Downloads · 30 days
5
17% of all-time downloads
MN-12B-Lyra-v3-FP8
mikudev/MN-12B-Lyra-v3-FP8
MN-12B-Lyra-v3-FP8 is a machine learning model from mikudev. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
Original Model: https://huggingface.co/Sao10K/MN-12B-Lyra-v3
All-time downloads
29
Public
Parameters
12.2B
13.6 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors13.6 GB · 100%
Parameter types
How the weights are stored.
F8_E4M310.9B · 89%
Base models
Model card
From the Hugging Face model README
Original Model: https://huggingface.co/Sao10K/MN-12B-Lyra-v3
Quantized with FP8 using https://github.com/neuralmagic/AutoFP8
Script:
from datasets import load_dataset
from transformers import AutoTokenizer
from auto_fp8 import AutoFP8ForCausalLM, BaseQuantizeConfig
pretrained_model_dir = "Sao10K/MN-12B-Lyra-v3"
quantized_model_dir = "MN-12B-Lyra-v3-FP8"
tokenizer = AutoTokenizer.from_pretrained(pretrained_model_dir, use_fast=True, model_max_length=4096)
tokenizer.pad_token = tokenizer.eos_token
ds = load_dataset("mgoin/ultrachat_2k", split="train_sft").select(range(512))
examples = [tokenizer.apply_chat_template(batch["messages"], tokenize=False) for batch in ds]
examples = tokenizer(examples, padding=True, truncation=True, return_tensors="pt").to("cuda")
quantize_config = BaseQuantizeConfig(
quant_method="fp8",
activation_scheme="static",
ignore_patterns=["re:.*lm_head"],
)
model = AutoFP8ForCausalLM.from_pretrained(
pretrained_model_dir, quantize_config=quantize_config
)
model.quantize(examples)
model.save_quantized(quantized_model_dir)
- Type
- mistral
- Languages
- en
- Created
- Sep 4, 2024
- Updated
- Sep 4, 2024