flatrun 0.1.0__tar.gz

flatrun-0.1.0/LICENSE

@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of tracking and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for describing the origin of the Work and
+      reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Support. While redistributing the Work or
+      Derivative Works thereof, You may choose to offer, and charge a
+      fee for, acceptance of support, warranty, indemnity, or other
+      liability obligations and/or rights consistent with this License.
+      However, in accepting such obligations, You may act only on Your
+      own behalf and on Your sole responsibility, not on behalf of any
+      other Contributor, and only if You agree to indemnify, defend,
+      and hold each Contributor harmless for any liability incurred by, or
+      claims asserted against, such Contributor by reason of your
+      accepting any such warranty or support.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright 2026 judotens <judotens@flatseek.io>
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+   implied. See the License for the specific language governing
+   permissions and limitations under the License.

flatrun-0.1.0/PKG-INFO

@@ -0,0 +1,442 @@
+Metadata-Version: 2.4
+Name: flatrun
+Version: 0.1.0
+Summary: The streaming inference runtime for open-weight LLMs.
+Author-email: Judotens Budiarto <judotens@flatseek.io>
+License: MIT
+Project-URL: Homepage, https://flatseek.io
+Project-URL: Documentation, https://flatseek.io/docs
+Project-URL: Repository, https://github.com/flatseek/flatrun
+Project-URL: Issues, https://github.com/flatseek/flatrun/issues
+Project-URL: Flatlens, https://github.com/flatseek/flatlens
+Project-URL: Flatvec, https://github.com/flatseek/flatvec
+Project-URL: Flatask, https://github.com/flatseek/flatask
+Project-URL: Flattune, https://github.com/flatseek/flattune
+Project-URL: Benchmarks, https://github.com/flatseek/flatbench
+Keywords: llm,inference,runtime,mmap,safetensors,memory-efficient,gguf,quantization,streaming
+Classifier: Development Status :: 3 - Alpha
+Classifier: Intended Audience :: Developers
+Classifier: Intended Audience :: Science/Research
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: numpy>=1.26.0
+Provides-Extra: dev
+Requires-Dist: pytest>=8.0.0; extra == "dev"
+Requires-Dist: pytest-cov>=4.1.0; extra == "dev"
+Requires-Dist: mypy>=1.8.0; extra == "dev"
+Requires-Dist: ruff>=0.2.0; extra == "dev"
+Provides-Extra: safetensors
+Requires-Dist: safetensors>=0.4.0; extra == "safetensors"
+Provides-Extra: native
+Requires-Dist: pybind11>=2.10; extra == "native"
+Dynamic: license-file
+
+<div align="center">
+
+<img src="logo.svg" alt="Flatrun" width="64" height="64">
+
+# Flatrun
+
+**The streaming inference runtime for open-weight LLMs.**
+
+<p align="center">
+  <em>
+Run GGUF, SafeTensors, and MLX models layer-by-layer directly from storage,
+enabling inference on models larger than available system RAM.
+  </em>
+</p>
+
+[![Python](https://img.shields.io/badge/python-3.10%2B-blue.svg)](https://www.python.org/downloads/)
+[![License](https://img.shields.io/badge/license-Apache%202.0-green.svg)](LICENSE)
+[![Tests](https://github.com/flatseek/flatrun/workflows/Test/badge.svg)](https://github.com/flatseek/flatrun/actions)
+[![PyPI version](https://img.shields.io/pypi/v/flatrun.svg)](https://pypi.org/project/flatrun/)
+
+
+**GitHub:** https://github.com/flatseek/flatrun
+&nbsp;&middot;&nbsp;
+**Organization:** https://github.com/flatseek
+
+<br>
+
+**Part of the Flatseek ecosystem**
+
+[Flatseek](https://github.com/flatseek/flatseek) (Keyword Search) •
+[Flatvec](https://github.com/flatseek/flatvec) (Vector Search) •
+[Flatask](https://github.com/flatseek/flatask) (RAG Runtime) •
+[Flatweight](https://github.com/flatseek/flatweight) (AI Model Storage) •
+[Flattune](https://github.com/flatseek/flattune) (LLM Fine-Tuning) •
+[Flatlens](https://github.com/flatseek/flatlens) (Data Visualization)
+
+</div>
+
+---
+## Try It
+
+Download a sample GGUF model from Hugging Face and start an interactive chat:
+
+```bash
+wget https://huggingface.co/lmstudio-community/SmolLM2-135M-Instruct-GGUF/resolve/main/SmolLM2-135M-Instruct-Q4_K_M.gguf -O smoll.gguf
+
+flatrun chat --model smoll.gguf
+```
+
+Example session:
+
+```text
+Detected format: gguf
+Building tokenizer from GGUF metadata (SmolLM2-135M-Instruct-Q4_K_M.gguf) ...
+Tokenizer vocab: 49152
+Chat template: Qwen2 ChatML
+Loaded model in 0.53 s; layers=30
+
+Chat mode (max_new=128/turn, history=True).
