attn.output.tensor of FYPE IQ3_M in IQ4_XS

If FTYPE IQ4_XS has attn.output.tensor in IQ4_XS (4.5BPW), there's no reason to have FTYPE IQ3_M to have attn.output.tensor in Q4_K (4.5BPW).
In terms of perplexity, on a Llama 3.1 70b model, the proposed change reduces the size by 1%, and increases the preplexity by 0.25%.
This commit is contained in:
Nexes the Old 2024-08-04 11:59:52 +02:00 committed by GitHub
parent d5779c27ba
commit 93c35f86a9
No known key found for this signature in database
GPG key ID: B5690EEEBB952194

View file

@ -15486,7 +15486,7 @@ static ggml_type llama_tensor_get_type(quantize_state_internal & qs, ggml_type n
else if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_XXS) new_type = GGML_TYPE_IQ3_S;
else if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_M ) new_type = GGML_TYPE_Q4_K;
else if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_L ) new_type = GGML_TYPE_Q5_K;
else if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_M ) new_type = GGML_TYPE_Q4_K;
else if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_M ) new_type = GGML_TYPE_IQ4_XS;
}
} else {
if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_L) new_type = GGML_TYPE_Q4_K;