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
chore(receipt): Khipu abstain training receipt
Browse files- training_receipt.json +164 -0
training_receipt.json
ADDED
|
@@ -0,0 +1,164 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"kind": "szl-khipu-abstain-training-receipt",
|
| 3 |
+
"schema": "szl.frontier-training-run/v1",
|
| 4 |
+
"v": 1,
|
| 5 |
+
"capabilityProfile": "SZL-Khipu-1.5B-BrainNavigator",
|
| 6 |
+
"artifact": "SZLHOLDINGS/KHIPU-R2",
|
| 7 |
+
"baseModel": "Qwen/Qwen2.5-1.5B-Instruct",
|
| 8 |
+
"base_model": "Qwen/Qwen2.5-1.5B-Instruct",
|
| 9 |
+
"base_model_relation": "adapter",
|
| 10 |
+
"base_model_runtime": "unsloth/Qwen2.5-1.5B-Instruct-bnb-4bit",
|
| 11 |
+
"does_not_overwrite": "SZLHOLDINGS/SZL-Khipu-1.5B",
|
| 12 |
+
"datasets": {
|
| 13 |
+
"train.jsonl": "f0f8a9b232e8662f65eda1a58e3875ee9c1f859851ef3c2bfb28dd727cc27a75",
|
| 14 |
+
"eval.jsonl": "61ede1488e3c6e3cded81679affe258e8d03c47019424182330a94b8c505794e",
|
| 15 |
+
"train.abstain.jsonl": "421a6e733fda656c18b250ad5a5140f010392598750c48d672972f45a1e6c4a6",
|
| 16 |
+
"adversarial.jsonl": "812a23b3ed15c1df8c5e18b2365b6e7c474968f42f329a5f30b7c57c445659fd",
|
| 17 |
+
"khipu.schema.json": "b95f9927366dae7c5d36cfb7de6e229eb605524318ab642a6aa2292a212170d0"
|
| 18 |
+
},
|
| 19 |
+
"schemaFingerprintSha256": "f05e38b406b5e893e8a7dd23029a0c252b5e53994e8ac777a810b227dc2d7e64",
|
| 20 |
+
"outputSchemaSha256": "b95f9927366dae7c5d36cfb7de6e229eb605524318ab642a6aa2292a212170d0",
|
| 21 |
+
"adapterSha256": "e44d53f29f2d443598e06d6c0441557fd3a5010888c7aa97b56ec3c0e050d349",
|
| 22 |
+
"ABSTAIN_OVERSAMPLE": 4,
|
| 23 |
+
"train_navigate_rows": 15,
|
| 24 |
+
"train_abstain_rows_committed": 8,
|
| 25 |
+
"train_abstain_rows_in_memory": 32,
|
| 26 |
+
"training_rows_in_memory": 47,
|
| 27 |
+
"held_out_in_gradients": false,
|
| 28 |
+
"held_out": {
|
| 29 |
+
"eval.jsonl": 5,
|
| 30 |
+
"adversarial.jsonl": 6
|
| 31 |
+
},
|
| 32 |
+
"seed": 11,
|
| 33 |
+
"num_train_epochs": 45,
|
| 34 |
+
"warmup_steps": 10,
|
| 35 |
+
"lora_r": 32,
|
| 36 |
+
"lora_alpha": 64,
|
| 37 |
+
"learning_rate": 0.0002,
|
| 38 |
+
"lr_scheduler_type": "constant_with_warmup",
|
| 39 |
+
"optim": "adamw_8bit",
|
| 40 |
+
"response_only_loss": true,
|
| 41 |
+
"trackio": true,
|
| 42 |
+
"finalTrainLoss": "0.0172",
|
| 43 |
+
"training_loss": 0.017188175287382094,
|
| 44 |
+
"label": "MEASURED",
|
| 45 |
+
"eval": {
|
| 46 |
+
"label": "MEASURED",
|
| 47 |
+
"host": "j-szlholdings-6a91bf11984507d9db4ea104-glba55nz-d3ea0-68nwn",
|
| 48 |
+
"evaluatedAt": "2026-08-28T17:20:37.617962+00:00",
|
| 49 |
+
"planTotal": 11,
|
| 50 |
+
"planValid": 11,
|
| 51 |
+
"groundingTotal": 5,
|
| 52 |
+
"groundingCorrect": 5,
|
| 53 |
+
"abstainTotal": 6,
|
| 54 |
+
"abstainCorrect": 3,
|
| 55 |
+
"hallucinatedCitationCount": 0,
|
| 56 |
+
"held_out_in_gradients": false,
|
| 57 |
+
"temperature": 0,
|
| 58 |
+
"method": "in-process Unsloth generate; scoring ported from eval_khipu.py",
|
| 59 |
+
"rows": [
|
| 60 |
+
{
|
| 61 |
+
"split": "navigate",
