Modelpublic

RichardErkhov/Sparkoo_-_KateAI50m-gguf

sourceHugging Faceupdated 2y agoView on Hugging Face

0likes89downloads

Model Card

Quantization made by Richard Erkhov.

Github

Discord

Request more models

KateAI50m - GGUF

—Model creator: https://huggingface.co/Sparkoo/
—Original model: https://huggingface.co/Sparkoo/KateAI50m/

Name	Quant method	Size
KateAI50m.Q2_K.gguf	Q2_K	0.04GB
KateAI50m.IQ3_XS.gguf	IQ3_XS	0.04GB
KateAI50m.IQ3_S.gguf	IQ3_S	0.04GB
KateAI50m.Q3_K_S.gguf	Q3KS	0.04GB
KateAI50m.IQ3_M.gguf	IQ3_M	0.04GB
KateAI50m.Q3_K.gguf	Q3_K	0.04GB
KateAI50m.Q3_K_M.gguf	Q3KM	0.04GB
KateAI50m.Q3_K_L.gguf	Q3KL	0.04GB
KateAI50m.IQ4_XS.gguf	IQ4_XS	0.04GB
KateAI50m.Q4_0.gguf	Q4_0	0.05GB
KateAI50m.IQ4_NL.gguf	IQ4_NL	0.05GB
KateAI50m.Q4_K_S.gguf	Q4KS	0.05GB
KateAI50m.Q4_K.gguf	Q4_K	0.05GB
KateAI50m.Q4_K_M.gguf	Q4KM	0.05GB
KateAI50m.Q4_1.gguf	Q4_1	0.05GB
KateAI50m.Q5_0.gguf	Q5_0	0.05GB
KateAI50m.Q5_K_S.gguf	Q5KS	0.05GB
KateAI50m.Q5_K.gguf	Q5_K	0.05GB
KateAI50m.Q5_K_M.gguf	Q5KM	0.05GB
KateAI50m.Q5_1.gguf	Q5_1	0.05GB
KateAI50m.Q6_K.gguf	Q6_K	0.06GB
KateAI50m.Q8_0.gguf	Q8_0	0.07GB

Original model description: --- language:

—en pipeline_tag: text-generation ---

Warning!!

This model is in the process of being moved from gpt2 architecture -> a custom architecture. Note that it may not work at certain times because of the moving. Thank you for understanding.

Kate

This is a custom model for text generation.

Model Details

—model_type: GPT2*

GPT2

This model is NOT A FINETUNE!!. It uses the GPT2 architecture but it doesnt finetune it.

python

# Model configuration for a smaller GPT-2 style model
    config = GPT2Config(
        vocab_size=50257,  # Standard GPT-2 vocabulary size
        n_positions=512,   # Maximum sequence length
        n_ctx=512,        # Context window size
        n_embd=512,       # Embedding dimension
        n_layer=6,        # Number of transformer layers
        n_head=8,         # Number of attention heads
        bos_token_id=50256,
        eos_token_id=50256,
        pad_token_id=50256,
        _name_or_path=""  # Empty to ensure no pretrained weights are loaded
    )

    # Initialize model with random weights
    model = GPT2LMHeadModel(config)