Text Generation
GGUF
English
llama.cpp
rocmfpx
rocm
amd
strix-halo
gfx1151
mtp
speculative-decoding
agent
tool-use
code
qwen3
vision
imatrix
conversational
Instructions to use philtheriver/Qwopus3.6-27B-Coder-MTP-ROCmFPX 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 philtheriver/Qwopus3.6-27B-Coder-MTP-ROCmFPX 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 philtheriver/Qwopus3.6-27B-Coder-MTP-ROCmFPX:BF16 # Run inference directly in the terminal: llama cli -hf philtheriver/Qwopus3.6-27B-Coder-MTP-ROCmFPX:BF16
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama serve -hf philtheriver/Qwopus3.6-27B-Coder-MTP-ROCmFPX:BF16 # Run inference directly in the terminal: llama cli -hf philtheriver/Qwopus3.6-27B-Coder-MTP-ROCmFPX:BF16
Build from source code
git clone https://github.com/ggml-org/llama.cpp.git cd llama.cpp cmake -B build cmake --build build -j --target llama-app # Start a local OpenAI-compatible server with a web UI: ./build/bin/llama serve -hf philtheriver/Qwopus3.6-27B-Coder-MTP-ROCmFPX:BF16 # Run inference directly in the terminal: ./build/bin/llama cli -hf philtheriver/Qwopus3.6-27B-Coder-MTP-ROCmFPX:BF16
Use Docker
docker model run hf.co/philtheriver/Qwopus3.6-27B-Coder-MTP-ROCmFPX:BF16
- Llama new
- LM Studio
- Jan
- vLLM
How to use philtheriver/Qwopus3.6-27B-Coder-MTP-ROCmFPX with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "philtheriver/Qwopus3.6-27B-Coder-MTP-ROCmFPX" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "philtheriver/Qwopus3.6-27B-Coder-MTP-ROCmFPX", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/philtheriver/Qwopus3.6-27B-Coder-MTP-ROCmFPX:BF16
- Ollama
How to use philtheriver/Qwopus3.6-27B-Coder-MTP-ROCmFPX with Ollama:
ollama run hf.co/philtheriver/Qwopus3.6-27B-Coder-MTP-ROCmFPX:BF16
- Unsloth Desktop
- Pi
How to use philtheriver/Qwopus3.6-27B-Coder-MTP-ROCmFPX with Pi:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf philtheriver/Qwopus3.6-27B-Coder-MTP-ROCmFPX:BF16
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:9931/v1", "api": "openai-completions", "apiKey": "none", "models": [ { "id": "philtheriver/Qwopus3.6-27B-Coder-MTP-ROCmFPX:BF16" } ] } } }Run Pi
# Start Pi in your project directory: pi
- Docker Model Runner
How to use philtheriver/Qwopus3.6-27B-Coder-MTP-ROCmFPX with Docker Model Runner:
docker model run hf.co/philtheriver/Qwopus3.6-27B-Coder-MTP-ROCmFPX:BF16
- Lemonade
How to use philtheriver/Qwopus3.6-27B-Coder-MTP-ROCmFPX with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull philtheriver/Qwopus3.6-27B-Coder-MTP-ROCmFPX:BF16
Run and chat with the model
lemonade run user.Qwopus3.6-27B-Coder-MTP-ROCmFPX-BF16
List all available models
lemonade list
- Hermes Agent
How to use philtheriver/Qwopus3.6-27B-Coder-MTP-ROCmFPX with Hermes Agent:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf philtheriver/Qwopus3.6-27B-Coder-MTP-ROCmFPX:BF16
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://127.0.0.1:9931/v1 hermes config set model.default philtheriver/Qwopus3.6-27B-Coder-MTP-ROCmFPX:BF16
Run Hermes
hermes
- Atomic Chat
- OpenClaw
How to use philtheriver/Qwopus3.6-27B-Coder-MTP-ROCmFPX with OpenClaw:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf philtheriver/Qwopus3.6-27B-Coder-MTP-ROCmFPX:BF16
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://127.0.0.1:9931/v1 \ --custom-model-id "philtheriver/Qwopus3.6-27B-Coder-MTP-ROCmFPX:BF16" \ --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"
Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -22,122 +22,89 @@ tags:
|
|
| 22 |
- vision
|
| 23 |
---
|
| 24 |
|
| 25 |
-
# Qwopus3.6-27B-Coder
|
| 26 |
|
| 27 |
-
|
| 28 |
|
| 29 |
-
|
| 30 |
-
> `ROCmFPX` is a family of **custom AMD quant types**; this build uses **enum IDs 110–115**. Those IDs are **not upstream-stable** — a newer fork revision could renumber them and silently mis-decode these weights, so **build from the pinned commit above**. The file will **NOT load** in stock `llama.cpp`, LM Studio, Ollama, or any other runtime — only the fork's `llama-quantize`/`llama-server` understand it.
