SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2310.13012 · EMNLP · 2023

H2O Open Ecosystem for State-of-the-art Large Language Models

Marcos Conde, Arno Candel, Jon Mckinney, Philipp Singer, Pascal Pfeiffer, Maximilian Jeblick, Chun Lee

arXiv · PDF · Open in the Atlas

Code that ran

7 of 12 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; 2 honoured the contract we drafted, 4 violated it and 0 have no contract verdict; 1 failure of our instrument; 5 no recorded run), measured 2026-11-22; 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/2310.13012.json · badge: /api/ran/2310.13012.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
h2oai/h2ogpt canonical (classifier label)
classifier basis: census
7 ran · 5 no recorded run
FunctionStatusWhere it lives
extra_stop_token_ids Ran
ran, honoured the contract we drafted
h2oai/h2ogpt/src/enums.py
code served (permissive licence) · get_code("bd66a06aa7050024")
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 list, ndim 1, dtype family int; 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 torch
import torch.nn as nn

class MockTokenizer:
    """Minimal tokenizer that implements encode method"""
    def encode(self, words, add_special_tokens=False):
        # Simple mock: return list of integers based on word count
        if isinstance(words, list):
            return list(range(len(words)))
        return [0]

# Test case 1: base_model is None
ARGS_1 = {
    'base_model': None,
    'tokenizer': None,
    'as_ids': False
}

# Test case 2: base_model is one of the InternVL models, as_ids=False
ARGS_2 = {
    'base_model': 'OpenGVLab/InternVL2-2B',
    'tokenizer': MockTokenizer(),
    'as_ids': False
}

# Test case 3: base_model is one of the InternVL models, as_ids=True
ARGS_3 = {
    'base_model': 'OpenGVLab/InternVL2-8B',
    'tokenizer': MockTokenizer(),
    'as_ids': True
}

# Test case 4: base_model is unknown model
ARGS_4 = {
    'base_model': 'some/unknown-model',
    'tokenizer': MockTokenizer(),
    'as_ids': False
}

# Use the most comprehensive test case
ARGS = ARGS_3
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 list, ndim 1, dtype family int — a list 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
get_dbid Ran
ran, violated the contract we drafted
h2oai/h2ogpt/src/db_utils.py
code served (permissive licence) · get_code("daf097bc0bd836d2")
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 int, ndim 0, dtype family int, shape () 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

torch.manual_seed(0)

# The function get_dbid accesses db1[1], so db1 needs to be indexable
# It returns the element at index 1
db1 = [10, 42, 100]

ARGS = {
    "db1": db1
}
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 int, ndim 0, dtype family int, shape () — shape () 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
get_mem_space Ran
ran, honoured the contract we drafted
h2oai/h2ogpt/models/gpu_mem_track.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from applied-machine-learning-lab/dt4ier, is pointer only (licence: NONE) · get_code("efa7095cdb85be6c")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (Apache-2.0), but get_code serves the stored copy, first harvested from applied-machine-learning-lab/dt4ier, 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.

is_gradio_vision_model Ran
ran, violated the contract we drafted
h2oai/h2ogpt/src/enums.py
code served (permissive licence) · get_code("cc1bf5a3113b7711")
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 bool, ndim 0, dtype family float, shape () 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 torch.nn as nn

# The function is_gradio_vision_model takes a base_model parameter
# and checks if it's a string that starts with certain prefixes.
# It calls .startswith() on base_model, so we need a string.

ARGS = {
    "base_model": "llava-7b"
}
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 bool, ndim 0, dtype family float, shape () — shape () 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
is_vision_model Ran
ran, violated the contract we drafted
h2oai/h2ogpt/src/enums.py
code served (permissive licence) · get_code("dae40f9ffdf8c10f")
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 bool, ndim 0, dtype family float, shape () 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

# Analyze the function: is_vision_model takes 3 parameters:
# 1. base_model: a string that is checked with .startswith() and 'in' operators
# 2. all_visible_models: a list that is checked with 'in' operator
# 3. visible_vision_models: a list that is checked with indexing [0] and 'in' operator

# The function logic:
# - base_model must be a string (uses .startswith() and 'in')
# - all_visible_models and visible_vision_models are lists
# - The condition checks if visible_vision_models[0] is in all_visible_models
# - Then checks various string prefixes and exact matches on base_model

# Set up test case 1: base_model is a vision model string
base_model_vision = "gpt-4o"
all_visible_models_1 = []
visible_vision_models_1 = []

# Set up test case 2: using the visible models lists
base_model_other = "some-model"
all_visible_models_2 = ["model-1", "model-2"]
visible_vision_models_2 = ["model-1"]

ARGS = {
    "base_model": "gpt-4o",
    "all_visible_models": [],
    "visible_vision_models": []
}
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 bool, ndim 0, dtype family float, shape () — shape () 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
prune_long_sequences Ran
ran, violated the contract we drafted
h2oai/h2ogpt/finetune.py
code served (permissive licence) · get_code("cf2ae51214a5eec8")
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 bool, ndim 0, dtype family float, shape () 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

torch.manual_seed(0)

