bit-ttt-engine 0.6.1__cp310-cp310-win_amd64.whl → 0.6.2__cp310-cp310-win_amd64.whl

{bit_ttt_engine-0.6.1.dist-info → bit_ttt_engine-0.6.2.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: bit-ttt-engine
-Version: 0.6.1
+Version: 0.6.2
 Classifier: Development Status :: 4 - Beta
 Classifier: Programming Language :: Rust
 Classifier: Programming Language :: Python :: Implementation :: CPython
@@ -22,14 +22,13 @@ Description-Content-Type: text/markdown; charset=UTF-8; variant=GFM
 [![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](LICENSE)
 [![PyPI](https://img.shields.io/pypi/v/bit-ttt-engine.svg)](https://pypi.org/project/bit-ttt-engine/)
 
-**1.58-bit Quantization + Test-Time Training (TTT)** Implementation in Pure Rust.
-
-This package provides Python bindings for the Bit-TTT Engine, allowing you to run ultra-light ternary LLMs with real-time adaptation.
+**4-bit Quantization + Test-Time Training (TTT)** Implementation in Pure Rust.
 
 ## ✨ Features
-1. **Ultra-Light**: Runs large LLMs on cheap hardware using **1.58-bit (ternary) weights**.
-2. **Adaptive (TTT)**: Learns *while* inferring, adapting to context in real-time.
+1. **Fast**: **40 tokens/second** on GPU (RTX 4060 Ti).
+2. **Adaptive (TTT)**: Learns *while* inferring - unique to Bit-TTT!
 3. **Pure Rust**: High performance with minimal dependencies.
+4. **Easy**: Load GGUF models directly.
 
 ## 🚀 Installation
 
@@ -37,13 +36,60 @@ This package provides Python bindings for the Bit-TTT Engine, allowing you to ru
 pip install bit-ttt-engine
 ```
 
-## 💻 Usage
+## 💻 Quick Start (GGUF Models)
+
+```python
+from cortex_rust import GgufModel
+
+# Load model
+model = GgufModel("model.gguf", tokenizer="tokenizer.json")
+
+# Generate text
+output = model.generate(
+    "Hello, how are you?",
+    max_tokens=50,
+    temperature=0.7
+)
+print(output)
+
+# Streaming output
+model.generate_with_callback(
+    "Tell me a story",
+    lambda t: print(t, end="", flush=True),
+    max_tokens=100
+)
+```
+
+## 🧠 TTT (Test-Time Training)
+
+**TTT makes the model adapt during inference** - something no other local LLM can do!
+
+```python
+from cortex_rust import GgufModel
+
+model = GgufModel("model.gguf", tokenizer="tokenizer.json")
+
+# Enable TTT
+model.enable_ttt(layers=4, learning_rate=0.1)
+
+# Without TTT: Pass 1 == Pass 2 (same output)
+# With TTT:    Pass 1 != Pass 2 (model is learning!)
+
+out1 = model.generate("My name is Alice.", max_tokens=20)
+out2 = model.generate("My name is Alice.", max_tokens=20)
+print(f"Different: {out1 != out2}")  # True!
+
+# TTT controls
+model.disable_ttt()
+model.reset_ttt_state()
+print(model.ttt_enabled)  # False
+```
+
+## 🏗️ Legacy API (BitLlama)
 
 ```python
 import cortex_rust
-import json
 
-# Initialize Configuration
 config = cortex_rust.BitLlamaConfig(
     vocab_size=32000,
     hidden_dim=512,
@@ -51,34 +97,14 @@ config = cortex_rust.BitLlamaConfig(
     inner_lr=0.001
 )
 
-# Initialize Model (Inference)
 model = cortex_rust.BitLlama(
     config=config,
     checkpoint_path="path/to/model.safetensors",
-    device="cpu", # or "cuda"
+    device="cpu",
     tokenizer_path="path/to/tokenizer.json"
 )
 
-# Generate Text
-output = model.generate(prompt="Hello, world!", max_tokens=50)
-print(output)
-```
-
-## 🏗️ Training (TTT)
-
-```python
-trainer = cortex_rust.PyTrainer(
-    config=config,
-    checkpoint_path="path/to/model.safetensors",
-    device="cuda"
-)
-
-# Single training step
-loss = trainer.train_step(input_ids=[...], targets=[...])
-print(f"Loss: {loss}")
-
-# Save checkpoint
-trainer.save_checkpoint("model_updated.safetensors")
+output = model.generate(prompt="Hello!", max_tokens=50)
 ```
 
 ## 📖 Documentation

bit_ttt_engine-0.6.2.dist-info/RECORD

@@ -0,0 +1,9 @@
+LICENSE,sha256=JJLZ3h6-sbZqpBjH8srqgZ40NaAKhVqh2oXOM6E2Mak,1088
+bit_ttt_engine-0.6.2.dist-info\METADATA,sha256=p4eog3pP24VjOz964Uu3Ji2V5i_ylwfrNhttA79o1p8,3268
+bit_ttt_engine-0.6.2.dist-info\WHEEL,sha256=GCQ19ZBvayuBQJpz6xNbc8p6I5GMQcns9k4vQBQ8VH8,97
+bit_ttt_engine-0.6.2.dist-info\licenses\LICENSE,sha256=JJLZ3h6-sbZqpBjH8srqgZ40NaAKhVqh2oXOM6E2Mak,1088
+cortex_rust\__init__.py,sha256=USDgfYs5sRU8A4mPwD_Ir4XXHDvnoM-FcUQRUoNVl_g,127
+cortex_rust\__init__.pyi,sha256=pKbj_GxWbSkFTKWmE_32O55dC6PCuDj9up0vRX0aNX4,3069
+cortex_rust\cortex_rust.cp310-win_amd64.pyd,sha256=NAo1Yh1NZNe-6LJBlEBtD1fg9g4zBfuadrXPvZ698Lo,8478208
+cortex_rust\py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+bit_ttt_engine-0.6.2.dist-info\RECORD,,

bit_ttt_engine-0.6.2.dist-info/licenses/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 imonoonoko
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

bit_ttt_engine-0.6.1.dist-info/RECORD

@@ -1,8 +0,0 @@
-bit_ttt_engine-0.6.1.dist-info\METADATA,sha256=huxZqMaXS_Pp4UFFps1vn_PVktllcux8T-ua-b4V5is,2763
-bit_ttt_engine-0.6.1.dist-info\WHEEL,sha256=GCQ19ZBvayuBQJpz6xNbc8p6I5GMQcns9k4vQBQ8VH8,97
-bit_ttt_engine-0.6.1.dist-info\licenses\LICENSE,sha256=JJLZ3h6-sbZqpBjH8srqgZ40NaAKhVqh2oXOM6E2Mak,1088
-cortex_rust\__init__.py,sha256=USDgfYs5sRU8A4mPwD_Ir4XXHDvnoM-FcUQRUoNVl_g,127
-cortex_rust\__init__.pyi,sha256=pKbj_GxWbSkFTKWmE_32O55dC6PCuDj9up0vRX0aNX4,3069
-cortex_rust\cortex_rust.cp310-win_amd64.pyd,sha256=v7NUfIbEVGnTu-ZmIa4yF4EUZ6YE_KKiUHhb_ctv2mE,8484864
-cortex_rust\py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-bit_ttt_engine-0.6.1.dist-info\RECORD,,