File size: 2,692 Bytes
ac967b8 933f78a ac967b8 933f78a 255bc45 ac967b8 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 |
---
inference: false
datasets:
- WizardLM/WizardLM_evol_instruct_V2_196k
---
# Model Card for Model ID
This is exl2 5.53bpw quant of Vicuna, specifically https://huggingface.co/lmsys/vicuna-13b-v1.5-16k
More notes on the original model can be found here [lmSys page](https://github.com/lm-sys/FastChat/blob/main/docs/vicuna_weights_version.md).
`python convert.py -i C:\webui\models\lmsys_vicuna-13b-v1.5-16k -o C:\webui\models\Processed
-nr -om lmsys_vicuna-13b-v1.5-16k_measurement.json -mr 15 -gr 15
-c "C:\webui\repositories\exllamav2\WizardLM_evol_instruct_V2_196k_0000.parquet"
&& python convert.py -i C:\webui\models\lmsys_vicuna-13b-v1.5-16k -o C:\webui\models\Processed
-nr -m lmsys_vicuna-13b-v1.5-16k_measurement.json -b 5.53 -gr 40
-c "C:\webui\repositories\exllamav2\WizardLM_evol_instruct_V2_196k_0000.parquet"
-cf lmsys_vicuna-13b-v1.5-16k-exl2-5.53bpw -ss 4000 `
## Original Model Details
# Vicuna Model Card
## Model Details
Vicuna is a chat assistant trained by fine-tuning LLaMA on user-shared conversations collected from ShareGPT.
- **Developed by:** [LMSYS](https://lmsys.org/)
- **Model type:** An auto-regressive language model based on the transformer architecture.
- **License:** Non-commercial license
- **Finetuned from model:** [LLaMA](https://arxiv.org/abs/2302.13971).
### Model Sources
- **Repository:** https://github.com/lm-sys/FastChat
- **Blog:** https://lmsys.org/blog/2023-03-30-vicuna/
- **Paper:** https://arxiv.org/abs/2306.05685
- **Demo:** https://chat.lmsys.org/
## Uses
The primary use of Vicuna is research on large language models and chatbots.
The primary intended users of the model are researchers and hobbyists in natural language processing, machine learning, and artificial intelligence.
## How to Get Started with the Model
Command line interface: https://github.com/lm-sys/FastChat#vicuna-weights.
APIs (OpenAI API, Huggingface API): https://github.com/lm-sys/FastChat/tree/main#api.
## Training Details
Vicuna v1.1 is fine-tuned from LLaMA with supervised instruction fine-tuning.
The training data is around 70K conversations collected from ShareGPT.com.
See more details in the "Training Details of Vicuna Models" section in the appendix of this [paper](https://arxiv.org/pdf/2306.05685.pdf).
## Evaluation
Vicuna is evaluated with standard benchmarks, human preference, and LLM-as-a-judge. See more details in this [paper](https://arxiv.org/pdf/2306.05685.pdf) and [leaderboard](https://huggingface.co/spaces/lmsys/chatbot-arena-leaderboard).
## Difference between different versions of Vicuna
See [vicuna_weights_version.md](https://github.com/lm-sys/FastChat/blob/main/docs/vicuna_weights_version.md) |