+Type your message; Ctrl-D (EOF) or 'exit' to quit.
+
+You: where is paris?
+
+Assistant:
+Paris is the capital and largest city of France. It is located in the
+north-central part of the country along the Seine River and is one of
+Europe's major cultural, historical, and economic centers.
+
+```
+---
+# Overview
+
+**A pure Python and NumPy runtime for AI inference research and optimization.**
+
+Flatrun is an experimental inference runtime built to explore how large language models can execute **without loading an entire checkpoint into memory**.
+
+Instead of treating a model as a single monolithic file, Flatrun streams weights layer-by-layer directly from disk. Only the layer currently being executed resides in memory while the remaining weights stay memory-mapped on storage.
+
+This streaming architecture allows models significantly larger than available system RAM to run on commodity hardware, trading throughput for dramatically lower memory usage.
+
+---
+# Why Flatrun?
+
+Flatrun is designed for developers and researchers who want to understand how LLM inference works.
+
+It provides a transparent streaming runtime where every stage—from tensor loading to KV cache management and forward execution—can be inspected, profiled, modified, and optimized.
+
+---
+# Highlights
+
+- Streaming layer-by-layer execution
+- Constant-memory architecture
+- Pure Python + NumPy runtime
+- Optional native C++ acceleration
+- GGUF, SafeTensors, and MLX support
+- Q1/Q4/Q5/Q6/Q8 quantization
+- Python API and interactive chat
+- Layer profiler and memory tracing
+---
+
+# Supported Models
+
+| Family | GGUF | SafeTensors | MLX |
+|---------|:---:|:-----------:|:---:|
+| Llama 1 / 2 / 3 | ✅ | ✅ | — |
+| Qwen2 | ✅ | ✅ | — |
+| Qwen2.5 | ✅ | ✅ | — |
+| Qwen2.5 Coder | ✅ | ✅ | — |
+| Qwen3 | ✅ | ✅ | — |
+| Qwen3.5 | ✅ | — | — |
+| SmolLM2 | ✅ | ✅ | — |
+| Gemma 3 | — | — | ✅ |
+| Bonsai Q1_0 | ✅ | — | — |
+
+Support for additional architectures is continuously expanding.
+
+---
+
+# Installation
+
+Install from PyPI:
+
+```bash
+pip install flatrun
+```
+Or clone + editable install:
+
+
+```bash
+git clone https://github.com/flatseek/flatrun.git
+
+cd flatrun
+
+make install
+```
+
+Requirements:
+
+- Python 3.10+
+- NumPy
+
+---
+
+# Quick Start
+
+Run a GGUF model:
+
+```bash
+flatrun \
+    --model model.gguf \
+    --prompt "The capital of France is" \
+    --max-new 32
+```
+
+Interactive chat:
+
+```bash
+flatrun chat \
+    --model model.gguf
+```
+
+Run a SafeTensors checkpoint:
+
+```bash
+flatrun \
+    --model /path/to/model \
+    --prompt "Explain quantum computing."
+```
+
+---
+
+## Backend Selection
+
+Flatrun provides two interchangeable execution backends.
+
+### Python Backend (Default)
+
+The reference runtime is implemented entirely in pure Python and NumPy, making every stage of inference easy to inspect, profile, and modify.
+
+```bash
+flatrun \
+    --backend python \
+    --model model.gguf \
+    --prompt "Hello"
+```
+
+This is the default backend, so `--backend python` can be omitted.
+
+### Native Backend
+
+For higher performance, Flatrun can execute performance-critical operations using the native C++ backend located in `src/flatrun_native`.
+
+```bash
+flatrun \
+    --backend native \
+    --model model.gguf \
+    --prompt "Hello"
+```
+
+The native backend preserves the same streaming execution pipeline while accelerating compute-intensive operations such as quantized GEMM and dequantization.
+
+Backend options:
+
+- `--backend python` *(default)* — Pure Python and NumPy runtime.
+- `--backend native` — Native C++ accelerated runtime.
+
+---
+
+# Python API
+
+Flatrun is fully scriptable. The same runtime that powers the CLI is
+exposed as a Python library — pick a local model file, hand it a
+prompt, and read the logits back. Nothing in the public surface
+requires downloading a model from a hub.
