A.X-K1-mic

Slim model weights, executable usage and framework expectations in one repository.

Contract

Slim model

A.X-K1-mic

Slim model weights, executable usage and framework expectations in one repository.

Contract

A.X K1 text generation

Runnable usage and the framework expectations directory are linked above.

Slim model

Slim fixture: A.X-K1-mic

A 21.0M-parameter fixture for the AXK1 MIC, covering skt/A.X-K1. It is a test artifact, not a usable model, tagged test-artifact to keep it out of candidate-model search.

  • Fixture key 6dc3907d0172. It covers every checkpoint whose processor files are byte-identical and whose config matches after shrinking.
  • Same architecture, smaller config: AXK1ForCausalLM, shrunk by the generic rule in sweep.py (num_hidden_layers 2).
  • Processor files copied byte for byte from skt/A.X-K1: chat_template.jinja, generation_config.json, special_tokens_map.json, tokenizer.json, tokenizer_config.json.
  • Random weights (seed 17). They passed the input-sensitivity gate: every perturbation of the contract's inputs moved the logits by at least 0.001 (weakest: attention_mask:all_ones, 0.228), so no training was needed.
  • Same tensor names as the real checkpoint. The real checkpoint lacks model.layers.1.mlp.gate.e_score_correction_bias, model.layers.10.mlp.gate.e_score_correction_bias, model.layers.11.mlp.gate.e_score_correction_bias, model.layers.12.mlp.gate.e_score_correction_bias, model.layers.13.mlp.gate.e_score_correction_bias, model.layers.14.mlp.gate.e_score_correction_bias, model.layers.15.mlp.gate.e_score_correction_bias, model.layers.16.mlp.gate.e_score_correction_bias, model.layers.17.mlp.gate.e_score_correction_bias, model.layers.18.mlp.gate.e_score_correction_bias, model.layers.19.mlp.gate.e_score_correction_bias, model.layers.2.mlp.gate.e_score_correction_bias, model.layers.20.mlp.gate.e_score_correction_bias, model.layers.21.mlp.gate.e_score_correction_bias, model.layers.22.mlp.gate.e_score_correction_bias, model.layers.23.mlp.gate.e_score_correction_bias, model.layers.24.mlp.gate.e_score_correction_bias, model.layers.25.mlp.gate.e_score_correction_bias, model.layers.26.mlp.gate.e_score_correction_bias, model.layers.27.mlp.gate.e_score_correction_bias, model.layers.28.mlp.gate.e_score_correction_bias, model.layers.29.mlp.gate.e_score_correction_bias, model.layers.3.mlp.gate.e_score_correction_bias, model.layers.30.mlp.gate.e_score_correction_bias, model.layers.31.mlp.gate.e_score_correction_bias, model.layers.32.mlp.gate.e_score_correction_bias, model.layers.33.mlp.gate.e_score_correction_bias, model.layers.34.mlp.gate.e_score_correction_bias, model.layers.35.mlp.gate.e_score_correction_bias, model.layers.36.mlp.gate.e_score_correction_bias, model.layers.37.mlp.gate.e_score_correction_bias, model.layers.38.mlp.gate.e_score_correction_bias, model.layers.39.mlp.gate.e_score_correction_bias, model.layers.4.mlp.gate.e_score_correction_bias, model.layers.40.mlp.gate.e_score_correction_bias, model.layers.41.mlp.gate.e_score_correction_bias, model.layers.42.mlp.gate.e_score_correction_bias, model.layers.43.mlp.gate.e_score_correction_bias, model.layers.44.mlp.gate.e_score_correction_bias, model.layers.45.mlp.gate.e_score_correction_bias, model.layers.46.mlp.gate.e_score_correction_bias, model.layers.47.mlp.gate.e_score_correction_bias, model.layers.48.mlp.gate.e_score_correction_bias, model.layers.49.mlp.gate.e_score_correction_bias, model.layers.5.mlp.gate.e_score_correction_bias, model.layers.50.mlp.gate.e_score_correction_bias, model.layers.51.mlp.gate.e_score_correction_bias, model.layers.52.mlp.gate.e_score_correction_bias, model.layers.53.mlp.gate.e_score_correction_bias, model.layers.54.mlp.gate.e_score_correction_bias, model.layers.55.mlp.gate.e_score_correction_bias, model.layers.56.mlp.gate.e_score_correction_bias, model.layers.57.mlp.gate.e_score_correction_bias, model.layers.58.mlp.gate.e_score_correction_bias, model.layers.59.mlp.gate.e_score_correction_bias, model.layers.6.mlp.gate.e_score_correction_bias, model.layers.60.mlp.gate.e_score_correction_bias, model.layers.7.mlp.gate.e_score_correction_bias, model.layers.8.mlp.gate.e_score_correction_bias, model.layers.9.mlp.gate.e_score_correction_bias, so the fixture omits it too, and loading initializes it as it does for users.
  • Saved in bfloat16, the dtype skt/A.X-K1 loads in by default.
  • Finite in float32, bfloat16, and float16 (forward, loss, gradients).

Built by build_fixtures.py; build.json records the pinned sources, gate results, and file hashes. This repository includes the slim weights, usage and framework expectations; rebuild the other files with python example/build_fixtures.py --fixture A.X-K1-mic.

Downloads last month
2,256
Safetensors
Model size
21M params
Tensor type
BF16
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for huggingface-mic/A.X-K1-mic

Base model

skt/A.X-K1
Finetuned
(1)
this model

Space using huggingface-mic/A.X-K1-mic 1