digvijay-llm 0.1.2

Run local LLMs on low-RAM machines with automatic hardware detection and streaming inference

digvijay_llm

Run big open-source LLMs — including 70B-class models — on machines with as little as 16 GB of RAM, by streaming weights from disk instead of loading the full model into RAM.

This version is more production-oriented: it can detect your hardware automatically, choose sensible defaults for backend/device/threading/context, expose the detected configuration through llm.config, and support streaming plus basic benchmarking.

pip install digvijay_llm

---

⚠️ Read this first (important honesty note)

A CPU/GPU can only compute on data sitting in RAM/VRAM — it is physically impossible to run matrix multiplications directly on a disk. What this library actually does, and what makes low-RAM 70B inference possible, is disk-backed weight streaming: weights live on disk and are pulled into RAM only for the layer currently being computed, then released.

Peak RAM usage becomes roughly "one layer's worth of weights + activations" instead of "the entire model's weights." The tradeoff is speed — disk I/O is much slower than RAM, so expect lower tokens/sec than a fully-in-RAM setup, especially on HDDs. An NVMe SSD is strongly recommended.

---

How it works

digvijay_llm ships two backends:

Backend	File format	Mechanism	Best for
GGUFStreamingEngine	.gguf (quantized)	Wraps llama.cpp via llama-cpp-python, using mmap() so the OS pages weights in/out of RAM on demand	Production use — fastest, most memory-efficient, supports 4-bit/5-bit quantization
SafetensorsLayerStreamer	raw HuggingFace .safetensors checkpoints	Builds the model with empty weights, then uses accelerate's disk-offload hooks to load one transformer layer at a time from disk right before it's needed	When you only have an unconverted HF checkpoint and don't want to quantize/convert to GGUF first

Both are wrapped behind one simple class, LowRAMLLM, so your code doesn't need to care which backend is active.

---

Install

pip install digvijay_llm

# then install the backend you need:
pip install digvijay_llm[gguf]   # GGUF backend only (recommended, lighter)
pip install digvijay_llm[hf]     # safetensors / HuggingFace backend only
pip install digvijay_llm[all]    # both backends

---

Step 1 — Automatic hardware detection

You no longer need to run a separate script first. digvijay_llm can detect your machine automatically when you create a model instance.

from digvijay_llm import LowRAMLLM

llm = LowRAMLLM.from_gguf("models/llama-70b.Q4_K_M.gguf")
print(llm.config)
print(llm.device_info)

The library will infer a sensible configuration for:

• backend selection
• GPU vs CPU placement
• n_gpu_layers
• n_threads
• n_ctx
• n_batch
• n_ram_gb

You can still override any of these manually when needed.

If you want the older CLI-style report, the standalone script is still available:

python detect_params.py

---

Step 2 — Get a model

Go to huggingface.co and search for llama 70b GGUF. Download the Q4_K_M version. Recommended repos:

• bartowski/Meta-Llama-3.1-70B-Instruct-GGUF
• TheBloke/Llama-2-70B-GGUF

Place the .gguf file anywhere on your machine, e.g. models/llama-70b.Q4_K_M.gguf

---

Step 3 — Write your code

Option A — GGUF model (recommended, fastest)

from digvijay_llm import LowRAMLLM

llm = LowRAMLLM.from_gguf(
    "models/llama-70b.Q4_K_M.gguf",
    detect=True,      # default: auto-detect and tune the config
    n_ram_gb=16,
    n_gpu_layers=0,    # override if you want to force a different value
    n_threads=8,
)

# Generate text
print(llm.generate("Explain quantum entanglement simply."))

# Stream token by token
for token in llm.stream("Write a short poem about the sea."):
    print(token, end="", flush=True)

# Chat interface
reply = llm.chat([{"role": "user", "content": "Hello!"}])
print(reply)

# Basic benchmark summary
print(llm.benchmark("Summarize the benefits of local inference."))

Option B — Raw HuggingFace safetensors checkpoint (no conversion needed)

from digvijay_llm import LowRAMLLM

llm = LowRAMLLM.from_safetensors("models/llama-3-70b-hf/", detect=True, n_ram_gb=16)
print(llm.generate("Write a haiku about the ocean."))

