SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2004.03497 · 2020

ProGen: Language Modeling for Protein Generation

arXiv · PDF · Open in the Atlas

Code that ran

12 of 14 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 3 honoured the contract we drafted, 1 violated it and 7 have no contract verdict; 1 failure of our instrument; 2 no recorded run), measured 2026-11-21; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2004.03497.json · badge: /api/ran/2004.03497.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
lucidrains/progen-jax community (classifier label)
classifier basis not recorded on the link
9 ran · 2 no recorded run
lucidrains/progen reimplementation (classifier label)
classifier basis: classifier_2025
3 ran · 0 no recorded run
FunctionStatusWhere it lives
apply_rotary_pos_emb Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
lucidrains/progen/progen_transformer/progen.py
code served (permissive licence) · get_code("7088a112a013d3a0")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned ArrayImpl, ndim 3, dtype family float, shape (2, 3, 8) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import jax.numpy as jnp
import torch

torch.manual_seed(0)

# Create small test tensors
# x: shape (batch=2, seq_len=3, dim=8)
# We need dim to be even for rotate_every_two to work
x = jnp.array(torch.randn(2, 3, 8).numpy(), dtype=jnp.float32)

# sincos: tuple of (sin, cos) tensors
# sin and cos should have shape (..., rot_dim) where rot_dim <= x.shape[-1]
# Let's use rot_dim=4 (half of 8)
rot_dim = 4
sin = jnp.array(torch.randn(2, 3, rot_dim).numpy(), dtype=jnp.float32)
cos = jnp.array(torch.randn(2, 3, rot_dim).numpy(), dtype=jnp.float32)
sincos = (sin, cos)

ARGS = {
    'x': x,
    'sincos': sincos,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
einops, jax, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, ndim 3, dtype family float, shape (2, 3, 8) — shape (2, 3, 8) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
collate_fn Ran
ran on a synthesized input
lucidrains/progen-jax/progen_transformer/data.py
code served (permissive licence) · get_code("8f3fc2fd28cc770c")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 2, dtype family int, shape (2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# The function expects:
# - batch: an object with a .numpy() method that returns a sequence of byte strings
# - pad_length: an integer
# - offset: an integer (default 0)

# Create a mock batch object that has a .numpy() method
class MockBatch:
    def __init__(self, data):
        self.data = data
    
    def numpy(self):
        # Return a list of byte strings
        return self.data

# Create sample byte data
# Each element should be convertible via np.frombuffer to uint8
byte_data_1 = b'\x01\x02\x03\x04\x05'
byte_data_2 = b'\x06\x07\x08\x09\x0a'

batch = MockBatch([byte_data_1, byte_data_2])

ARGS = {
    'batch': batch,
    'pad_length': 4,
    'offset': 0
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family int, shape (2, 4) — shape (2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
exists Ran
ran, violated the contract we drafted
lucidrains/progen-jax/progen_transformer/utils.py
pointer only (licence: MIT); the stored copy get_code serves, from ThomasMrY/VCT, is pointer only (licence: NONE) · get_code("aa5486a3650902d8")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from ThomasMrY/VCT, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

file_get_last_checkpoint Ran
ran on a synthesized input
lucidrains/progen-jax/progen_transformer/checkpoint.py
code served (permissive licence) · get_code("03563ad6fe313230")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import tempfile
from pathlib import Path
import pickle
from cloudpickle import pickle as cloudpickle

# Create a temporary directory to act as the checkpoint path
temp_dir = tempfile.TemporaryDirectory()
path = Path(temp_dir.name)

# Create some dummy checkpoint files
checkpoint_data_1 = {"model_state": "checkpoint_1", "epoch": 1}
checkpoint_data_2 = {"model_state": "checkpoint_2", "epoch": 2}
checkpoint_data_3 = {"model_state": "checkpoint_3", "epoch": 3}

