vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
770 commits 380 branches 3056 tags 365 MiB
Commit graph

770 commits

This branch
This branch
All branches
Author SHA1 Message Date
Iwan Kawrakow
ce19b965f0 k_quants: switch Q4_K to 4-bit scales when QK_K = 64 
Here the loss in accuracy is greater than for Q3_K,
 but the Q4_K points still move further to the left on
 the perplexity vs size curve.
 2023-06-26 12:58:32 +03:00
Iwan Kawrakow
aeefd4e781 k_quants: swicth Q3_K to 4-bit scales when QK_K = 64 
Otherwise there isn't much benefit from this
quantization type. There is some very slight loss
in accuracy, but we reduce size by ~7%.
E.g., for OpenLLaMA-3B, Q3_K_S perplexity is
8.6131 with 8-bit scales and 8.6352 with 4-bit,
while file size decreases from 1.53G to 1.44G.
 2023-06-26 12:58:32 +03:00
Iwan Kawrakow
88412a1aa0 Simplify via lambda 2023-06-26 12:58:32 +03:00
Iwan Kawrakow
333ffcc5ba Fixed bug in q4_K quantization added with the 64-block addition 2023-06-26 12:58:32 +03:00
Iwan Kawrakow
558a19427b k_quants: correctly define QK_K in llama.cpp 2023-06-26 12:58:32 +03:00
Iwan Kawrakow
8b98d01e31 k_quants: call them _K, not _k, also on Metal 2023-06-26 12:58:32 +03:00
Iwan Kawrakow
285eeb1531 k_quants: WIP super-blocks with 64 weights 
Q5_K works on Metal and is slightly faster
than QK_K = 256 (23.7 ms vs 26.3 ms).
 2023-06-26 12:58:32 +03:00
Iwan Kawrakow
ff83e32c6a k_quants: WIP super-blocks with 64 weights 
Q3_K works on Metal and is slightly faster
than QK_K = 256 (26.6 ms vs 28.3 ms).
 2023-06-26 12:58:32 +03:00
Iwan Kawrakow
6081a65527 k_quants: WIP super-blocks with 64 weights 
Q2_K works on Metal and is very slightly faster
than QK_K = 256 (23.8 ms vs 24.2 ms).
 2023-06-26 12:58:32 +03:00
Iwan Kawrakow
167a0bbe34 k_quants: WIP super-blocks with 64 weights 
Q4_K works on Metal and is actually slightly faster
than QK_K = 256 (21.95 ms vs 24.0 ms).
 2023-06-26 12:58:32 +03:00
Iwan Kawrakow
e1bbcfc5cb k_quants: WIP super-blocks with 64 weights 
* We are able to pass preprocessor macros to the Metal
  compiler
* Q6_K works and is actually slightly more efficient than
  the QK_K = 256 version (25.2 ms vs 25.8 ms)
 2023-06-26 12:58:32 +03:00
Iwan Kawrakow
fae24afd01 k_quants: WIP super-blocks with 64 weights 
Yet another speedup for Q5_K on ARM_NEON.
We are now within 10% of the QK_K = 256 version.
 2023-06-26 12:58:32 +03:00
Iwan Kawrakow
d92c5a9e29 k_quants: WIP super-blocks with 64 weights 
Another small improvement for Q3_K and Q5_K on ARM_NEON
 2023-06-26 12:58:32 +03:00
Iwan Kawrakow
2ff543c147 k_quants: WIP super-blocks with 64 weights 
Slightly more efficient Q3_K and Q5_K
 2023-06-26 12:58:32 +03:00
Iwan Kawrakow
9d27d8d0ea k_quants: WIP super-blocks with 64 weights 
Q5_K working on ARM_NEON, but quite a bit slower than 256 weights.

