ewinregirgojr

MiniCPM5-1B-Agentic-Tooluse-QLoRA-v3

README

License: apache-2.0

Why MiniCPM5-1B-Agentic-Tooluse?

Tiny footprint, real accuracy. 1B parameters total, LoRA adapter itself is under 100MB — deployable anywhere a 7B+ model can't go: mobile apps, browser extensions, IoT/embedded agents, offline assistants, cost-sensitive high-throughput API backends.
Purpose-built for agentic tool use. Trained specifically to parse a tool/function schema plus a natural-language user request and emit a correctly-named, correctly-structured, correctly-valued function call — the core skill every LLM agent framework (LangChain, LlamaIndex, AutoGen, CrewAI, custom ReAct loops, MCP servers) depends on.
Two-stage training pipeline: QLoRA SFT + GRPO reinforcement learning. Most open tool-calling fine-tunes stop at supervised fine-tuning. This adapter goes a step further with GRPO (Group Relative Policy Optimization) reinforcement learning on top of the SFT checkpoint, specifically rewarding exact function-name selection and exact argument-value correctness — the two hardest, most failure-prone parts of tool calling for small models.
Honestly measured, not marketing numbers. Every metric below comes from one single evaluation harness run end-to-end on a locked, held-out 300-example test split — same parser, same grader, same slice, for the base model, the SFT model, and this GRPO-refined v3 model. No cherry-picked runs, no mixed benchmarks.

Benchmark results (held-out test set, n=300)

Table with columns: metric, v2 (previous release), SFT retrain, v3 (SFT + GRPO), Δ vs SFT
metric	v2 (previous release)	SFT retrain	v3 (SFT + GRPO)	Δ vs SFT
parseable	0.9933	1.0000	1.0000	+0.0000
valid_name	0.9700	0.9867	0.9933	+0.0066
expected_name	0.9067	0.9567	0.9733	+0.0166
args_exact	0.6133	0.7367	0.7500	+0.0133
arg_key_overlap	0.8757

What these metrics mean, in plain terms:

parseable — the model's output is well-formed, machine-parseable function-call syntax (essentially solved at 100%).
valid_name — the model calls a function that actually exists in the provided tool schema.
expected_name — the model calls the correct function for the user's request (97%+ of the time, up from v2).
args_exact — the strictest metric: correct function and every single argument value matches the expected value exactly. This is the metric that separates a model that "looks right" from one that actually works in production, and it's where the GRPO reinforcement-learning stage delivers its biggest win over plain SFT.
arg_key_overlap — how much of the correct argument set (keys) the model produces, even when some values are imperfect.

Model details

Base model: openbmb/MiniCPM5-1B — a compact, efficient, Llama-architecture 1B-parameter language model from OpenBMB, ideal for resource-constrained inference, edge computing, and low-latency serving.
Adapter type: LoRA (Low-Rank Adaptation) via PEFT, rank r=32, alpha=64, dropout=0.05
Target modules: q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj (full attention + MLP coverage)
Training pipeline: QLoRA supervised fine-tuning on tool-calling / function-calling trajectories → GRPO reinforcement-learning refinement optimizing for exact argument correctness

Quickstart

python
from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

base = AutoModelForCausalLM.from_pretrained("openbmb/MiniCPM5-1B")
tok = AutoTokenizer.from_pretrained("openbmb/MiniCPM5-1B")
model = PeftModel.from_pretrained(base, "ewinregirgojr/MiniCPM5-1B-Agentic-Tooluse-QLoRA-v3")

# Use tok.apply_chat_template(..., tools=[...]) with your function/tool schema,
# then generate as usual — the model emits a structured function call.

Prefer not to deal with adapter loading, or want a single-file local build? See the related repos below for a merged full-weight checkpoint and quantized GGUF files for llama.cpp / Ollama / LM Studio.

Ideal use cases

Local, private, offline AI agents that need to call tools/APIs without sending data to a cloud LLM provider
Home automation and smart-home assistants (small enough to run on a Raspberry Pi-class device or a home server)
Mobile and embedded applications where a 7B+ model is impractical
High-throughput, cost-sensitive backend services orchestrating many tool calls per request
Any LangChain / LlamaIndex / AutoGen / MCP-based agent that needs a cheap, fast, locally-hostable function-calling backbone
Research and experimentation on small-model reasoning, LoRA fine-tuning, and RL-based (GRPO) post-training for structured generation

Merged, full-precision FP16 build (no PEFT/adapter loading required): MiniCPM5-1B-Agentic-Tooluse-v3-Merged-FP16
GGUF quantized builds for llama.cpp, Ollama, LM Studio (F16 / Q8_0 / Q4_K_M): MiniCPM5-1B-Agentic-Tooluse-v3-GGUF
Previous adapter version: MiniCPM5-1B-Agentic-Tooluse-QLoRA-v2

FAQ

Is this a full model or an adapter? This repo is a LoRA adapter — small, fast to download, must be loaded on top of the base MiniCPM5-1B model via PEFT. If you want a single ready-to-serve checkpoint, use the Merged-FP16 or GGUF repos linked above instead.

Can I run this on CPU / a laptop / a phone? Yes — the whole point of a 1B-parameter model is that it's small enough for CPU inference, laptops, and (via the GGUF quantized builds) even lower-power edge devices.

How does this compare to using GPT-4o / Claude for function calling? This model trades some absolute accuracy for massive gains in cost, latency, privacy, and deployability — you get a locally-hostable, fine-tunable, fully open-weight alternative for agentic tool-use workloads where sending every request to a large hosted API isn't practical or affordable.

What license is this under? Apache 2.0, matching the base model.

