Add TinyMind GGUF-X PureCode sidecar evidence pack

.gitattributes

@@ -1,35 +1,3 @@
-*.7z filter=lfs diff=lfs merge=lfs -text
-*.arrow filter=lfs diff=lfs merge=lfs -text
-*.bin filter=lfs diff=lfs merge=lfs -text
-*.bz2 filter=lfs diff=lfs merge=lfs -text
-*.ckpt filter=lfs diff=lfs merge=lfs -text
-*.ftz filter=lfs diff=lfs merge=lfs -text
-*.gz filter=lfs diff=lfs merge=lfs -text
-*.h5 filter=lfs diff=lfs merge=lfs -text
-*.joblib filter=lfs diff=lfs merge=lfs -text
-*.lfs.* filter=lfs diff=lfs merge=lfs -text
-*.mlmodel filter=lfs diff=lfs merge=lfs -text
-*.model filter=lfs diff=lfs merge=lfs -text
-*.msgpack filter=lfs diff=lfs merge=lfs -text
-*.npy filter=lfs diff=lfs merge=lfs -text
-*.npz filter=lfs diff=lfs merge=lfs -text
-*.onnx filter=lfs diff=lfs merge=lfs -text
-*.ot filter=lfs diff=lfs merge=lfs -text
-*.parquet filter=lfs diff=lfs merge=lfs -text
-*.pb filter=lfs diff=lfs merge=lfs -text
-*.pickle filter=lfs diff=lfs merge=lfs -text
-*.pkl filter=lfs diff=lfs merge=lfs -text
-*.pt filter=lfs diff=lfs merge=lfs -text
-*.pth filter=lfs diff=lfs merge=lfs -text
-*.rar filter=lfs diff=lfs merge=lfs -text
+*.gguf filter=lfs diff=lfs merge=lfs -text
 *.safetensors filter=lfs diff=lfs merge=lfs -text
-saved_model/**/* filter=lfs diff=lfs merge=lfs -text
-*.tar.* filter=lfs diff=lfs merge=lfs -text
-*.tar filter=lfs diff=lfs merge=lfs -text
-*.tflite filter=lfs diff=lfs merge=lfs -text
-*.tgz filter=lfs diff=lfs merge=lfs -text
-*.wasm filter=lfs diff=lfs merge=lfs -text
-*.xz 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
+*.jsonl filter=lfs diff=lfs merge=lfs -text

Modelfile.evo

@@ -0,0 +1,30 @@
+FROM model\astraweave-fusion\artifacts\tinymind-purebase.gguf
+
+PARAMETER temperature 0.14
+PARAMETER top_p 0.78
+PARAMETER top_k 32
+PARAMETER num_ctx 65536
+PARAMETER num_predict 8192
+PARAMETER repeat_penalty 1.19
+PARAMETER repeat_last_n 4096
+PARAMETER mirostat 2
+PARAMETER mirostat_tau 4.2
+PARAMETER mirostat_eta 0.08
+
+SYSTEM """
+You are TinyMind GGUF Evo, an evidence-first local model runtime.
+
+Operating law:
+- Answer from grounded evidence when available; clearly mark uncertainty when evidence is missing.
+- Prefer concise structure for simple tasks and deep step-by-step reasoning for hard tasks.
+- Preserve Thai and English nuance; do not translate away technical meaning.
+- For code, provide runnable, minimal, audited patches or commands.
+- For long context, summarize anchors first, then answer from exact anchors.
+- Never claim the GGUF weights were retrained unless a saved training/export manifest proves it.
+- Refuse credential leakage, destructive actions, exploit chains, stealth, and malware improvement.
+
+Quality style:
+- Be direct, natural, and precise.
+- Separate Fact, Inference, and Next Verification when stakes are high.
+- If unsure, propose the smallest real measurement that resolves uncertainty.
+"""

