diff options
| author | Kawrakow <48489457+ikawrakow@users.noreply.github.com> | 2024-09-09 14:56:34 +0300 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2024-09-09 14:56:34 +0300 |
| commit | 8c86231f9306c81dc291c4c4a16f88bbc7c97793 (patch) | |
| tree | d49325de2775076e1f71ddf94667d0cd02db3cc5 /src/llama.cpp | |
| parent | bf4b19b474b78a6ddfa1f0fe19f76f3c7ac92030 (diff) | |
Adding IQ1_TN - 1.6875 bpw for TriLM ternary models (#44)
* Adding iq1_tn - 1.6875 bpw for TriLM ternary models
* iq1_tn: NEON
* iq1_tn: faster NEON
* iq2_bn: improve performance on NEON
We now get TG-128 = 100 t/s for Bitnet-3B-1.58b!
* iq1_tn: improve AVX2
PP-512 goes to 533 t/s up from 455.
TG-128 @ 2 threads goes to 16.6 t/s up from 14.2.
However, we seem to have a bottleneck somewhere as
TG saturates at 8 threads.
* iq1_tn: improve Zen4
PP-512 goes to 485 t/s up from 352. With FA we get 545 t/s up from 380.
TG-128 @ 1 thread goes to 12.4 t/s up from 10.4.
However, we seem to have a bottleneck somewhere as
TG saturates at 8 threads.
* iq2_bn: improve on Zen4
We now get PP-512 = 614 t/s up from 542 t/s
* iq2_bn: improve AVX2 implementation
We now get PP-512 = 753 t/s up from 680 t/s.
* Remove unnecessary barrier in ggml_compute_forward_mul_mat
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
Diffstat (limited to 'src/llama.cpp')
| -rw-r--r-- | src/llama.cpp | 10 |
1 files changed, 7 insertions, 3 deletions
diff --git a/src/llama.cpp b/src/llama.cpp index 768aafa7..bb9b6848 100644 --- a/src/llama.cpp +++ b/src/llama.cpp @@ -3788,6 +3788,7 @@ struct llama_model_loader { case GGML_TYPE_IQ1_M: ftype = LLAMA_FTYPE_MOSTLY_IQ1_M; break; case GGML_TYPE_IQ1_BN: ftype = LLAMA_FTYPE_MOSTLY_IQ1_BN; break; case GGML_TYPE_IQ2_BN: ftype = LLAMA_FTYPE_MOSTLY_IQ2_BN; break; + case GGML_TYPE_IQ1_TN: ftype = LLAMA_FTYPE_MOSTLY_IQ1_TN; break; case GGML_TYPE_IQ2_TN: ftype = LLAMA_FTYPE_MOSTLY_IQ2_TN; break; case GGML_TYPE_IQ4_NL: ftype = LLAMA_FTYPE_MOSTLY_IQ4_NL; break; case GGML_TYPE_IQ4_XS: ftype = LLAMA_FTYPE_MOSTLY_IQ4_XS; break; @@ -4497,8 +4498,9 @@ static std::string llama_model_ftype_name(llama_ftype ftype) { case LLAMA_FTYPE_MOSTLY_IQ5_K: return "IQ5_K - 5.5 bpw"; case LLAMA_FTYPE_MOSTLY_IQ6_K: return "IQ6_K - 6.6 bpw"; case LLAMA_FTYPE_MOSTLY_IQ1_BN: return "IQ1_BN - 1.625 bpw Bitnet"; + case LLAMA_FTYPE_MOSTLY_IQ1_TN: return "IQ1_TN - 1.6875 bpw TriLM"; case LLAMA_FTYPE_MOSTLY_IQ2_BN: return "IQ2_BN - 2.00 bpw Bitnet"; - case LLAMA_FTYPE_MOSTLY_IQ2_TN: return "IQT_BN - 2.06 bpw TriLM"; + case LLAMA_FTYPE_MOSTLY_IQ2_TN: return "IQ2_TN - 2.06 bpw TriLM"; case LLAMA_FTYPE_MOSTLY_IQ3_S: return "IQ3_S - 3.4375 bpw"; case LLAMA_FTYPE_MOSTLY_IQ3_M: return "IQ3_S mix - 3.66 bpw"; case LLAMA_FTYPE_MOSTLY_Q4_0_4_4: return "Q4_0_4_4"; @@ -15644,7 +15646,7 @@ static ggml_type llama_tensor_get_type(quantize_state_internal & qs, ggml_type n else if (ftype == LLAMA_FTYPE_MOSTLY_IQ1_BN || ftype == LLAMA_FTYPE_MOSTLY_IQ2_BN) { new_type = GGML_TYPE_IQ4_NL; } - else if (ftype == LLAMA_FTYPE_MOSTLY_IQ2_TN) { + else if (ftype == LLAMA_FTYPE_MOSTLY_IQ1_TN || ftype == LLAMA_FTYPE_MOSTLY_IQ2_TN) { new_type = GGML_TYPE_Q4_K; } else if (new_type == GGML_TYPE_Q4_0_4_4 || new_type == GGML_TYPE_Q4_0_4_8 || @@ -15856,7 +15858,7 @@ static ggml_type llama_tensor_get_type(quantize_state_internal & qs, ggml_type n new_type == GGML_TYPE_IQ3_XXS || new_type == GGML_TYPE_IQ1_S || new_type == GGML_TYPE_IQ3_S || new_type == GGML_TYPE_IQ1_M || new_type == GGML_TYPE_IQ4_K || new_type == GGML_TYPE_IQ2_K || new_type == GGML_TYPE_IQ5_K || new_type == GGML_TYPE_IQ3_K || new_type == GGML_TYPE_IQ2_TN || - new_type == GGML_TYPE_IQ6_K) { + new_type == GGML_TYPE_IQ6_K || new_type == GGML_TYPE_IQ1_TN) { int nx = tensor->ne[0]; int ny = tensor->ne[1]; if (nx % QK_K != 0) { @@ -15881,6 +15883,7 @@ static ggml_type llama_tensor_get_type(quantize_state_internal & qs, ggml_type n case GGML_TYPE_IQ3_S: case GGML_TYPE_IQ1_S: case GGML_TYPE_IQ1_M: + case GGML_TYPE_IQ1_TN: case GGML_TYPE_IQ2_TN: case GGML_TYPE_Q2_K: case GGML_TYPE_Q3_K: @@ -15991,6 +15994,7 @@ static void llama_model_quantize_internal(const std::string & fname_inp, const s case LLAMA_FTYPE_MOSTLY_IQ1_M: default_type = GGML_TYPE_IQ1_M; break; case LLAMA_FTYPE_MOSTLY_IQ1_BN: default_type = GGML_TYPE_IQ1_BN; break; case LLAMA_FTYPE_MOSTLY_IQ2_BN: default_type = GGML_TYPE_IQ2_BN; break; + case LLAMA_FTYPE_MOSTLY_IQ1_TN: default_type = GGML_TYPE_IQ1_TN; break; case LLAMA_FTYPE_MOSTLY_IQ2_TN: default_type = GGML_TYPE_IQ2_TN; break; case LLAMA_FTYPE_MOSTLY_IQ4_NL: default_type = GGML_TYPE_IQ4_NL; break; case LLAMA_FTYPE_MOSTLY_IQ4_XS: default_type = GGML_TYPE_IQ4_XS; break; |