You can deploy this model and start earning money today!

RichardErkhov/
mrm8488_-_limstral-7B-v0.1-gguf

Name: mrm8488_-_limstral-7B-v0.1-gguf
Availability: InStock

endpoints_compatible

gguf

template

Views

No views yet

Quantization made by Richard Erkhov.

Github

Discord

Request more models

limstral-7B-v0.1 - GGUF

Model creator: https://huggingface.co/mrm8488/
Original model: https://huggingface.co/mrm8488/limstral-7B-v0.1/

Name	Quant method	Size
limstral-7B-v0.1.Q2_K.gguf	Q2_K	2.53GB
limstral-7B-v0.1.IQ3_XS.gguf	IQ3_XS	2.81GB
limstral-7B-v0.1.IQ3_S.gguf	IQ3_S	2.96GB
limstral-7B-v0.1.Q3_K_S.gguf	Q3_K_S	2.95GB
limstral-7B-v0.1.IQ3_M.gguf	IQ3_M	3.06GB
limstral-7B-v0.1.Q3_K.gguf	Q3_K	3.28GB
limstral-7B-v0.1.Q3_K_M.gguf	Q3_K_M	3.28GB
limstral-7B-v0.1.Q3_K_L.gguf	Q3_K_L	3.56GB
limstral-7B-v0.1.IQ4_XS.gguf	IQ4_XS	3.67GB
limstral-7B-v0.1.Q4_0.gguf	Q4_0	3.83GB
limstral-7B-v0.1.IQ4_NL.gguf	IQ4_NL	3.87GB
limstral-7B-v0.1.Q4_K_S.gguf	Q4_K_S	3.86GB
limstral-7B-v0.1.Q4_K.gguf	Q4_K	4.07GB
limstral-7B-v0.1.Q4_K_M.gguf	Q4_K_M	4.07GB
limstral-7B-v0.1.Q4_1.gguf	Q4_1	4.24GB
limstral-7B-v0.1.Q5_0.gguf	Q5_0	4.65GB
limstral-7B-v0.1.Q5_K_S.gguf	Q5_K_S	4.65GB
limstral-7B-v0.1.Q5_K.gguf	Q5_K	4.78GB
limstral-7B-v0.1.Q5_K_M.gguf	Q5_K_M	4.78GB
limstral-7B-v0.1.Q5_1.gguf	Q5_1	5.07GB
limstral-7B-v0.1.Q6_K.gguf	Q6_K	5.53GB
limstral-7B-v0.1.Q8_0.gguf	Q8_0	7.17GB

Original model description:

license: apache-2.0 datasets:

GAIR/lima language:
en pipeline_tag: text-generation thumbnail: https://huggingface.co/mrm8488/limstral-7B-v0.1/resolve/main/limstral_logo.png

LIMSTRAL 🇲🍋

limstral logo

Mistral 7B fine-tuned on LIMA

This model is a fine-tuned version of mistralai/Mistral-7B-v0.1 on the LIMA dataset for instruction following downstream task.

Training procedure

The model was loaded on 8 bits and fine-tuned on the LIMA dataset using the LoRA PEFT technique with the huggingface/peft library and trl/sft for 2 epochs on 1 x A100 (40GB) GPU.

SFT Trainer params:

trainer = SFTTrainer(
    model=model,
    train_dataset=train_ds,
    eval_dataset=test_ds,
    peft_config=peft_config,
    dataset_text_field="text",
    max_seq_length=2048,
    tokenizer=tokenizer,
    args=training_arguments,
    packing=False
)

LoRA config:

config = LoraConfig(
        lora_alpha=16,
        lora_dropout=0.1,
        r=64,
        bias="none",
        task_type="CAUSAL_LM",
        target_modules = ['q_proj', 'k_proj', 'down_proj', 'v_proj', 'o_proj', 'gate_proj', 'up_proj']
    )

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0002
train_batch_size: 2
eval_batch_size: 8
seed: 66
gradient_accumulation_steps: 64
total_train_batch_size: 128
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: constant
lr_scheduler_warmup_ratio: 0.03
num_epochs: 2
mixed_precision_training: Native AMP

Training results

Step	Training Loss	Validation Loss
5	1.802800	1.848371
10	1.605800	1.803416
15	1.844800	1.762276
20	1.752600	1.754042
25	1.512400	1.750550

Usage

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline

repo_id = "mrm8488/limstral-7B-v0.1"

model = AutoModelForCausalLM.from_pretrained(repo_id, torch_dtype=torch.bfloat16)
tokenizer = AutoTokenizer.from_pretrained(repo_id)

gen = pipeline("text-generation", model=model, tokenizer=tokenizer, device=0)

instruction = "[INST] Write an email to say goodbye to me boss [\INST]"
res = gen(instruction, max_new_tokens=512, temperature=0.3, top_p=0.75, top_k=40, repetition_penalty=1.2)
print(res[0]['generated_text'])

Framework versions

Transformers 4.35.0.dev0
Pytorch 2.1.0+cu118
Datasets 2.14.6
Tokenizers 0.14.1

RichardErkhov/mrm8488_-_limstral-7B-v0.1-gguf