You can deploy this model and start earning money today!

RichardErkhov/
Kukedlc_-_NeuralKrishna-7B-V2-DPO-gguf

Name: Kukedlc_-_NeuralKrishna-7B-V2-DPO-gguf
Availability: InStock

endpoints_compatible

gguf

template

Views

No views yet

Quantization made by Richard Erkhov.

Github

Discord

Request more models

NeuralKrishna-7B-V2-DPO - GGUF

Model creator: https://huggingface.co/Kukedlc/
Original model: https://huggingface.co/Kukedlc/NeuralKrishna-7B-V2-DPO/

Name	Quant method	Size
NeuralKrishna-7B-V2-DPO.Q2_K.gguf	Q2_K	2.53GB
NeuralKrishna-7B-V2-DPO.IQ3_XS.gguf	IQ3_XS	2.81GB
NeuralKrishna-7B-V2-DPO.IQ3_S.gguf	IQ3_S	2.96GB
NeuralKrishna-7B-V2-DPO.Q3_K_S.gguf	Q3_K_S	2.95GB
NeuralKrishna-7B-V2-DPO.IQ3_M.gguf	IQ3_M	3.06GB
NeuralKrishna-7B-V2-DPO.Q3_K.gguf	Q3_K	3.28GB
NeuralKrishna-7B-V2-DPO.Q3_K_M.gguf	Q3_K_M	3.28GB
NeuralKrishna-7B-V2-DPO.Q3_K_L.gguf	Q3_K_L	3.56GB
NeuralKrishna-7B-V2-DPO.IQ4_XS.gguf	IQ4_XS	3.67GB
NeuralKrishna-7B-V2-DPO.Q4_0.gguf	Q4_0	3.83GB
NeuralKrishna-7B-V2-DPO.IQ4_NL.gguf	IQ4_NL	3.87GB
NeuralKrishna-7B-V2-DPO.Q4_K_S.gguf	Q4_K_S	3.86GB
NeuralKrishna-7B-V2-DPO.Q4_K.gguf	Q4_K	4.07GB
NeuralKrishna-7B-V2-DPO.Q4_K_M.gguf	Q4_K_M	4.07GB
NeuralKrishna-7B-V2-DPO.Q4_1.gguf	Q4_1	4.24GB
NeuralKrishna-7B-V2-DPO.Q5_0.gguf	Q5_0	4.65GB
NeuralKrishna-7B-V2-DPO.Q5_K_S.gguf	Q5_K_S	4.65GB
NeuralKrishna-7B-V2-DPO.Q5_K.gguf	Q5_K	4.78GB
NeuralKrishna-7B-V2-DPO.Q5_K_M.gguf	Q5_K_M	4.78GB
NeuralKrishna-7B-V2-DPO.Q5_1.gguf	Q5_1	5.07GB
NeuralKrishna-7B-V2-DPO.Q6_K.gguf	Q6_K	5.53GB
NeuralKrishna-7B-V2-DPO.Q8_0.gguf	Q8_0	7.17GB

Original model description:

license: apache-2.0 model-index:

name: NeuralKrishna-7B-V2-DPO 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: 74.06 name: normalized accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Kukedlc/NeuralKrishna-7B-V2-DPO 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: 88.97 name: normalized accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Kukedlc/NeuralKrishna-7B-V2-DPO 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: 64.41 name: accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Kukedlc/NeuralKrishna-7B-V2-DPO 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: 76.19 source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Kukedlc/NeuralKrishna-7B-V2-DPO 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: 84.29 name: accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Kukedlc/NeuralKrishna-7B-V2-DPO 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: 68.08 name: accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Kukedlc/NeuralKrishna-7B-V2-DPO name: Open LLM Leaderboard

Neural Krishna DPO

Fine-tuning + lnegth(choose)

Training Args:

python

# LoRA configuration
peft_config = LoraConfig(
    r=16,
    lora_alpha=16,
    lora_dropout=0.05,
    bias="none",
    task_type="CAUSAL_LM",
    target_modules=['k_proj', 'gate_proj', 'v_proj', 'up_proj', 'q_proj', 'o_proj', 'down_proj']
)

# Model to fine-tune
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype=torch.float16,
    load_in_4bit=True
)
model.config.use_cache = False



# Training arguments
training_args = TrainingArguments(
    per_device_train_batch_size=4,
    gradient_accumulation_steps=4,
    gradient_checkpointing=True,
    learning_rate=5e-5,
    lr_scheduler_type="cosine",
    max_steps=120,
    save_strategy="no",
    logging_steps=1,
    output_dir=new_model,
    optim="paged_adamw_32bit",
    warmup_steps=50,
    bf16=True,
    report_to="wandb",
)

# Create DPO trainer
dpo_trainer = DPOTrainer(
    model,
    args=training_args,
    train_dataset=dataset,
    tokenizer=tokenizer,
    peft_config=peft_config,
    beta=0.1,
    max_prompt_length=1024,
    max_length=1536,
)

# Fine-tune model with DPO
dpo_trainer.train()

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	76.00
AI2 Reasoning Challenge (25-Shot)	74.06
HellaSwag (10-Shot)	88.97
MMLU (5-Shot)	64.41
TruthfulQA (0-shot)	76.19
Winogrande (5-shot)	84.29
GSM8k (5-shot)	68.08

RichardErkhov/Kukedlc_-_NeuralKrishna-7B-V2-DPO-gguf

Original model description:

Neural Krishna DPO

Fine-tuning + lnegth(choose)

Open LLM Leaderboard Evaluation Results

RichardErkhov/
Kukedlc_-_NeuralKrishna-7B-V2-DPO-gguf