Model Card for Model ID
This model is a finetuning of other models based on mistralai/Mistral-7B-v0.1.
Model Details
Model Description
The model has been generated from the merging of the models viethq188/LeoScorpius-7B-Chat-DPO and GreenNode/GreenNodeLM-7B-v1olet and a later finetuning with a Platypus dataset garage-bAInd/Open-Platypus.
- Developed by: Ignos
- Model type: Mistral
- License: Apache-2.0
Uses
The model aims to have good overall comparative results on HuggingFace metrics, improving reasoning.
Bias, Risks, and Limitations
The same bias, risks and limitations from base models.
Training Details
Training Data
Training Procedure
- Training with QLoRA approach and merging with base model.
Results
- Huggingface evaluation pending
Summary
Technical Specifications
Model Architecture and Objective
- Models based on Mistral Architecture
Compute Infrastructure
- Training on RunPod
Hardware
- 4 x Nvidia RTX 4090
- 64 vCPU 503 GB RAM
Software
- Mergekit (main)
- Axolotl 0.3.0
Training procedure
The following bitsandbytes quantization config was used during training:
- quant_method: bitsandbytes
- load_in_8bit: False
- load_in_4bit: True
- llm_int8_threshold: 6.0
- llm_int8_skip_modules: None
- llm_int8_enable_fp32_cpu_offload: False
- llm_int8_has_fp16_weight: False
- bnb_4bit_quant_type: nf4
- bnb_4bit_use_double_quant: True
- bnb_4bit_compute_dtype: bfloat16
Framework versions
- PEFT 0.6.0