Instructions to use kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF:Q4_0
# Run inference directly in the terminal:
llama cli -hf kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF:Q4_0

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF:Q4_0
# Run inference directly in the terminal:
llama cli -hf kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF:Q4_0

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF:Q4_0
# Run inference directly in the terminal:
./llama-cli -hf kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF:Q4_0

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF:Q4_0
# Run inference directly in the terminal:
./build/bin/llama-cli -hf kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF:Q4_0

Use Docker

docker model run hf.co/kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF:Q4_0

LM Studio
Jan

vLLM

How to use kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF:Q4_0

Ollama
How to use kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF with Ollama:
```
ollama run hf.co/kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF:Q4_0
```
Unsloth Desktop

How to use kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF:Q4_0

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF:Q4_0"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF with Docker Model Runner:
```
docker model run hf.co/kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF:Q4_0
```

Lemonade

How to use kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF:Q4_0

Run and chat with the model

lemonade run user.Nex-N2.5-mini-ROCmFP4-GGUF-Q4_0

List all available models

lemonade list

Hermes Agent

How to use kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF:Q4_0

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF:Q4_0

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF:Q4_0

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "kingjones777/Nex-N2.5-mini-ROCmFP4-GGUF:Q4_0" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

Nex-N2.5-mini-ROCmFP4-GGUF / recipe /pipeline /nex_phase1.sh

kingjones777

Add files using upload-large-folder tool

4294de3 verified 8 days ago

Raw

History Blame Contribute Delete

2.9 kB

	#!/bin/bash
	# Nex-N2.5-mini phase 1 (CPU): verified download -> BF16 GGUF + vision projector -> the three standard 4-bit tiers
	# (King: STRIX_LEAN + COHERENT + FAST; no Q8/Q6) with head protection, read back by exact tensor name.
	# Runs in the capped scope `nex-conv` (Agnes memory sizing waits for nex-* scopes). Waits until the Agnes BF16
	# re-grade (60 GiB on the GPU) is finished so the two large memory users never overlap.
	set -uo pipefail
	W=/mnt/models/nex-n2.5-mini; T=/opt/llama-rocm/rocmfpx-724; B=$T/build-hipvk/bin; N=Nex-N2.5-mini
	A=/mnt/models/agnes-3.0-flash
	export LD_LIBRARY_PATH=$B:/opt/rocm-7.2.4/lib TMPDIR=/mnt/models/.tmp PYTHONUNBUFFERED=1
	cd $W; mkdir -p gguf out logs
	log(){ echo "[$(date -u +%FT%TZ)] $*"; }
	until [ -f logs/DOWNLOAD_RC ]; do sleep 20; done
	if [ "$(cat logs/DOWNLOAD_RC)" != 0 ] \|\| [ "$(cat logs/DOWNLOAD_VERIFY_RC 2>/dev/null)" != 0 ]; then
	log "download or verify failed -> stop"; log "NEX_PHASE1_FAILED"; exit 1
	fi
	log "download verified; waiting for the Agnes BF16 re-grade to leave the GPU"
	until grep -q "R3 grade" $A/logs/regrade.log 2>/dev/null; do sleep 20; done

	log "C1 convert BF16 (the checkpoint has no mtp.* tensors, so no MTP block is emitted)"
	python3 $T/convert_hf_to_gguf.py hf --outtype bf16 --model-name "$N" --outfile gguf/$N-BF16.gguf > logs/C1_convert.log 2>&1
	rc=$?; log "C1 exit=$rc"; [ $rc -eq 0 ] \|\| { tail -30 logs/C1_convert.log; log "NEX_PHASE1_FAILED"; exit 2; }
	log "C2 convert vision projector"
	python3 $T/convert_hf_to_gguf.py hf --outtype bf16 --mmproj --model-name "$N" --outfile out/mmproj-$N-BF16.gguf > logs/C2_mmproj.log 2>&1
	rc=$?; log "C2 exit=$rc"; [ $rc -eq 0 ] \|\| { tail -30 logs/C2_mmproj.log; log "NEX_PHASE1_FAILED"; exit 3; }
	python3 readback.py - - gguf/$N-BF16.gguf \| tee logs/C_readback.log
	python3 readback.py - - out/mmproj-$N-BF16.gguf \| tee -a logs/C_readback.log

	BF=gguf/$N-BF16.gguf; Q=$B/llama-quantize
	log "Q1 standard tiers"
	$Q --output-tensor-type q6_K $BF out/$N-Q4_0_ROCMFP4_STRIX_LEAN.gguf Q4_0_ROCMFP4_STRIX_LEAN 16 > logs/Q1_q106.log 2>&1; log " q106 exit=$?"
	$Q --output-tensor-type q6_K --token-embedding-type q6_K $BF out/$N-Q4_0_ROCMFP4_COHERENT.gguf Q4_0_ROCMFP4_COHERENT 16 > logs/Q1_q102.log 2>&1; log " q102 exit=$?"
	$Q --output-tensor-type q6_K $BF out/$N-Q4_0_ROCMFP4_FAST.gguf Q4_0_ROCMFP4_FAST 16 > logs/Q1_q103.log 2>&1; log " q103 exit=$?"
	python3 readback.py Q6_K Q5_K out/$N-Q4_0_ROCMFP4_STRIX_LEAN.gguf \| tee logs/Q_readback.log
	python3 readback.py Q6_K Q6_K out/$N-Q4_0_ROCMFP4_COHERENT.gguf \| tee -a logs/Q_readback.log
	python3 readback.py Q6_K - out/$N-Q4_0_ROCMFP4_FAST.gguf \| tee -a logs/Q_readback.log
	for l in Q1_q106 Q1_q102 Q1_q103; do printf "%-8s " $l; grep -oE "quant size\s=\s[0-9.]+ MiB $[0-9.]+ BPW$" logs/$l.log; done \| tee logs/Q_sizes.log
	log "NEX_PHASE1_DONE"