|
| 31 |
|
| 32 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 33 |
|
| 34 |
-
-
|
| 35 |
|
| 36 |
-
##
|
| 37 |
|
| 38 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 39 |
|
| 40 |
-
|
| 41 |
|
| 42 |
-
|
| 43 |
-
|-------|-------|
|
| 44 |
-
| Base preset | `Q6_0_ROCMFPX_AGENT` (ROCmFPX enum 114 — *agent/tool-call coherent ROCmFPx Q6 routing*) |
|
| 45 |
-
| Token embeddings | `f16` (`--token-embedding-type f16`) |
|
| 46 |
-
| Output head | `Q6_K` (`--output-tensor-type q6_K`) |
|
| 47 |
-
| Importance matrix | code-calibrated (~1.3 MB corpus: C#/.NET, Python, JavaScript, SQL) |
|
| 48 |
-
| Source | `Qwopus3.6-27B-Coder-MTP-BF16.gguf` (54.66 GB BF16, MTP head spliced in) |
|
| 49 |
-
| Output size | ~25 GiB |
|
| 50 |
|
| 51 |
-
Exact build:
|
| 52 |
```bash
|
| 53 |
-
|
| 54 |
-
--imatrix qwopus-coder.imatrix \
|
| 55 |
-
--token-embedding-type f16 --output-tensor-type q6_K \
|
| 56 |
-
Qwopus3.6-27B-Coder-MTP-BF16.gguf \
|
| 57 |
-
Qwopus3.6-27B-Coder-MTP-STRIX-embF16-headQ6-Q6_0_ROCMFPX_AGENT.gguf \
|
| 58 |
-
Q6_0_ROCMFPX_AGENT
|
| 59 |
-
```
|
| 60 |
-
|
| 61 |
-
## Quant tiers (see the **Files** tab)
|
| 62 |
-
|
| 63 |
-
Every tier carries the MTP head (`draft=2`) and pairs with the bundled `mmproj/mmproj-F32.gguf` (vision).
|
| 64 |
-
f16 token embeddings throughout; `_AGENT` presets route attn/FFN at higher precision for tool-call /
|
| 65 |
-
JSON coherence. Filenames are prefixed `Qwopus3.6-27B-Coder-MTP-STRIX-`.
|
| 66 |
-
|
| 67 |
-
| File suffix | Preset | Enum | ~bpw | Size | Role |
|
| 68 |
-
|-------------|--------|------|------|------|------|
|
| 69 |
-
| `embF16-headQ6-Q6_0_ROCMFPX_AGENT.gguf` | `Q6_0_ROCMFPX_AGENT` | 114 | ~6.5 | 26 GB | **flagship** — f16 emb + Q6_K head + code imatrix |
|
| 70 |
-
| `embF16-Q8_0_ROCMFPX_AGENT.gguf` | `Q8_0_ROCMFPX_AGENT` | 115 | ~8.25 | 30 GB | highest-fidelity agent |
|
| 71 |
-
| `embF16-Q8_0_ROCMFPX.gguf` | `Q8_0_ROCMFPX` | 111 | 8.25 | 29 GB | highest-fidelity (straight) |
|
| 72 |
-
| `embF16-Q6_0_ROCMFPX.gguf` | `Q6_0_ROCMFPX` | 110 | 6.50 | 24 GB | balanced (straight) |
|
| 73 |
-
| `embF16-Q3_0_ROCMFPX_AGENT.gguf` | `Q3_0_ROCMFPX_AGENT` | 113 | ~3.5 | 21 GB | smallest agent |
|
| 74 |
-
| `embF16-Q3_0_ROCMFPX.gguf` | `Q3_0_ROCMFPX` | 112 | 3.50 | 17 GB | smallest |
|
| 75 |
-
| `embF16-Q4_0_ROCMFP4.gguf` | `Q4_0_ROCMFP4` | 100 | ~5.7¹ | 19 GB | fastest decode (4-bit body, ~13 tok/s raw) |
|
| 76 |
-
|
| 77 |
-
> Tiers beyond the flagship are uploaded as they finish building — the **Files** tab is the source of truth for what's live.
