fix: re-export 5m weights; true int8/int4 keys; honest model card

Browse files

Files changed (7) hide show

README.md +51 -59
model.safetensors +2 -2
model_fp16.safetensors +1 -1
model_int4.safetensors +2 -2
model_int8.safetensors +2 -2
model_int8_weight.safetensors +3 -0
pytorch_model.bin +2 -2

README.md CHANGED Viewed

@@ -5,89 +5,81 @@ library_name: transformers
 license: apache-2.0
 pipeline_tag: feature-extraction
 tags:
-- jev
 - hastejev
-- hastejev-5m
 - decision-engine
 - system-1
-- pica
-- zero-bias
-- low-latency
 - non-generative
-- autonomous-agents
-- browser-control
-- web-automation
-- agentic-ai
-- fast-inference
-- decision-making
-- calibration
 - safetensors
 - pytorch
 - quantized
-- int8
-- int4
-- fp16
 ---
-# ⚡ Haste Jev 5M (Medium) (~5.0M Parameters)
-[![Hugging Face Model](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-noffy%2Fhastejev-5m-yellow)](https://huggingface.co/noffy/hastejev-5m)
-[![GitHub Repository](https://img.shields.io/badge/GitHub-racstan%2Fhastejev-black?logo=github)](https://github.com/racstan/hastejev)
-[![License](https://img.shields.io/badge/License-Apache_2.0-blue.svg)](https://opensource.org/licenses/Apache-2.0)
-[![Parameters](https://img.shields.io/badge/Parameters-~5.0M-brightgreen.svg)]()
-[![Memory](https://img.shields.io/badge/RAM_Footprint-~19_MB-purple.svg)]()
-> **Haste Jev 5M (Medium)** is part of the **Haste Jev** family of open-weights, zero-bias **System-1 Decision Engines**. Medium System-1 decision engine tailored for complex financial routing, AML compliance, and multi-criteria risk scoring.
----
-## 🔬 Model Specifications
-- **Total Parameters**: 5,003,971 (~5.0M)
-- **Trainable Parameters**: 3,369,667
-- **Buffer / Projection Table**: 1,634,304
-- **Hidden Dimension ($d_{\text{model}}$)**: 224
-- **Transformer Layers**: 5
-- **Attention Heads**: 4
-- **Target Deployment**: Financial & KYC routing
-- **Quantization Formats Available**: `FP32`, `FP16` (`model_fp16.safetensors`), `INT8` (`model_int8.safetensors`), `INT4` (`model_int4.safetensors`)
----
-## ⚡ Quickstart
-```python
-from hastejev import HasteJevEngine
-# 1. Load standard weights directly from Hugging Face Hub
-engine = HasteJevEngine.from_pretrained("noffy/hastejev-5m")
-# 2. Or load with INT8 / INT4 quantization
-engine_int8 = HasteJevEngine.from_pretrained("noffy/hastejev-5m", quantization="int8")
-# 3. Execute Decision Primitives
-state = "Account balance is $14,850.50 with pending transaction of $3,200.00."
-options = ["Approve Transaction", "Flag for Review", "Decline"]
-res = engine.choice(state, options)
-print(f"Decision: {res.decision} (Confidence: {res.confidence:.3f})")
 ```
----
-## 📊 Complete Haste Jev Model Family
-| Model | Parameters | Hidden Dim | Layers | Heads | RAM (FP32) | RAM (INT8) | Target Use Case |
-| :--- | :---: | :---: | :---: | :---: | :---: | :---: | :--- |
-| [`hastejev-100k`](https://huggingface.co/noffy/hastejev-100k) | **~98k** | 48 | 2 | 2 | ~0.4 MB | ~0.1 MB | Microcontrollers, WASM, IoT |
-| [`hastejev-500k`](https://huggingface.co/noffy/hastejev-500k) | **~500k** | 96 | 3 | 4 | ~2.0 MB | ~0.5 MB | Mobile CPU, in-browser workers |
-| [`hastejev-1m`](https://huggingface.co/noffy/hastejev-1m) | **~1.1M** | 128 | 4 | 4 | ~4.4 MB | ~1.1 MB | High-throughput API sidecars |
-| [`hastejev-2m`](https://huggingface.co/noffy/hastejev-2m) | **~1.8M** | 160 | 4 | 4 | ~7.3 MB | ~1.8 MB | Browser automation & bots |
-| [`hastejev-5m`](https://huggingface.co/noffy/hastejev-5m) | **~5.0M** | 224 | 5 | 4 | ~20.0 MB | ~5.0 MB | Financial & KYC routing |
-| [`hastejev-10m`](https://huggingface.co/noffy/hastejev-10m) | **~10.0M** | 320 | 5 | 4 | ~40.0 MB | ~10.0 MB | Multimodal agent kernels |
-| [`hastejev-20m`](https://huggingface.co/noffy/hastejev) | **~20.4M** | 256 | 4 | 4 | ~81.5 MB | ~20.4 MB | Enterprise decision engine |
----
-## 📄 License
-Apache License 2.0.

 license: apache-2.0
 pipeline_tag: feature-extraction
 tags:
 - hastejev
+- jev
 - decision-engine
 - system-1
+- agent-routing
+- tool-routing
 - non-generative
+- pica
 - safetensors
 - pytorch
 - quantized
 ---
+# Haste Jev 5m (Medium)
+**Role:** Finance / KYC workflow branch
+Open-weights **System-1 decision engine** for software paths that need typed
+decisions under a time budget (agent routing, tool routing, intent classification,
+pre-flight guardrails, browser action selection). Not a chat model.
+## Measured specification
+| Field | Value |
+|---|---|
+| Total parameters | 5,003,971 |
+| Trainable parameters | 3,369,667 |
+| Hash-table buffers | 1,634,304 |
+| d_model | 224 |
+| Layers | 5 |
+| Heads | 4 |
+Parameter counts match the GitHub README table (verified with `verify_claims.py`).
+## Honest claims
+| Claim | Status |
+|---|---|
+| PICA option-order bias = 0.0% | **Verified** architecturally |
+| Exact parameter table | **Verified** |
+| FP32 ~0.4 MB for 100k weights | **Verified** (weight storage only) |
+| p99 &lt; 15ms / ECE &lt; 0.009 / 99.4% arithmetic | **Not verified** — do not cite from this card |
+| Published latency / accuracy on your workload | **Measure yourself** |
+Weights may be lightly or untrained prototypes depending on export; treat behavioral
+accuracy as unknown until you evaluate on labeled data.
+## Quickstart
+```python
+from hastejev import HasteJevEngine
+eng = HasteJevEngine.from_pretrained("noffy/hastejev-5m")
+# eng = HasteJevEngine.from_pretrained("noffy/hastejev-5m", quantization="int4")
+r = eng.choice(
+    "Request: reset password for user@corp.example",
+    ["auth_self_service", "billing", "security_review"],
+)
+print(r.decision, r.confidence)
 ```
+Install: `pip install git+https://github.com/racstan/hastejev.git`
+## Files
+| File | Contents |
+|---|---|
+| `model.safetensors` / `pytorch_model.bin` | FP32 state dict |
+| `model_fp16.safetensors` | FP16 |
+| `model_int8.safetensors` | True weight-only int8 (`weight_q`) when re-exported with ≥1.1.0 |
+| `model_int4.safetensors` | True packed int4 (`weight_packed`) when re-exported with ≥1.1.0 |
+Older revisions of `model_int8`/`model_int4` may be mislabeled FP32; re-export or
+re-download after this commit.
+## License
+Apache-2.0

