Quantization made by Richard Erkhov.

ChemDFM-13B-v1.0 - GGUF

Model creator: https://huggingface.co/OpenDFM/
Original model: https://huggingface.co/OpenDFM/ChemDFM-13B-v1.0/

Name	Quant method	Size
ChemDFM-13B-v1.0.Q2_K.gguf	Q2_K	4.56GB
ChemDFM-13B-v1.0.IQ3_XS.gguf	IQ3_XS	5.04GB
ChemDFM-13B-v1.0.IQ3_S.gguf	IQ3_S	5.31GB
ChemDFM-13B-v1.0.Q3_K_S.gguf	Q3_K_S	5.31GB
ChemDFM-13B-v1.0.IQ3_M.gguf	IQ3_M	5.62GB
ChemDFM-13B-v1.0.Q3_K.gguf	Q3_K	5.95GB
ChemDFM-13B-v1.0.Q3_K_M.gguf	Q3_K_M	5.95GB
ChemDFM-13B-v1.0.Q3_K_L.gguf	Q3_K_L	6.5GB
ChemDFM-13B-v1.0.IQ4_XS.gguf	IQ4_XS	6.59GB
ChemDFM-13B-v1.0.Q4_0.gguf	Q4_0	6.91GB
ChemDFM-13B-v1.0.IQ4_NL.gguf	IQ4_NL	6.95GB
ChemDFM-13B-v1.0.Q4_K_S.gguf	Q4_K_S	6.96GB
ChemDFM-13B-v1.0.Q4_K.gguf	Q4_K	7.37GB
ChemDFM-13B-v1.0.Q4_K_M.gguf	Q4_K_M	7.37GB
ChemDFM-13B-v1.0.Q4_1.gguf	Q4_1	7.66GB
ChemDFM-13B-v1.0.Q5_0.gguf	Q5_0	8.41GB
ChemDFM-13B-v1.0.Q5_K_S.gguf	Q5_K_S	8.41GB
ChemDFM-13B-v1.0.Q5_K.gguf	Q5_K	8.65GB
ChemDFM-13B-v1.0.Q5_K_M.gguf	Q5_K_M	8.65GB
ChemDFM-13B-v1.0.Q5_1.gguf	Q5_1	9.16GB
ChemDFM-13B-v1.0.Q6_K.gguf	Q6_K	10.0GB
ChemDFM-13B-v1.0.Q8_0.gguf	Q8_0	12.96GB

Original model description:

license: agpl-3.0 pipeline_tag: text-generation tags: - chemistry language: - en - zh

ChemDFM-13B: An LLM-Based Dialogue Foundation Model for Chemistry

ChemDFM is the pioneering open-sourced dialogue foundation model for Chemistry and molecule science, which is build based on LLaMa-13B. ChemDFM outperforms the open-sourced LLMs in all the typical tasks of chemistry, and even reach comparable or higher performances of GPT-4. For more details, please refer to our paper.

News

2024-06-13: The results on the comprehensive science benchmark SciKnowEval show that "ChemDFM emerged as one of the top open-source models by continuing pre-training and fine-tuning on a vast corpus of scientific literature".
2024-04-17: The evaluation data (including instructions) we used in our paper is released on GitHub
2024-03-12: The parameter of ChemDFM-13B is open-sourced!
2024-01-26: The paper of ChemDFM-13B is released on arXiv: ChemDFM: Dialogue Foundation Model for Chemistry

Usage Details

The online demo of ChemDFM will be up soon!