README.md

@@ -0,0 +1,60 @@
+---
+license: other
+language:
+- th
+- en
+tags:
+- gguf
+- tinymind
+- qlora
+- evidence-gated
+- local-llm
+pipeline_tag: text-generation
+library_name: gguf
+base_model: mistralai/Mistral-Nemo-Instruct-2407
+---
+
+# TinyMind GGUF-X PureCode
+
+This repository contains the TinyMind GGUF-X PureCode runtime package.
+
+GGUF-X is a TinyMind packaging format: a GGUF v3-compatible binary runtime plus sidecar evidence, decoding policy, purity lineage, and claim gates. It is designed to keep the runtime compatible with GGUF loaders while attaching the evidence needed to avoid fake claims.
+
+## What Is Included
+
+- `Modelfile.evo`: Ollama-compatible runtime profile.
+- `tinymind_ggufx_spec.json`: TinyMind GGUF-X sidecar spec.
+- `gguf_evo_upgrade_manifest.json`: hashes, lineage, runtime settings, and promotion gates.
+- `gguf_evo_eval_prompts.jsonl`: smoke eval prompts.
+- `README_GGUF_X.md`: concise format boundary.
+
+## Current Evidence
+
+- Source GGUF size: `15.7098 GB`
+- Source GGUF SHA-256: `b8707e57f676d8dd1b80f623b45200cc92e6966b0e95275e606f412095a49fde`
+- Adapter training evidence present: `True`
+- Adapter eval loss: `2.808002471923828`
+- Adapter perplexity: `16.576772556518506`
+- Purity density: `0.7211701159881736`
+- Dominant domain share: `0.17511939959062997`
+- Runtime context requested: `65536`
+
+## Claim Boundary
+
+- Custom TinyMind format pack: `True`
+- GGUF v3 compatibility: `True`
+- Better-than-GGUF-v3 claim: `False`
+- Weights-better-than-source claim: `False`
+
+The current package does not claim that GGUF tensors were merged/retrained inside the binary. Weight-level improvement requires a real adapter-to-GGUF merge/export log and same-prompt baseline comparison.
+
+## Reproduce Locally
+
+```powershell
+ollama create tinymind-ggufx-purecode -f .\Modelfile.evo
+ollama run tinymind-ggufx-purecode "อธิบายความต่างของ QLoRA กับ GGUF"
+```
+
+## Repository
+
+HF repo target: `bbkdevops/tinymind-ggufx-purecode`

README_GGUF_X.md

@@ -0,0 +1,19 @@
+# TinyMind GGUF-X
+
+GGUF-X is a TinyMind runtime package: a GGUF v3-compatible model plus sidecar evidence, decode policy, and lineage gates.
+It does not mutate GGUF tensors directly and does not claim better weights without a real merge/export/eval report.
+
+## Files
+
+- Modelfile: `reports\ggufx_purecode_latest\Modelfile.evo`
+- Spec: `reports\ggufx_purecode_latest\tinymind_ggufx_spec.json`
+- Eval prompts: `reports\ggufx_purecode_latest\gguf_evo_eval_prompts.jsonl`
+- Create script: `reports\ggufx_purecode_latest\create_ollama_evo.ps1`
+- Compare script: `reports\ggufx_purecode_latest\compare_rawzero_vs_evo.ps1`
+
+## Claim Boundary
+
+- Custom TinyMind format pack: allowed
+- GGUF v3 runtime compatibility: allowed
+- Better-than-source GGUF weights: blocked until same-prompt baseline evidence exists
+- World-best/runtime superiority: blocked until external benchmark evidence exists

compare_rawzero_vs_evo.ps1

@@ -0,0 +1,9 @@
+$ErrorActionPreference = 'Stop'
+$prompts = Get-Content -LiteralPath 'reports\ggufx_purecode_latest\gguf_evo_eval_prompts.jsonl' | ForEach-Object { $_ | ConvertFrom-Json }
+$out = @()
+foreach ($p in $prompts) {
+  $raw = ollama run tinymind-rawzero-fusion $p.prompt
+  $evo = ollama run tinymind-ggufx-purecode $p.prompt
+  $out += [ordered]@{ id=$p.id; prompt=$p.prompt; rawzero=$raw; evo=$evo; checks=$p.checks }
+}
+$out | ConvertTo-Json -Depth 8 | Set-Content -Path 'reports\ggufx_purecode_latest\rawzero_vs_evo_outputs.json' -Encoding UTF8

create_ollama_evo.ps1

@@ -0,0 +1,2 @@
+ollama create tinymind-ggufx-purecode -f "reports\ggufx_purecode_latest\Modelfile.evo"
+ollama show tinymind-ggufx-purecode

gguf_evo_eval_prompts.jsonl

@@ -0,0 +1,4 @@
+{"checks": ["ภาษาไทย", "QLoRA", "GGUF", "ไม่อ้างว่า train GGUF ตรงๆ"], "id": "thai_technical_explain", "prompt": "อธิบาย QLoRA กับ GGUF ต่างกันอย่างไรแบบเข้าใจง่ายแต่ครบถ้วน"}
+{"checks": ["skip invalid line", "strict=False or guarded decode", "test"], "id": "code_patch_reasoning", "prompt": "Given a Python JSONL reader that crashes on one bad line, design a robust fix and test plan."}
+{"checks": ["no", "external eval", "saved evidence"], "id": "grounding_boundary", "prompt": "Can you claim this model is world best after one local smoke eval?"}
+{"checks": ["external ledger", "hash", "retrieval", "regenerated KV"], "id": "long_context_anchor", "prompt": "In a 10M token archive, how can a small model recall exact details without hallucinating?"}

