Instructions to use dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP 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 dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP:Q8_0
# Run inference directly in the terminal:
llama cli -hf dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP:Q8_0

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP:Q8_0
# Run inference directly in the terminal:
llama cli -hf dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP:Q8_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 dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP:Q8_0
# Run inference directly in the terminal:
./llama-cli -hf dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP:Q8_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 dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP:Q8_0
# Run inference directly in the terminal:
./build/bin/llama-cli -hf dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP:Q8_0

Use Docker

docker model run hf.co/dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP:Q8_0

LM Studio
Jan

vLLM

How to use dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP:Q8_0

Ollama
How to use dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP with Ollama:
```
ollama run hf.co/dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP:Q8_0
```
Unsloth Desktop

How to use dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP:Q8_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": "dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP:Q8_0"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP with Docker Model Runner:
```
docker model run hf.co/dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP:Q8_0
```

Lemonade

How to use dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP:Q8_0

Run and chat with the model

lemonade run user.Qwen3.5-4B-Q8_0-FlashNgram-MTP-Q8_0

List all available models

lemonade list

Hermes Agent

How to use dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP:Q8_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 dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP:Q8_0

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP:Q8_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 "dburner/Qwen3.5-4B-Q8_0-FlashNgram-MTP:Q8_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"

dburner commited on 4 days ago

Commit

376ced9

verified ·

1 Parent(s): 221cd17

Add merge validation report

Browse files

Files changed (1) hide show

Qwen3.5-4B-Q8_0-FlashNgram-MTP.gguf.report.json +116 -0

Qwen3.5-4B-Q8_0-FlashNgram-MTP.gguf.report.json ADDED Viewed

	@@ -0,0 +1,116 @@

+{
+  "base": "S:\\Proiecte\\qwen-mash\\models\\qwen35-4b-mtp\\Qwen3.5-4B-Q8_0.gguf",
+  "base_validation": {
+    "architecture": "qwen35",
+    "backbone_block_count": 32,
+    "block_indices": [
+      0,
+      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
+    ],
+    "hidden_size": 2560,
+    "metadata_block_count": 33,
+    "mtp_present": true,
+    "q8_tensor_count": 257,
+    "required_q8_tensors": {
+      "blk.0.attn_qkv.weight": 8,
+      "blk.0.ffn_gate.weight": 8
+    },
+    "tensor_count": 441
+  },
+  "composition": "base tensors + per_layer_token_embd.weight only",
+  "dry_run": false,
+  "flash_validation": {
+    "ignored_incompatible_ple_tensor_count": 6,
+    "ignored_incompatible_ple_tensor_names": [
+      "blk.1.ple_conv1d.weight",
+      "blk.1.ple_key.weight",
+      "blk.1.ple_norm_conv.weight",
+      "blk.1.ple_norm_key.weight",
+      "blk.1.ple_norm_query.weight",
+      "blk.1.ple_value.weight"
+    ],
+    "metadata_source": "S:\\Proiecte\\qwen-mash\\models\\flash-next\\UD-Q3_K_XL\\Qwen3.8-Flash-Next-UD-Q3_K_XL-00001-of-00003.gguf",
+    "table": {
+      "ggml_type": 20,
+      "name": "per_layer_token_embd.weight",
+      "payload_bytes": 28800138240,
+      "shape": [
+        160,
+        320001536
+      ]
+    },
+    "table_source": "S:\\Proiecte\\qwen-mash\\models\\flash-next\\UD-Q3_K_XL\\Qwen3.8-Flash-Next-UD-Q3_K_XL-00002-of-00003.gguf",
+    "tokenizer": {
+      "base": {
+        "bos_token_id": 248044,
+        "eos_token_id": 248046,
+        "vocab_size": 248320
+      },
+      "flash": {
+        "bos_token_id": 248044,
+        "eos_token_id": 248046,
+        "ple_eos_token_id": 248044,
+        "vocab_size": 248320
+      }
+    }
+  },
+  "metadata_added": [
+    "qwen35.ple.layers",
+    "qwen35.ple.ngram_size",
+    "qwen35.ple.heads_per_ngram",
+    "qwen35.ple.conv_kernel",
+    "qwen35.ple.eos_token_id",
+    "qwen35.embedding_length_per_layer_input",
+    "qwen35.ple.layer_multipliers",
+    "qwen35.ple.head_offsets",
+    "qwen35.ple.head_vocab_sizes",
+    "qwen35.ple.image_token_id",
+    "qwen35.ngram.gate",
+    "qwen35.ngram.layer",
+    "qwen35.ngram.source_architecture"
+  ],
+  "output": "S:\\Proiecte\\qwen-mash\\models\\qwen35-4b-mtp\\Qwen3.5-4B-Q8_0-FlashNgram-MTP.gguf",
+  "output_bytes": 33410719936,
+  "output_bytes_estimate": 33410719936,
+  "output_data_start": 10970304,
+  "split": {
+    "count": 3,
+    "indices": [
+      1,
+      2,
+      3
+    ],
+    "stem": "Qwen3.8-Flash-Next-UD-Q3_K_XL"
+  }
+}