# Write checkpoints with proper naming convention (ckpt_*)
with open(str(path / "ckpt_001"), "wb") as f:
    cloudpickle.dump(checkpoint_data_1, f)

with open(str(path / "ckpt_002"), "wb") as f:
    cloudpickle.dump(checkpoint_data_2, f)

with open(str(path / "ckpt_003"), "wb") as f:
    cloudpickle.dump(checkpoint_data_3, f)

ARGS = {
    "path": path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cloudpickle — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
fixed_pos_embedding Ran
ran, honoured the contract we drafted
lucidrains/progen/progen_transformer/progen.py
code served (permissive licence) · get_code("c2e3f2e143e65142")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned tuple, ndim 4, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import jax.numpy as np
from einops import rearrange, repeat

ARGS = {
    "seq": 8,
    "dim": 4,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
einops, jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 4, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
gcs_get_last_checkpoint Ran
ran on a synthesized input
lucidrains/progen-jax/progen_transformer/checkpoint.py
code served (permissive licence) · get_code("bad2c74c90d94219")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np
from unittest.mock import Mock, MagicMock
from cloudpickle import pickle
import tempfile
import os

torch.manual_seed(0)

# Create a mock blob object
mock_blob = Mock()
mock_blob.name = "checkpoint_001.pkl"

# Create a temporary pickle file to simulate the checkpoint
temp_dir = tempfile.gettempdir()
temp_checkpoint_path = os.path.join(temp_dir, "checkpoint_001.pkl")
checkpoint_data = {"model_state": torch.randn(2, 4), "epoch": 1, "loss": 0.5}
with open(temp_checkpoint_path, 'wb') as f:
    pickle.dump(checkpoint_data, f)

# Read the file back to set up mock download
with open(temp_checkpoint_path, 'rb') as f:
    checkpoint_bytes = f.read()

# Define a download_to_file function that writes the checkpoint
def mock_download_to_file(file_obj, timeout=None):
    file_obj.write(checkpoint_bytes)

mock_blob.download_to_file = mock_download_to_file

# Create a mock bucket
mock_bucket = Mock()
mock_bucket.list_blobs = Mock(return_value=[mock_blob])

ARGS = {
    "bucket": mock_bucket
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cloudpickle, numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_annotations_from_description Ran
ran on a synthesized input
lucidrains/progen-jax/generate_data.py
code served (permissive licence) · get_code("17886a9cb6efdb96")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import re

def get_annotations_from_description(config, description):
    taxonomy_matches = re.findall('Tax=([a-zA-Z\\s]*)\\s[a-zA-Z\\=]', description)
    annotations = dict()
    if len(taxonomy_matches) > 0:
        annotations['tax'] = taxonomy_matches[0]
    return annotations

config = {
    'version': 1,
    'name': 'test_config'
}

description = 'Tax=Bacteria Archaea D=test_domain'

ARGS = {
    'config': config,
    'description': description
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
log Ran
ran on a synthesized input
lucidrains/progen-jax/progen_transformer/utils.py
code served (permissive licence) · get_code("0bd02f1529386045")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ArrayImpl, ndim 1, dtype family float, shape (4,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import jax.numpy as np

ARGS = {
    "t": np.array([0.5, 1.0, 2.0, 0.1]),
    "eps": 1e-20
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, ndim 1, dtype family float, shape (4,) — shape (4,) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
noop Ran
ran, honoured the contract we drafted
lucidrains/progen-jax/progen_transformer/utils.py
pointer only (licence: MIT); the stored copy get_code serves, from Abhimanyu08/Fixup_Initialization, is pointer only (licence: NONE) · get_code("5e2ed23acb71ee11")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from Abhimanyu08/Fixup_Initialization, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

order_dict_by Ran
ran on a synthesized input
lucidrains/progen-jax/generate_data.py
code served (permissive licence) · get_code("c601205d9879134a")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# The function order_dict_by takes:
# - d: a dictionary
# - fn: a function that takes keys and returns an ordered sequence of keys

# Create a simple dictionary with string keys
test_dict = {
    'zebra': 1,
    'apple': 2,
    'mango': 3,
    'banana': 4
}

