Distilbert Base Uncased Go Emotions Student GGUF

--- license: mit base_model: joeddav/distilbert-base-uncased-go-emotions-student tags: - gguf - llama.cpp - text-classification - emotion pipeline_tag: text-classification --- # distilbert-base-uncased-go-emotions-student-GGUF GGUF conversion of [joeddav/distilbert-base-uncased-go-emotions-student](https://huggingface.co/joeddav/distilbert-base-uncased-go-emotions-student), a DistilBERT model distilled from a zero-shot classification pipeline into the 28 GoEmotions labels, for use with [llama.cpp](https://github.com/ggml-org/llama.cpp). ## Files | File | Quant | Size | |---|---|---| | goemotions-f16.gguf | F16 | 129 MiB | | goemotions-Q8_0.gguf | Q8_0 | 70 MiB | | goemotions-Q6_K.gguf | Q6_K | 54 MiB | | goemotions-Q4_K_M.gguf | Q4_K_M | 45 MiB | ## Usage Run with `llama-embedding` using rank pooling to get the 28 classifier logits: ``` llama-embedding -m goemotions-Q8_0.gguf --pooling rank --embd-normalize -1 -p "your text here" ``` Note: `--embd-normalize` defaults to 2 (L2 normalization), which is meaningless for classifier logits — always pass `--embd-normalize -1` to get raw scores.