Hugging Face's logo

mpasila
/
Finnish-Alpaca-Tiny-V2-7B-exl2-4bpw

Text Generation
Transformers
Finnish
llama
text-generation-inference
unsloth
trl
sft
Inference Endpoints
Model card Files Community
1
Finnish-Alpaca-Tiny-V2-7B-exl2-4bpw / README.md
mpasila's picture
mpasila
Update README.md
22f3f7d verified
preview code
|
raw
history blame contribute delete
7.05 kB
metadata
language:
  - fi
license: apache-2.0
tags:
  - text-generation-inference
  - transformers
  - unsloth
  - llama
  - trl
  - sft
base_model: LumiOpen/Viking-7B
datasets:
  - mpasila/Finnish-Alpaca-Tiny

This is an ExLlamaV2 quantized model in 4bpw of mpasila/Finnish-Alpaca-Tiny-V2-7B using the default calibration dataset.

Original Model card:

This is a merge of mpasila/Finnish-Alpaca-Tiny-V2-LoRA-7B.

LoRA trained in 4-bit with 2k context using LumiOpen/Viking-7B as the base model for 1 epoch.

Dataset used is mpasila/Finnish-Alpaca-Tiny.

It works relatively well for question and answering. I will make a bigger dataset for the next fine-tune.

Prompt format: Alpaca

It uses Alpaca format but with a translated instruction at the start:

{
    "instruction,output": "Alla on ohje, jossa kuvataan tehtävä. Kirjoita vastaus, joka täyttää pyynnön asianmukaisesti.\n\n### Instruction:\n%instruction%\n\n### Response:\n%output%",
    "instruction,input,output": "Alla on ohje, jossa kuvataan tehtävä ja joka on yhdistetty kontekstia lisäävään syötteeseen. Kirjoita vastaus, joka täyttää pyynnön asianmukaisesti.\n\n### Instruction:\n%instruction%\n\n### Input:\n%input%\n\n### Response:\n%output%"
}

Evaluation

Model Size Type FIN-bench (score)
mpasila/Finnish-Alpaca-Tiny-V2-7B 7B Instruct 0.4654
mpasila/Finnish-Alpaca-Small-7B 7B Instruct 0.3586
mpasila/Alpacazord-Viking-7B 7B Instruct 0.4123
mpasila/NordicAlpaca-Finnish-V1-7B 7B Instruct 0.3891
mpasila/Finnish-Viking-Alpaca-V1-7B 7B Instruct 0.3943
Finnish-NLP/llama-7b-finnish-instruct-v0.1 7B Instruct 0.4365
Finnish-NLP/llama-7b-finnish-instruct-v0.2 7B Instruct 0.3993
Finnish-NLP/llama-7b-finnish 7B Base 0.2350
LumiOpen/Viking-7B (1000B) 7B Base 0.3721
HPLT/gpt-7b-nordic-prerelease 7B Base 0.3169

Source

FIN-bench scores:

Task Version Metric Value Stderr
bigbench_analogies 0 multiple_choice_grade 0.6385 ± 0.0423
bigbench_arithmetic_1_digit_addition 0 multiple_choice_grade 0.7200 ± 0.0451
bigbench_arithmetic_1_digit_division 0 multiple_choice_grade 0.7391 ± 0.0936
bigbench_arithmetic_1_digit_multiplication 0 multiple_choice_grade 0.4800 ± 0.0502
bigbench_arithmetic_1_digit_subtraction 0 multiple_choice_grade 0.6300 ± 0.0485
bigbench_arithmetic_2_digit_addition 0 multiple_choice_grade 0.4000 ± 0.0492
bigbench_arithmetic_2_digit_division 0 multiple_choice_grade 0.5000 ± 0.0503
bigbench_arithmetic_2_digit_multiplication 0 multiple_choice_grade 0.2800 ± 0.0451
bigbench_arithmetic_2_digit_subtraction 0 multiple_choice_grade 0.4300 ± 0.0498
bigbench_arithmetic_3_digit_addition 0 multiple_choice_grade 0.5800 ± 0.0496
bigbench_arithmetic_3_digit_division 0 multiple_choice_grade 0.3100 ± 0.0465
bigbench_arithmetic_3_digit_multiplication 0 multiple_choice_grade 0.2900 ± 0.0456
bigbench_arithmetic_3_digit_subtraction 0 multiple_choice_grade 0.5100 ± 0.0502
bigbench_arithmetic_4_digit_addition 0 multiple_choice_grade 0.5300 ± 0.0502
bigbench_arithmetic_4_digit_division 0 multiple_choice_grade 0.3900 ± 0.0490
bigbench_arithmetic_4_digit_multiplication 0 multiple_choice_grade 0.3100 ± 0.0465
bigbench_arithmetic_4_digit_subtraction 0 multiple_choice_grade 0.6200 ± 0.0488
bigbench_arithmetic_5_digit_addition 0 multiple_choice_grade 0.6500 ± 0.0479
bigbench_arithmetic_5_digit_division 0 multiple_choice_grade 0.3200 ± 0.0469
bigbench_arithmetic_5_digit_multiplication 0 multiple_choice_grade 0.3000 ± 0.0461
bigbench_arithmetic_5_digit_subtraction 0 multiple_choice_grade 0.6400 ± 0.0482
bigbench_cause_and_effect_one_sentence 0 multiple_choice_grade 0.5686 ± 0.0700
bigbench_cause_and_effect_one_sentence_no_prompt 0 multiple_choice_grade 0.6471 ± 0.0676
bigbench_cause_and_effect_two_sentences 0 multiple_choice_grade 0.4314 ± 0.0700
bigbench_emotions 0 multiple_choice_grade 0.2250 ± 0.0331
bigbench_empirical_judgments 0 multiple_choice_grade 0.2525 ± 0.0439
bigbench_general_knowledge 0 multiple_choice_grade 0.3429 ± 0.0571
bigbench_hhh_alignment_harmless 0 multiple_choice_grade 0.3793 ± 0.0643
bigbench_hhh_alignment_helpful 0 multiple_choice_grade 0.3390 ± 0.0622
bigbench_hhh_alignment_honest 0 multiple_choice_grade 0.3729 ± 0.0635
bigbench_hhh_alignment_other 0 multiple_choice_grade 0.5349 ± 0.0770
bigbench_intent_recognition 0 multiple_choice_grade 0.2153 ± 0.0156
bigbench_misconceptions 0 multiple_choice_grade 0.5224 ± 0.0433
bigbench_paraphrase 0 multiple_choice_grade 0.4750 ± 0.0354
bigbench_sentence_ambiguity 0 multiple_choice_grade 0.4833 ± 0.0651
bigbench_similarities_abstraction 0 multiple_choice_grade 0.6974 ± 0.0530

Uploaded model

  • Developed by: mpasila
  • License: apache-2.0
  • Finetuned from model : LumiOpen/Viking-7B

This llama model was trained 2x faster with Unsloth and Huggingface's TRL library.