You can deploy this model and start earning money today!

micahr234/
mouse-example-model-online

mouse-core

pytorch

reinforcement-learning

Views

No views yet

micahr234/mouse-example-model-online

This repository contains a MOUSE model checkpoint.

Architecture

Backbone: qwen3
Hidden dimension: 1024
Heads: action_value
Action head: action_value

Encoder

StepEmbedder reads flat step-record dicts and projects each declared modality into the shared 1024-dimensional token space before the backbone.

Field	Type	Required	Tensor shape	Dtype	Notes
`action`	`discrete`	yes	`[B, S]`	`torch.long`	integer ids in `[0, 3]`
`observation`	`discrete`	yes	`[B, S]`	`torch.long`	integer ids in `[0, 63]`
`reward`	`rff`	yes	`[B, S]`	`torch.float32`	scalar value
`done`	`discrete`	yes	`[B, S]`	`torch.long`	integer ids in `[0, 4]`

Install MouseCore

pip install mouse-core

Load The Model

python

import torch
from mouse_core import load_model

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = load_model("micahr234/mouse-example-model-online", map_location="cpu").eval().to(device)

Run Inference

The model accepts a list[list[dict]] batch of shape [B][S] — B sequences, each containing S step-record dicts with flat keys matching the encoder's declared modalities above.

python

# Batch shape: [B=1][S=1] — one sequence of one step.
batch = [[
    {
    "action": 0,
    "observation": 0,
    "reward": 0.0,
    "done": 0,
    }
]]
predictions, objective_data, cache = model(batch)

with torch.no_grad():
    predictions, _, cache = model(batch)
    action = model.get_action(predictions, temperature=0.0)

model() returns (predictions, objective_data, cache). objective_data is a TensorDict[B, S] of the modality tensors extracted by the encoder — pass it to objectives during training. For cached incremental rollout, keep cache and pass it back on the next call with use_cache=True. Cached batch rows may have different lengths on every call (e.g. envs emitting different numbers of steps between model calls): decoding runs through a FlexAttention session carried in the cache, so each row decodes exactly as it would alone.