Accuracy

TP TN
FP 447 141
FN 39 373

정확도: 82%(447+373/1000)

-Train λ°μ΄ν„°λŠ” nsmc['train']을 shuffleν•œ λ‹€μŒμ— μƒμœ„ 5000개λ₯Ό ν™œμš©ν•˜μ˜€μŠ΅λ‹ˆλ‹€

-Test λ°μ΄ν„°λŠ” nsmc['train']을 shuffleν•œ λ‹€μŒμ— μƒμœ„ 1000개λ₯Ό ν™œμš©ν•˜μ˜€μŠ΅λ‹ˆλ‹€.

-1000개의 train 데이터 쀑 820개λ₯Ό μ •ν™•ν•˜κ²Œ λΆ„λ₯˜ν•˜μ˜€μŠ΅λ‹ˆλ‹€.

-Base Model은 meta-llama/Llama-2-7b-chat-hf둜, base model에 Lora μ–΄λŒ‘ν„°λ₯Ό λΆ™μ—¬μ„œ SFTtrainerλ₯Ό ν†΅ν•˜μ—¬ μƒˆλ‘œμš΄ 데이터셋 nsmc에 λŒ€ν•΄μ„œ λ―Έμ„ΈνŠœλ‹μ„ ν•˜μ˜€μŠ΅λ‹ˆλ‹€.

-λ―Έμ„ΈνŠœλ‹ν•œ 둜라 μ–΄λŒ‘ν„°λ₯Ό ν—ˆκΉ…νŽ˜μ΄μŠ€μ— μ—…λ‘œλ“œν•œ ν›„ 4λΉ„νŠΈλ‘œ μ–‘μžν™”λœ base model에 뢙인 ν›„ 좔둠을 ν•˜μ˜€μŠ΅λ‹ˆλ‹€.

-μƒμœ„ 2000개의 train_dataset에 λŒ€ν•΄μ„œ ν›ˆλ ¨μ„ ν•œ ν›„ μƒμœ„ 1000개의 test_dataset에 λŒ€ν•΄μ„œ 좔둠은 ν•œ κ²°κ³ΌλŠ” μœ„μ˜ ν‘œμ™€ κ°™μŠ΅λ‹ˆλ‹€.

-정확도 ν–₯상을 μœ„ν•˜μ—¬ test λ°μ΄ν„°μ…‹μ˜ 크기λ₯Ό 늘리렀 ν•˜μ˜€μœΌλ‚˜, GPU μš©λŸ‰ μ œν•œμœΌλ‘œ μΈν•˜μ—¬ μ§€μ†μ μœΌλ‘œ 였λ₯˜κ°€ λ°œμƒν•˜μ˜€μŠ΅λ‹ˆλ‹€.

Downloads last month
13
Inference Providers NEW
This model is not currently available via any of the supported Inference Providers.
The model cannot be deployed to the HF Inference API: The model has no pipeline_tag.

Model tree for kayla0913/hw-llama2-7B-nsmc

Adapter
(1113)
this model