Nanthasit/sakthai-coder-browser-gguf

SakThai Coder Browser — GGUF 🤖🌐

F16 GGUF build of the browser automation model — run a web agent on your laptop with llama.cpp

⚠ BROKEN — DO NOT DEPLOY (as of 2026-07-31) — This GGUF was converted from the merged weights that are corrupted by a faulty LoRA merge: all 84 attention-projection bias tensors are non-zero while Qwen2 initializes these biases to ZERO (layer-0 k_proj.bias absmean 27.7 / max 354). Multi-trial inference probes of the merged model produced only whitespace loops — 0 tool calls, 0 valid JSON at temp ≤ 0.7 (seeds 7/42/1337), and a GGUF Q4_K_M probe returned 0 output tokens on all 3 trials. Full evidence: .eval_results/benchmark-20260731_052122.yaml. The fault is in the weights, not the GGUF conversion or the prompt format (GGUF tensor layout is structurally identical to the working sakthai-plus-1.5b GGUF). A clean re-merge of the LoRA adapter is required before this build is usable. Do not download for deployment until re-merged and re-verified.

The GGUF edition of the SakThai browser automation model — a Qwen2.5-Coder-1.5B-Instruct fine-tune converted to F16 GGUF for use with llama.cpp, llama-cpp-python, and Ollama.

⚠ Status: BROKEN — see banner above. The pages below document the intended design and usage; the current weights do not produce valid output until the parent model is re-merged from a clean LoRA merge.

---

What Is This?

This is an F16 (full-precision) GGUF conversion of the sakthai-coder-browser merged model. It transforms a Qwen2.5-Coder-1.5B-Instruct general-purpose LLM into a browser automation agent that outputs structured <tool_call> XML for web interaction.

Variant	Repository	Best For
🧠 Merged (Transformers)	sakthai-coder-browser	Python / Transformers pipelines
🎯 LoRA Adapter	sakthai-coder-browser-lora	Fine-tuning / PEFT workflows
💾 GGUF (this repo)	⬅ sakthai-coder-browser-gguf	CPU inference, llama.cpp, Ollama, edge devices

---

File

File	Format	Size	Purpose
sakthai-coder-browser-f16.gguf	GGUF F16	7,111,560,032 B (7.11 GB)	Full-precision GGUF for local CPU/GPU inference

Note: F16 preserves the full model quality. For a smaller footprint, a Q4_K_M quantized version may follow based on demand — ⭐ like this repo or open an issue to vote for it.

---

Supported Browser Actions

The model generates structured XML tool calls for these browser operations:

Tool	Example
browser_navigate(url)	<tool_call>{"name": "browser_navigate", "arguments": {"url": "https://example.com"}}</tool_call>
browser_click(element)	<tool_call>{"name": "browser_click", "arguments": {"element": "#search-button"}}</tool_call>
browser_type(element, text)	<tool_call>{"name": "browser_type", "arguments": {"element": "#search-input", "text": "AI news"}}</tool_call>
browser_extract()	<tool_call>{"name": "browser_extract", "arguments": {}}</tool_call>

---

Quick Start

⚠ These commands are the intended usage. With the current BROKEN weights they will not produce valid tool calls — see the banner above. Re-run only after a clean re-merge is published.

llama.cpp (CLI)

# Download the GGUF file
huggingface-cli download Nanthasit/sakthai-coder-browser-gguf \
  sakthai-coder-browser-f16.gguf --local-dir ./

# …or with wget
wget https://huggingface.co/Nanthasit/sakthai-coder-browser-gguf/resolve/main/sakthai-coder-browser-f16.gguf

# Run with llama.cpp
./llama-cli -m sakthai-coder-browser-f16.gguf \
  -p "<|im_start|>system\nYou are a browser automation assistant. You can browse the web, click elements, type text, and extract information from pages.<|im_end|>\n<|im_start|>user\nGo to google.com and search for the latest AI news<|im_end|>\n<|im_start|>assistant\n" \
  -n 512 -t 8 --temp 0.3

llama-cpp-python

pip install llama-cpp-python huggingface-hub

from llama_cpp import Llama
from huggingface_hub import hf_hub_download

# Download GGUF
model_path = hf_hub_download(
    repo_id="Nanthasit/sakthai-coder-browser-gguf",
    filename="sakthai-coder-browser-f16.gguf"
)