|
| 78 |
-
> ¹ `Q4_0_ROCMFP4` is a true 4-bit body; the ~5.7 *effective* bpw is inflated by the f16 token embeddings (a lookup — it doesn't slow the fast 4-bit decode).
|
| 79 |
-
|
| 80 |
-
## Running (ROCmFPX fork required)
|
| 81 |
-
|
| 82 |
-
```bash
|
| 83 |
-
# 1) build the fork at the pinned commit (ROCm/HIP toolchain, gfx1151)
|
| 84 |
git clone https://github.com/charlie12345/ROCmFPX.git && cd ROCmFPX
|
| 85 |
-
git checkout 553504b
|
| 86 |
-
JOBS=16 scripts/build-strix-rocmfp4-mtp.sh # -> build-strix-rocmfp4/bin/
|
| 87 |
|
| 88 |
-
#
|
| 89 |
-
HSA_OVERRIDE_GFX_VERSION=11.5.1 \
|
| 90 |
-
build-strix-rocmfp4/bin/llama-server \
|
| 91 |
-m Qwopus3.6-27B-Coder-MTP-STRIX-embF16-headQ6-Q6_0_ROCMFPX_AGENT.gguf \
|
| 92 |
-
--alias qwopus-coder \
|
| 93 |
-dev ROCm0 -ngl 999 -fa on -c 32768 \
|
| 94 |
--spec-type draft-mtp --spec-draft-ngl all --spec-draft-n-max 2 \
|
| 95 |
-
--jinja --mmproj mmproj/mmproj-F32.gguf
|
| 96 |
-
--host 0.0.0.0 --port 8080
|
| 97 |
```
|
| 98 |
|
| 99 |
-
|
| 100 |
-
- **Vision (Qwen3-VL):** bundled `mmproj/mmproj-F32.gguf`; pass `--mmproj` (add `--image-min-tokens 1024` for high-detail images).
|
| 101 |
-
- **Tool calls:** point your client at the **`qwen3_coder`** tool-call parser, or the model narrates code instead of emitting structured calls.
|
| 102 |
-
- The chat template is embedded — `--jinja` reads it.
|
| 103 |
-
|
| 104 |
-
## Verification
|
| 105 |
|
| 106 |
-
|
|
|
|
| 107 |
|
| 108 |
-
|
| 109 |
-
|:----:|:------:|:-----------:|:--------------:|:---------:|
|
| 110 |
-
| ✅ | ✅ | ✅ | ✅ | ✅¹ |
|
| 111 |
|
| 112 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 113 |
|
| 114 |
-
|
| 115 |
-
|--------|-------|
|
| 116 |
-
| MTP decode (`draft-mtp`, `--spec-draft-n-max 2`) | ✅ active — ~18.6 tok/s decode on gfx1151 |
|
| 117 |
-
| PPL — code corpus (~110 chunks @ ctx 512) | flagship **2.922** vs stock `Q6_K` **2.917** → **Δ +0.17%** (both ±0.04 — within error) |
|
| 118 |
|
| 119 |
-
|
| 120 |
-
> All numbers are hardware/driver/prompt-sensitive — reproduce before relying.
|
| 121 |
|
| 122 |
-
|
|
|
|
|
|
|
|
|
|
| 123 |
|
| 124 |
-
|
| 125 |
|
| 126 |
-
| Context |
|
| 127 |
|---|---|---|---|
|
| 128 |
| pp512 | 204 | 269 | **+32%** |
|
| 129 |
-
| pp2048 | 193 | 254 |
|
| 130 |
-
| pp10k | 181 | 236 |
|
| 131 |
-
| pp16k | 174 | 225 |
|
| 132 |
-
| pp32k | 158 | 199 |
|
| 133 |
| pp64k | 134 | 163 | **+22%** |
|
| 134 |
|
| 135 |
-
|
| 136 |
-
|
| 137 |
-
|
| 138 |
|
| 139 |
## Credits & license
|
| 140 |
|
| 141 |
-
|
| 142 |
-
- **Chain:** Jackrong + Kyle Hessling (fine-tune) → Qwen3.6-27B (base) → [charlie12345 / ROCmFPX](https://github.com/charlie12345/ROCmFPX) (quant fork). ROCmFPX quantization by this repo's author.
