Downloads · 30 days
149
7% of all-time downloads
RichardErkhov/adamo1139_-_Mistral-7B-AEZAKMI-v1-gguf
adamo1139_-_Mistral-7B-AEZAKMI-v1-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
149
7% of all-time downloads
All-time downloads
2K
Public
Repo size
82.2 GB
Likes
0
Public
Click a slice to open those files.
.gguf95.2 GB · 100%
From the Hugging Face model README
Quantization made by Richard Erkhov.
Mistral-7B-AEZAKMI-v1 - GGUF
| Name | Quant method | Size |
|---|---|---|
| Mistral-7B-AEZAKMI-v1.Q2_K.gguf | Q2_K | 2.53GB |
| Mistral-7B-AEZAKMI-v1.IQ3_XS.gguf | IQ3_XS | 2.81GB |
| Mistral-7B-AEZAKMI-v1.IQ3_S.gguf | IQ3_S | 2.96GB |
| Mistral-7B-AEZAKMI-v1.Q3_K_S.gguf | Q3_K_S | 2.95GB |
| Mistral-7B-AEZAKMI-v1.IQ3_M.gguf | IQ3_M | 3.06GB |
| Mistral-7B-AEZAKMI-v1.Q3_K.gguf | Q3_K | 3.28GB |
| Mistral-7B-AEZAKMI-v1.Q3_K_M.gguf | Q3_K_M | 3.28GB |
| Mistral-7B-AEZAKMI-v1.Q3_K_L.gguf | Q3_K_L | 3.56GB |
| Mistral-7B-AEZAKMI-v1.IQ4_XS.gguf | IQ4_XS | 3.67GB |
| Mistral-7B-AEZAKMI-v1.Q4_0.gguf | Q4_0 | 3.83GB |
| Mistral-7B-AEZAKMI-v1.IQ4_NL.gguf | IQ4_NL | 3.87GB |
| Mistral-7B-AEZAKMI-v1.Q4_K_S.gguf | Q4_K_S | 3.86GB |
| Mistral-7B-AEZAKMI-v1.Q4_K.gguf | Q4_K | 4.07GB |
| Mistral-7B-AEZAKMI-v1.Q4_K_M.gguf | Q4_K_M | 4.07GB |
| Mistral-7B-AEZAKMI-v1.Q4_1.gguf | Q4_1 | 4.24GB |
| Mistral-7B-AEZAKMI-v1.Q5_0.gguf | Q5_0 | 4.65GB |
| Mistral-7B-AEZAKMI-v1.Q5_K_S.gguf | Q5_K_S | 4.65GB |
| Mistral-7B-AEZAKMI-v1.Q5_K.gguf | Q5_K | 4.78GB |
| Mistral-7B-AEZAKMI-v1.Q5_K_M.gguf | Q5_K_M | 4.78GB |
| Mistral-7B-AEZAKMI-v1.Q5_1.gguf | Q5_1 | 5.07GB |
| Mistral-7B-AEZAKMI-v1.Q6_K.gguf | Q6_K | 5.53GB |
| Mistral-7B-AEZAKMI-v1.Q8_0.gguf | Q8_0 | 7.17GB |
Mistral 7B model fine-tuned on AEZAKMI v1 dataset that is derived from airoboros 2.2.1 and airoboros 2.2. Finetuned with axolotl, using qlora and nf4 double quant, around 2 epochs, batch size 8, lr 0.00008, lr scheduler cosine. Scheduled training was 5 epochs, but loss seemed fine after 2 so I finished it quicker. Training took around 10 hours on single RTX 3090 Ti.
Main feature of this model is that it's output is free of refusals and it feels somehow more natural. Prompt format is standard chatml. Don't expect it to be good at math, riddles or be crazy smart. My end goal with AEZAKMI is to create a cozy free chatbot.
Not sure what license it needs to have, given license of airoboros dataset. I'll leave it as other for now.