Beyonder-4x7B-v2
This model is a Mixture of Experts (MoE) made with mergekit (mixtral branch). It uses the following base models:
- openchat/openchat-3.5-1210
- beowolx/CodeNinja-1.0-OpenChat-7B
- maywell/PiVoT-0.1-Starling-LM-RP
- WizardLM/WizardMath-7B-V1.1
The recommended context length is 8k.
โก Quantized models
Thanks to TheBloke for the quantized models:
- GGUF: https://huggingface.co/TheBloke/Beyonder-4x7B-v2-GGUF
- AWQ: https://huggingface.co/TheBloke/Beyonder-4x7B-v2-AWQ
- GPTQ: https://huggingface.co/TheBloke/Beyonder-4x7B-v2-GPTQ
- EXL2: https://huggingface.co/bartowski/Beyonder-4x7B-v2-exl2
๐ Evaluation
Beyonder-4x7B-v2 is competitive with Mixtral-8x7B-Instruct-v0.1 on the Open LLM Leaderboard, while only having 4 experts instead of 8.
It also displays a significant improvement over the individual experts.
It also performs very well compared to other models on Nous benchmark suite. It's almost as good as the best Yi-34B fine-tune, which is a much bigger model: 24.2B parameters + only two experts are selected during inference (so ~12B) vs. 34B param.
Model | AGIEval | GPT4All | TruthfulQA | Bigbench | Average |
---|---|---|---|---|---|
Beyonder-4x7B-v2 | 45.29 | 75.95 | 60.86 | 46.4 | 57.13 |
NeuralHermes-2.5-Mistral-7B | 43.67 | 73.24 | 55.37 | 41.76 | 53.51 |
OpenHermes-2.5-Mistral-7B | 42.75 | 72.99 | 52.99 | 40.94 | 52.42 |
Nous-Hermes-2-SOLAR-10.7B | 47.79 | 74.69 | 55.92 | 44.84 | 55.81 |
Nous-Hermes-2-Yi-34B | 50.27 | 76.00 | 60.34 | 46.69 | 58.33 |
AGIEval
Task | Version | Metric | Value | Stderr | |
---|---|---|---|---|---|
agieval_aqua_rat | 0 | acc | 23.62 | ยฑ | 2.67 |
acc_norm | 23.62 | ยฑ | 2.67 | ||
agieval_logiqa_en | 0 | acc | 41.47 | ยฑ | 1.93 |
acc_norm | 43.01 | ยฑ | 1.94 | ||
agieval_lsat_ar | 0 | acc | 23.04 | ยฑ | 2.78 |
acc_norm | 23.48 | ยฑ | 2.80 | ||
agieval_lsat_lr | 0 | acc | 51.57 | ยฑ | 2.22 |
acc_norm | 52.94 | ยฑ | 2.21 | ||
agieval_lsat_rc | 0 | acc | 64.31 | ยฑ | 2.93 |
acc_norm | 64.68 | ยฑ | 2.92 | ||
agieval_sat_en | 0 | acc | 79.13 | ยฑ | 2.84 |
acc_norm | 79.13 | ยฑ | 2.84 | ||
agieval_sat_en_without_passage | 0 | acc | 43.20 | ยฑ | 3.46 |
acc_norm | 43.20 | ยฑ | 3.46 | ||
agieval_sat_math | 0 | acc | 34.55 | ยฑ | 3.21 |
acc_norm | 32.27 | ยฑ | 3.16 |
GPT4All
Task | Version | Metric | Value | Stderr | |
---|---|---|---|---|---|
arc_challenge | 0 | acc | 61.86 | ยฑ | 1.42 |
acc_norm | 64.51 | ยฑ | 1.40 | ||
arc_easy | 0 | acc | 85.06 | ยฑ | 0.73 |
acc_norm | 82.45 | ยฑ | 0.78 | ||
boolq | 1 | acc | 88.35 | ยฑ | 0.56 |
hellaswag | 0 | acc | 68.04 | ยฑ | 0.47 |
acc_norm | 85.12 | ยฑ | 0.36 | ||
openbookqa | 0 | acc | 37.80 | ยฑ | 2.17 |
acc_norm | 48.60 | ยฑ | 2.24 | ||
piqa | 0 | acc | 83.08 | ยฑ | 0.87 |
acc_norm | 83.95 | ยฑ | 0.86 | ||
winogrande | 0 | acc | 78.69 | ยฑ | 1.15 |
TruthfulQA
Task | Version | Metric | Value | Stderr | |
---|---|---|---|---|---|
truthfulqa_mc | 1 | mc1 | 44.55 | ยฑ | 1.74 |
mc2 | 60.86 | ยฑ | 1.57 |
Bigbench
Task | Version | Metric | Value | Stderr | |
---|---|---|---|---|---|
bigbench_causal_judgement | 0 | multiple_choice_grade | 58.95 | ยฑ | 3.58 |
bigbench_date_understanding | 0 | multiple_choice_grade | 66.40 | ยฑ | 2.46 |
bigbench_disambiguation_qa | 0 | multiple_choice_grade | 48.84 | ยฑ | 3.12 |
bigbench_geometric_shapes | 0 | multiple_choice_grade | 22.56 | ยฑ | 2.21 |
exact_str_match | 13.37 | ยฑ | 1.80 | ||
bigbench_logical_deduction_five_objects | 0 | multiple_choice_grade | 30.40 | ยฑ | 2.06 |
