tags:
- merge
- mergekit
- lazymergekit
- Trappu/Nemo-Picaro-fixed
- anthracite-org/magnum-v2-12b
base_model:
- Trappu/Nemo-Picaro-fixed
- anthracite-org/magnum-v2-12b
model-index:
- name: Magnum-Picaro-0.7-v2-12b
results:
- task:
type: text-generation
name: Text Generation
dataset:
name: IFEval (0-Shot)
type: HuggingFaceH4/ifeval
args:
num_few_shot: 0
metrics:
- type: inst_level_strict_acc and prompt_level_strict_acc
value: 30.03
name: strict accuracy
source:
url: >-
https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=Trappu/Magnum-Picaro-0.7-v2-12b
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: BBH (3-Shot)
type: BBH
args:
num_few_shot: 3
metrics:
- type: acc_norm
value: 35.75
name: normalized accuracy
source:
url: >-
https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=Trappu/Magnum-Picaro-0.7-v2-12b
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: MATH Lvl 5 (4-Shot)
type: hendrycks/competition_math
args:
num_few_shot: 4
metrics:
- type: exact_match
value: 4.76
name: exact match
source:
url: >-
https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=Trappu/Magnum-Picaro-0.7-v2-12b
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: GPQA (0-shot)
type: Idavidrein/gpqa
args:
num_few_shot: 0
metrics:
- type: acc_norm
value: 9.73
name: acc_norm
source:
url: >-
https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=Trappu/Magnum-Picaro-0.7-v2-12b
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: MuSR (0-shot)
type: TAUR-Lab/MuSR
args:
num_few_shot: 0
metrics:
- type: acc_norm
value: 19.56
name: acc_norm
source:
url: >-
https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=Trappu/Magnum-Picaro-0.7-v2-12b
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: MMLU-PRO (5-shot)
type: TIGER-Lab/MMLU-Pro
config: main
split: test
args:
num_few_shot: 5
metrics:
- type: acc
value: 28.67
name: accuracy
source:
url: >-
https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=Trappu/Magnum-Picaro-0.7-v2-12b
name: Open LLM Leaderboard
Details
This model is a merge between Trappu/Nemo-Picaro-fixed, a model trained on my own little dataset free of synthetic data, which focuses solely on storywriting and scenrio prompting (Example: [ Scenario: bla bla bla; Tags: bla bla bla ]), and anthracite-org/magnum-v2-12b.
The reason why I decided to merge it with Magnum (and don't recommend Picaro alone) is because that model, aside from its obvious flaws (rampant impersonation, stupid, etc...), is a one-trick pony and will be really rough for the average LLM user to handle. The idea was to have Magnum work as some sort of stabilizer to fix the issues that emerge from the lack of multiturn/smart data in Picaro's dataset. It worked, I think. I enjoy the outputs and it's smart enough to work with.
But yeah the goal of this merge was to make a model that's both good at storytelling/narration but also fine when it comes to other forms of creative writing such as RP or chatting. I don't think it's quite there yet but it's something for sure.
Prompting
As explained before, Picaro is a model that functions mainly through scenario prompting but merging it with Magnum has made it a lot more versatile so you can use it however you see fit. Both models were trained on chatml so below is the recommended prompt formatting.
<|im_start|>system
system prompt<|im_end|>
<|im_start|>user
bla bla bla<|im_end|>
<|im_start|>assistant
bla bla bla you!<|im_end|>
For SillyTavern users:
The above settings are the ones I recommend.
Temp = 1.2
Min P = 0.1
DRY Rep Pen: Multiplier = 0.8, Base = 1.75, Allowed Length = 2, Penalty Range = 1024
Little guide on useful samplers and how to import settings presets and instruct/context templates and other stuff people might find useful here
Every other sampler neutralized.
Magnum-Picaro-0.7-v2-12b
Magnum-Picaro-0.7-v2-12b is a merge of the following models using LazyMergekit:
🧩 Configuration
models:
- model: Trappu/Nemo-Picaro-fixed
parameters:
density: 0.7
weight: 0.5
- model: anthracite-org/magnum-v2-12b
parameters:
density: 0.3
weight: 0.5
merge_method: ties
base_model: Trappu/Nemo-Picaro-fixed
parameters:
normalize: true
int8_mask: true
dtype: bfloat16
💻 Usage
!pip install -qU transformers accelerate
from transformers import AutoTokenizer
import transformers
import torch
model = "Trappu/Magnum-Picaro-0.7-v2-12b"
messages = [{"role": "user", "content": "What is a large language model?"}]
tokenizer = AutoTokenizer.from_pretrained(model)
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
pipeline = transformers.pipeline(
"text-generation",
model=model,
torch_dtype=torch.float16,
device_map="auto",
)
outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
print(outputs[0]["generated_text"])
Open LLM Leaderboard Evaluation Results
Detailed results can be found here
| Metric | Value |
|---|---|
| Avg. | 21.42 |
| IFEval (0-Shot) | 30.03 |
| BBH (3-Shot) | 35.75 |
| MATH Lvl 5 (4-Shot) | 4.76 |
| GPQA (0-shot) | 9.73 |
| MuSR (0-shot) | 19.56 |
| MMLU-PRO (5-shot) | 28.67 |