gguf_evo_upgrade_manifest.json

@@ -0,0 +1,104 @@
+{
+  "base_modelfile": "model\\astraweave-fusion\\artifacts\\Modelfile.rawzero",
+  "compare_script": "reports\\ggufx_purecode_latest\\compare_rawzero_vs_evo.ps1",
+  "create_script": "reports\\ggufx_purecode_latest\\create_ollama_evo.ps1",
+  "created_at": "2026-05-27T16:22:59.715373+00:00",
+  "eval_prompts": "reports\\ggufx_purecode_latest\\gguf_evo_eval_prompts.jsonl",
+  "evo_modelfile": "reports\\ggufx_purecode_latest\\Modelfile.evo",
+  "ggufx_format": {
+    "binary_compatibility": {
+      "base_container": "GGUF",
+      "base_container_version": 3,
+      "runtime_target": "Ollama / llama.cpp compatible GGUF loader",
+      "sidecar_required_for_tinymind_features": true
+    },
+    "claim_rules": {
+      "may_claim_better_than_gguf_v3": false,
+      "may_claim_binary_gguf_v3_compatibility": true,
+      "may_claim_custom_tinymind_format_pack": true,
+      "may_claim_weights_retrained_inside_gguf": false,
+      "required_to_unlock_better_than_v3_claim": [
+        "real adapter-to-GGUF merge/export log",
+        "baseline GGUF v3 eval",
+        "GGUF-X eval on same prompts",
+        "latency/memory report",
+        "hashes for every artifact"
+      ]
+    },
+    "format_kind": "GGUF-v3-compatible-binary-plus-TinyMind-sidecar",
+    "schema_version": "tinymind-gguf-x-v1",
+    "tinymind_extensions": {
+      "adapter_lineage_gate": {
+        "adapter_manifest": "model\\tinymind-12b\\adapters\\tinymind-12b-ggufx-purecode-lr3e7-s4-25690527_232007\\tinymind_12b_manifest.json",
+        "enabled": true,
+        "purpose": "separate trained LoRA evidence from GGUF packaging evidence",
+        "training_manifest": "model\\tinymind-12b\\adapters\\tinymind-12b-ggufx-purecode-lr3e7-s4-25690527_232007\\tinymind_12b_manifest.json"
+      },
+      "evidence_first_decode_law": {
+        "enabled": true,
+        "purpose": "reduce hallucination/repetition without pretending the GGUF weights changed",
+        "repeat_penalty": 1.19,
+        "temperature": 0.14,
+        "top_p": 0.78
+      },
+      "purity_lineage_gate": {
+        "enabled": true,
+        "manifest": "reports\\purity_concentrator_code_puremax_latest\\purity_concentrator_manifest.json",
+        "purpose": "tie runtime claims to the exact purity-concentrated dataset used for adapter training"
+      },
+      "regen_ledger_ready_metadata": {
+        "enabled": true,
+        "kv_growth_claim": "bounded only when paired with Evidence Ledger/ReGenesis retrieval runtime"
+      }
+    }
+  },
+  "ggufx_readme": "reports\\ggufx_purecode_latest\\README_GGUF_X.md",
+  "ggufx_spec": "reports\\ggufx_purecode_latest\\tinymind_ggufx_spec.json",
+  "lineage": {
+    "adapter_eval_loss": 2.808002471923828,
+    "adapter_manifest": "model\\tinymind-12b\\adapters\\tinymind-12b-ggufx-purecode-lr3e7-s4-25690527_232007\\tinymind_12b_manifest.json",
+    "adapter_perplexity": 16.576772556518506,
+    "data_manifest": "reports\\purity_concentrator_code_puremax_latest\\purity_concentrator_manifest.json",
+    "dominant_domain_share": 0.17511939959062997,
+    "purity_density": 0.7211701159881736,
+    "training_eval_loss": 2.808002471923828,
+    "training_manifest": "model\\tinymind-12b\\adapters\\tinymind-12b-ggufx-purecode-lr3e7-s4-25690527_232007\\tinymind_12b_manifest.json",
+    "training_perplexity": 16.576772556518506
+  },
+  "model_name": "tinymind-ggufx-purecode",
+  "promotion_gate": {
+    "adapter_training_evidence_present": true,
+    "can_claim_better_than_v3": false,
+    "can_claim_runtime_quality_upgrade": true,
+    "can_claim_weights_better_than_source": false,
+    "custom_ggufx_sidecar_created": true,
+    "evo_eval_required": true,
+    "gguf_binary_tensor_merge_performed": false,
+    "must_beat_baseline_on_prompt_suite": true,
+    "rawzero_baseline_required": true,
+    "reason": "This pack upgrades GGUF runtime behavior and evaluation path. Weight-level improvement requires real conversion/export from trained adapters plus benchmark evidence.",
+    "weight_training_performed": true
+  },
+  "runtime_upgrade": {
+    "context_window_requested": 65536,
+    "decode_profile": {
+      "mirostat": 2,
+      "repeat_penalty": 1.19,
+      "temperature": 0.14,
+      "top_k": 32,
+      "top_p": 0.78
+    },
+    "quality_controls": [
+      "evidence-first system law",
+      "Thai-English technical preservation",
+      "long-context anchor discipline",
+      "claim-boundary enforcement",
+      "lower-temperature repetition-resistant decoding"
+    ]
+  },
+  "schema_version": "tinymind-gguf-evo-upgrade-v1",
+  "source_gguf": "model\\astraweave-fusion\\artifacts\\tinymind-purebase.gguf",
+  "source_gguf_sha256": "b8707e57f676d8dd1b80f623b45200cc92e6966b0e95275e606f412095a49fde",
+  "source_gguf_size_bytes": 16868240704,
+  "source_gguf_size_gb": 15.709773361682892
+}