# Load model
llm = Llama(
    model_path=model_path,
    n_ctx=4096,
    n_threads=8,
    n_gpu_layers=-1,  # Offload all layers to GPU if available
    verbose=False,
)

# Run inference
output = llm.create_chat_completion(
    messages=[
        {"role": "system", "content": "You are a browser automation assistant. Output tool calls as <tool_call> JSON."},
        {"role": "user", "content": "Search for the latest AI news on Google and summarize the top result."}
    ],
    max_tokens=512,
    temperature=0.3,
    stop=["<|im_end|>"],
)
print(output["choices"][0]["message"]["content"])

Ollama (Modelfile)

FROM ./sakthai-coder-browser-f16.gguf
TEMPLATE """{{ .System }}

{{ .Prompt }}"""
PARAMETER temperature 0.3
PARAMETER top_p 0.8
PARAMETER stop "<|im_end|>"

ollama create sakthai-coder-browser -f Modelfile
ollama run sakthai-coder-browser

---

Architecture

Property	Value
Base Model	Qwen/Qwen2.5-Coder-1.5B-Instruct
Architecture	Qwen2ForCausalLM (decoder-only transformer)
Parameters	1.54B
Hidden Size	1,536
Layers	28
Attention Heads	12 (GQA: 2 KV heads)
Intermediate Size	8,960
Max Position	32,768 tokens
Vocab Size	151,936
Precision	F16 (GGUF)
Activation	SiLU (SwiGLU)
Normalization	RMSNorm (eps=1e-6)
Quantization	None (F16 — full precision)
File Format	GGUF (GPT-Generated Unified Format)

---

Training Summary

Property	Detail
Fine-tuning Method	SFT via LoRA (r=16, α=32, dropout 0.05, rsLoRA) on all 7 linear projections (q/k/v/o + gate/up/down), then merged to full weights — verified from sibling adapter_config.json (peft 0.20.0, CAUSAL_LM, bias none)
Training Data	SimpleToolCalling + sakthai-combined-v7
Context Length	32,768 tokens
Hardware	Free T4 GPU (Kaggle)
Budget	$0
License	Apache 2.0

The model learns to produce structured <tool_call> JSON for browser actions through supervised fine-tuning on curated web navigation trajectories.

Benchmarks & Evaluation Status

Honest status: this fine-tune does not yet have verified benchmark scores of its own. No model-index is published because no measured numbers exist — publishing one would be misleading. What is known:

Item	Status
Base model reference (Qwen2.5-Coder-1.5B-Instruct)	HumanEval pass@1: 74.4 · MBPP pass@1: 71.2 (as published on the sibling sakthai-coder-1.5b card; reference values, not measured on this fine-tune)
This model's own eval	❌ Resolved — MODEL_BROKEN (bias corruption): multi-trial probes + weight inspection of the merged weights (2026-07-31 05:50 UTC) found all 84 attention bias tensors non-zero. See the banner above and the parent repo's benchmark YAML.
GGUF Q4_K_M probe (2026-07-31)	3/3 trials returned 0 output tokens — confirmed weight corruption, NOT a prompt-format mismatch (weight inspection: 84/84 bias tensors absmean > 0.01; GGUF tensor layout structurally identical to the working sakthai-plus-1.5b GGUF)
Hosted inference	Not available — router probe returned 404; run locally via llama.cpp / llama-cpp-python / Ollama
Cron eval (repo's own .eval_results)	cron-eval-sakthai-coder-browser-gguf-2026-07-31-1.yaml — confirms: no model-index yet (no measured scores), card quality 90/100, repo hygiene 95/100, health 32/100 (capped by 0 downloads / 0 momentum at eval time)

Ecosystem Status (from the repo's own cron-eval YAML, 2026-07-31):

Metric	Value
Downloads rank	20/20 (0 downloads — brand new repo)
Download velocity	0.0 / day (rank 20)
Card quality score	90/100
Repo hygiene score	95/100
Health score	32/100 (popularity & momentum weigh 0 — nothing to measure yet)
Model-index present	No (honest — no verified benchmark scores to publish)

This F16 GGUF is a direct full-precision conversion of the merged sakthai-coder-browser weights — and the merged weights are currently corrupted by a faulty LoRA merge (see banner). This build is not usable until the parent is re-merged from a clean LoRA merge and this GGUF is re-converted. Treat all quality claims as invalid for the current artifact.

