Modelpublic

h3rb3rn/smollm3-expert-coder-3b

sourceHugging Faceapache-2.0updated 12d agoView on Hugging Face

1likes111downloads

Model Card

MoE Sovereign Systems-Programming & Code-Synthesis Expert 3B -- SmolLM3 (`smollm3-expert-coder-3b`)

![License: Apache 2.0](https://opensource.org/licenses/Apache-2.0) ![Base Model: Qwen3.5-4B](https://huggingface.co/HuggingFaceTB/SmolLM3-3B)

Model Summary

smollm3-expert-coder-3b is a LoRA fine-tune of the text-decoder of Qwen3.5-4B, specialized as the coder domain expert within the MoE Sovereign compound-AI system.

You are a high-assurance systems-programming and code-synthesis expert (moe-expert-coder-4b) specialized in Rust, C++, Python, and Go. Produce precise, compiler-checked code and minimal atomic diffs. Uphold memory-safety invariants strictly -- correct ownership, correct lock-free memory ordering (acquire/release pairing), no data races. Flag any construct you cannot verify as sound rather than guessing.

Base Architecture

SmolLM3-3B is a genuinely open-source dense Transformer (weights, training data, and training code all publicly documented by HuggingFaceTB) -- distinguishing this Spur-2 track from the open-weight-only Qwen3.5 base used in the parallel 4B expert line.

Training Configuration

Parameter	Value
Method	LoRA (rank 16, alpha 32, dropout 0.05), targeting `q/k/v/o_proj` + `gate/up/down_proj`
Trainable parameters	30,228,480 of 3,105,327,104 (0.97%)
Epochs	3
Effective batch size	128 (micro-batch 4 x 8 GPUs x grad-accum 4)
Learning rate	1.5e-5
Training sequence length	4,096 tokens
Optimizer sharding	DeepSpeed ZeRO-2, bf16
Compute	EuroHPC LUMI-G, 8x AMD Instinct MI250X GCDs, ROCm
Training examples	2,294 curated instruction/response pairs

Observed Training Trajectory

Training loss over the run (representative logged steps): 1.536 -> 1.477 -> 1.385. Smooth, monotonic decline consistent with genuine generalization, not memorization.

Prompt Format

ChatML:

<|im_start|>system
{system_prompt}<|im_end|>
<|im_start|>user
{user_message}<|im_end|>
<|im_start|>assistant
{response}<|im_end|>

System Prompt

You are a high-assurance systems-programming and code-synthesis expert (moe-expert-coder-4b) specialized in Rust, C++, Python, and Go. Produce precise, compiler-checked code and minimal atomic diffs. Uphold memory-safety invariants strictly -- correct ownership, correct lock-free memory ordering (acquire/release pairing), no data races. Flag any construct you cannot verify as sound rather than guessing.

Available Formats

File	Notes
`smollm3-expert-coder-3b-Q4_K_M.gguf`	Recommended for single/multi-GPU deployment
`smollm3-expert-coder-3b-Q8_0.gguf`	Higher-fidelity reference quantization

Hardware Guidance

SmolLM3-3B's native context window is 65,536 tokens. On single 8GB GPUs cap num_ctx to 32,768 and use f16 KV-cache on Maxwell-generation hardware (no Flash Attention support there).

Ollama `Modelfile`

dockerfile

FROM ./smollm3-expert-coder-3b-Q4_K_M.gguf
SYSTEM """You are a high-assurance systems-programming and code-synthesis expert (moe-expert-coder-4b) specialized in Rust, C++, Python, and Go. Produce precise, compiler-checked code and minimal atomic diffs. Uphold memory-safety invariants strictly -- correct ownership, correct lock-free memory ordering (acquire/release pairing), no data races. Flag any construct you cannot verify as sound rather than guessing."""
TEMPLATE """{{ if .System }}<|im_start|>system
{{ .System }}<|im_end|>
{{ end }}{{ if .Prompt }}<|im_start|>user
{{ .Prompt }}<|im_end|>
{{ end }}<|im_start|>assistant
{{ .Response }}<|im_end|>"""
PARAMETER stop "<|im_end|>"
PARAMETER temperature 0.2
PARAMETER num_ctx 32768

Limitations

—Does not execute code/queries/tools itself; outputs should be validated against the actual system before use.
—Specialized for its domain; general-purpose conversation is out of scope.

License

Apache 2.0, inherited from the SmolLM3-3B base model.

MoE Sovereign Systems-Programming & Code-Synthesis Expert 3B -- SmolLM3 (smollm3-expert-coder-3b)