Quantization made by Richard Erkhov.

L-MChat-7b - GGUF

Model creator: https://huggingface.co/Artples/
Original model: https://huggingface.co/Artples/L-MChat-7b/

Name	Quant method	Size
L-MChat-7b.Q2_K.gguf	Q2_K	2.53GB
L-MChat-7b.IQ3_XS.gguf	IQ3_XS	2.81GB
L-MChat-7b.IQ3_S.gguf	IQ3_S	2.96GB
L-MChat-7b.Q3_K_S.gguf	Q3_K_S	2.95GB
L-MChat-7b.IQ3_M.gguf	IQ3_M	3.06GB
L-MChat-7b.Q3_K.gguf	Q3_K	3.28GB
L-MChat-7b.Q3_K_M.gguf	Q3_K_M	3.28GB
L-MChat-7b.Q3_K_L.gguf	Q3_K_L	3.56GB
L-MChat-7b.IQ4_XS.gguf	IQ4_XS	3.67GB
L-MChat-7b.Q4_0.gguf	Q4_0	3.83GB
L-MChat-7b.IQ4_NL.gguf	IQ4_NL	3.87GB
L-MChat-7b.Q4_K_S.gguf	Q4_K_S	3.86GB
L-MChat-7b.Q4_K.gguf	Q4_K	4.07GB
L-MChat-7b.Q4_K_M.gguf	Q4_K_M	4.07GB
L-MChat-7b.Q4_1.gguf	Q4_1	4.24GB
L-MChat-7b.Q5_0.gguf	Q5_0	4.65GB
L-MChat-7b.Q5_K_S.gguf	Q5_K_S	4.65GB
L-MChat-7b.Q5_K.gguf	Q5_K	4.78GB
L-MChat-7b.Q5_K_M.gguf	Q5_K_M	4.78GB
L-MChat-7b.Q5_1.gguf	Q5_1	5.07GB
L-MChat-7b.Q6_K.gguf	Q6_K	5.53GB
L-MChat-7b.Q8_0.gguf	Q8_0	7.17GB

Original model description:

license: apache-2.0 tags: - merge - mergekit - Nexusflow/Starling-LM-7B-beta - FuseAI/FuseChat-7B-VaRM base_model: - Nexusflow/Starling-LM-7B-beta - FuseAI/FuseChat-7B-VaRM model-index: - name: L-MChat-7b 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: 65.61 name: normalized accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Artples/L-MChat-7b 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: 84.59 name: normalized accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Artples/L-MChat-7b 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: 65.44 name: accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Artples/L-MChat-7b 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: 50.94 source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Artples/L-MChat-7b 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: 81.37 name: accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Artples/L-MChat-7b 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: 69.45 name: accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Artples/L-MChat-7b name: Open LLM Leaderboard

L-MChat-7b

L-MChat-7b is a merge of the following models:

Configuration

slices:
  - sources:
      - model: Nexusflow/Starling-LM-7B-beta
        layer_range: [0, 32]
      - model: FuseAI/FuseChat-7B-VaRM
        layer_range: [0, 32]
merge_method: slerp
base_model: FuseAI/FuseChat-7B-VaRM
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
dtype: bfloat16

Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "Artples/M-LChat-7b"
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"])

License

Apache 2.0 but you cannot use this model to directly compete with OpenAI.

How?

Usage of LazyMergekit.

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	69.57
AI2 Reasoning Challenge (25-Shot)	65.61
HellaSwag (10-Shot)	84.59
MMLU (5-Shot)	65.44
TruthfulQA (0-shot)	50.94
Winogrande (5-shot)	81.37
GSM8k (5-shot)	69.45