off using a hardened SignalLatch-style methodology (the discriminator is a 160k-token real-world multi-file build, scored multi-seed — single-shot coding gates saturate and were rejected):
+## What changed vs the SignalLatch release
 
-| candidate | build (KritaLite) | long-context | discipline | autonomy-loop | size |
-|---|---|---|---|---|---|
-| **IQ4_NL (reasoning imatrix)** | 0.920 (±0.067) | 0.975 | 0.842 (±0.333) | 100% / 0% halluc | **16 G** |
-| BF16 (control) | 0.867 | 0.942 | 0.931 | — | 58 G |
-| Q8_0 | 0.867 | 0.969 | 0.742 | 100% | 29 G |
-| mixed-recipe (promoted attn) | 0.907–0.933 | 0.935 | 0.69–0.90 | 100% | 20–24 G |
+The previous SignalLatch file is the base deployment target this is meant to improve:
 
-- **IQ4_NL ties BF16 within noise** on build / long-context / discipline / autonomous-loop convergence (8 held-out agentic tasks × 5 seeds: 100% convergence, 0% hallucinated-success, 0% stall for every quant). Near-lossless Q8 and promoted-attention mixed recipes bought **no measurable edge** and cost 2–6× the size.
-- **Key methodology lesson:** on this suite, KritaLite build has ±0.067–0.13 run-to-run variance and discipline ±0.3. A 3-seed single-condition run shipped a *false* winner (a mixed recipe scored 0.933 once, never reproduced). Only 5+ seed same-condition head-to-heads + non-noise axes (size, recipe safety) reliably tiebreak. IQ4_NL wins on size + the plain-quant recipe (no promotion risk).
+`Qwen3.6-27B-AEON-RYS-SignalLatch-ckpt386-s010-IQ4_NL.gguf`
 
----
+hosted at `https://huggingface.co/jackasda211233/Qwen3.6-27B-AEON-RYS-SignalLatch-GGUF`.
 
-## Lineage
+This upload merges an agentic-coder joint behaviour LoRA into that already-strong SignalLatch line before exporting to `IQ4_NL`. The goal is not to make a new general-purpose model family. The goal is to improve practical code-agent behaviour while preserving the practical small-file deployment path: following repo-edit instructions, handling tool-shaped context, finishing concrete patches, and avoiding repeated timeout-like failures.
 
+Training summary:
+- dataset: ~`58.5k` agentic-coding behaviour examples (coding execution traces + action-first style traces)
+- training completion: checkpoint `3661`, one epoch
+- LoRA rank: `32`
+- LoRA alpha: `64`
+- LoRA dropout: `0.05`
+- target modules: all-linear, incl. the hybrid self-attn + linear-attn/SSM + MLP projections
+- selected merge strength: `0.5`
+
+## Recommended runtime
+
+Use the custom AEON ik-llama fork:
+
+`https://github.com/noonr48/qwen36-aeon-ik-llama`
+
+Tested server shape (single GPU / single slot):
+
+```bash
+./build/bin/llama-server \
+  -m /path/to/Qwen3.6-27B-AEON-RYS-Agentic-Coder-PatchCode.IQ4_NL.gguf \
+  -c 65536 \
+  -ngl 999 \
+  -np 1 \
+  -fa on \
+  -sm none \
+  --temp 0.7 \
+  --jinja \
+  --reasoning-format deepseek \
+  --reasoning-budget 0
 ```
-Qwen/Qwen3.6-27B                      (Apache-2.0)
-  └─ AEON-RYS abliteration            (uncensored; safety rails surgically removed)
-       └─ + prior behaviour LoRA @ light strength   →  ckpt386_s010
-            └─ + Agentic-Coder joint LoRA @ λ=0.5   →  this model (merged_lam0.5)
+
+Long-context single-slot reference:
+
+```bash
+./build/bin/llama-server \
+  -m /path/to/Qwen3.6-27B-AEON-RYS-Agentic-Coder-PatchCode.IQ4_NL.gguf \
+  -c 163840 \
+  -np 1 \
+  -ngl 999 \
+  -b 512 \
+  -ub 128 \
+  -fa on \
+  -sm none \
+  -ctk f16 \
+  -ctv f16 \
+  --temp 0.7 \
+  --jinja \
+  --reasoning-format deepseek \
+  --reasoning-budget 0
 ```
 
-`base_model` lineage: `Qwen/Qwen3.6-27B` → AEON-RYS → `ckpt386_s010` (AEON-RYS + a light prior behaviour LoRA) → this model (joint agentic-coder LoRA, checkpoint-3661, merged at λ=0.5).
+Runtime notes:
+- `<think>` is emitted as a separate `reasoning_content` field. Use `--reasoning-format deepseek` (or fold `reasoning_content` back into `<think>…</think>` in your harness) so tool-action parsing sees the action, not the chain-of-thought.
+- use the merged GGUF as the deployment artifact
+- prefer the `IQ4_NL` file for practical deployment
+- use the single-file `BF16` GGUF as the source-quality merged artifact for downstream quantization or further work
+- the tested profile uses flash attention; `-sm none` for one visible GPU, `-sm layer` for multi-GPU RAM-cache parallel lanes
+- live LoRA loading is not the production path for this release
+- the chat/runtime format should use Jinja plus DeepSeek reasoning formatting
+- for one visible GPU use `-sm none`. `-sm graph` requires at least two visible GPU devices and will fail during model load if the process is pinned to one GPU.
 