bigbench_logical_deduction_seven_objects | 0 | multiple_choice_grade | 20.57 | ยฑ | 1.53 |
bigbench_logical_deduction_three_objects | 0 | multiple_choice_grade | 52.00 | ยฑ | 2.89 |
bigbench_movie_recommendation | 0 | multiple_choice_grade | 44.40 | ยฑ | 2.22 |
bigbench_navigate | 0 | multiple_choice_grade | 52.10 | ยฑ | 1.58 |
bigbench_reasoning_about_colored_objects | 0 | multiple_choice_grade | 69.75 | ยฑ | 1.03 |
bigbench_ruin_names | 0 | multiple_choice_grade | 55.36 | ยฑ | 2.35 |
bigbench_salient_translation_error_detection | 0 | multiple_choice_grade | 23.65 | ยฑ | 1.35 |
bigbench_snarks | 0 | multiple_choice_grade | 77.35 | ยฑ | 3.12 |
bigbench_sports_understanding | 0 | multiple_choice_grade | 73.02 | ยฑ | 1.41 |
bigbench_temporal_sequences | 0 | multiple_choice_grade | 46.80 | ยฑ | 1.58 |
bigbench_tracking_shuffled_objects_five_objects | 0 | multiple_choice_grade | 22.08 | ยฑ | 1.17 |
bigbench_tracking_shuffled_objects_seven_objects | 0 | multiple_choice_grade | 19.03 | ยฑ | 0.94 |
bigbench_tracking_shuffled_objects_three_objects | 0 | multiple_choice_grade | 52.00 | ยฑ | 2.89 |
๐งฉ Configuration
base_model: mlabonne/Marcoro14-7B-slerp
experts:
- source_model: openchat/openchat-3.5-1210
positive_prompts:
- "chat"
- "assistant"
- "tell me"
- "explain"
- source_model: beowolx/CodeNinja-1.0-OpenChat-7B
positive_prompts:
- "code"
- "python"
- "javascript"
- "programming"
- "algorithm"
- source_model: maywell/PiVoT-0.1-Starling-LM-RP
positive_prompts:
- "storywriting"
- "write"
- "scene"
- "story"
- "character"
- source_model: WizardLM/WizardMath-7B-V1.1
positive_prompts:
- "reason"
- "math"
- "mathematics"
- "solve"
- "count"
๐ป Usage
Here's a notebook to run this model in 4-bit precision using a free T4 GPU on Google Colab.
!pip install -qU transformers bitsandbytes accelerate
from transformers import AutoTokenizer
import transformers
import torch
model = "mlabonne/Beyonder-4x7B-v2"
tokenizer = AutoTokenizer.from_pretrained(model)
pipeline = transformers.pipeline(
"text-generation",
model=model,
model_kwargs={"torch_dtype": torch.float16, "load_in_4bit": True},
)
messages = [{"role": "user", "content": "Explain what a Mixture of Experts is in less than 100 words."}]
prompt = pipeline.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
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"])
Output:
A Mixture of Experts (ME) is a machine learning technique that combines multiple expert models to make predictions or decisions. Each expert model is specialized in a different aspect of the problem, and their outputs are combined to produce a more accurate and robust solution. This approach allows the model to leverage the strengths of individual experts and compensate for their weaknesses, improving overall performance.
- Downloads last month
- 1,353
Model tree for mlabonne/Beyonder-4x7B-v2
Collection including mlabonne/Beyonder-4x7B-v2
Evaluation results
- normalized accuracy on AI2 Reasoning Challenge (25-Shot)test set Open LLM Leaderboard68.770
- normalized accuracy on HellaSwag (10-Shot)validation set Open LLM Leaderboard86.800
- accuracy on MMLU (5-Shot)test set Open LLM Leaderboard65.100
- mc2 on TruthfulQA (0-shot)validation set Open LLM Leaderboard60.680
- accuracy on Winogrande (5-shot)validation set Open LLM Leaderboard80.900
- accuracy on GSM8k (5-shot)test set Open LLM Leaderboard71.720