metadata

library_name: peft
base_model: meta-math/MetaMath-Mistral-7B
license: apache-2.0
pipeline_tag: text2text-generation
language:
  - en

Model Card for Model ID

Model Details

Model Description

Developed by: Timofej Kiselev (tfshaman)
Model type: Mistral finetuned for solving MWPs using symbolic expressions with SymPy
Language(s) (NLP): English, Python with SymPy
License: Apache-2.0
Finetuned from model [optional]: meta-math/MetaMath-Mistral-7B

Model Sources [optional]

Repository: [More Information Needed]
Paper [optional]: https://dspace.cvut.cz/bitstream/handle/10467/115466/F3-BP-2024-Kiselev-Timofej-Thesis_Timofej_Kiselev.pdf

Uses

Input format: f"Question {your_math_word_problem}\n\nAnswer: "

Direct Use

bnb_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_use_double_quant=True,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_compute_dtype=torch.bfloat16, 
)
config = PeftConfig.from_pretrained("tfshaman/SymPy-Mistral")
base_model = AutoModelForCausalLM.from_pretrained("meta-math/MetaMath-Mistral-7B", quantization_config=bnb_config)
tokenizer = AutoTokenizer.from_pretrained("tfshaman/SymPy-Mistral-tokenizer", use_fast=False, padding_side="left")
base_model.resize_token_embeddings(len(tokenizer))
tokenizer.pad_token = "<s>"
tokenizer.padding_side='left'
model = PeftModel.from_pretrained(base_model, "tfshaman/SymPy-Mistral", quantization_config=bnb_config)
model = model.to("cuda")

[More Information Needed]

Downstream Use [optional]

[More Information Needed]

Citation

@mastersthesis{timofej2024velke, title={Velk{'e} jazykov{'e} modely pro numerick{'e} dotazy}, author={Timofej, Kiselev}, type={{B.S.} thesis}, year={2024}, school={{\v{C}}esk{'e} vysok{'e} u{\v{c}}en{'\i} technick{'e} v Praze. Vypo{\v{c}}etn{'\i} a informa{\v{c}}n{'\i} centrum.} }

Framework versions

PEFT 0.7.1