Refresh model card for qwen3_5_4b_runtime_dialogue_unsloth_wsl_full_gatefix_20260310

README.md

@@ -1,33 +1,47 @@
----
-base_model: Qwen/Qwen3.5-4B
-tags:
-- acidnet
-- lora
-- gguf
-- qwen
----
-
-# AcidNet Model Artifacts
-
-This repo stores portable AcidNet model artifacts.
-
-It is a registry, not the live runtime mount. AcidNet still serves and evaluates local files after you restore them into the repo working tree.
-
-## Latest Published Run
-
-- run: `qwen3_5_4b_runtime_dialogue_unsloth_wsl_full_headsync_20260308`
-- base model: `Qwen/Qwen3.5-4B`
-- adapter repo path: `runs/qwen3_5_4b_runtime_dialogue_unsloth_wsl_full_headsync_20260308/adapter`
-- publish manifest: `runs/qwen3_5_4b_runtime_dialogue_unsloth_wsl_full_headsync_20260308/manifests/publish_manifest.json`
-- repo card: `README.md`
-
-## Restore Map
-
-- restore `runs/qwen3_5_4b_runtime_dialogue_unsloth_wsl_full_headsync_20260308/adapter` into a local directory such as `data/training/qwen3_5_4b_runtime_dialogue_unsloth_wsl_full_headsync_20260308_adapter/` for `local_peft` dev/eval parity
-- restore the GGUF files below into `data/gguf/` for `llama-server`
-- keep the base `Q4_K_M` model in a separate local path such as `models/`; this repo stores only the fine-tuned adapter artifacts
-
-## GGUF Files
-
-- `data/gguf/qwen3_5_4b_runtime_dialogue_unsloth_wsl_full_headsync_20260308_adapter-f16.gguf` -> `runs/qwen3_5_4b_runtime_dialogue_unsloth_wsl_full_headsync_20260308/gguf/adapter-f16.gguf`
-- `data/gguf/qwen3_5_4b_runtime_dialogue_unsloth_wsl_full_headsync_20260308_adapter_manifest.json` -> `runs/qwen3_5_4b_runtime_dialogue_unsloth_wsl_full_headsync_20260308/gguf/adapter_manifest.json`
+---
+base_model: Qwen/Qwen3.5-4B
+tags:
+- acidnet
+- lora
+- gguf
+- qwen
+---
+
+# AcidNet Model Artifacts
+
+This repo stores portable AcidNet model artifacts.
+
+It is a registry, not the live runtime mount. AcidNet still serves and evaluates local files after you restore them into the repo working tree.
+
+## Latest Uploaded Run
+
+- run: `qwen3_5_4b_runtime_dialogue_unsloth_wsl_full_gatefix_20260310`
+- promotion status: `failed_gate`
+- base model: `Qwen/Qwen3.5-4B`
+- adapter repo path: `runs/qwen3_5_4b_runtime_dialogue_unsloth_wsl_full_gatefix_20260310/adapter`
+- publish manifest: `runs/qwen3_5_4b_runtime_dialogue_unsloth_wsl_full_gatefix_20260310/manifests/publish_manifest.json`
+- repo card: `README.md`
+
+## Promoted Alias
+
+- not updated by this publish
+
+## Gate Summary
+
+- gate passed: `False`
+- prompt average score: `0.979`
+- prompt rows with failures: `4`
+- prompt average latency ms: `2444.762`
+- circulation score: `0.807`
+- starving npc count: `0`
+- source report: `data/eval/model_gate_runtime_dialogue_unsloth_wsl_full_gatefix_20260310_report.json`
+
+## Restore Map
+
+- restore `runs/qwen3_5_4b_runtime_dialogue_unsloth_wsl_full_gatefix_20260310/adapter` into a local directory such as `data/training/qwen3_5_4b_runtime_dialogue_unsloth_wsl_full_gatefix_20260310_adapter/` for `local_peft` dev/eval parity
+- restore the GGUF files below into `data/gguf/` for `llama-server`
+- keep the base `Q4_K_M` model in a separate local path such as `models/`; this repo stores only the fine-tuned adapter artifacts
+
+## GGUF Files
+
+- none