+
+## Full minimal example
+
+```python
+from flatrun import KVCache, StreamingExecutor, load_huggingface
+from flatrun.model import make_qwen2_forwarder
+from flatrun.model.sampling import Sampler
+from flatrun.runtime.backend import get_backend
+from flatrun.tokenizer import auto_load
+
+# 1. Open a local GGUF
+loaded = load_huggingface("/Users/me/models/qwen3-0.6b-q4_k_m.gguf")
+
+# 2. Tokenize
+tokenizer = auto_load("/Users/me/models/qwen3-0.6b-q4_k_m.gguf")
+prompt = "The capital of France is"
+tokens = list(tokenizer.encode(prompt))
+
+# 3. Build the executor
+backend = get_backend("python")
+forwarder = make_qwen2_forwarder(
+    loaded.config, enable_dequant_cache=True, backend=backend,
+)
+kv_cache = KVCache(capacity=4096)
+scheduler = loaded.runtime.build_scheduler(loaded.manifest.layers)
+executor = StreamingExecutor(scheduler, forwarder, kv_cache=kv_cache)
+
+# 4. Prefill + decode
+sampler = Sampler(temperature=0.0001, top_k=1)
+step = executor.step(tokens)
+next_id = sampler.sample(step.last_hidden[-1])
+out = list(tokens) + [next_id]
+for _ in range(32):
+    step = executor.step([next_id])
+    next_id = sampler.sample(step.last_hidden[-1])
+    out.append(next_id)
+
+# 5. Decode
+print(tokenizer.decode(out))
+```
+
+---
+# CLI
+
+```bash
+flatrun run
+flatrun chat
+```
+
+Supports:
+
+- Text generation and sampling
+- Python and native backends
+- Runtime and cache configuration
+- Quantization overrides
+- Layer selection
+- Profiling and debugging
+
+See `flatrun run --help` for the complete command reference.
+
+---
+# How It Works
+
+```text
+Model on Disk
+      │
+      ▼
+ Load Layer
+      │
+      ▼
+ Execute
+      │
+      ▼
+ Update KV
+      │
+      ▼
+Release Layer
+      │
+      ▼
+ Next Layer
+```
+
+Instead of loading an entire checkpoint into memory, Flatrun streams one decoder layer at a time. Only the active layer and KV cache remain resident in memory, keeping peak RAM nearly constant regardless of model size.
+
+---
+# Performance
+
+Flatrun prioritizes memory efficiency over maximum throughput.
+
+Typical performance on modern Apple Silicon:
+
+| Model | Peak RAM | Speed |
+|--------|----------:|------:|
+| SmolLM2-360M Q8 | ~400 MB | ~40 tok/s |
+| Qwen3-0.6B Q4 | ~512 MB | ~12 tok/s |
+| Qwen3-14B Q4 | ~2–30 GB | ~0.7–1.5 tok/s |
+
+Performance depends on:
+
+- Model size
+- Quantization
+- Storage bandwidth
+- Cache configuration
+
+Profile every stage of execution using:
+
+```bash
+flatrun \
+    --profile-detailed \
+    --profile-save profile.json
+```
+
+---
+
+# Native Backend
+
+While the reference runtime is intentionally written in pure Python, Flatrun also includes an optional native backend located in `src/flatrun_native`.
+
+The native backend accelerates performance-critical components using C++ while preserving the exact same streaming architecture and runtime API.
+
+Current focus includes:
+
+- Quantized GEMM kernels (fused dequant + matmul)
+- Per-quant coverage: **Q4_K**, **Q6_K**, **Q8_0**
+- SIMD vectorization (ARM NEON 128-bit intrinsics)
+- Multi-threaded row-parallel matmul (std::thread pool, one per call)
+- Batched matmul API for prefill (no per-token Python overhead)
+- Reduced Python overhead via pybind11 marshalling
+
+The native backend is selected with `--backend native`. When the C++
+extension is not built (no pybind11 toolchain available), or when a
+specific kernel rejects an unusual layout, the call falls back to
+the numpy path transparently — `--backend native` is always safe.
+
+Each supported quant type has a parity test in
+`src/flatrun_native/tests/`:
+
+| Quant | Test | Tolerance |
+|-------|------|----------:|
+| Q4_K  | `test_q4_k_parity.py` | ~3e-6 max rel |
+| Q6_K  | `test_q6_k_parity.py` | ~3e-4 max rel |
+| Q8_0  | `test_q8_0_parity.py` | ~2e-5 max rel |
+
+See [`docs/native.md`](docs/native.md) for the kernel design notes
+and the multi-threading / batched-API rationale.
+
+---
+
+# Research Focus
+
+Flatrun serves as a platform for exploring new ideas in AI runtime design.
+
+Current research areas include:
+
+- Streaming inference
+- Layer scheduling
+- Memory-aware execution
+- Quantized inference
+- Dequantization optimization
+- KV cache architectures
+- Native kernel acceleration
+- Runtime instrumentation
+- Pure NumPy performance
+
+The codebase is intentionally designed to be readable, modular, and easy to modify, making it suitable for experimentation, benchmarking, and runtime research.
+
+---
+
+# Limitations
+
+Current limitations include:
+
+- CPU-only execution
+- Pure NumPy reference runtime
+- No CUDA backend
+- No Metal kernels
+- Lower throughput than GPU runtimes
+- Some architectures are still under active development
+
+Flatrun is built primarily for experimentation, research, and runtime innovation rather than large-scale production serving.
+
+---
+
+# License
+
+Apache License 2.0
+
+See `LICENSE`.