You can deploy this model and start earning money today!

RichardErkhov/
Weyaxi_-_Neural-una-cybertron-7b-gguf

Name: Weyaxi_-_Neural-una-cybertron-7b-gguf
Availability: InStock

Views

No views yet

Quantization made by Richard Erkhov.

Neural-una-cybertron-7b - GGUF

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

Original model description:

license: apache-2.0 datasets:

Neural-una-cybertron-7b is an fblgit/una-cybertron-7b-v2-bf16 model that has been further fine-tuned with Direct Preference Optimization (DPO) using the Intel/orca_dpo_pairs dataset.

This model was created after examining the procedure of mlabonne/NeuralHermes-2.5-Mistral-7B model. Special thanks to @mlabonne.

This model was fine-tuned on Nvidia A100-SXM4-40GB GPU.

The total training time was 1 hour and 10 minutes.

<|im_start|>system
{system}<|im_end|>
<|im_start|>user
{user}<|im_end|>
<|im_start|>assistant
{asistant}<|im_end|>

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:

DPOTrainer: