danelcsb/openlocalagent-10m-matrix

la-10m — LocalAgent (10.52M params)

A from-scratch, byte-fallback BPE tool-calling agent model from LocalAgent. Pure PyTorch, 10.52M params, trained on CPU. It pairs a tiny decoder (GQA + RoPE + SwiGLU) with a dual head (tool-selection classifier + pointer/copy argument head) and prompt-grounded constrained decoding for reliable tool calls across 0 tools (general assistant, the Claude Code / Codex coding surface, and computer-use / productivity tools), including parallel two-call turns.

Architecture

• vocab 16384 (byte-fallback BPE), d_model 384, layers 4, heads 6/1 (GQA), ffn 512
• factorized embeddings: False
• tokenizer: tokenizer.json (SHA-256 2f49a6444eff01bf89f9bd85f895e22efcfebc830cd3f97bc362ad0803570620)

Files

• config.json — ModelConfig
• model.safetensors — decoder weights
• agent_heads.bin — trained tool/pointer/route/dispatch heads plus dispatch metadata (optional)

What it can do (use cases)

One byte-fallback BPE model that turns a natural-language turn into a grounded tool call — across an assistant, a coding agent, computer-use/productivity apps, and parallel two-call turns:

you say	it calls
"What's the weather in Cusco?"	get_weather(city="Cusco")
"What is 19 * 19 * 5?"	calculator(expression="19*19*5")
"Open the file bin/run.sh."	read_file(path="bin/run.sh")
"Grep for 'TODO'."	grep_search(pattern="TODO")
"Run the tests."	run_tests()
"Commit with message 'fix bug'."	git_commit(message="fix bug")
"Send an email to Greta."	send_email(recipient="Greta")
"Go to figma.com."	open_url(url="figma.com")
"Send a Slack message saying 'ship it'."	slack_send(message="ship it")
"Create a Jira ticket titled 'broken link'."	jira_issue(summary="broken link")
"Compose an email to Judy and search for how tall is Everest."	send_email(recipient="Judy") + web_search(query="how tall is Everest")

Multi-turn coding (grounds a follow-up arg from a tool response): read_file(tests/test_api.py) → result → run_tests() → "FAILED…" → fix. At catalog scale (100s–1000s of tools) selection is done by retrieval (top-k) instead of a fixed head. See the LocalAgent repo.

Load (pure PyTorch, no transformers)

import json, torch
from huggingface_hub import hf_hub_download
from openlocalagent.model import LocalAgentLM, ModelConfig

cfg_d = json.load(open(hf_hub_download("danelcsb/openlocalagent-10m-matrix", "config.json")))
cfg = ModelConfig(**{k: v for k, v in cfg_d.items() if k in ModelConfig.__dataclass_fields__})
model = LocalAgentLM(cfg)
from safetensors.torch import load_file
model.load_state_dict(load_file(hf_hub_download("danelcsb/openlocalagent-10m-matrix", "model.safetensors")))
model.eval()
# BPE checkpoints also ship tokenizer.json; byte-tier checkpoints use ByteTokenizer().
tokenizer_meta = cfg_d.get("tokenizer", {})
if tokenizer_meta.get("kind") == "bpe":
    from openlocalagent.model.tokenizer import load_tokenizer
    tokenizer = load_tokenizer("bpe", hf_hub_download("danelcsb/openlocalagent-10m-matrix", tokenizer_meta["filename"]))

See the LocalAgent repo for the grounded decoder / agent runtime (tool head, pointer head, retrieval, parallel-call decode).