Synthyra/ESM2-150M

This checkpoint packages the FastPLMs ESM2 implementation.

Accepted inputs are amino-acid sequences tokenized to residue IDs. Supported Transformers entry points are AutoConfig, AutoModel, AutoModelForMaskedLM, AutoModelForSequenceClassification, AutoModelForTokenClassification.

Install and platform requirements

Install the current FastPLMs package:

python -m pip install \
  "fastplms @ git+https://github.com/Synthyra/FastPLMs.git"

Python 3.11-3.14, PyTorch 2.13, and Transformers 5.13 are required. Eager, SDPA, and Flex use the core install. FlashAttention requires the flash extra, compatible CUDA hardware, and BF16 execution. The Hub quick start below requires network access on first download. For an air-gapped run, first build the manifest-pinned local artifact and use the offline form shown in the example.

Quick start

from transformers import AutoModel

model_id = "Synthyra/ESM2-150M"
model = AutoModel.from_pretrained(
    model_id,
    trust_remote_code=True,
).eval()

This example uses the published Hub repository. For offline validation, build the manifest-pinned artifact and replace model_id with its local dist/hub/ESM2-150M path, then pass local_files_only=True.

Leave attention unspecified for the Transformers default. Supported explicit choices are eager, sdpa, flex_attention, flash_attention_2, flash_attention_3. Pass the selected name through attn_implementation. When an optimized backend cannot return full attention tensors, output_attentions=True emits one explicit runtime warning and uses a correctly masked eager implementation for that call only. The warning identifies the configured backend, effective backend, and reason. Configuration and later calls are unchanged. For BF16 execution, this family uses FP32 parameters with CUDA BF16 autocast.

Tokenization and forward inference

Load the tokenizer from the same artifact as the model. Padding is represented explicitly by the attention mask:

import torch
from transformers import AutoTokenizer

model_id = "Synthyra/ESM2-150M"
tokenizer = AutoTokenizer.from_pretrained(
    model_id,
    trust_remote_code=True,
)
batch = tokenizer(
    ["MSTNPKPQRKTKRNT", "MKTIIALSYIFCLVFA"],
    padding=True,
    return_tensors="pt",
)

with torch.inference_mode():
    output = model(**batch)

print(output.last_hidden_state.shape)

Dataset embeddings

The shared embedding API accepts sequences, (id, sequence) pairs, EmbeddingInput records, insertion-ordered {id: sequence} mappings, or a FASTA path. Results preserve order and duplicate identifiers:

result = model.embed_dataset(
    ["MSTNPKPQRKTKRNT", "MKTIIALSYIFCLVFA"],
    batch_size=2,
    pooling=("mean", "std"),
)

for record in result:
    print(record.id, record.sequence, record.tensor.shape)

Set full_embeddings=True for one residue tensor with shape (l, d) per sequence. Set output to a directory for bounded-memory, transactional safetensors with ordered-prefix resume, or choose format="sqlite" for batch-level database commits and exact resume. Pooling excludes boundary, padding, and other non-biological positions.

For a long FASTA run, stream completed batches into SQLite:

persisted = model.embed_dataset(
    "proteins.fasta",
    batch_size=64,
    pooling=("mean",),
    output="protein-embeddings.sqlite",
    format="sqlite",
    resume=True,
)

Resume verifies the input order, model state, tokenizer policy, backend, dtype, and pooling configuration. It never appends incompatible records to an existing run.

Masked language modeling and contacts

Use the masked-language-model AutoClass when logits are required:

import torch
from transformers import AutoModelForMaskedLM, AutoTokenizer

model_id = "Synthyra/ESM2-150M"
tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
masked_model = AutoModelForMaskedLM.from_pretrained(
    model_id,
    trust_remote_code=True,
).eval()
batch = tokenizer("MSTNPKPQRKTKRNT", return_tensors="pt")

with torch.inference_mode():
    logits = masked_model(**batch).logits
    contacts = masked_model.predict_contacts(
        batch["input_ids"],
        batch["attention_mask"],
    )

print(logits.shape, contacts.shape)

Contact prediction materializes attention maps and should not be enabled in a high-throughput embedding path unless those maps are required.

Plain AutoModel omits the optional ESM pooler because this masked-language- model checkpoint contains no trained pooler weights. Pass add_pooling_layer=True only when intentionally initializing and training that head.

Runtime contract

• Public input: Amino-acid sequences tokenized to residue IDs
• Advertised AutoClasses: AutoConfig, AutoModel, AutoModelForMaskedLM, AutoModelForSequenceClassification, AutoModelForTokenClassification
• AutoClass weight status: AutoConfig = FastPLMs extension, AutoModel = pretrained, AutoModelForMaskedLM = pretrained, AutoModelForSequenceClassification = base weights + untrained task head, AutoModelForTokenClassification = base weights + untrained task head
• Attention implementations: eager, sdpa, flex_attention, flash_attention_2, flash_attention_3
• Precision policies: default
• BF16 execution: fp32_parameters_autocast
• Generation contract: not_applicable
• Optional dependency group: core
• Weight publication allowed: true
• Weight license status: resolved
• Redistributable: true
• Complete weight publication required: false

Release record

• FastPLMs weights: Synthyra/ESM2-150M
• Runtime revision: recorded separately in the built artifact and published commit
• Source-tree and runtime-bundle SHA-256: recorded in provenance.json
• Generator/schema version and complete/runtime-only attestations: recorded in provenance.json
• Official checkpoint: facebook/esm2_t30_150M_UR50D
• Artifact source: fast
• State transform: esm2_hf_to_fastplms_v1
• BF16 execution: fp32_parameters_autocast
• Pinned upstreams: fair-esm
• Reference container: reference-esm2
• Release tiers: check, compliance, feature, artifact, benchmark
• Unresolved required file identities: 0

The local artifact records exact file identities, conversion details, source revisions, and legal texts in provenance.json. A nonzero unresolved count is a release blocker.

Validation boundary

For tiers declared by the manifest, the release contract compares applicable semantic configuration, tokenizer behavior, state keys, shapes, dtypes, values, aliases, and representative inference with the pinned official implementation. This metadata does not by itself claim that a particular build passed, that one backend is faster, or that an output has biological or therapeutic validity.

License

Checkpoint terms: MIT. The Hub model-card identifier is mit. Applicable source licenses, notices, attribution, and conversion records are distributed with the local artifact. Review them before use.