language:
- en
license: cc-by-nc-4.0
model-index:
- name: mistral-ft-optimized-1218
results:
- task:
type: text-generation
name: Text Generation
dataset:
name: AI2 Reasoning Challenge (25-Shot)
type: ai2_arc
config: ARC-Challenge
split: test
args:
num_few_shot: 25
metrics:
- type: acc_norm
value: 67.92
name: normalized accuracy
source:
url: >-
https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=OpenPipe/mistral-ft-optimized-1218
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: HellaSwag (10-Shot)
type: hellaswag
split: validation
args:
num_few_shot: 10
metrics:
- type: acc_norm
value: 86.26
name: normalized accuracy
source:
url: >-
https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=OpenPipe/mistral-ft-optimized-1218
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: MMLU (5-Shot)
type: cais/mmlu
config: all
split: test
args:
num_few_shot: 5
metrics:
- type: acc
value: 64.99
name: accuracy
source:
url: >-
https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=OpenPipe/mistral-ft-optimized-1218
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: TruthfulQA (0-shot)
type: truthful_qa
config: multiple_choice
split: validation
args:
num_few_shot: 0
metrics:
- type: mc2
value: 59.48
source:
url: >-
https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=OpenPipe/mistral-ft-optimized-1218
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: Winogrande (5-shot)
type: winogrande
config: winogrande_xl
split: validation
args:
num_few_shot: 5
metrics:
- type: acc
value: 80.74
name: accuracy
source:
url: >-
https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=OpenPipe/mistral-ft-optimized-1218
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: GSM8k (5-shot)
type: gsm8k
config: main
split: test
args:
num_few_shot: 5
metrics:
- type: acc
value: 72.25
name: accuracy
source:
url: >-
https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=OpenPipe/mistral-ft-optimized-1218
name: Open LLM Leaderboard
Update 12/27/2023: We have released an updated version of this model with similar performance and a more permissive license at https://huggingface.co/OpenPipe/mistral-ft-optimized-1227. We recommend that model over this one for most users.
This model is intended to be a strong base suitable for downstream fine-tuning on a variety of tasks. Based on our internal evaluations, we believe it's one of the strongest models for most down-stream tasks. You can read more about our development and evaluation process here.
Mergekit config used to create this model:
slices:
- sources:
- model: Weyaxi/OpenHermes-2.5-neural-chat-v3-3-Slerp
layer_range: [0, 32]
- model: Q-bert/MetaMath-Cybertron-Starling
layer_range: [0, 32]
merge_method: slerp
base_model: mistralai/Mistral-7B-v0.1
parameters:
t:
- filter: self_attn
value: [0, 0.5, 0.3, 0.7, 1]
- filter: mlp
value: [1, 0.5, 0.7, 0.3, 0]
- value: 0.5 # fallback for rest of tensors
dtype: bfloat16
Note: It appears that https://huggingface.co/Weyaxi/Seraph-7B was merged from the same base models using the same mergekit defaults as this model. So major credit goes to @Weyaxi both for creating one of the base merges this model was merged from, as well as being the first one to perform this exact merge as well!
Open LLM Leaderboard Evaluation Results
Detailed results can be found here
| Metric | Value |
|---|---|
| Avg. | 71.94 |
| AI2 Reasoning Challenge (25-Shot) | 67.92 |
| HellaSwag (10-Shot) | 86.26 |
| MMLU (5-Shot) | 64.99 |
| TruthfulQA (0-shot) | 59.48 |
| Winogrande (5-shot) | 80.74 |
| GSM8k (5-shot) | 72.25 |