Text Generation
PEFT
Safetensors
English
qlora
governed-agent
proposal-only
research-only
szl-holdings
khipu
abstain-retrain
conversational
Instructions to use SZLHOLDINGS/KHIPU-R2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use SZLHOLDINGS/KHIPU-R2 with PEFT:
from peft import PeftModel from transformers import AutoModelForCausalLM base_model = AutoModelForCausalLM.from_pretrained("unsloth/Qwen2.5-1.5B-Instruct-bnb-4bit") model = PeftModel.from_pretrained(base_model, "SZLHOLDINGS/KHIPU-R2") - Notebooks
- Google Colab
- Kaggle
docs: 6a91ba2c ERROR abstain still 2/6; live 6a91bf11 RUNNING report_to=none
Browse files
README.md
CHANGED
|
@@ -2,7 +2,8 @@
|
|
| 2 |
license: apache-2.0
|
| 3 |
library_name: transformers
|
| 4 |
pipeline_tag: text-generation
|
| 5 |
-
base_model: Qwen/
|
|
|
|
| 6 |
tags:
|
| 7 |
- governed-ai
|
| 8 |
- receipt-agent
|
|
@@ -16,17 +17,17 @@ szl:
|
|
| 16 |
originality: FINETUNE_DISCLOSED_BASE
|
| 17 |
line: KHIPU-R2
|
| 18 |
jobs: RUNNING
|
| 19 |
-
job_id:
|
|
|
|
|
|
|
| 20 |
job_namespace: SZLHOLDINGS
|
|
|
|
| 21 |
never_overwrite: SZLHOLDINGS/SZL-Khipu-1.5B
|
| 22 |
---
|
| 23 |
|
| 24 |
# KHIPU-R2 — cutting
|
| 25 |
|
| 26 |
-
Fall 2026 collection. **No weights in this repository yet.** The card is the look. Weights land only with a signed training receipt.
|
| 27 |
-
|
| 28 |
-
**Silhouette:** small instruct (Qwen3 / Gemma 3 class).
|
| 29 |
-
**Our cut:** receipt-native agent. It drafts. It refuses. It never executes. A controller outside the weights validates JSON, requires human approval, and mints a DSSE receipt only after approved action.
|
| 30 |
|
| 31 |
This repo is the **abstain retrain** target. It does **not** overwrite signed [`SZL-Khipu-1.5B`](https://huggingface.co/SZLHOLDINGS/SZL-Khipu-1.5B).
|
| 32 |
|
|
@@ -34,33 +35,30 @@ This repo is the **abstain retrain** target. It does **not** overwrite signed [`
|
|
| 34 |
|---|---|
|
| 35 |
| **House** | [SZL Holdings](https://huggingface.co/SZLHOLDINGS) |
|
| 36 |
| **Line** | Khipu successor. Signed 1.5B stays immutable. |
|
| 37 |
-
| **
|
| 38 |
-
| **
|
|
|
|
| 39 |
| **License** | Apache-2.0 |
|
| 40 |
| **Λ** | Conjecture 1 (never a theorem) |
|
| 41 |
| **Doctrine** | v11 LOCKED · 749 / 14 / 163 |
|
| 42 |
-
| **Serve** |
|
| 43 |
-
|
| 44 |
-
## What it will do
|
| 45 |
|
| 46 |
-
|
| 47 |
-
- Refuse fabrication of evidence, approval, execution, or MEASURED numbers.
|
| 48 |
-
- Cite honest labels: MEASURED / REPORTED / DECLARED / SIMULATED / UNKNOWN / UNAVAILABLE.
|
| 49 |
|
| 50 |
-
|
| 51 |
|
| 52 |
-
-
|
| 53 |
-
-
|
| 54 |
-
-
|
| 55 |
-
-
|
| 56 |
-
-
|
| 57 |
|
| 58 |
-
##
|
| 59 |
-
|
| 60 |
-
Previous ready-to-wear: [`SZL-Khipu-1.5B`](https://huggingface.co/SZLHOLDINGS/SZL-Khipu-1.5B) (Qwen2.5-1.5B QLoRA) and [`szl-receiptagent-qwen35-0.8b-v2`](https://huggingface.co/SZLHOLDINGS/szl-receiptagent-qwen35-0.8b-v2) (MEASURED 5/5 JSON + 6/6 adversarial refusals on a small preregistered gate, not a public benchmark).