# Create a function that sorts the keys
def sort_keys_fn(keys):
    return sorted(keys)

ARGS = {
    'd': test_dict,
    'fn': sort_keys_fn
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
parse_fn Ran
ran on a synthesized input
lucidrains/progen-jax/progen_transformer/data.py
code served (permissive licence) · get_code("48bc0d4ccdd99810")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import tensorflow as tf

# Create a valid TFRecord example with a 'seq' field containing string data
feature = {
    'seq': tf.train.Feature(bytes_list=tf.train.BytesList(value=[b'example_sequence']))
}
example = tf.train.Example(features=tf.train.Features(feature=feature))
serialized_sample = example.SerializeToString()

ARGS = {
    'sample': serialized_sample
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
rotate_every_two Ran
ran, honoured the contract we drafted
lucidrains/progen/progen_transformer/progen.py
code served (permissive licence) · get_code("59c0bfea601f828d")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned ArrayImpl, ndim 2, dtype family float, shape (2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import jax.numpy as jnp
import torch

torch.manual_seed(0)

# The function rotate_every_two takes a single parameter 'x'
# It expects x to be an array-like that can be rearranged
# The function rearranges it to split the last dimension into pairs,
# then rotates those pairs, then rearranges back
# So x should have a last dimension that is even

# Create a small test tensor with shape (2, 4)
# Last dimension is 4, which is even and can be split into 2 pairs
x = jnp.array([
    [1.0, 2.0, 3.0, 4.0],
    [5.0, 6.0, 7.0, 8.0]
])

ARGS = {
    'x': x
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
einops, jax, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
fasta_row_to_sequence_strings Not yet run
unverified (harvested, no recorded run)
lucidrains/progen-jax/generate_data.py
code served (permissive licence) · get_code("82e10e377256f226")
How it ran

no invocation recorded.

iterator_from_tfrecords_folder Not yet run
unverified (harvested, no recorded run)
lucidrains/progen-jax/progen_transformer/data.py
code served (permissive licence) · get_code("db2d8c3632b26569")
How it ran

no invocation recorded.

Also present in other papers' code

Of 14 samples lifted from this paper's linked repositories, 14 carry a function identity and 0 do not. Of the 14 distinct functions with an identity, 3 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 0 the same month, 3 later) and 11 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
exists 486 other papers
17 dated earlier, 467 later, 2 undated · 78 only as an assembly brick
1603.08155 (2016-03)
Perceptual Losses for Real-Time Style Transfer and Super-Resolution
1611.07004 (2016-11)
Image-to-Image Translation with Conditional Adversarial Networks
1804.04637 (2018-04)
EMBER: An Open Dataset for Training Static PE Malware Machine Learning Models
and 483 more not named here
noop 9 other papers
4 dated earlier, 5 later
1805.08318 (2018-05)
Self-Attention Generative Adversarial Networks
1901.09321 (2019-01)
Fixup Initialization: Residual Learning Without Normalization
1906.03028 (2019-06)
Automatic Reparameterisation of Probabilistic Programs
and 6 more not named here
rotate_every_two 1 other paper
1 later
2204.02311 (2022-04)
PaLM: Scaling Language Modeling with Pathways

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-21.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Generative modeling for protein engineering is key to solving fundamental problems in synthetic biology, medicine, and material science. We pose protein engineering as an unsupervised sequence generation problem in order to leverage the exponentially growing set of proteins that lack costly, structural annotations. We train a 1.2B-parameter language model, ProGen, on ~280M protein sequences conditioned on taxonomic and keyword tags such as molecular function and cellular component. This provides ProGen with an unprecedented range of evolutionary sequence diversity and allows it to generate with fine-grained control as demonstrated by metrics based on primary sequence similarity, secondary structure accuracy, and conformational energy.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2004.03497")
get_code_for_paper("2004.03497")
have("2004.03497")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2004.03497.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2004.03497.svg)](https://syntology.ai/paper/2004.03497)

Connect an agent — have() is free.