Downloads · 30 days
102
9% of all-time downloads
RichardErkhov/chlee10_-_T3Q-ko-solar-sft-v3.0-gguf
chlee10_-_T3Q-ko-solar-sft-v3.0-gguf is a machine learning model from RichardErkhov. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
Downloads · 30 days
102
9% of all-time downloads
All-time downloads
1.1K
Public
Repo size
122 GB
Likes
0
Public
Click a slice to open those files.
.gguf141 GB · 100%
From the Hugging Face model README
Quantization made by Richard Erkhov.
T3Q-ko-solar-sft-v3.0 - GGUF
| Name | Quant method | Size |
|---|---|---|
| T3Q-ko-solar-sft-v3.0.Q2_K.gguf | Q2_K | 3.73GB |
| T3Q-ko-solar-sft-v3.0.IQ3_XS.gguf | IQ3_XS | 4.14GB |
| T3Q-ko-solar-sft-v3.0.IQ3_S.gguf | IQ3_S | 4.37GB |
| T3Q-ko-solar-sft-v3.0.Q3_K_S.gguf | Q3_K_S | 4.34GB |
| T3Q-ko-solar-sft-v3.0.IQ3_M.gguf | IQ3_M | 4.51GB |
| T3Q-ko-solar-sft-v3.0.Q3_K.gguf | Q3_K | 4.84GB |
| T3Q-ko-solar-sft-v3.0.Q3_K_M.gguf | Q3_K_M | 4.84GB |
| T3Q-ko-solar-sft-v3.0.Q3_K_L.gguf | Q3_K_L | 5.26GB |
| T3Q-ko-solar-sft-v3.0.IQ4_XS.gguf | IQ4_XS | 5.43GB |
| T3Q-ko-solar-sft-v3.0.Q4_0.gguf | Q4_0 | 5.66GB |
| T3Q-ko-solar-sft-v3.0.IQ4_NL.gguf | IQ4_NL | 5.72GB |
| T3Q-ko-solar-sft-v3.0.Q4_K_S.gguf | Q4_K_S | 5.7GB |
| T3Q-ko-solar-sft-v3.0.Q4_K.gguf | Q4_K | 6.02GB |
| T3Q-ko-solar-sft-v3.0.Q4_K_M.gguf | Q4_K_M | 6.02GB |
| T3Q-ko-solar-sft-v3.0.Q4_1.gguf | Q4_1 | 6.27GB |
| T3Q-ko-solar-sft-v3.0.Q5_0.gguf | Q5_0 | 6.89GB |
| T3Q-ko-solar-sft-v3.0.Q5_K_S.gguf | Q5_K_S | 6.89GB |
| T3Q-ko-solar-sft-v3.0.Q5_K.gguf | Q5_K | 7.08GB |
| T3Q-ko-solar-sft-v3.0.Q5_K_M.gguf | Q5_K_M | 7.08GB |
| T3Q-ko-solar-sft-v3.0.Q5_1.gguf | Q5_1 | 7.51GB |
| T3Q-ko-solar-sft-v3.0.Q6_K.gguf | Q6_K | 8.2GB |
| T3Q-ko-solar-sft-v3.0.Q8_0.gguf | Q8_0 | 10.62GB |
pipeline_tag: text-generation license: apache-2.0 language:
Update @ 2024.03.25

This model is a SFT fine-tuned version of chihoonlee10/T3Q-ko-solar-dpo-v3.0
Model Developers Chihoon Lee(chlee10), T3Q
The following hyperparameters were used during training:
# 데이터셋과 훈련 횟수와 관련된 하이퍼 파라미터
batch_size = 16
num_epochs = 1
micro_batch = 1
gradient_accumulation_steps = batch_size // micro_batch
# 훈련 방법에 대한 하이퍼 파라미터
cutoff_len = 4096
lr_scheduler = 'cosine'
warmup_ratio = 0.06 # warmup_steps = 100
learning_rate = 5e-5
optimizer = 'paged_adamw_32bit'
weight_decay = 0.01
max_grad_norm = 1.0
# LoRA config
lora_r = 16
lora_alpha = 16
lora_dropout = 0.05
lora_target_modules = ["k_proj", "v_proj","gate_proj", "down_proj", "up_proj"]
# Tokenizer에서 나오는 input값 설정 옵션
train_on_inputs = False
add_eos_token = False
# NEFTune params
neftune_noise_alpha = 5