Gemma3-1B-tools

Gemma3 1B Tools

Gemma 3 1B fine-tuned with LoRA for tool calling and agent-style interactions.

Base model

This model was fine-tuned from:

google/gemma-3-1b-it

Training

Training was performed using NVIDIA NeMo AutoModel with LoRA/PEFT.

LoRA configuration

• LoRA dimension: 32
• LoRA alpha: 32
• Dropout: 0.05
• Target modules: *.proj (all *_proj linear layers)

Training configuration

• Max sequence length: 4096
• Learning rate: 5e-5 (cosine decay, 15 warmup steps, min 1e-6)
• Weight decay: 0.01
• Global batch size: 64 (micro batch 2 x 32 accumulation)
• Training steps: 336 (4 epochs)
• Mixed precision: bf16
• Validation loss: 0.579 → 0.4715 (final epoch)

Dataset

Training used the sft_tools split of the r0b0tlab/qwen3.8-max-glm5.2-kimi-k3-distillation dataset.

Tool-calling format

This model was trained with a custom chat template (bundled as chat_template.jinja). It renders the tool schemas into a developer turn and emits tool calls as:

<tool_call>
{"name": <function-name>, "arguments": <args-json-object>}
</tool_call>

Serving stacks must render prompts with this template (e.g. via the tokenizer's apply_chat_template) for tool calling to work.

Intended use

• Structured tool/function calling
• Agent-style multi-step interactions
• Small-footprint on-device or edge deployment

It is not intended to be a general replacement for larger Gemma models.

Usage

Serve the model:

trtllm-serve serve iromu/Gemma3-1B-tools --port 8000

Load it with llama.cpp:

llama-cli -hf iromu/Gemma3-1B-tools-GGUF:Q4_K_M

<!-- VALIDATION:BEGIN (auto-generated, do not edit) -->

Validation matrix

Tool-calling validation on the sft_tools validation split (greedy decoding, 384 max new tokens). Throughput is single-stream greedy decode, not serving throughput.

Pretrained base (google/gemma-3-1b-it): 2.0% exact-args match (1/50). Fine-tuned (BF16): 66.0% exact-args match (33/50) (+64pp vs base).

• GGUF-BF16: 20/50 (40.0%) exact, 66.0 tok/s — 61% of BF16.
• GGUF-Q4_K_M: 10/50 (20.0%) exact, 89.5 tok/s — 30% of BF16.
• GGUF-Q5_K_M: 24/50 (48.0%) exact, 60.7 tok/s — 73% of BF16.
• GGUF-Q8_0: 22/50 (44.0%) exact, 50.2 tok/s — 67% of BF16.
• NVFP4: 2/50 (4.0%) exact, 169.0 tok/s — 6% of BF16.

<!-- VALIDATION:END -->