diff options
| author | Kawrakow <iwankawrakow@gmail.com> | 2024-10-02 15:22:13 +0300 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2024-10-02 15:22:13 +0300 |
| commit | cce49832c1b81b4e535e78ff308417ef3a386b18 (patch) | |
| tree | 33b10f9344f4656d58cd3ea068233ba75888498d /src/llama.cpp | |
| parent | d6909ed6f00f91f20c9ef628085a1a1a6a55c453 (diff) | |
Adding Q6_0 (#77)
* Adding q6_0 - basics + AVX2/Zen4 working
* Adding q6_0: CUDA dequantize works, but not mmvq
* Adding q6_0: CUDA mmvq works
* Adding q6_0: CUDA cpy, so Q6_0 can be used for KV-cache
* Add q6_0 to CPU flash attention
Disappointing result: for LlaMA-3.2-1B, q6_0 K- and V-cache
gives about the same PPL as q8_0 K-cache and q4_0 V-cache,
while needing the exact same RAM.
I.e., what was the point?
* q6_0: slightly better kv-cache result
Better than q8_0+q4_0, but not as good as q8_0+iq4_nl
* q6_0: works on ARM_NEON
* q6_0: dequantize works on Metal, but not vector dot product
* q6_0: it now works on Metal
Outperforms q5_0 by a significant margin. E.g.
| model | size | params | backend | ngl | threads | test | t/s |
| ------------------------------ | ---------: | ---------: | ---------- | --: | ------: | ------------: | ---------------: |
| llama 8B Q6_0 | 6.08 GiB | 8.03 B | Metal | 100 | 4 | tg128 | 44.02 ± 0.08 |
| llama 8B Q5_0 | 5.21 GiB | 8.03 B | Metal | 100 | 4 | tg128 | 40.13 ± 0.12 |
| llama 8B Q6_0 | 6.08 GiB | 8.03 B | Metal | 100 | 4 | pp512 | 500.55 ± 0.32 |
| llama 8B Q5_0 | 5.21 GiB | 8.03 B | Metal | 100 | 4 | pp512 | 448.02 ± 0.27 |
* q6_0: can now be used for kv-cache on Metal
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
Diffstat (limited to 'src/llama.cpp')
| -rw-r--r-- | src/llama.cpp | 3 |
1 files changed, 3 insertions, 0 deletions
diff --git a/src/llama.cpp b/src/llama.cpp index dca03ade..eb982125 100644 --- a/src/llama.cpp +++ b/src/llama.cpp @@ -3774,6 +3774,7 @@ struct llama_model_loader { case GGML_TYPE_Q4_1: ftype = LLAMA_FTYPE_MOSTLY_Q4_1; break; case GGML_TYPE_Q5_0: ftype = LLAMA_FTYPE_MOSTLY_Q5_0; break; case GGML_TYPE_Q5_1: ftype = LLAMA_FTYPE_MOSTLY_Q5_1; break; + case GGML_TYPE_Q6_0: ftype = LLAMA_FTYPE_MOSTLY_Q6_0; break; case GGML_TYPE_Q8_0: ftype = LLAMA_FTYPE_MOSTLY_Q8_0; break; case GGML_TYPE_Q2_K: ftype = LLAMA_FTYPE_MOSTLY_Q2_K; break; case GGML_TYPE_Q3_K: ftype = LLAMA_FTYPE_MOSTLY_Q3_K_M; break; @@ -4471,6 +4472,7 @@ static std::string llama_model_ftype_name(llama_ftype ftype) { case LLAMA_FTYPE_MOSTLY_Q4_1: return "Q4_1"; case LLAMA_FTYPE_MOSTLY_Q5_0: return "Q5_0"; case LLAMA_FTYPE_MOSTLY_Q5_1: return "Q5_1"; + case LLAMA_FTYPE_MOSTLY_Q6_0: return "Q6_0"; case LLAMA_FTYPE_MOSTLY_Q8_0: return "Q8_0"; case LLAMA_FTYPE_MOSTLY_Q2_K: return "Q2_K - Medium"; case LLAMA_FTYPE_MOSTLY_Q2_K_S: return "Q2_K - Small"; @@ -15967,6 +15969,7 @@ static void llama_model_quantize_internal(const std::string & fname_inp, const s case LLAMA_FTYPE_MOSTLY_Q4_1: default_type = GGML_TYPE_Q4_1; break; case LLAMA_FTYPE_MOSTLY_Q5_0: default_type = GGML_TYPE_Q5_0; break; case LLAMA_FTYPE_MOSTLY_Q5_1: default_type = GGML_TYPE_Q5_1; break; + case LLAMA_FTYPE_MOSTLY_Q6_0: default_type = GGML_TYPE_Q6_0; break; case LLAMA_FTYPE_MOSTLY_Q8_0: default_type = GGML_TYPE_Q8_0; break; case LLAMA_FTYPE_MOSTLY_F16: default_type = GGML_TYPE_F16; break; case LLAMA_FTYPE_MOSTLY_BF16: default_type = GGML_TYPE_BF16; break; |