Feature Extraction
Transformers
PyTorch
Safetensors
English
hastejev
jev
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
quantized
int8
int4
fp16
Instructions to use noffy/hastejev-5m with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use noffy/hastejev-5m with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("feature-extraction", model="noffy/hastejev-5m")# Load model directly from transformers import HasteJevEngine model = HasteJevEngine.from_pretrained("noffy/hastejev-5m", device_map="auto") - Notebooks
- Google Colab
- Kaggle
β‘ Haste Jev 5M (Medium) (~5.0M Parameters)
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.
π¬ 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)
β‘ Quickstart
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})")
π Complete Haste Jev Model Family
| Model | Parameters | Hidden Dim | Layers | Heads | RAM (FP32) | RAM (INT8) | Target Use Case |
|---|---|---|---|---|---|---|---|
hastejev-100k |
~98k | 48 | 2 | 2 | ~0.4 MB | ~0.1 MB | Microcontrollers, WASM, IoT |
hastejev-500k |
~500k | 96 | 3 | 4 | ~2.0 MB | ~0.5 MB | Mobile CPU, in-browser workers |
hastejev-1m |
~1.1M | 128 | 4 | 4 | ~4.4 MB | ~1.1 MB | High-throughput API sidecars |
hastejev-2m |
~1.8M | 160 | 4 | 4 | ~7.3 MB | ~1.8 MB | Browser automation & bots |
hastejev-5m |
~5.0M | 224 | 5 | 4 | ~20.0 MB | ~5.0 MB | Financial & KYC routing |
hastejev-10m |
~10.0M | 320 | 5 | 4 | ~40.0 MB | ~10.0 MB | Multimodal agent kernels |
hastejev-20m |
~20.4M | 256 | 4 | 4 | ~81.5 MB | ~20.4 MB | Enterprise decision engine |
π License
Apache License 2.0.
- Downloads last month
- -