itazap/MOSS-Transcribe-Diarize-HF
Original908.5Mparameters
Fixture10.0Mparameters
Layer stacks2compared
Findings0to review
✓ Same architecture, smaller
Every module class, layer kind and checked property of the original is in the fixture.
Layer stacks
One square per layer, colored by kind: the layer's module classes and type attributes. Hover a square for its index and kind.
model.audio_tower.layers
Original · 24
Fixture · 2
| Kind | Original | Fixture | |
|---|---|---|---|
| AudioLayer | 24 | 2 | ✓ kept |
model.language_model.layers
Original · 28
Fixture · 2
| Kind | Original | Fixture | |
|---|---|---|---|
| Qwen3DecoderLayer | 28 | 2 | ✓ kept |
Where the parameters are
Share of each model's parameters per component. Vocabulary and patch sizes are kept, so embeddings usually take a larger share of the fixture.
Original
Fixture
| Component | Original | Share | Fixture | Share |
|---|---|---|---|---|
| model.language_model.layers | 440.5M | 48.5% | 74.0K | 0.7% |
| model.audio_tower.layers | 302.3M | 33.3% | 66.8K | 0.7% |
| model.language_model | 155.6M | 17.1% | 9.7M | 97.1% |
| model.multi_modal_projector | 5.2M | 0.6% | 20.7K | 0.2% |
| model.audio_tower | 4.9M | 0.5% | 123.9K | 1.2% |
Preserved properties
| Config | Property | Original | Fixture | |
|---|---|---|---|---|
| model | model_type | moss_transcribe_diarize | moss_transcribe_diarize | ✓ kept |
| model | tie_word_embeddings | True | True | ✓ kept |
| audio_config | model_type | moss_transcribe_diarize_audio | moss_transcribe_diarize_audio | ✓ kept |
| audio_config | num_mel_bins | 80 | 80 | ✓ kept |
| text_config | attention style | GQA (16 heads, 8 KV) | GQA (4 heads, 2 KV) | ✓ kept |
| text_config | RoPE type | default | default | ✓ kept |
| text_config | model_type | qwen3 | qwen3 | ✓ kept |
| text_config | vocab_size | 151936 | 151936 | ✓ kept |
| text_config | tie_word_embeddings | True | True | ✓ kept |
| text_config | hidden_act | silu | silu | ✓ kept |
| text_config | max_position_embeddings | 131072 | 131072 | ✓ kept |
Module classes
Instances of each class in the built model (prefix MossTranscribeDiarize omitted).
| Class | Original | Fixture |
|---|---|---|
| Linear | 343 | 29 |
| Qwen3RMSNorm | 113 | 9 |
| LayerNorm | 50 | 6 |
| SiLUActivation | 29 | 3 |
| Qwen3Attention | 28 | 2 |
| Qwen3DecoderLayer | 28 | 2 |
| Qwen3MLP | 28 | 2 |
| GELUActivation | 24 | 2 |
| AudioAttention | 24 | 2 |
| AudioLayer | 24 | 2 |
| Conv1d | 2 | 2 |
| Embedding | 2 | 2 |
| ModuleList | 2 | 2 |
| AudioModel | 1 | 1 |
| ForConditionalGeneration | 1 | 1 |
| Model | 1 | 1 |
| MultiModalProjector | 1 | 1 |
| Qwen3Model | 1 | 1 |
| Qwen3RotaryEmbedding | 1 | 1 |
Config changes
11 fields changed, 60 unchanged.
| Field | Original | Fixture |
|---|---|---|
| audio_config.hidden_size | 1024 | 64 |
| audio_config.intermediate_size | 4096 | 128 |
| audio_config.num_attention_heads | 16 | 4 |
| audio_config.num_hidden_layers | 24 | 2 |
| text_config.head_dim | 128 | 16 |
| text_config.hidden_size | 1024 | 64 |
| text_config.intermediate_size | 3072 | 128 |
| text_config.layer_types | 28 entries: 28× full_attention | ["full_attention", "full_attention"] |
| text_config.num_attention_heads | 16 | 4 |
| text_config.num_hidden_layers | 28 | 2 |
| text_config.num_key_value_heads | 8 | 2 |