← MIC coverage of Transformers model types

facebook/esm2_t6_8M_UR50D

huggingface-mic/esm2_t6_8M_UR50D-mic@d3655a4ee0f1 ← facebook/esm2_t6_8M_UR50D@c731040fcd8d · EsmForMaskedLM · real weights

CI: no failure on main in 30 days · MIC failures dashboard

Original7.5Mparameters
Fixture7.5Mparameters
Layer stacks1compared
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.

esm.encoder.layer

Original · 6
Fixture · 6
KindOriginalFixture
Layer66✓ 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
ComponentOriginalShareFixtureShare
esm.encoder.layer7.4M98.5%7.4M98.5%
lm_head.dense102.7K1.4%102.7K1.4%
esm.embeddings10.6K0.1%10.6K0.1%
lm_head.layer_norm6400.0%6400.0%
esm.encoder6400.0%6400.0%
esm.contact_head1210.0%1210.0%
lm_head330.0%330.0%

Preserved properties

ConfigPropertyOriginalFixture
modelmodel_typeesmesm✓ kept
modelvocab_size3333✓ kept
modeltie_word_embeddingsTrueTrue✓ kept
modelhidden_actgelugelu✓ kept
modelmax_position_embeddings10261026✓ kept
modelrope_theta10000.010000.0✓ kept

Module classes

Instances of each class in the built model (prefix Esm omitted).

ClassOriginalFixture
Linear3939
LayerNorm1414
Dropout1313
Attention66
Intermediate66
Layer66
Output66
SelfAttention66
SelfOutput66
Embedding11
ContactPredictionHead11
Embeddings11
Encoder11
ForMaskedLM11
LMHead11
Model11
RotaryEmbedding11
ModuleList11
Sigmoid11

Config changes

0 fields changed, 36 unchanged.

FieldOriginalFixture