With that, we have full support for ARM_NEON, although
performance is not quite there.
 2023-06-26 12:58:32 +03:00
Iwan Kawrakow
2b2a13c4f9 k_quants: WIP super-blocks with 64 weights 
Q3_K working on ARM_NEON, but quite a bit slower than 256 weights.
 2023-06-26 12:58:32 +03:00
Iwan Kawrakow
80c75fe821 k_quants: WIP super-blocks with 64 weights 
Q2_K working on ARM_NEON, but quite a bit slower than 256 weights
 2023-06-26 12:58:32 +03:00
Iwan Kawrakow
cda47a6b2f k_quants: WIP super-blocks with 64 weights 
Q4_K working on ARM_NEON, but quite a bit slower than 256 weights
 2023-06-26 12:58:32 +03:00
Iwan Kawrakow
03f30c8eca k_quants: WIP super-blocks with 64 weights 
Q6_K working on ARM_NEON
 2023-06-26 12:58:32 +03:00
Iwan Kawrakow
3bd9ae79d8 k_quants: WIP super-blocks with 64 weights 
Q5_K working on CUDA, and with this CUDA is done.
 2023-06-26 12:58:32 +03:00
Iwan Kawrakow
460dd841b1 k_quants: WIP super-blocks with 64 weights 
Q3_K working on CUDA.
 2023-06-26 12:58:29 +03:00
Iwan Kawrakow
41e46ec1c2 k_quants: WIP super-blocks with 64 weights 
Q2_K working on CUDA. ~3% slower on GTX-1660,
10% slower on 4080.
 2023-06-26 12:55:35 +03:00
Iwan Kawrakow
5aae4b8d4f k_quants: WIP super-blocks with 64 weights 
Q4_K working on CUDA. ~10% slower on GTX-1660,
16% slower on 4080.
 2023-06-26 12:52:57 +03:00
Iwan Kawrakow
c6c35366bf k_quants: WIP super-blocks with 64 weights 
Q6_K working on CUDA. Cannot make it run quite as gast as
with super-blocks with 256 weigths: 8% slower on 4080,
20% slower on the 1660 (but there we fit 1 less layer on the
GPU because pf the larger model size), so some fraction of
these 20% is due to that,
 2023-06-26 12:42:36 +03:00
Iwan Kawrakow
bcf8c5c384 k_quants: WIP super-blocks with 64 weights 
Q5_K scalar and AVX2 works, and with that all
k_quants are done on AVX2 and scalar
 2023-06-26 12:42:36 +03:00
Iwan Kawrakow
2b2ab31a89 k_quants: WIP super-blocks with 64 weights 
Q3_K scalar and AVX2 works.
 2023-06-26 12:42:36 +03:00
Iwan Kawrakow
aebd5471e9 k_quants: WIP super-blocks with 64 weights 
Q2_K scalar and AVX2 works. Q2_K is way too slow (it is actually slower
than the scalar implementation)
 2023-06-26 12:42:36 +03:00
Iwan Kawrakow
1f6195c2f2 k_quants: WIP super-blocks with 64 weights 
Q4_K scalar and AVX2 works
 2023-06-26 12:42:36 +03:00
Iwan Kawrakow
9fe2a2b1db k_quants: WIP super-blocks with 64 weights 
Q6_K scalar and AVX2 works
 2023-06-26 12:42:36 +03:00
Iwan Kawrakow
d2f12ac354 k_quants: WIP super-blocks with 64 weights 2023-06-26 12:42:36 +03:00
Georgi Gerganov
447ccbe8c3
readme : add new roadmap + manifesto 2023-06-25 16:08:12 +03:00
Georgi Gerganov
bd34cdde38
ggml : sync latest ggml (custom operators) 2023-06-25 14:25:08 +03:00
anon998
c2a08f87b8
fix server sampling: top k sampler first (#1977) 
Co-authored-by: anon <anon@example.org>
 2023-06-25 10:48:36 +02:00
Georgi Gerganov
66a2555ba6
readme : add Azure CI discussion link 2023-06-25 09:07:03 +03:00
sjinzh
e65ca7e14a
zig : upgrade build system support (#1981) 
* upgrade zig build system support

* zig : add new line at the end of the file

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
 2023-06-25 08:45:44 +03:00
Robyn
5ec8dd5a3c
#1869 Fix null reference errors when training from scratch with CUDA (#1907) 
* #1869 Fix null reference errors when training from scratch with CUDA build

Calling ggml_compute_forward when node->src0 was null was causing train-text-from-scratch.exe to terminate unexpectedly.

* ggml : do not dereference src0 if NULL

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
 2023-06-24 20:10:29 +02:00
Georgi Gerganov
65bdd52a86
tests : sync test-grad0 from ggml 2023-06-24 19:40:18 +03:00
Rowan Hart
fdd1860911
flake : fix ggml-metal.metal path and run nixfmt (#1974) 2023-06-24 14:07:08 +03:00
AN Long
c943d823c1
convert : fix invalid params in write_vocab_only (#1975) 2023-06-24 14:02:06 +03:00
slaren
f2c754e1c3
ggml : improve ggml_graph_dump_dot, add ggml_format_name (#1978) 
* Improve ggml_graph_dump_dot, add ggml_format_name

* add more automatic names to view ops

* fix name of copies
 2023-06-24 13:57:18 +03:00
Georgi Gerganov
11da1a85cd
readme : fix whitespaces 2023-06-24 13:38:18 +03:00
Alberto
235b610d65
readme : fixed termux instructions (#1973) 2023-06-24 13:32:13 +03:00
Alex Renda
b061ba9e2a
llama : fix top-p sampling to match the canonical definition (#1953) 
* Fix top-p sampling to match the standard definition (smallest set that has probability mass at least p, not largest set with probability mass less than p)

* top-p: correct gt to gte

* add test for correct top-p behavior
 2023-06-24 13:15:01 +03:00
Didzis Gosko
527b6fba1d
llama : make model stateless and context stateful (llama_state) (#1797) 
* llama : make model stateless and context stateful

* llama : minor cleanup

* llama : update internal API declaration

* Apply suggestions from code review

fix style

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

* Missing model memory release

* Fix style

* Add deprecated warning for public API function llama_init_from_file

* Update public API use cases: move away from deprecated llama_init_from_file

* Deprecate public API function llama_apply_lora_from_file

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
 2023-06-24 11:47:58 +03:00
eiery
d7b7484f74
Add OpenLLaMA instructions to the README (#1954) 
* add openllama to readme
 2023-06-23 10:38:01 +02:00
Erik Scholz
7487137227
rework convert.py to read hyper-parameters from config.json (#1958) 
* Read hyper-parameters from HuggingFace-transformer config.json, if they exist, and fall back to guessing, like before otherwise.
  This allows converting open_llama 3B and other non-standard model designs.
 2023-06-22 14:20:47 +02:00
Johannes Gäßler
bbca06e269
cmake: revert CUDA arch default to 52, 61 if f16 (#1959) 2023-06-21 23:49:25 +02:00
Rahul Vivek Nair
fb98254f99
Fix typo in README.md (#1961) 2023-06-21 23:48:43 +02:00
Georgi Gerganov
049aa16b8c
readme : add link to p1 2023-06-20 19:05:54 +03:00
Xiake Sun
2322ec223a
Fix typo (#1949) 2023-06-20 15:42:40 +03:00