Option C — Auto-detect format

from digvijay_llm import LowRAMLLM

llm = LowRAMLLM.auto("models/llama-3-70b.Q4_K_M.gguf", n_ram_gb=16)
# or
llm = LowRAMLLM.auto("models/llama-3-70b-hf/", n_ram_gb=16)

The same auto-detection logic applies here; the library will inspect the current machine and choose sensible runtime settings.

---

New in this version

• Built-in hardware detection for CPU, CUDA, ROCm, Metal, and Intel XPU-style backends
• Automatic tuning for n_gpu_layers, n_threads, n_ctx, n_batch, and n_ram_gb
• llm.config and llm.device_info for inspecting the detected settings
• Streaming with callback support via llm.stream(...)
• Basic benchmark helpers via llm.benchmark(...)
• Clearer runtime error handling for missing optional dependencies

---

GPU support

digvijay_llm works on every machine — no GPU, small GPU, or big GPU. Just set n_gpu_layers accordingly (or let detect_params.py pick it for you):

Your GPU VRAM	n_gpu_layers
No GPU	0
4 GB	10–15
6 GB	20–25
8 GB	30–35
12 GB	45–50
16 GB	60–65
24 GB+	80 (all layers, no disk streaming needed)

Layers that fit in VRAM run at full GPU speed. Layers that don't fit are streamed from disk automatically. You always get the best of both.

# No GPU
llm = LowRAMLLM.from_gguf("model.gguf", n_ram_gb=16, n_gpu_layers=0)

# Small GPU (8GB VRAM)
llm = LowRAMLLM.from_gguf("model.gguf", n_ram_gb=16, n_gpu_layers=35)

# Big GPU (24GB+ VRAM) — everything fits, no disk streaming needed
llm = LowRAMLLM.from_gguf("model.gguf", n_ram_gb=16, n_gpu_layers=80)

---

Optional — Check your RAM budget

from digvijay_llm import plan_for_budget

plan = plan_for_budget(
    model_path           = "models/llama-3-70b-hf/",
    total_params_billion = 70,
    n_layers             = 80,
    bytes_per_param      = 2.0,   # fp16; use 0.6 for Q4 quantized
    n_ram_gb             = 16,
)
print(plan)
for w in plan.warnings:
    print("WARNING:", w)

Warns you if your disk doesn't have enough free space or your RAM is too tight before you even try to load the model.

---

Minimum requirements

	Minimum	Recommended
RAM	16 GB	32 GB
Disk	45 GB free (for 70B Q4)	NVMe SSD
Python	3.9+	3.11+
OS	Windows / Mac / Linux	any
GPU	not required	any CUDA GPU helps

---

Practical tips for 70B on 16GB

1. Use Q4_K_M quantization — shrinks the model from ~140 GB (fp16) to ~38 GB on disk. Biggest single lever.
2. Use an NVMe SSD, not a spinning HDD — random-access read speed directly sets your tokens/sec.
3. Keep context length modest — the KV cache also eats RAM. digvijay_llm auto-shrinks context on tight budgets.
4. Close other RAM-heavy apps while running — the OS needs free RAM as page cache for the mmap'd weights.
5. Any GPU helps — even offloading a few layers to a small GPU (n_gpu_layers=20) meaningfully improves speed.

---

Project layout

digvijay_llm/               ← repo root
├── digvijay_llm/           ← the package
│   ├── __init__.py
│   ├── api.py
│   ├── engine_gguf.py
│   ├── engine_safetensors.py
│   ├── hardware.py
│   └── ram_planner.py
├── examples/
│   ├── run_gguf_example.py
│   └── run_safetensors_example.py
├── tests/
│   ├── test_hardware_detection.py
│   └── test_ram_planner.py
├── .github/
│   └── workflows/
│       └── publish.yml
├── detect_params.py        ← run this first!
├── .gitignore
├── LICENSE
├── pyproject.toml
├── requirements.txt
├── README.md
└── setup.py

---

License

MIT — see LICENSE

---

Author

Built by Digvijay Phutane