RichardErkhov/sonthenguyen_-_OpenHermes-2.5-Mistral-7B-mt-bench-DPO-recovered-gguf

Quantization made by Richard Erkhov.

OpenHermes-2.5-Mistral-7B-mt-bench-DPO-recovered - GGUF

Model creator: https://huggingface.co/sonthenguyen/
Original model: https://huggingface.co/sonthenguyen/OpenHermes-2.5-Mistral-7B-mt-bench-DPO-recovered/

Name	Quant method	Size
OpenHermes-2.5-Mistral-7B-mt-bench-DPO-recovered.Q2_K.gguf	Q2_K	2.53GB
OpenHermes-2.5-Mistral-7B-mt-bench-DPO-recovered.IQ3_XS.gguf	IQ3_XS	2.81GB
OpenHermes-2.5-Mistral-7B-mt-bench-DPO-recovered.IQ3_S.gguf	IQ3_S	2.96GB
OpenHermes-2.5-Mistral-7B-mt-bench-DPO-recovered.Q3_K_S.gguf	Q3_K_S	2.95GB
OpenHermes-2.5-Mistral-7B-mt-bench-DPO-recovered.IQ3_M.gguf	IQ3_M	3.06GB
OpenHermes-2.5-Mistral-7B-mt-bench-DPO-recovered.Q3_K.gguf	Q3_K	3.28GB
OpenHermes-2.5-Mistral-7B-mt-bench-DPO-recovered.Q3_K_M.gguf	Q3_K_M	3.28GB
OpenHermes-2.5-Mistral-7B-mt-bench-DPO-recovered.Q3_K_L.gguf	Q3_K_L	3.56GB
OpenHermes-2.5-Mistral-7B-mt-bench-DPO-recovered.IQ4_XS.gguf	IQ4_XS	3.67GB
OpenHermes-2.5-Mistral-7B-mt-bench-DPO-recovered.Q4_0.gguf	Q4_0	3.83GB
OpenHermes-2.5-Mistral-7B-mt-bench-DPO-recovered.IQ4_NL.gguf	IQ4_NL	3.87GB
OpenHermes-2.5-Mistral-7B-mt-bench-DPO-recovered.Q4_K_S.gguf	Q4_K_S	3.86GB
OpenHermes-2.5-Mistral-7B-mt-bench-DPO-recovered.Q4_K.gguf	Q4_K	4.07GB
OpenHermes-2.5-Mistral-7B-mt-bench-DPO-recovered.Q4_K_M.gguf	Q4_K_M	4.07GB
OpenHermes-2.5-Mistral-7B-mt-bench-DPO-recovered.Q4_1.gguf	Q4_1	4.24GB
OpenHermes-2.5-Mistral-7B-mt-bench-DPO-recovered.Q5_0.gguf	Q5_0	4.65GB
OpenHermes-2.5-Mistral-7B-mt-bench-DPO-recovered.Q5_K_S.gguf	Q5_K_S	4.65GB
OpenHermes-2.5-Mistral-7B-mt-bench-DPO-recovered.Q5_K.gguf	Q5_K	4.78GB
OpenHermes-2.5-Mistral-7B-mt-bench-DPO-recovered.Q5_K_M.gguf	Q5_K_M	4.78GB
OpenHermes-2.5-Mistral-7B-mt-bench-DPO-recovered.Q5_1.gguf	Q5_1	5.07GB
OpenHermes-2.5-Mistral-7B-mt-bench-DPO-recovered.Q6_K.gguf	Q6_K	5.53GB
OpenHermes-2.5-Mistral-7B-mt-bench-DPO-recovered.Q8_0.gguf	Q8_0	7.17GB

Original model description:

license: apache-2.0

Training hyperparameters LoRA: r=16 lora_alpha=16 lora_dropout=0.05 bias="none" task_type="CAUSAL_LM" target_modules=['k_proj', 'gate_proj', 'v_proj', 'up_proj', 'q_proj', 'o_proj', 'down_proj']

Training arguments: auto_find_batch_size=True gradient_checkpointing=True learning_rate=5e-7 lr_scheduler_type="cosine" max_steps=3922 optim="paged_adamw_32bit" warmup_steps=100

DPOTrainer: beta=0.1 max_prompt_length=1024 max_length=1536

Arxiv link: https://arxiv.org/abs/2403.02745