Ollaya package for Cloudflare/clef-flash

README.md

@@ -0,0 +1,44 @@
+---
+license: apache-2.0
+base_model:
+- Cloudflare/clef-flash
+library_name: onnx
+tags:
+- ollaya
+- onnx
+- decision-model
+- system-one
+pipeline_tag: text-classification
+---
+
+# clef for Ollaya
+
+[Ollaya](https://github.com/ollaya-dev/ollaya) package of **[Cloudflare/clef-flash](https://huggingface.co/Cloudflare/clef-flash)** by Cloudflare (post-trained model and joint schema head) and the Qwen team (base model).
+Ollaya runs open decision models locally, the way Ollama runs LLMs: typed questions in,
+calibrated answers out, behind a TypeSafe-compatible API.
+
+```sh
+ollaya run clef
+```
+
+## What is in this repository
+
+This repository holds only the files Ollaya derives, with no weights. Each graph is an ONNX export of the
+original model whose weights **reference the authors' own weight files by byte offset**,
+so `ollaya pull` downloads the weights from the upstream repositories, unmodified and pinned to a
+commit, and verifies their sha256.
+
+| Tag | Upstream | Files |
+|---|---|---|
+| `clef:flash` | [Cloudflare/clef-flash@17f0b0a](https://huggingface.co/Cloudflare/clef-flash/tree/17f0b0ad64efb65d273590632833508766b2aae6) | `flash/model-fp32.onnx`, `flash/decision.json`, `flash/calibration.json` |
+
+Each tag has an fp32 graph, used on CPU and GPU. Each tag also has `decision.json` (sequence layout, special tokens) and
+`calibration.json` (temperatures).
+
+## Parity
+
+Ollaya's Rust runtime matches the authors' own code (joint_schema_model.py: their encoder, Qwen3.5 model and joint schema head, fp32) on 571 questions from 131 requests, on CUDA: identical token ids and spans, the same 13 rejected requests, the same decision on every question, logits within 4.3e-5 and probabilities within 6.3e-6.
+
+## License
+
+Same as the upstream model (Apache-2.0). Ollaya itself is Apache-2.0.

flash/calibration.json

@@ -0,0 +1,9 @@
+{
+  "temperature": [
+    1.0,
+    1.0,
+    1.0
+  ],
+  "temperature_by_options": {},
+  "source": "none: upstream answers with a plain softmax of the head's logits"
+}

flash/decision.json

@@ -0,0 +1,100 @@
+{
+  "engine": "onnx",
+  "family": "clef",
+  "layout": "clef-joint-v1",
+  "upstream": {
+    "repo": "Cloudflare/clef-flash",
+    "revision": "17f0b0ad64efb65d273590632833508766b2aae6",
+    "code": "joint_schema_model.py in the model repository (encode_record, JointSchemaHead)"
+  },
+  "contract": {
+    "inputs": {
+      "input_ids": {
+        "dtype": "int64",
+        "shape": [
+          1,
+          "seq"
+        ],
+        "note": "the request's ids; seq a multiple of 64; right-pad with the pad id"
+      },
+      "token_positions": {
+        "dtype": "int64",
+        "shape": [
+          "tokens"
+        ],
+        "note": "0..tokens-1"
+      },
+      "question_spans": {
+        "dtype": "int64",
+        "shape": [
+          "questions",
+          2
+        ],
+        "note": "instruction tokens [start, end) per question"
+      },
+      "question_types": {
+        "dtype": "int64",
+        "shape": [
+          "questions"
+        ],
+        "note": "type_index"
+      },
+      "option_spans": {
+        "dtype": "int64",
+        "shape": [
+          "options",
+          2
+        ],
+        "note": "option tokens [start, end), every question's options in order"
+      },
+      "option_question": {
+        "dtype": "int64",
+        "shape": [
+          "options"
+        ],
+        "note": "the question index of each option"
+      }
+    },
+    "outputs": {
+      "logits": {
+        "dtype": "float32",
+        "shape": [
+          "options"
+        ],
+        "note": "raw option logits"
+      }
+    },
+    "seq_multiple": 64,
+    "positions": "0..seq-1, implicit",
+    "attention": "causal; no mask input"
+  },
+  "prompt": {
+    "prefix": "<|im_start|>system\nRead the complete state and schema. Decide every field jointly. Each answer must be exactly one of that field's allowed options.<|im_end|>\n<|im_start|>user\nSTATE:\n",
+    "schema": "\n\nSCHEMA FIELDS:\n",
+    "field": "\nFIELD {n}\nID: {id}\nTYPE: {type}\nINSTRUCTION: ",
+    "options": "\nALLOWED OPTIONS:\n",
+    "option": "OPTION {n}: ",
+    "option_end": "\n",
+    "field_end": "END FIELD\n",
+    "suffix": "\n<|im_end|>\n<|im_start|>assistant\n<think>\n\n</think>\n\nJOINT SCHEMA DECISIONS:"
+  },
+  "noul_criteria": {
+    "true": "The proposition is true or the answer is yes.",
+    "false": "The proposition is false or the answer is no."
+  },
+  "max_tokens": 4096,
+  "pad": 248044,
+  "type_index": {
+    "noul": 0,
+    "choice": 1,
+    "score": 2
+  },
+  "option_logits": {
+    "noul": "[true, false]",
+    "choice": "labels sorted by code point",
+    "score": "levels in order"
+  },
+  "opset": 20,
+  "precision": "fp32 compute; weights BF16 (widened by Cast)",
+  "weights_in_memory": "bf16"
+}

flash/model-fp32.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5cc3f10d3ad4158c34411e2e29247d914fe6b70a2dc92b53458ee7bff6b62a57
+size 11991785