Text Generation
GGUF
Japanese
English
llama.cpp
Mixture of Experts
expert-pruning
intel-mac
cpu
local-agent
imatrix
conversational
Instructions to use miutti/intel-mac-local-llm with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- llama.cpp
How to use miutti/intel-mac-local-llm 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 miutti/intel-mac-local-llm:UD-Q2_K_XL # Run inference directly in the terminal: llama cli -hf miutti/intel-mac-local-llm:UD-Q2_K_XL
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama serve -hf miutti/intel-mac-local-llm:UD-Q2_K_XL # Run inference directly in the terminal: llama cli -hf miutti/intel-mac-local-llm:UD-Q2_K_XL
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 miutti/intel-mac-local-llm:UD-Q2_K_XL # Run inference directly in the terminal: ./llama-cli -hf miutti/intel-mac-local-llm:UD-Q2_K_XL
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 miutti/intel-mac-local-llm:UD-Q2_K_XL # Run inference directly in the terminal: ./build/bin/llama-cli -hf miutti/intel-mac-local-llm:UD-Q2_K_XL
Use Docker
docker model run hf.co/miutti/intel-mac-local-llm:UD-Q2_K_XL
- LM Studio
- Jan
- vLLM
How to use miutti/intel-mac-local-llm with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "miutti/intel-mac-local-llm" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "miutti/intel-mac-local-llm", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/miutti/intel-mac-local-llm:UD-Q2_K_XL
- Ollama
How to use miutti/intel-mac-local-llm with Ollama:
ollama run hf.co/miutti/intel-mac-local-llm:UD-Q2_K_XL
- Unsloth Desktop
- Pi
How to use miutti/intel-mac-local-llm with Pi:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf miutti/intel-mac-local-llm:UD-Q2_K_XL
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": "miutti/intel-mac-local-llm:UD-Q2_K_XL" } ] } } }Run Pi
# Start Pi in your project directory: pi
- Docker Model Runner
How to use miutti/intel-mac-local-llm with Docker Model Runner:
docker model run hf.co/miutti/intel-mac-local-llm:UD-Q2_K_XL
- Lemonade
How to use miutti/intel-mac-local-llm with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull miutti/intel-mac-local-llm:UD-Q2_K_XL
Run and chat with the model
lemonade run user.intel-mac-local-llm-UD-Q2_K_XL
List all available models
lemonade list
- Hermes Agent
How to use miutti/intel-mac-local-llm with Hermes Agent:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf miutti/intel-mac-local-llm:UD-Q2_K_XL
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 miutti/intel-mac-local-llm:UD-Q2_K_XL
Run Hermes
hermes
- Atomic Chat
- OpenClaw
How to use miutti/intel-mac-local-llm with OpenClaw:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf miutti/intel-mac-local-llm:UD-Q2_K_XL
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 "miutti/intel-mac-local-llm:UD-Q2_K_XL" \ --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"
File size: 6,003 Bytes
df41178 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 | """1ビット未満まで含む重み圧縮コーデック群。
共通の約束:
encode(W, ...) -> (W_hat, bpw, label)
W_hat : 復元された重み (float32, Wと同形)
bpw : 1重みあたりの実効ビット数(スケール等のオーバーヘッド込み)
label : 表示用の名前
前提: W は (out, in) の行優先。量子化グループは「同じ行の連続 G 要素」。
"""
import numpy as np
FP16_BITS = 16
# ---------- 補助 ----------
def _grouped(W, G):
"""(out,in) -> (ngroup, G) に切る。in が G で割り切れる前提。"""
out, inn = W.shape
assert inn % G == 0, f"in={inn} が group={G} で割り切れない"
return W.reshape(-1, G)
def _scales(X):
s = np.abs(X).max(axis=1, keepdims=True)
s[s == 0] = 1.0
return s
# ---------- 1. RTN(普通の整数量子化。比較の基準線)----------
def rtn(W, bits=4, G=128):
X = _grouped(W, G)
s = _scales(X)
qmax = 2 ** (bits - 1) - 1