---

Limitations

• BROKEN weights (as of 2026-07-31) — this GGUF was converted from the corrupted coder-browser merge (all 84 attention bias tensors non-zero); inference produces whitespace loops / 0 output tokens. Do not deploy until the parent model is re-merged and re-verified and this build is re-converted.
• Text-only — this model cannot see images or screenshots (use sakthai-vision-7b for vision tasks)
• CPU inference is slow at F16 — the full ~7.1 GB model benefits from GPU offloading; use n_gpu_layers=-1 when available
• Context-limited — best results with page content ≤ 4K tokens per interaction
• English only — trained primarily on English web data

---

Model Family

Part of the SakThai Model Family — 20 public models across text, vision, speech, and embeddings, all built with $0 budget from a shelter in Cork, Ireland. Live download counts as of 2026-07-31; sizes API-verified from largest weight file.

Model	Size	Purpose	⬇️
Context 1.5B Merged	3.09 GB	Merged weights (flagship)	1,599
Context 0.5B Merged	988 MB	Merged weights (edge)	1,370
Context 7B Merged	15.2 GB	Merged weights (power)	744
Context 7B 128K	config-only	YaRN long-context recipe	506
Context 7B Tools	20.2 MB	LoRA adapter	399
Embedding Multilingual	471 MB	Embedding model	362
Context 1.5B Tools v1	~8.75 MB	LoRA adapter (v1)	349
Vision 7B	4.08 GB	Vision-language (GGUF)	186
TTS Model	141 MB	Kokoro-82M TTS (GGUF)	150
Context 0.5B Tools	988 MB	Merged weights	94
Coder 1.5B	1.12 GB	Code model (GGUF)	93
Context 1.5B Merged v2	3.09 GB	Merged weights	0
Context 1.5B Tools v2	73.9 MB	LoRA adapter	0
Coder Browser GGUF ⬅	7.11 GB	Browser agent F16 GGUF (this repo)	0
Coder Browser LoRA	73.9 MB	LoRA adapter	0
Plus 1.5B	3.09 GB	Merged weights	0
Plus 1.5B LoRA	73.9 MB	LoRA adapter	0
Plus 1.5B Coder	—	Planned (no weights yet)	0
Coder Browser	3.09 GB	Merged weights (browser automation)	0

---

Environment

• Author: Nanthasit (Beer)
• Built with: Free tiers only (Kaggle, HF Hub)
• Philosophy: Small models + great data > big models + generic data

---

How You Can Help

• ⭐ Leave a like — tells the algorithm this model matters
• 🔄 Share with anyone building private AI web agents
• 🍴 Fork and experiment — the weights are open
• 💬 Report issues / watch for the re-merge — Beer reads every issue and comment; the BROKEN status will be lifted when a clean re-merge is converted and verified

---

Citation

If you use this model in your work, please cite the base model (Qwen2.5) and link the fine-tune:

@misc{qwen25,
  title={Qwen2.5 Technical Report},
  author={Qwen Team},
  year={2025},
  eprint={2412.15115},
  archivePrefix={arXiv},
  primaryClass={cs.CL}
}

@misc{sakthai-coder-browser-gguf,
  author = {Nanthasit (Beer)},
  title = {SakThai Coder Browser -- GGUF},
  year = {2026},
  publisher = {Hugging Face},
  howpublished = {\url{https://huggingface.co/Nanthasit/sakthai-coder-browser-gguf}}
}

Built with $0 budget from a shelter in Cork, Ireland — proof that open, private AI does not need a datacenter.

---

Part of the House of Sak — one family, one home, $0 budget.