local inference

To load and run ChemDFM locally, here is an example:

import torch
from transformers import LlamaTokenizer, LlamaForCausalLM, GenerationConfig

model_name_or_id = "X-LANCE/ChemDFM-13B-v1.0"
tokenizer = LlamaTokenizer.from_pretrained(model_name_or_id)
model = LlamaForCausalLM.from_pretrained(model_name_or_id, torch_dtype=torch.float16, device_map="auto")

input_text = "Can you please give detailed descriptions of the molecule below?\nCl.O=C1c2c(O)cccc2-c2nn(CCNCCO)c3ccc(NCCNCCO)c1c23"
input_text = f"[Round 0]\nHuman: {input_text}\nAssistant:"

inputs = tokenizer(input_text, return_tensors="pt").to("cuda")
generation_config = GenerationConfig(
    do_sample=True,
    top_k=20,
    top_p=0.9,
    temperature=0.9,
    max_new_tokens=1024,
    repetition_penalty=1.05,
    eos_token_id=tokenizer.eos_token_id
)

outputs = model.generate(**inputs, generation_config=generation_config)
generated_text = tokenizer.batch_decode(outputs, skip_special_tokens=True)[0][len(input_text):]
print(generated_text.strip())

input format

To get better responses, we recommend to preprocess your input and history with the dialogue templates which are used during instruction tuning of ChemDFM. Specifically, for an input queries

{'current_query': current_query, 'history': [(query1, answer1), (query2, answer2), ...]}

, you can use the following code to preprocess the input and history:

def formatting_input(current_query, history):
    input_text = ''
    for idx, (query, answer) in history:
        input_text += f"[Round {idx}]\nHuman: {query}\nAssistant: {answer}\n"
    input_text += f"[Round {len(history)}]\nHuman: {current_query}\nAssistant:"
    return input_text

SMILES preprocess

When there involves SMILES notation in your input, we recommend to preprocess the SMILES with the rdkit package to canonicalize the SMILES. Here is an example:

from rdkit import Chem
def canonicalize_smiles(smiles):
    mol = Chem.MolFromSmiles(smiles)
    if mol is None:
        return None
    return Chem.MolToSmiles(mol, isomericSmiles=True, kekuleSmiles=False)

or directly:

from rdkit import Chem
def canonicalize_smiles(smiles):
    return Chem.CanonSmiles(smiles, useChiral=True)

Performance

Chemical Benchmarks

We evaluate the performance of ChemDFM-13B on multiple widely-used benchmarks in chemistry. The detail introduction of the benchmarks can be found in our paper. The overall performance of ChemDFM-13B is shown below:

Human Evaluation

We mark the correct and relevant information in the replies in green, the correct but irrelevant information in yellow, and the wrong information in red. In addition, the key points of the answer are marked in bold if they appear in the reply.

The results show that while open-sourced LLMs perform well when asked about existing knowledge (Q1), only ChemDFM can provide correct and comprehensive answers when questions involve new molecules and reactions (Q2 [Yin et al., 2023] & Q3 [Dargo et al., 2023])

The above conversation is also inspired by Yin et al.[2023]. During the dialogue, the researcher wants to selectively oxidize one of the two carbonyl groups of a molecule. However, the initial solution given by ChemDFM results in both carbonyl groups being oxidized. Through the correction given by the researcher, ChemDFM adjusts its proposal and provides two possible solutions. Finally, the researcher chooses to use protecting groups and ChemDFM further details its advice. In the dialogue, ChemDFM shows promising capabilities regarding error correction (Round 2) and detailing (Round 3) when handling real-world research scenarios.

For more examples and analysis, please refer to our paper.

Citation

@misc{zhao2024chemdfm,
    title={ChemDFM: Dialogue Foundation Model for Chemistry},
    author={Zihan Zhao and Da Ma and Lu Chen and Liangtai Sun and Zihao Li and Hongshen Xu and Zichen Zhu and Su Zhu and Shuai Fan and Guodong Shen and Xin Chen and Kai Yu},
    year={2024},
    eprint={2401.14818},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}

Disclaimer

Current version of ChemDFM may generate incorrect or misleading information. Please use it with caution and verify the results with domain experts before making any decisions based on the results.

Contact

If you have any questions or further requests, please contact Zihan Zhao and Lu Chen.