Edit model card

SabbatH 2x7B

drawing

Model Description

SabbatH 2x7B is a Japanese language model that has been created by combining two models: Antler-RP-ja-westlake-chatvector and Hameln-japanese-mistral-7B, using a Mixture of Experts (MoE) approach. It also used chatntq-ja-7b-v1.0 as a base model.

Usage

Ensure you are using Transformers 4.34.0 or newer.

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Elizezen/SabbatH-2x7B")
model = AutoModelForCausalLM.from_pretrained(
  "Elizezen/SabbatH-2x7B",
  torch_dtype="auto",
)
model.eval()

if torch.cuda.is_available():
    model = model.to("cuda")

input_ids = tokenizer.encode(
    "ๅพ่ผฉใฏ็Œซใงใ‚ใ‚‹ใ€‚ๅๅ‰ใฏใพใ ใชใ„ใ€‚ใใ‚“ใชๅพ่ผฉใฏไปŠใ€",
    add_special_tokens=True, 
    return_tensors="pt"
)

tokens = model.generate(
    input_ids.to(device=model.device),
    max_new_tokens=512,
    temperature=1,
    top_p=0.95,
    do_sample=True,
)

out = tokenizer.decode(tokens[0][input_ids.shape[1]:], skip_special_tokens=True).strip()
print(out)

"""
output example:
็ช“ใ‹ใ‚‰ๅค–ใ‚’่ฆ‹ใฆใ„ใŸใ€‚
ใ€ŒใŠใ‚„ใ€้ณฅใŒ้ฃ›ใ‚“ใงใ„ใ‚‹ใ€
็›ฎใซใคใใฎใฏ้’็ฉบใจใ€ๅคงใใ‚ใฎ็™ฝใ„็พฝใฐใŸใ้Ÿณใ ใฃใŸใ€‚ใใ‚Œไปฅๅค–ใซไฝ•ใ‚‚ใชใ„ใ€‚ๅพ่ผฉใฏ็ช“ใ‹ใ‚‰้™ใ‚Šใฆ้ƒจๅฑ‹ใ‚’่ฆ‹ๅ›žใ—ใŸใŒใ€็‰นๅˆฅๅค‰ใ‚ใฃใŸๆง˜ๅญใ‚‚ใชใ„ใ‚ˆใ†ใงใ‚ใ‚‹ใ€‚
ใ€Œใ•ใฆใจโ€ฆใ€
ๆš‡ใ ใ€‚ใ“ใฎใพใพใšใฃใจๅค–ใซๅฑ…็ถšใ‘ใ‚Œใฐ้คŒใใ‚‰ใ„ใฏ่ฒฐใˆใ‚‹ใ‹ใ‚‚็Ÿฅใ‚Œใ‚“ใŒใ€ใใ‚ŒไปฅไธŠใซ็‰นๅˆฅใ™ใใซๅฟ…่ฆใช็‰ฉใฏ็„กใ„ใ€‚ๅฝผๅฅณใŸใกใฏๅฎถไบ‹ใ‚’ใ—ใฆใ„ใ‚‹ๆง˜ๅญใ ใŒใ€้ฃŸๆ–™ๅบซใฎไธญ่บซใพใงๆŠŠๆกใ—ใใ‚Œใฆใ„ใชใ„็‚บใ€่‡ชๅˆ†ใ‹ใ‚‰้คŒใญใ ใ‚Šใซ่กŒใใ“ใจใ‚‚ๅ‡บๆฅใชใ„ใ—ใ€ใใ†ใ™ใ‚‹ๅฟ…่ฆๆ€งใ‚‚ๆ„Ÿใ˜ใชใ‹ใฃใŸใ€‚
ใ€Œโ€ฆโ€ฆโ€ฆใ€
ๅพ่ผฉใฏ่€ƒใˆ่พผใ‚“ใงใ—ใพใฃใŸใ€‚ไฝ•ใ‚’ใ™ใ‚Œใฐ่‰ฏใ„ใฎใ ใ‚ใ†ใ‹๏ผŸ็œ ใ‚Šใซใคใใซใฏๆ—ฉ้ŽใŽใ‚‹ๆ™‚้–“ใงใ‚ใ‚‹็‚บใ€ใใ‚Œใ‚‚ๅ‡บๆฅใชใ„ใ€‚ใ“ใฎๅฎถใง็”Ÿๆดปใ—ใฆ๏ผ“ๆ—ฅ็›ฎใซ็ชๅ…ฅใ—ใ‚ˆใ†ใจใ—ใฆใ„ใŸๅพ่ผฉใฏ้€”ๆ–นใซๆšฎใ‚Œใฆใ„ใŸใ€‚
ใตใจๆ€ใฃใŸไบ‹ใŒใ‚ใ‚‹ใ€‚ไบบ้–“็•Œใ‚’่ฆ‹ๅญฆใ™ใ‚‹ใฎใ‚‚ไธ€ใคใฎๆ‰‹ใงใฏใชใ„ใ ใ‚ใ†ใ‹๏ผŸใใ†่€ƒใˆใ€ๅพ่ผฉใฏ็ช“ใ‹ใ‚‰ๅค–ใซๅ‡บใฆใฟใŸใ€‚ๅœฐ้ขใซ้™ใ‚Š็ซ‹ใกใ€ๅ‘จใ‚Šใ‚’่ฆ‹ๆธกใ—ใฆใฟใ‚‹ใจโ€ฆ
ใ€ŒใŠใ‚„ใ€ใ“ใ‚Œใฏใ€
็›ฎใฎๅ‰ใซๅฐใ•ใชไบบใŒๅฑ…ใŸใ€‚่ƒŒไธญใŒๆ›ฒใŒใฃใฆใ„ใ‚‹ใ‚ˆใ†ใงใ‚ใ‚‹็‚บใ€ๅนดๅฏ„ใ‚Šใ‹ใ‚‚็Ÿฅใ‚Œใชใ„ใ€‚ใใ‚“ใช
"""

Intended Use

The primary purpose of this language model is to assist in generating novels. While it can handle various prompts, it may not excel in providing instruction-based responses. Note that the model's responses are not censored, and occasionally sensitive content may be generated.

Downloads last month
1
Safetensors
Model size
12.9B params
Tensor type
BF16
ยท