model.safetensors CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bdb4b46af8454b99ff789c091482e4eeb0d32b8f420c0fb701b99e1c01eba83f
-size 20025212

 version https://git-lfs.github.com/spec/v1
+oid sha256:eccf6fd4153f73090f0ea0ed32d92802c2c18fbd46d410cb28585226c0c727c7
+size 11489144

model_fp16.safetensors CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:08714b10fae52fe32e75f8dd0aac7239fd83709d10acf4992e79ff36ae1e398a
 size 10017158

 version https://git-lfs.github.com/spec/v1
+oid sha256:461cae8529fef45b0f8cd08b8b930ec13c4f3a9691e133904a22c37728e35fe4
 size 10017158

model_int4.safetensors CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bdb4b46af8454b99ff789c091482e4eeb0d32b8f420c0fb701b99e1c01eba83f
-size 20025212

 version https://git-lfs.github.com/spec/v1
+oid sha256:eccf6fd4153f73090f0ea0ed32d92802c2c18fbd46d410cb28585226c0c727c7
+size 11489144

model_int8.safetensors CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9607c56c31d51a21aafd064aa449d11ddb3208da43e31fe4b695560bd287c79e
-size 20025212

 version https://git-lfs.github.com/spec/v1
+oid sha256:56903a6231524f6d8178335240aaaec05ebb1503d09b04fb4ab290d2b5d714d7
+size 12713248

model_int8_weight.safetensors ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:56903a6231524f6d8178335240aaaec05ebb1503d09b04fb4ab290d2b5d714d7
+size 12713248

pytorch_model.bin CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:154475d6226bffaf81501804b604df7b11eeaa9cdb8e5a93507c9f0674178012
-size 20044303

 version https://git-lfs.github.com/spec/v1
+oid sha256:4d70b77d4f7a24937815c17743eafdc54e493c9988750ca701b3ebb8b1603901
+size 11513119