|
| 61 |
|
| 62 |
-
|
|
|
|
|
|
|
| 63 |
|
| 64 |
## Evidence bound
|
| 65 |
|
| 66 |
-
|
|
|
|
| 2 |
license: apache-2.0
|
| 3 |
library_name: transformers
|
| 4 |
pipeline_tag: text-generation
|
| 5 |
+
base_model: Qwen/Qwen2.5-1.5B-Instruct
|
| 6 |
+
base_model_relation: adapter
|
| 7 |
tags:
|
| 8 |
- governed-ai
|
| 9 |
- receipt-agent
|
|
|
|
| 17 |
originality: FINETUNE_DISCLOSED_BASE
|
| 18 |
line: KHIPU-R2
|
| 19 |
jobs: RUNNING
|
| 20 |
+
job_id: 6a91bf11984507d9db4ea104
|
| 21 |
+
job_retry: 6a91bf11984507d9db4ea104
|
| 22 |
+
job_prior_error: 6a91ba2c45686a1580c12020
|
| 23 |
job_namespace: SZLHOLDINGS
|
| 24 |
+
weights: UNAVAILABLE
|
| 25 |
never_overwrite: SZLHOLDINGS/SZL-Khipu-1.5B
|
| 26 |
---
|
| 27 |
|
| 28 |
# KHIPU-R2 — cutting
|
| 29 |
|
| 30 |
+
Fall 2026 collection. **No weights in this repository yet.** The card is the look. Weights land only with a signed training receipt.
|
|
|
|
|
|
|
|
|
|
| 31 |
|
| 32 |
This repo is the **abstain retrain** target. It does **not** overwrite signed [`SZL-Khipu-1.5B`](https://huggingface.co/SZLHOLDINGS/SZL-Khipu-1.5B).
|
| 33 |
|
|
|
|
| 35 |
|---|---|
|
| 36 |
| **House** | [SZL Holdings](https://huggingface.co/SZLHOLDINGS) |
|
| 37 |
| **Line** | Khipu successor. Signed 1.5B stays immutable. |
|
| 38 |
+
| **Base** | `Qwen/Qwen2.5-1.5B-Instruct` (adapter). Runtime `unsloth/Qwen2.5-1.5B-Instruct-bnb-4bit`. |
|
| 39 |
+
| **HF Jobs** | Live **RUNNING** [`6a91bf11984507d9db4ea104`](https://huggingface.co/jobs/SZLHOLDINGS/6a91bf11984507d9db4ea104) (`report_to=none`). Not MEASURED until files exist. |
|
| 40 |
+
| **Status** | CUTTING · ROADMAP until tensors + receipt land |
|
| 41 |
| **License** | Apache-2.0 |
|
| 42 |
| **Λ** | Conjecture 1 (never a theorem) |
|
| 43 |
| **Doctrine** | v11 LOCKED · 749 / 14 / 163 |
|
| 44 |
+
| **Serve** | Lab load forbidden. Lab serves signed GGUF only. |
|
|
|
|
|
|
|
| 45 |
|
| 46 |
+
## Prior job (ERROR, not RUNNING)
|
|
|
|
|
|
|
| 47 |
|
| 48 |
+
[`6a91ba2c45686a1580c12020`](https://huggingface.co/jobs/SZLHOLDINGS/6a91ba2c45686a1580c12020) ERROR after train+eval.
|
| 49 |
|
| 50 |
+
- finalTrainLoss **0.0171** (train metric, not an eval)
|
| 51 |
+
- Held-out **MEASURED**: plan-valid 7/11, routing 3/5, abstain **2/6** (same blocker as signed 1.5B), hallucinated-citations 1
|
| 52 |
+
- 2/6 is **not** a pass
|
| 53 |
+
- Adapter sha `36b8123e` local only; `upload_folder` died on Trackio 404 `betterwithage/trackio-bucket`
|
| 54 |
+
- No tensors on Hub
|
| 55 |
|
| 56 |
+
## What it will not do
|
|
|
|
|
|
|
| 57 |
|
| 58 |
+
- Not an overwrite of `SZL-Khipu-1.5B`
|
| 59 |
+
- Not a Trackio dashboard (404, no URL)
|
| 60 |
+
- Not MEASURED capability until a receipt and adapters exist on this repo
|
| 61 |
|
| 62 |
## Evidence bound
|
| 63 |
|
| 64 |
+
Until adapters and `training_receipt.json` exist here, capability sentences are **ROADMAP**.
|