DeepSeek-V3.2-mic

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

Contract

DeepSeek-V3.2 causal LM

Runnable usage and the framework expectations directory are linked above.

Slim model

DeepSeek-V3.2-mic

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

Contract

DeepSeek-V3.2 causal LM

Runnable usage and the framework expectations directory are linked above.

Slim model

Slim fixture: DeepSeek-V3.2-mic

A 16.9M-parameter fixture for the DeepSeekV32 MIC, covering deepseek-ai/DeepSeek-V3.2. It is a test artifact, not a usable model, tagged test-artifact to keep it out of candidate-model search.

  • Fixture key c6a205d62f54. It covers every checkpoint whose processor files are byte-identical and whose config matches after shrinking.
  • Same architecture, smaller config: DeepseekV32ForCausalLM, shrunk by the generic rule in sweep.py (num_hidden_layers 2).
  • Processor files copied byte for byte from deepseek-ai/DeepSeek-V3.2: LICENSE, generation_config.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: input_ids:roll_batch, 1.12), so no training was needed.
  • Same tensor names as the real checkpoint. The real checkpoint has every parameter, and so does the fixture.
  • Saved in bfloat16, the dtype deepseek-ai/DeepSeek-V3.2 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 DeepSeek-V3.2-mic.

Downloads last month
241
Safetensors
Model size
16.9M 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/DeepSeek-V3.2-mic

Finetuned
(35)
this model

Space using huggingface-mic/DeepSeek-V3.2-mic 1