# Create a data_point dict with 'input_ids' key
# The function checks if len(data_point['input_ids']) < cutoff_len
data_point = {
    'input_ids': [1, 2, 3, 4, 5]  # A list of 5 token IDs
}

cutoff_len = 10  # Cutoff length longer than the input_ids, so pruning returns True

ARGS = {
    'data_point': data_point,
    'cutoff_len': cutoff_len
}
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 bool, ndim 0, dtype family float, shape () — shape () 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
tokenize Ran
ran; our contract draft was wrong, not the code
h2oai/h2ogpt/finetune.py
code served (permissive licence) · get_code("38cae2d1ced7906a")
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
from transformers import AutoTokenizer

# Create a minimal stand-in tokenizer that mimics HuggingFace tokenizer behavior
class StandInTokenizer:
    def __init__(self):
        self.eos_token_id = 2
    
    def __call__(self, prompt, truncation=True, max_length=None, padding=False, return_tensors=None):
        # Simulate tokenization: convert prompt to list of token ids
        # Return a dict with 'input_ids' and 'attention_mask' as lists
        token_ids = [101, 2054, 2003, 1045]  # Example token sequence
        attention_mask = [1, 1, 1, 1]
        
        # Apply truncation
        if truncation and max_length:
            token_ids = token_ids[:max_length]
            attention_mask = attention_mask[:max_length]
        
        return {
            'input_ids': token_ids,
            'attention_mask': attention_mask
        }

# Define the arguments dict
ARGS = {
    'prompt': "Hello, how are you?",
    'tokenizer': StandInTokenizer(),
    'cutoff_len': 10,
    'add_eos_token': True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, transformers — 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
covert_prompt_to_input_ids_with_history Not yet run
unverified (harvested, no recorded run)
h2oai/h2ogpt/models/predict_aquila.py
code served (permissive licence) · get_code("ad6c3608d441e75a")
How it ran

no invocation recorded.

get_conv_template Not yet run
unverified (harvested, no recorded run)
h2oai/h2ogpt/models/predict_aquila.py
code served (permissive licence) · get_code("f393883643980cc1")
How it ran

no invocation recorded.

get_conversation_template Not yet run
unverified (harvested, no recorded run)
h2oai/h2ogpt/models/predict_aquila.py
code served (permissive licence) · get_code("96e06a37fa7389cf")
How it ran

no invocation recorded.

get_sentences Not yet run
unverified (harvested, no recorded run)
h2oai/h2ogpt/src/create_data.py
code served (permissive licence) · get_code("aca03ba436af1dd6")
How it ran

no invocation recorded.

parse_rst_file Not yet run
unverified (harvested, no recorded run)
h2oai/h2ogpt/src/create_data.py
code served (permissive licence) · get_code("4a3ce15d96849109")
How it ran

no invocation recorded.

Also present in other papers' code

Of 12 samples lifted from this paper's linked repositories, 12 carry a function identity and 0 do not. Of the 12 distinct functions with an identity, 3 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 2 the same month, 3 later) and 9 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
get_conv_template 69 other papers
7 dated earlier, 1 the same month, 60 later, 1 undated
1911.03705 (2019-11)
CommonGen: A Constrained Text Generation Challenge for Generative Commonsense Reasoning
2112.05682 (2021-12)
Self-attention Does Not Need $O(n^2)$ Memory
2204.07705 (2022-04)
SUPER-NATURALINSTRUCTIONS: Generalization via Declarative Instructions on 1600+ NLP Tasks
and 66 more not named here
get_mem_space 7 other papers
1 dated earlier, 1 the same month, 5 later
2306.06446 (2023-06)
ShiftAddViT: Mixture of Multiplication Primitives Towards Efficient Vision Transformers
2310.20199 (2023-10)
In Search of Lost Online Test-time Adaptation: A Survey
2311.08756 (2023-11)
Accelerating Toeplitz Neural Network with Constant-time Inference Complexity
and 4 more not named here
get_conversation_template 1 other paper
1 later
2403.00862 (2024-03)
NewsBench: A Systematic Evaluation Framework for Assessing Editorial Capabilities of Large Language Models in Chinese Journalism

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 2026-11-22.

Repositories linked to this paper

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

Abstract

Large Language Models (LLMs) represent a revolution in AI. However, they also pose many significant risks, such as the presence of biased, private, copyrighted or harmful text. For this reason we need open, transparent and safe solutions. We introduce a complete open-source ecosystem for developing and testing LLMs. The goal of this project is to boost open alternatives to closed-source approaches. We release h2oGPT, a family of fine-tuned LLMs from 7 to 70 Billion parameters. We also introduce H2O LLM Studio, a framework and no-code GUI designed for efficient fine-tuning, evaluation, and deployment of LLMs using the most recent state-of-the-art techniques. Our code and models are licensed under fully permissive Apache 2.0 licenses. We believe open-source language models help to boost AI development and make it more accessible and trustworthy.

For agents

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

get_harvested_code_for_paper("2310.13012")
get_code_for_paper("2310.13012")
have("2310.13012")

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

curl https://syntology.ai/api/ran/2310.13012.json

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

[![Syntology run record](https://syntology.ai/api/ran/2310.13012.svg)](https://syntology.ai/paper/2310.13012)

Connect an agent — have() is free.