q = np.clip(np.rint(X / s * qmax), -qmax, qmax)
Xh = q / qmax * s
bpw = bits + FP16_BITS / G
return Xh.reshape(W.shape).astype(np.float32), bpw, f"RTN-{bits}bit(G={G})"
# ---------- 2. 三値(BitNet風。約1.58bit)----------
def ternary(W, G=128):
X = _grouped(W, G)
s = np.abs(X).mean(axis=1, keepdims=True)
s[s == 0] = 1.0
q = np.clip(np.rint(X / s), -1, 1)
Xh = q * s
bpw = np.log2(3) + FP16_BITS / G
return Xh.reshape(W.shape).astype(np.float32), bpw, f"三値(G={G})"
# ---------- 3. 学習コードブックの積量子化(1bit未満の本命)----------
def _kmeans(data, ncode, iters=8, seed=0, sample=120_000):
rng = np.random.default_rng(seed)
if len(data) > sample:
data = data[rng.choice(len(data), sample, replace=False)]
C = data[rng.choice(len(data), ncode, replace=False)].copy()
for _ in range(iters):
idx = _assign(data, C)
dead = []
for c in range(ncode):
m = idx == c
if m.any():
C[c] = data[m].mean(0)
else:
dead.append(c)
if dead:
# 誰にも選ばれなかったコードは初期値のまま居座り、語彙を無駄にする。
# 最も表現できていない点(現コードから最も遠い点)へ置き直して回収する。
d2 = ((data - C[idx]) ** 2).sum(1)
far = np.argsort(-d2)[:len(dead)]
C[np.array(dead)] = data[far]
return C
def _assign(X, C, chunk=65536):
"""最近傍コード番号。||x-c||^2 = ||x||^2 -2x·c + ||c||^2 の展開で高速化。"""
cn = (C * C).sum(1)
out = np.empty(len(X), np.int32)
for i in range(0, len(X), chunk):
x = X[i:i + chunk]
d = cn[None, :] - 2.0 * (x @ C.T)
out[i:i + chunk] = d.argmin(1)
return out
def pvq(W, k=8, cb_bits=8, G=128, seed=0, random_codebook=False):
"""k次元ごとに 2^cb_bits 個のコードへ割り当てる積ベクトル量子化。
bpw = cb_bits/k + スケール分。k=8, cb_bits=4 なら 0.5bit/重み。
random_codebook=True なら学習せず乱数で作る(コードブックの保存が不要=seed だけ)。
"""
assert G % k == 0
X = _grouped(W, G)
s = _scales(X)
Xn = (X / s).reshape(-1, k) # 正規化済みサブベクトル
ncode = 2 ** cb_bits
if random_codebook:
rng = np.random.default_rng(seed)
C = rng.normal(0, Xn.std(), size=(ncode, k)).astype(np.float32)
tag = "乱数CB"
else:
C = _kmeans(Xn.astype(np.float32), ncode, seed=seed)
tag = "学習CB"
idx = _assign(Xn.astype(np.float32), C)
Xh = (C[idx].reshape(X.shape) * s)
bpw = cb_bits / k + FP16_BITS / G
return Xh.reshape(W.shape).astype(np.float32), bpw, f"PVQ-{tag} k={k},{cb_bits}bit(G={G})"
# ---------- 4. 重要度ハイブリッド(少数の列だけ厚く、残りを極薄に)----------
def hybrid(W, keep_frac=0.01, keep_bits=8, sub=None, G=128, importance=None):
"""入力次元(列)の重要度上位 keep_frac だけ高精度、残りを sub コーデックで潰す。
importance: 長さ in の配列(キャリブレーションから来る活性化スケール等)。
None なら列ノルムで代用。
"""
if sub is None:
sub = lambda M: pvq(M, k=8, cb_bits=4, G=G)
out, inn = W.shape
imp = importance if importance is not None else np.linalg.norm(W, axis=0)
nkeep = max(1, int(round(inn * keep_frac)))
keep = np.argsort(-imp)[:nkeep]
mask = np.zeros(inn, bool)
mask[keep] = True
Wh = np.empty_like(W, dtype=np.float32)
# 高精度側(列数が G で割り切れないので行ごとに1グループ扱い)
Wk = W[:, mask]
s = _scales(Wk)
qmax = 2 ** (keep_bits - 1) - 1
Wh[:, mask] = np.clip(np.rint(Wk / s * qmax), -qmax, qmax) / qmax * s
Wr = np.ascontiguousarray(W[:, ~mask])
rest_in = Wr.shape[1]
pad = (-rest_in) % G
if pad:
Wr = np.pad(Wr, ((0, 0), (0, pad)))
Wrh, sub_bpw, sub_label = sub(Wr)
Wh[:, ~mask] = Wrh[:, :rest_in]
f = nkeep / inn
bpw = f * (keep_bits + FP16_BITS / min(G, nkeep)) + (1 - f) * sub_bpw + 1.0 / out
return Wh, bpw, f"ハイブリッド {keep_frac*100:.1f}%@{keep_bits}bit + {sub_label}"
# ---------- 評価 ----------
def evaluate(W, Wh, X=None, seed=0):
"""重み誤差と、実際に効く「出力誤差」を測る。"""
E = W - Wh
rel_w = np.linalg.norm(E) / np.linalg.norm(W)
if X is None:
rng = np.random.default_rng(seed)
n = W.shape[1]
# 正規分布+外れ値チャンネル(実際の活性化は少数の次元が突出する)
X = rng.normal(size=(n, 64)).astype(np.float32)
big = rng.choice(n, max(1, n // 100), replace=False)
X[big] *= 20.0
Y, Yh = W @ X, Wh @ X
rel_y = np.linalg.norm(Y - Yh) / np.linalg.norm(Y)
return rel_w, rel_y
|