|
| 143 |
-
- Experimental community release; no full safety eval. The base card's 67% SWE-bench figure is a no-thinking Q5_K_M number — treat these FPX numbers as independent.
|
|
|
|
| 22 |
- vision
|
| 23 |
---
|
| 24 |
|
| 25 |
+
# Qwopus3.6-27B-Coder · ROCmFPX
|
| 26 |
|
| 27 |
+
### Stock `Q6_K` quality, ~30% faster prompt-processing on AMD Strix Halo (`gfx1151`)
|
| 28 |
|
| 29 |
+
ROCmFPX 4→8-bit quants of [`Jackrong/Qwopus3.6-27B-Coder-MTP-GGUF`](https://huggingface.co/Jackrong/Qwopus3.6-27B-Coder-MTP-GGUF) — 27B, MTP speculative decoding + Qwen3-VL vision, agent/tool-use tuned.
|
|
|
|
| 30 |
|
| 31 |
+
| | |
|
| 32 |
+
|---|---|
|
| 33 |
+
| **Quality** | ≈ stock `Q6_K` — PPL **+0.17%** (within error) |
|
| 34 |
+
| **Prompt processing** | **+32%** vs Q6_K (short ctx) → +22% at 64k |
|
| 35 |
+
| **Decode** | ~18 tok/s with MTP (~9 raw) |
|
| 36 |
+
| **Vision** | Qwen3-VL — bundled `mmproj/` |
|
| 37 |
+
| **Context** | up to 128k |
|
| 38 |
|
| 39 |
+
> ⚠️ **Requires the [ROCmFPX fork](https://github.com/charlie12345/ROCmFPX) @ [`553504b`](https://github.com/charlie12345/ROCmFPX/commit/553504bd5d16fe1374e9b1508a1f395c9a35d746)** — custom AMD quant types (enum IDs 110–115), not upstream-stable. **Won't load** in stock llama.cpp / LM Studio / Ollama. HF's precision badge is wrong — **pick the file by name**.
|
| 40 |
|
| 41 |
+
## Pick a tier
|
| 42 |
|
| 43 |
+
| File suffix | Size | Best for |
|
| 44 |
+
|---|---|---|
|
| 45 |
+
| `…embF16-headQ6-Q6_0_ROCMFPX_AGENT.gguf` ★ | 26 GB | **best overall** — the flagship |
|
| 46 |
+
| `…embF16-Q8_0_ROCMFPX.gguf` | 29 GB | maximum fidelity |
|
| 47 |
+
| `…embF16-Q4_0_ROCMFP4.gguf` | 19 GB | fastest decode (4-bit) |
|
| 48 |
+
| `…embF16-Q3_0_ROCMFPX.gguf` | 17 GB | smallest |
|
| 49 |
|
| 50 |
+
Agent-routed `_AGENT` tiers + the full enum/bpw table are in the details below and the **Files** tab. All filenames prefixed `Qwopus3.6-27B-Coder-MTP-STRIX-`.
