🥨 Brezn-7B
This is right now our best performing german speaking 7B model with an apache license, with an average of 7.49 on mt-bench-de. You can test this model here: mayflowergmbh/Brezn-7B-GGUF-Chat.
Brezn-7B is a dpo aligned merge of the following models using LazyMergekit:
- FelixChao/WestSeverus-7B-DPO-v2
- mayflowergmbh/Wiedervereinigung-7b-dpo-laser
- cognitivecomputations/openchat-3.5-0106-laser
💻 Usage
In order to leverage instruction fine-tuning, your prompt should be surrounded by [INST] and [/INST] tokens. The very first instruction should begin with a begin of sentence id. The next instructions should not. The assistant generation will be ended by the end-of-sentence token id.
E.g.
text = "<s>[INST] What is your favourite condiment? [/INST]"
"Well, I'm quite partial to a good squeeze of fresh lemon juice. It adds just the right amount of zesty flavour to whatever I'm cooking up in the kitchen!</s> "
"[INST] Do you have mayonnaise recipes? [/INST]"
This format is available as a chat template via the apply_chat_template() method:
from transformers import AutoModelForCausalLM, AutoTokenizer
device = "cuda" # the device to load the model onto
model = AutoModelForCausalLM.from_pretrained("mayflowergmbh/Brezn-7b")
tokenizer = AutoTokenizer.from_pretrained("mayflowergmbh/Brezn-7b")
messages = [
{"role": "user", "content": "Was ist dein Lieblingsgewürz??"},
{"role": "assistant", "content": "Nun, ich mag besonders gerne einen guten Spritzer frischen Zitronensaft. Er fügt genau die richtige Menge an würzigem Geschmack hinzu, egal was ich gerade in der Küche zubereite!"},
{"role": "user", "content": "Hast du Mayonnaise-Rezepte?"}
]
encodeds = tokenizer.apply_chat_template(messages, return_tensors="pt")
model_inputs = encodeds.to(device)
model.to(device)
generated_ids = model.generate(model_inputs, max_new_tokens=1000, do_sample=True)
decoded = tokenizer.batch_decode(generated_ids)
print(decoded[0])
mt-bench-de
{
"first_turn": 7.6625,
"second_turn": 7.31875,
"categories": {
"writing": 8.75,
"roleplay": 8.5,
"reasoning": 6.1,
"math": 5.05,
"coding": 5.4,
"extraction": 7.975,
"stem": 9,
"humanities": 9.15
},
"average": 7.490625
}
🧩 Configuration
models:
- model: mistralai/Mistral-7B-v0.1
# no parameters necessary for base model
- model: FelixChao/WestSeverus-7B-DPO-v2
parameters:
density: 0.60
weight: 0.30
- model: mayflowergmbh/Wiedervereinigung-7b-dpo-laser
parameters:
density: 0.65
weight: 0.40
- model: cognitivecomputations/openchat-3.5-0106-laser
parameters:
density: 0.6
weight: 0.3
merge_method: dare_ties
base_model: mistralai/Mistral-7B-v0.1
parameters:
int8_mask: true
dtype: bfloat16
random_seed: 0