YAML Metadata
Warning:
empty or missing yaml metadata in repo card
(https://huggingface.co/docs/hub/model-cards#model-card-metadata)
Quantization made by Richard Erkhov.
K2S3-SOLAR-11b-v4.0 - GGUF
- Model creator: https://huggingface.co/Changgil/
- Original model: https://huggingface.co/Changgil/K2S3-SOLAR-11b-v4.0/
Name | Quant method | Size |
---|---|---|
K2S3-SOLAR-11b-v4.0.Q2_K.gguf | Q2_K | 3.73GB |
K2S3-SOLAR-11b-v4.0.IQ3_XS.gguf | IQ3_XS | 4.14GB |
K2S3-SOLAR-11b-v4.0.IQ3_S.gguf | IQ3_S | 4.37GB |
K2S3-SOLAR-11b-v4.0.Q3_K_S.gguf | Q3_K_S | 4.34GB |
K2S3-SOLAR-11b-v4.0.IQ3_M.gguf | IQ3_M | 4.51GB |
K2S3-SOLAR-11b-v4.0.Q3_K.gguf | Q3_K | 4.84GB |
K2S3-SOLAR-11b-v4.0.Q3_K_M.gguf | Q3_K_M | 4.84GB |
K2S3-SOLAR-11b-v4.0.Q3_K_L.gguf | Q3_K_L | 5.26GB |
K2S3-SOLAR-11b-v4.0.IQ4_XS.gguf | IQ4_XS | 5.43GB |
K2S3-SOLAR-11b-v4.0.Q4_0.gguf | Q4_0 | 5.66GB |
K2S3-SOLAR-11b-v4.0.IQ4_NL.gguf | IQ4_NL | 5.72GB |
K2S3-SOLAR-11b-v4.0.Q4_K_S.gguf | Q4_K_S | 5.7GB |
K2S3-SOLAR-11b-v4.0.Q4_K.gguf | Q4_K | 6.02GB |
K2S3-SOLAR-11b-v4.0.Q4_K_M.gguf | Q4_K_M | 6.02GB |
K2S3-SOLAR-11b-v4.0.Q4_1.gguf | Q4_1 | 6.27GB |
K2S3-SOLAR-11b-v4.0.Q5_0.gguf | Q5_0 | 6.89GB |
K2S3-SOLAR-11b-v4.0.Q5_K_S.gguf | Q5_K_S | 6.89GB |
K2S3-SOLAR-11b-v4.0.Q5_K.gguf | Q5_K | 7.08GB |
K2S3-SOLAR-11b-v4.0.Q5_K_M.gguf | Q5_K_M | 7.08GB |
K2S3-SOLAR-11b-v4.0.Q5_1.gguf | Q5_1 | 7.51GB |
K2S3-SOLAR-11b-v4.0.Q6_K.gguf | Q6_K | 8.2GB |
K2S3-SOLAR-11b-v4.0.Q8_0.gguf | Q8_0 | 10.62GB |
Original model description:
license: cc-by-nc-4.0 language: - ko
Developed by :
- K2S3
Model Number:
- K2S3-SOLAR-11b-v4.0
Base Model :
Training Data
- The training data for this model includes the Standard Korean Dictionary, training data from KULLM at Korea University, abstracts of master's and doctoral theses, Korean language samples from AI Hub, alpaca-gpt4-data, and samples from The OpenOrca Dataset.
- ์ด ๋ชจ๋ธ์ ํ๋ จ ๋ฐ์ดํฐ์๋ ํ์ค๊ตญ์ด๋์ฌ์ , ๊ณ ๋ ค๋ํ๊ต KULLM์์ ์ ๊ณตํ ํ๋ จ ๋ฐ์ดํฐ, ์์ฌ ๋ฐ ๋ฐ์ฌํ์ ๋ ผ๋ฌธ์ ์ด๋ก, AI Hub์์ ์ ๊ณตํ ํ๊ตญ์ด ๋ฐ์ดํฐ ์ํ, alpaca-gpt4-data, ๊ทธ๋ฆฌ๊ณ OpenOrca Dataset์์ ์ ๊ณตํ ์ํ๋ค์ด ํฌํจ๋ฉ๋๋ค.
Training Method
- This model was fine-tuned on the "upstage/SOLAR-10.7B-v1.0" base model using a full parameter tuning method with SFT (Supervised Fine-Tuning).
- ์ด ๋ชจ๋ธ์ "upstage/SOLAR-10.7B-v1.0" ๊ธฐ๋ฐ ๋ชจ๋ธ์ SFT๋ฅผ ์ฌ์ฉํ์ฌ ์ ์ฒด ํ๋ผ๋ฏธํฐ ์กฐ์ ๋ฐฉ๋ฒ์ผ๋ก ๋ฏธ์ธ์กฐ์ ๋์์ต๋๋ค.
Hardware
- Hardware: Utilized two A100 (80G*2EA) GPUs for training.
- Training Factors: This model was fine-tuned with SFT, using the HuggingFace SFTtrainer and applied fsdp.
- ์ด ๋ชจ๋ธ์ SFT๋ฅผ ์ฌ์ฉํ์ฌ HuggingFace SFTtrainer์ fsdp๋ฅผ ์ ์ฉํ์ฌ ๋ฏธ์ธ์กฐ์ ๋์์ต๋๋ค.
- Downloads last month
- 260