----
+## Practical eval snapshot — quant bake-off (5-seed, same-condition)
+
+These numbers are from an internal practical coding-agent build matrix. This is not an academic benchmark. The discriminator is a 160k-token real-world multi-file build, scored multi-seed (single-shot coding gates saturate on this model family and were rejected).
+
+| candidate | build (KritaLite) | long-context | discipline (style) | autonomy-loop | size |
+|---|---:|---:|---:|---|---:|
+| **PatchCode IQ4_NL (reasoning imatrix)** | `0.920` (±0.067) | `0.975` | `0.842` (±0.333) | 100% / 0% halluc | `16.6 G` |
+| BF16 (control) | `0.867` | `0.942` | `0.931` | — | `57.6 G` |
+| Q8_0 | `0.867` | `0.969` | `0.742` | 100% | `29 G` |
+| mixed-recipe (promoted attn) | `0.907–0.933` | `0.935` | `0.69–0.90` | 100% | `20–24 G` |
+
+Read:
+- PatchCode `IQ4_NL` ties `BF16` within noise on build, long-context, discipline, and autonomous-loop convergence (8 held-out agentic tasks × 5 seeds: 100% convergence, 0% hallucinated-success, 0% stall for every quant).
+- Near-lossless `Q8_0` and promoted-attention mixed recipes bought **no measurable edge** and cost 2–6× the size.
+- This is a tentative long-term default, not a claim that the fine-tune is solved. Build has ±0.067–0.13 run-to-run variance and discipline ±0.3 on this suite; a 3-seed single-condition run shipped a *false* winner (a mixed recipe scored 0.933 once, never reproduced). Only 5+ seed same-condition head-to-heads + non-noise axes (size, recipe safety) reliably tiebreak. `IQ4_NL` wins on size + the plain-quant recipe (no promotion risk).
+
+Autonomy-loop detail (`agent_loop_eval`, 8 held-out tasks × 5 seeds):
+
+| quant | convergence | mean turns | recovery | halluc-success | stall |
+|---|---:|---:|---:|---:|---:|
+| PatchCode IQ4_NL | 100% (40/40) | 7.2 | 0.4 | 0% | 0% |
+| Q8_0 | 100% (40/40) | 7.0 | 0.5 | 0% | 0% |
+| mixed (c76) | 100% (40/40) | 6.6 | 0.4 | 0% | 0% |
+
+Non-discriminating (0pp spread): the action-first, verify-before-claim discipline is preserved across all quants.
+
+## BF16 vs released IQ4_NL
+
+For this release:
+
+| item | value |
+|---|---|
+| BF16 size | `57.6 G` |
+| released IQ4_NL size | `16.6 G` |
+| build (5-seed) | `0.867` BF16 → `0.920` IQ4_NL |
+| long-context | `0.942` BF16 → `0.975` IQ4_NL |
+
+Short read:
+- about `70%` smaller on disk
+- IQ4_NL at or above BF16 on every practical axis (tied within run-to-run noise — not a quality cliff)
+
+## Why no stock `llama.cpp` / `vLLM` file
+
+We are not publishing a separate standard `llama.cpp` or `vLLM` model file as part of this release.
+
+Why:
+- the model needs the forked `ik-llama` runtime (Qwen3.6 hybrid/recurrent loader + graph-split long-context fixes + the custom mixed GGUF tensor layout)
+- stock upstream runtimes hit real load failures on the `qwen3_5` triple-hybrid architecture
+- because a special runtime was required either way, we did not think it was worth presenting a second public file as if plain `llama.cpp` / `vLLM` support were the point of the project
+
+So the intended path is:
+- use the fork: `https://github.com/noonr48/qwen36-aeon-ik-llama`
+- use the released `IQ4_NL` GGUF (or the `BF16` source artifact)
+- do not present these as stock `llama.cpp` / `vLLM` targets
+
+## Optional MTP speed patch
+
+The bundled `qwen36-mtp-rys_delta.patch` is an optional ik-llama MTP speculative-decoding **speed** patch.
+
+- it is **not** required to load or serve the model — without it the server uses normal autoregressive decode
+- in our tests the MTP path was technically interesting but **not** the better default (the non-MTP file was faster and cleaner in practical evals)
+- use it only if you are testing MTP behaviour or want the experimental decode speed-up on the fork
+
+## Hyper-focused project
+
+This was a deliberately narrow project.
+
+The target was not "best general chat model". The target was:
+- strongest Q4-class English-first model we could get for coding, reasoning, and academic work
+- derived from the AEON uncensored branch
+- distilled/calibrated toward agentic coding execution and tool use
 
 ## License
 
-Apache-2.0, inherited from `Qwen/Qwen3.6-27B` (and the AEON-RYS abliteration). The Qwen3.6 base license permits derivative redistribution; attribute the base model and the AEON-RYS abliteration.
+Apache-2.0, inherited from `Qwen/Qwen3.6-27B` via the AEON-RYS abliteration. The base license permits derivative redistribution; attribute the base model and the AEON-RYS abliteration.
 
 > Uncensored / abliterated: this derivative has had refusal/safety steering removed at the base. Use responsibly and in accordance with your local laws and platform policies.