language:
- en
library_name: transformers
pipeline_tag: text-generation
datasets:
- jondurbin/airoboros-2.2.1
- Open-Orca/OpenOrca
- garage-bAInd/Open-Platypus
- ehartford/samantha-data
- CollectiveCognition/chats-data-2023-09-27
- stingning/ultrachat
tags:
- llama-2
- code
license: llama2
model-index:
- name: SpeechlessCoder
results:
- task:
type: text-generation
dataset:
type: openai_humaneval
name: HumanEval
metrics:
- name: pass@1
type: pass@1
value: 0
verified: false
speechless-mistral-six-in-one-7b
This model is a merge of 6 SOTA Mistral-7B based models:
- ehartford/dolphin-2.1-mistral-7b
- Open-Orca/Mistral-7B-OpenOrca
- bhenrym14/mistral-7b-platypus-fp16
- ehartford/samantha-1.2-mistral-7b
- iteknium/CollectiveCognition-v1.1-Mistral-7B
- HuggingFaceH4/zephyr-7b-alpha
Model benchmark by sethuiyer . Thanks a lot.
I tested the Q6_0 version of the model against LLaMa2 70B chat and here are the results - Scoring as per ChatGPT and Bard's average. Named this model Mixtral. Questions taken from MT-Benchmark.
On a scale of 0 to 100, I would rate Mixtral at 98. Here's why:
- Intellect (100/100) - Mixtral has demonstrated immense intellectual abilities through its comprehensive knowledge and logical reasoning skills.
- Creativity (98/100) - In addition to being highly intelligent, Mixtral also displays impressive creative talents through its unique, nuanced responses.
- Adaptability (98/100) - Mixtral can converse flexibly on a wide variety of topics, adapting smoothly based on contextual cues.
- Communication (97/100) - Mixtral communicates clearly and eloquently through written language, thoroughly answering questions.
- Problem-Solving (98/100) - Questions are addressed comprehensively, considering multiple perspectives to arrive at well-thought solutions.
- Personability (97/100) - Responses are warm, inviting and non-threatening due to Mixtral's kindness and thoughtfulness.
Overall, a very capable model for it's size.
Code: https://github.com/uukuguy/speechless
HumanEval
| Metric | Value |
|---|---|
| humaneval-python |
CodeLlama-34B-Python: 53.29
CodeLlama-34B-Instruct: 50.79
CodeLlama-13B-Instruct: 50.6
CodeLlama-34B: 45.11
CodeLlama-13B-Python: 42.89
CodeLlama-13B: 35.07
Mistral-7B-v0.1: 30.488
LM-Evaluation-Harness
| Metric | Value |
|---|---|
| ARC | 62.97 |
| HellaSwag | 84.6 |
| MMLU | 63.29 |
| TruthfulQA | 57.77 |
| Winogrande | 77.51 |
| GSM8K | 18.42 |
| DROP | 9.13 |
| Average | 53.38 |
Model Card for Mistral-7B-v0.1
The Mistral-7B-v0.1 Large Language Model (LLM) is a pretrained generative text model with 7 billion parameters. Mistral-7B-v0.1 outperforms Llama 2 13B on all benchmarks we tested.
For full details of this model please read our paper and release blog post.
Model Architecture
Mistral-7B-v0.1 is a transformer model, with the following architecture choices:
- Grouped-Query Attention
- Sliding-Window Attention
- Byte-fallback BPE tokenizer
Troubleshooting
- If you see the following error:
KeyError: 'mistral' - Or:
NotImplementedError: Cannot copy out of meta tensor; no data!
Ensure you are utilizing a stable version of Transformers, 4.34.0 or newer.
Notice
Mistral 7B is a pretrained base model and therefore does not have any moderation mechanisms.
The Mistral AI Team
Albert Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lélio Renard Lavaud, Lucile Saulnier, Marie-Anne Lachaux, Pierre Stock, Teven Le Scao, Thibaut Lavril, Thomas Wang, Timothée Lacroix, William El Sayed.`
Open LLM Leaderboard Evaluation Results
Detailed results can be found here
| Metric | Value |
|---|---|
| Avg. | 53.38 |
| ARC (25-shot) | 62.97 |
| HellaSwag (10-shot) | 84.6 |
| MMLU (5-shot) | 63.29 |
| TruthfulQA (0-shot) | 57.77 |
| Winogrande (5-shot) | 77.51 |
| GSM8K (5-shot) | 18.42 |
| DROP (3-shot) | 9.13 |