Views
No views yet
| Name | Quant method | Size |
|---|---|---|
| MetaMath-Mistral-2x7B.Q2_K.gguf | Q2_K | 2.53GB |
| MetaMath-Mistral-2x7B.IQ3_XS.gguf | IQ3_XS | 2.81GB |
| MetaMath-Mistral-2x7B.IQ3_S.gguf | IQ3_S | 2.96GB |
| MetaMath-Mistral-2x7B.Q3_K_S.gguf | Q3_K_S | 2.95GB |
| MetaMath-Mistral-2x7B.IQ3_M.gguf | IQ3_M | 3.06GB |
| MetaMath-Mistral-2x7B.Q3_K.gguf | Q3_K | 3.28GB |
| MetaMath-Mistral-2x7B.Q3_K_M.gguf | Q3_K_M | 3.28GB |
| MetaMath-Mistral-2x7B.Q3_K_L.gguf | Q3_K_L | 3.56GB |
| MetaMath-Mistral-2x7B.IQ4_XS.gguf | IQ4_XS | 3.67GB |
| MetaMath-Mistral-2x7B.Q4_0.gguf | Q4_0 | 3.83GB |
| MetaMath-Mistral-2x7B.IQ4_NL.gguf | IQ4_NL | 3.87GB |
| MetaMath-Mistral-2x7B.Q4_K_S.gguf | Q4_K_S | 3.86GB |
| MetaMath-Mistral-2x7B.Q4_K.gguf | Q4_K | 4.07GB |
| MetaMath-Mistral-2x7B.Q4_K_M.gguf | Q4_K_M | 4.07GB |
| MetaMath-Mistral-2x7B.Q4_1.gguf | Q4_1 | 4.24GB |
| MetaMath-Mistral-2x7B.Q5_0.gguf | Q5_0 | 4.65GB |
| MetaMath-Mistral-2x7B.Q5_K_S.gguf | Q5_K_S | 4.65GB |
| MetaMath-Mistral-2x7B.Q5_K.gguf | Q5_K | 4.78GB |
| MetaMath-Mistral-2x7B.Q5_K_M.gguf | Q5_K_M | 4.78GB |
| MetaMath-Mistral-2x7B.Q5_1.gguf | Q5_1 | 5.07GB |
| MetaMath-Mistral-2x7B.Q6_K.gguf | Q6_K | 5.53GB |
| MetaMath-Mistral-2x7B.Q8_0.gguf | Q8_0 | 7.17GB |
1
2slices:
3 - sources:
4 - model: meta-math/MetaMath-Mistral-7B
5 layer_range: [0, 32]
6 - model: meta-math/MetaMath-Mistral-7B
7 layer_range: [0, 32]
8merge_method: slerp
9base_model: meta-math/MetaMath-Mistral-7B
10parameters:
11 t:
12 - filter: self_attn
13 value: [0, 0.5, 0.3, 0.7, 1]
14 - filter: mlp
15 value: [1, 0.5, 0.7, 0.3, 0]
16 - value: 0.5
17dtype: bfloat16
181
2## install dependencies
3## !pip install -q -U git+https://github.com/huggingface/transformers.git
4## !pip install -q -U git+https://github.com/huggingface/accelerate.git
5## !pip install -q -U sentencepiece
6
7## load model
8import torch
9from transformers import AutoModelForCausalLM, AutoTokenizer, TextStreamer
10
11model_name = "harshitv804/MetaMath-Mistral-2x7B"
12
13# load the model and tokenizer
14model = AutoModelForCausalLM.from_pretrained(
15 model_name,
16 device_map="auto",
17)
18
19tokenizer = AutoTokenizer.from_pretrained(
20 model_name,
21 trust_remote_code=True
22)
23
24tokenizer.pad_token = tokenizer.eos_token
25
26## inference
27
28query = "Maximoff's monthly bill is $60 per month. His monthly bill increased by thirty percent when he started working at home. How much is his total monthly bill working from home?"
29prompt =f"""
30Below is an instruction that describes a task. Write a response that appropriately completes the request.\n
31### Instruction:\n
32{query}\n
33### Response: Let's think step by step.
34"""
35
36# tokenize the input string
37inputs = tokenizer(
38 prompt,
39 return_tensors="pt",
40 return_attention_mask=False
41)
42
43# generate text using the model
44streamer = TextStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
45outputs = model.generate(**inputs, max_length=2048, streamer=streamer)
46
47# decode and print the output
48text = tokenizer.batch_decode(outputs)[0]
491@article{yu2023metamath,
2 title={MetaMath: Bootstrap Your Own Mathematical Questions for Large Language Models},
3 author={Yu, Longhui and Jiang, Weisen and Shi, Han and Yu, Jincheng and Liu, Zhengying and Zhang, Yu and Kwok, James T and Li, Zhenguo and Weller, Adrian and Liu, Weiyang},
4 journal={arXiv preprint arXiv:2309.12284},
5 year={2023}
6}1@article{jiang2023mistral,
2 title={Mistral 7B},
3 author={Jiang, Albert Q and Sablayrolles, Alexandre and Mensch, Arthur and Bamford, Chris and Chaplot, Devendra Singh and Casas, Diego de las and Bressand, Florian and Lengyel, Gianna and Lample, Guillaume and Saulnier, Lucile and others},
4 journal={arXiv preprint arXiv:2310.06825},
5 year={2023}
6}