|
| 51 |
|
| 52 |
+
## Quick start
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 53 |
|
|
|
|
| 54 |
```bash
|
| 55 |
+
# build the fork once (pinned commit so the enum IDs match)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 56 |
git clone https://github.com/charlie12345/ROCmFPX.git && cd ROCmFPX
|
| 57 |
+
git checkout 553504b && JOBS=16 scripts/build-strix-rocmfp4-mtp.sh
|
|
|
|
| 58 |
|
| 59 |
+
# serve the flagship — MTP + vision
|
| 60 |
+
HSA_OVERRIDE_GFX_VERSION=11.5.1 build-strix-rocmfp4/bin/llama-server \
|
|
|
|
| 61 |
-m Qwopus3.6-27B-Coder-MTP-STRIX-embF16-headQ6-Q6_0_ROCMFPX_AGENT.gguf \
|
|
|
|
| 62 |
-dev ROCm0 -ngl 999 -fa on -c 32768 \
|
| 63 |
--spec-type draft-mtp --spec-draft-ngl all --spec-draft-n-max 2 \
|
| 64 |
+
--jinja --mmproj mmproj/mmproj-F32.gguf --host 0.0.0.0 --port 8080
|
|
|
|
| 65 |
```
|
| 66 |
|
| 67 |
+
Tool calls: point your client at the **`qwen3_coder`** parser, or the model narrates code instead of emitting structured calls.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 68 |
|
| 69 |
+
<details>
|
| 70 |
+
<summary><b>All tiers · recipe · benchmarks</b></summary>
|
| 71 |
|
| 72 |
+
### All tiers
|
|
|
|
|
|
|
| 73 |
|
| 74 |
+
| File suffix | Preset | Enum | Size | Role |
|
| 75 |
+
|---|---|---|---|---|
|
| 76 |
+
| `embF16-headQ6-Q6_0_ROCMFPX_AGENT.gguf` | `Q6_0_ROCMFPX_AGENT` | 114 | 26 GB | flagship — f16 emb + Q6_K head + imatrix |
|
| 77 |
+
| `embF16-Q8_0_ROCMFPX_AGENT.gguf` | `Q8_0_ROCMFPX_AGENT` | 115 | 30 GB | highest-fidelity agent |
|
| 78 |
+
| `embF16-Q8_0_ROCMFPX.gguf` | `Q8_0_ROCMFPX` | 111 | 29 GB | highest fidelity |
|
| 79 |
+
| `embF16-Q6_0_ROCMFPX.gguf` | `Q6_0_ROCMFPX` | 110 | 24 GB | balanced |
|
| 80 |
+
| `embF16-Q3_0_ROCMFPX_AGENT.gguf` | `Q3_0_ROCMFPX_AGENT` | 113 | 21 GB | smallest agent |
|
| 81 |
+
| `embF16-Q3_0_ROCMFPX.gguf` | `Q3_0_ROCMFPX` | 112 | 17 GB | smallest |
|
| 82 |
+
| `embF16-Q4_0_ROCMFP4.gguf` | `Q4_0_ROCMFP4` | 100 | 19 GB | fastest decode (4-bit body) |
|
| 83 |
|
| 84 |
+
f16 token embeddings throughout; `_AGENT` presets keep attention/FFN routing at higher precision for tool-call/JSON coherence. (HF labels `Q4`/`Q8` but not `Q6`/`Q3` — the latter aren't standard llama.cpp quant names.)
|
|
|
|
|
|
|
|
|
|
| 85 |
|
| 86 |
+
### Verification (Strix Halo gfx1151)
|
|
|
|
| 87 |
|
| 88 |
+
| Metric | Value |
|
| 89 |
+
|---|---|
|
| 90 |
+
| Functional smoke (`check-rocmfpx-agentic-smoke.sh`) | chat / coding / JSON / tool-call ✅ |
|
| 91 |
+
| PPL vs `Q6_K` (code corpus) | flagship **2.922** vs Q6_K **2.917** → **+0.17%** (within ±0.04) |
|
| 92 |
|
| 93 |
+
### Performance — prompt-processing throughput (t/s) vs `Q6_K`
|
| 94 |
|
| 95 |
+
| Context | `Q6_K` | flagship | Δ |
|
| 96 |
|---|---|---|---|
|
| 97 |
| pp512 | 204 | 269 | **+32%** |
|
| 98 |
+
| pp2048 | 193 | 254 | +31% |
|
| 99 |
+
| pp10k | 181 | 236 | +30% |
|
| 100 |
+
| pp16k | 174 | 225 | +29% |
|
| 101 |
+
| pp32k | 158 | 199 | +26% |
|
| 102 |
| pp64k | 134 | 163 | **+22%** |
|
| 103 |
|
| 104 |
+
The gfx1151-tuned kernels win the compute-bound prefill; the edge is largest at short context and narrows toward +22% at 64k as O(n²) attention takes over. Decode is bandwidth-bound (≈ Q6_K raw), and **MTP (`--spec-type draft-mtp`) ~doubles it** in serving. `Q4_0_ROCMFP4` is the decode king (~13 tok/s raw). *Single-rep `llama-bench`; treat absolutes as ±a few %.*
|
| 105 |
+
|
| 106 |
+
</details>
|
| 107 |
|
| 108 |
## Credits & license
|
| 109 |
|
| 110 |
+
Apache-2.0 (inherited). Jackrong + Kyle Hessling (fine-tune) → Qwen3.6-27B (base) → [charlie12345 / ROCmFPX](https://github.com/charlie12345/ROCmFPX) (quant fork). ROCmFPX quantization by this repo's author.
|
|
|
|
|
|