Seraph-7B

This is the model for Seraph-7B. I used mergekit to merge models.

Prompt Templates

You can use these prompt templates, but I recommend using ChatML.

ChatML:

<|im_start|>system
{system}<|im_end|>
<|im_start|>user
{user}<|im_end|>
<|im_start|>assistant
{asistant}<|im_end|>

System, User, Asistant Alpaca Style:

### System:
{system}
### User:
{user}
### Assistant:

Yaml Config

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

Quantizationed versions

Quantizationed versions of this model is available thanks to TheBloke.

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	71.86
ARC (25-shot)	67.83
HellaSwag (10-shot)	86.22
MMLU (5-shot)	65.07
TruthfulQA (0-shot)	59.49
Winogrande (5-shot)	80.66
GSM8K (5-shot)	71.87

If you would like to support me:

☕ Buy Me a Coffee

Downloads last month: 219

Safetensors

Model size

7B params

Tensor type

BF16

Model tree for Weyaxi/Seraph-7B

Quantizations

7 models

Evaluation results

normalized accuracy on AI2 Reasoning Challenge (25-Shot)
test set Open LLM Leaderboard

67.830
normalized accuracy on HellaSwag (10-Shot)
validation set Open LLM Leaderboard

86.220
accuracy on MMLU (5-Shot)
test set Open LLM Leaderboard

65.070
mc2 on TruthfulQA (0-shot)
validation set Open LLM Leaderboard

59.490
accuracy on Winogrande (5-shot)
validation set Open LLM Leaderboard

80.660
accuracy on GSM8k (5-shot)
test set Open LLM Leaderboard

71.870