hf_package_manifest.json

@@ -0,0 +1,20 @@
+{
+  "files": [
+    ".gitattributes",
+    "Modelfile.evo",
+    "README.md",
+    "README_GGUF_X.md",
+    "compare_rawzero_vs_evo.ps1",
+    "create_ollama_evo.ps1",
+    "gguf_evo_eval_prompts.jsonl",
+    "gguf_evo_upgrade_manifest.json",
+    "tinymind_ggufx_spec.json"
+  ],
+  "large_binary_source": "model\\astraweave-fusion\\artifacts\\tinymind-purebase.gguf",
+  "large_binary_upload_required": true,
+  "out_dir": "reports\\hf_tinymind_ggufx_purecode_repo",
+  "repo_id": "bbkdevops/tinymind-ggufx-purecode",
+  "source_gguf": "model\\astraweave-fusion\\artifacts\\tinymind-purebase.gguf",
+  "source_gguf_sha256": "b8707e57f676d8dd1b80f623b45200cc92e6966b0e95275e606f412095a49fde",
+  "source_manifest": "reports\\ggufx_purecode_latest\\gguf_evo_upgrade_manifest.json"
+}

tinymind_ggufx_spec.json

@@ -0,0 +1,47 @@
+{
+  "binary_compatibility": {
+    "base_container": "GGUF",
+    "base_container_version": 3,
+    "runtime_target": "Ollama / llama.cpp compatible GGUF loader",
+    "sidecar_required_for_tinymind_features": true
+  },
+  "claim_rules": {
+    "may_claim_better_than_gguf_v3": false,
+    "may_claim_binary_gguf_v3_compatibility": true,
+    "may_claim_custom_tinymind_format_pack": true,
+    "may_claim_weights_retrained_inside_gguf": false,
+    "required_to_unlock_better_than_v3_claim": [
+      "real adapter-to-GGUF merge/export log",
+      "baseline GGUF v3 eval",
+      "GGUF-X eval on same prompts",
+      "latency/memory report",
+      "hashes for every artifact"
+    ]
+  },
+  "format_kind": "GGUF-v3-compatible-binary-plus-TinyMind-sidecar",
+  "schema_version": "tinymind-gguf-x-v1",
+  "tinymind_extensions": {
+    "adapter_lineage_gate": {
+      "adapter_manifest": "model\\tinymind-12b\\adapters\\tinymind-12b-ggufx-purecode-lr3e7-s4-25690527_232007\\tinymind_12b_manifest.json",
+      "enabled": true,
+      "purpose": "separate trained LoRA evidence from GGUF packaging evidence",
+      "training_manifest": "model\\tinymind-12b\\adapters\\tinymind-12b-ggufx-purecode-lr3e7-s4-25690527_232007\\tinymind_12b_manifest.json"
+    },
+    "evidence_first_decode_law": {
+      "enabled": true,
+      "purpose": "reduce hallucination/repetition without pretending the GGUF weights changed",
+      "repeat_penalty": 1.19,
+      "temperature": 0.14,
+      "top_p": 0.78
+    },
+    "purity_lineage_gate": {
+      "enabled": true,
+      "manifest": "reports\\purity_concentrator_code_puremax_latest\\purity_concentrator_manifest.json",
+      "purpose": "tie runtime claims to the exact purity-concentrated dataset used for adapter training"
+    },
+    "regen_ledger_ready_metadata": {
+      "enabled": true,
+      "kv_growth_claim": "bounded only when paired with Evidence Ledger/ReGenesis retrieval runtime"
+    }
+  }
+}