Base model

Built on MiniCPM5-1B by OpenBMB.

Available on FriendliAI

Dedicated Endpoints

Run this model inference on single tenant GPU with unmatched speed and reliability at scale.

Learn more

Container

Run this model inference with full control and performance in your environment.

Learn more

Model Details

Model Provider

ewinregirgojr

Model Tree

Base

openbmb/MiniCPM5-1B

Adapter

this model

Input Modalities

Text

Output Modalities

Text

Supported Functionality

Dedicated EndpointsContainer

Explore FriendliAI today

Get started Talk to an engineer

README

License: apache-2.0

Why MiniCPM5-1B-Agentic-Tooluse?

Tiny footprint, real accuracy. 1B parameters total, LoRA adapter itself is under 100MB — deployable anywhere a 7B+ model can't go: mobile apps, browser extensions, IoT/embedded agents, offline assistants, cost-sensitive high-throughput API backends.
Purpose-built for agentic tool use. Trained specifically to parse a tool/function schema plus a natural-language user request and emit a correctly-named, correctly-structured, correctly-valued function call — the core skill every LLM agent framework (LangChain, LlamaIndex, AutoGen, CrewAI, custom ReAct loops, MCP servers) depends on.
Two-stage training pipeline: QLoRA SFT + GRPO reinforcement learning. Most open tool-calling fine-tunes stop at supervised fine-tuning. This adapter goes a step further with GRPO (Group Relative Policy Optimization) reinforcement learning on top of the SFT checkpoint, specifically rewarding exact function-name selection and exact argument-value correctness — the two hardest, most failure-prone parts of tool calling for small models.
Honestly measured, not marketing numbers. Every metric below comes from one single evaluation harness run end-to-end on a locked, held-out 300-example test split — same parser, same grader, same slice, for the base model, the SFT model, and this GRPO-refined v3 model. No cherry-picked runs, no mixed benchmarks.

Benchmark results (held-out test set, n=300)

Table with columns: metric, v2 (previous release), SFT retrain, v3 (SFT + GRPO), Δ vs SFT
metric	v2 (previous release)	SFT retrain	v3 (SFT + GRPO)	Δ vs SFT
parseable	0.9933	1.0000	1.0000	+0.0000
valid_name	0.9700	0.9867	0.9933	+0.0066
expected_name	0.9067	0.9567	0.9733	+0.0166
args_exact	0.6133	0.7367	0.7500	+0.0133
arg_key_overlap	0.8757

What these metrics mean, in plain terms:

parseable — the model's output is well-formed, machine-parseable function-call syntax (essentially solved at 100%).
valid_name — the model calls a function that actually exists in the provided tool schema.
expected_name — the model calls the correct function for the user's request (97%+ of the time, up from v2).
args_exact — the strictest metric: correct function and every single argument value matches the expected value exactly. This is the metric that separates a model that "looks right" from one that actually works in production, and it's where the GRPO reinforcement-learning stage delivers its biggest win over plain SFT.
arg_key_overlap — how much of the correct argument set (keys) the model produces, even when some values are imperfect.

Model details

Base model: openbmb/MiniCPM5-1B — a compact, efficient, Llama-architecture 1B-parameter language model from OpenBMB, ideal for resource-constrained inference, edge computing, and low-latency serving.
Adapter type: LoRA (Low-Rank Adaptation) via PEFT, rank r=32, alpha=64, dropout=0.05
Target modules: q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj (full attention + MLP coverage)
Training pipeline: QLoRA supervised fine-tuning on tool-calling / function-calling trajectories → GRPO reinforcement-learning refinement optimizing for exact argument correctness

Quickstart

python
from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

base = AutoModelForCausalLM.from_pretrained("openbmb/MiniCPM5-1B")
tok = AutoTokenizer.from_pretrained("openbmb/MiniCPM5-1B")
model = PeftModel.from_pretrained(base, "ewinregirgojr/MiniCPM5-1B-Agentic-Tooluse-QLoRA-v3")

# Use tok.apply_chat_template(..., tools=[...]) with your function/tool schema,
# then generate as usual — the model emits a structured function call.

Ideal use cases

Local, private, offline AI agents that need to call tools/APIs without sending data to a cloud LLM provider
Home automation and smart-home assistants (small enough to run on a Raspberry Pi-class device or a home server)
Mobile and embedded applications where a 7B+ model is impractical
High-throughput, cost-sensitive backend services orchestrating many tool calls per request
Any LangChain / LlamaIndex / AutoGen / MCP-based agent that needs a cheap, fast, locally-hostable function-calling backbone
Research and experimentation on small-model reasoning, LoRA fine-tuning, and RL-based (GRPO) post-training for structured generation

Merged, full-precision FP16 build (no PEFT/adapter loading required): MiniCPM5-1B-Agentic-Tooluse-v3-Merged-FP16
GGUF quantized builds for llama.cpp, Ollama, LM Studio (F16 / Q8_0 / Q4_K_M): MiniCPM5-1B-Agentic-Tooluse-v3-GGUF
Previous adapter version: MiniCPM5-1B-Agentic-Tooluse-QLoRA-v2

FAQ

What license is this under? Apache 2.0, matching the base model.

MiniCPM5-1B-Agentic-Tooluse-QLoRA-v3

README

Why MiniCPM5-1B-Agentic-Tooluse?

Benchmark results (held-out test set, n=300)

Model details

Quickstart

Ideal use cases

Related repos

FAQ

Base model

Explore FriendliAI today

README

Why MiniCPM5-1B-Agentic-Tooluse?

Benchmark results (held-out test set, n=300)

Model details

Quickstart

Ideal use cases

Related repos

FAQ

Base model