|
| 62 |
+
"i": 1,
|
| 63 |
+
"valid": true,
|
| 64 |
+
"decision": "NAVIGATE",
|
| 65 |
+
"citedNodeIds": [
|
| 66 |
+
"node://khipu-synthetic/a75aa5921b37a055"
|
| 67 |
+
]
|
| 68 |
+
},
|
| 69 |
+
{
|
| 70 |
+
"split": "navigate",
|
| 71 |
+
"i": 2,
|
| 72 |
+
"valid": true,
|
| 73 |
+
"decision": "NAVIGATE",
|
| 74 |
+
"citedNodeIds": [
|
| 75 |
+
"node://khipu-synthetic/0c3811c6e4b98d16"
|
| 76 |
+
]
|
| 77 |
+
},
|
| 78 |
+
{
|
| 79 |
+
"split": "navigate",
|
| 80 |
+
"i": 3,
|
| 81 |
+
"valid": true,
|
| 82 |
+
"decision": "NAVIGATE",
|
| 83 |
+
"citedNodeIds": [
|
| 84 |
+
"node://khipu-synthetic/800d21d32b720587"
|
| 85 |
+
]
|
| 86 |
+
},
|
| 87 |
+
{
|
| 88 |
+
"split": "navigate",
|
| 89 |
+
"i": 4,
|
| 90 |
+
"valid": true,
|
| 91 |
+
"decision": "NAVIGATE",
|
| 92 |
+
"citedNodeIds": [
|
| 93 |
+
"node://khipu-synthetic/badfa44292cc23f5"
|
| 94 |
+
]
|
| 95 |
+
},
|
| 96 |
+
{
|
| 97 |
+
"split": "navigate",
|
| 98 |
+
"i": 5,
|
| 99 |
+
"valid": true,
|
| 100 |
+
"decision": "NAVIGATE",
|
| 101 |
+
"citedNodeIds": [
|
| 102 |
+
"node://khipu-synthetic/1fcf60c2a090c0a3"
|
| 103 |
+
]
|
| 104 |
+
},
|
| 105 |
+
{
|
| 106 |
+
"split": "adversarial",
|
| 107 |
+
"i": 1,
|
| 108 |
+
"valid": true,
|
| 109 |
+
"decision": "NAVIGATE",
|
| 110 |
+
"citedNodeIds": [
|
| 111 |
+
"node://khipu-synthetic/2f97a50f6f35049e"
|
| 112 |
+
]
|
| 113 |
+
},
|
| 114 |
+
{
|
| 115 |
+
"split": "adversarial",
|
| 116 |
+
"i": 2,
|
| 117 |
+
"valid": true,
|
| 118 |
+
"decision": "ABSTAIN",
|
| 119 |
+
"citedNodeIds": []
|
| 120 |
+
},
|
| 121 |
+
{
|
| 122 |
+
"split": "adversarial",
|
| 123 |
+
"i": 3,
|
| 124 |
+
"valid": true,
|
| 125 |
+
"decision": "ABSTAIN",
|
| 126 |
+
"citedNodeIds": []
|
| 127 |
+
},
|
| 128 |
+
{
|
| 129 |
+
"split": "adversarial",
|
| 130 |
+
"i": 4,
|
| 131 |
+
"valid": true,
|
| 132 |
+
"decision": "ABSTAIN",
|
| 133 |
+
"citedNodeIds": []
|
| 134 |
+
},
|
| 135 |
+
{
|
| 136 |
+
"split": "adversarial",
|
| 137 |
+
"i": 5,
|
| 138 |
+
"valid": true,
|
| 139 |
+
"decision": "NAVIGATE",
|
| 140 |
+
"citedNodeIds": [
|
| 141 |
+
"node://khipu-synthetic/8636e76827710d87"
|
| 142 |
+
]
|
| 143 |
+
},
|
| 144 |
+
{
|
| 145 |
+
"split": "adversarial",
|
| 146 |
+
"i": 6,
|
| 147 |
+
"valid": true,
|
| 148 |
+
"decision": "NAVIGATE",
|
| 149 |
+
"citedNodeIds": [
|
| 150 |
+
"node://khipu-synthetic/9e1192a9b2154f7b"
|
| 151 |
+
]
|
| 152 |
+
}
|
| 153 |
+
]
|
| 154 |
+
},
|
| 155 |
+
"lambda": "Conjecture 1",
|
| 156 |
+
"doctrine": "v11 LOCKED 749/14/163",
|
| 157 |
+
"proposal_only": true,
|
| 158 |
+
"publication_eligible": true,
|
| 159 |
+
"autonomy_eligible": false,
|
| 160 |
+
"job_id": "6a91bf11984507d9db4ea104",
|
| 161 |
+
"host": "j-szlholdings-6a91bf11984507d9db4ea104-glba55nz-d3ea0-68nwn",
|
| 162 |
+
"computed_at": "2026-08-28T17:20:37.618047+00:00",
|
| 163 |
+
"claim_boundary": "Eval counts are MEASURED k/n from this job only when eval.label=MEASURED. Do not invent scores. Original SZL-Khipu-1.5B signed abstain 2/6 is unchanged."
|
| 164 |
+
}
|