Commit Graph

57 Commits

Author SHA1 Message Date
Georgi Gerganov
812787cbc5 ggml : generalize GGML_OP_CONCAT (llama/7563)
* ggml : generalize GGML_OP_CONCAT (WIP)

ggml-ci

* tests : add dim != 2 tests

* metal : generalize concat kernel

* tests : naming

* cuda : generalize concat kernel

ggml-ci

* sycl : add warning and assert

* ggml : fix op params handling

* metal : bugfix kernel

ggml-ci

* ggml : reimplement CPU and Metal

* cuda : add asserts

ggml-ci

* ggml : fix ptrs

ggml-ci
2024-06-16 18:19:48 +03:00
Georgi Gerganov
7091c7ab5a metal : add GGML_OP_REPEAT kernels (llama/7557)
ggml-ci
2024-06-16 18:19:48 +03:00
Georgi Gerganov
d70ccb75f5 metal : disable FA kernel for HS=256 (llama/7556)
ggml-ci
2024-06-16 18:19:48 +03:00
Georgi Gerganov
45ddda8e0c ggml : drop support for QK_K=64 (llama/7473)
* ggml : drop support for QK_K=64

ggml-ci

* opencl : restore QK_K=256 define
2024-06-16 18:19:48 +03:00
liuwei-git
c9dcb75118 llama : add phi3 128K model support (llama/7225)
* add phi3 128k support in convert-hf-to-gguf

* add phi3 128k support in cuda

* address build warnings on llama.cpp

* adjust index value in cuda long rope freq factors

* add long rope support in ggml cpu backend

* make freq factors only depend on ctx size

* remove unused rope scaling type 'su' frin gguf converter

* fix flint warnings on convert-hf-to-gguf.py

* set to the short freq factor when context size is small than trained context size

* add one line of comments

* metal : support rope freq_factors

* ggml : update ggml_rope_ext API to support freq. factors

* backends : add dev messages to support rope freq. factors

* minor : style

* tests : update to use new rope API

* backends : fix pragma semicolons

* minor : cleanup

* llama : move rope factors from KV header to tensors

* llama : remove tmp assert

* cuda : fix compile warning

* convert : read/write n_head_kv

* llama : fix uninitialized tensors

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-06-16 18:19:48 +03:00
Georgi Gerganov
bbdbc3fc62 metal : handle F16 inf values, fix FA partial offload (llama/7434)
ggml-ci
2024-06-16 18:19:48 +03:00
John Balis
c4de1e19df ggml : add ggml_upscale_ext (ggml/814)
* initial commit with CPU implementation of upscale to shape and test, cuda implementation next

* experimental commit to see if dst shape is correct

* test version

* test

* removed unnecessary params

* refactor

* fixed tests

* ggml : metal impl + cleanup + sycl dev warnings

* patched ggml_upscale cuda op to handle non-contiguous tensors, added test for non-contiguous behavior

* metal : fix upsacle op to support nb00 + style

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-06-16 18:19:48 +03:00
Georgi Gerganov
1056ad762c
metal : support FA without mask + add asserts (llama/7278)
* ggml : fa without mask + add asserts

ggml-ci

* metal : support non-contiguous KV

ggml-ci
2024-05-14 19:16:29 +03:00
Georgi Gerganov
accada542a ggml : resolve merge (ggml/0)
ggml-ci
2024-05-13 11:02:26 +03:00
Georgi Gerganov
e54329da7b ggml : full ALiBi support (llama/7192)
* ggml : full ALiBi support

* ggml : update ggml_soft_max_ext() CUDA, SYCL

* ggml : ggml_flash_attn_ext() support ALiBi (CPU)

* ggml : ggml_flash_attn_ext() support ALiBi (Metal)

* ggml : fix warning

* ggml : ggml_flash_attn_ext() support ALiBi (CUDA)

ggml-ci

* ggml : fix assert message

* vulkan : add dev notes

* ggml : require mask when using ALiBi

ggml-ci

* convert : fix convert for refact models
2024-05-13 11:02:26 +03:00
Georgi Gerganov
1ae1a9cd56 metal : fix unused warning 2024-05-13 11:02:26 +03:00
Georgi Gerganov
156a33a990 ggml : add Flash Attention (llama/5021)
* ggml : add ggml_flash_attn_ext API

* ggml : fix GQA support in ggml_flash_attn_ext

* ggml : online attention (CPU)

* metal : initial implementation

* metal : f16 precision

* metal : reduce branches

* metal : specialize for head size

* wip : 8 rows per simd group

* wip : 4 rows per simd group

* wip : template for rows per warp

* metal : parallelize across KV size

* metal : parallel reduce across heads

* metal : efficient flash_attn_f16 implementation

* metal : avoid redundant loads of the attention

* metal : scale and mask in matrix form

* metal : fix comment

* llama : avoid ggml_cast, use F32 query

* metal : add parallel reduce version (disabled)

* metal : move output into local memory + optimize

- the result from each simdgroup now stays in the registers
- significantly reduced SRAM usage
- more efficient skipping of -INF blocks
- avoid simdgroup barrier in hot loop
- add comments

* metal : add tests, fix scaling, support C > 32

* metal : improve precision

* ggml : fix f16 mad

* metal : minor

* metal : support Q > 8

* tests : add ATTN tests

* metal : disable buffer allocation logs

* tests : more

* metal : faster inner loop for C == 32

* metal : fix array initialization

* tests : ifdef

* ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext

* ggml : fix ggml_soft_max mask requirement

* cuda : fix soft_max to use correct mask size

* cuda : add flash_attn kernel (wip)

* metal : optimize softmax for C > 32

* metal : optimize softmax

* tests : minor fix

* cuda : avoid zeroing fragments

* tests : update dims

* cuda : fix __hisinf() result check

* cuda : avoid warp_reduce for smax

* cuda : use int instead of int64_t

Noticeably improves performance (thanks to Johannes)

* cuda : make loops use the same loop values

Thanks Johannes again for the tip

* cuda : unroll some of the loops

* cuda : avoid __hisinf branches

* cuda : use half2 in softmax

* cuda : switch to 1 warp for bs > 16

* cuda : speed-up reduce part of the kernel

* cuda : unroll Q*K^T loop

* cuda : fix -INF block check

* cuda : simplify softmax

* cuda : fix matrix names

* cuda : minor

* llama : adapt to F16 KQ_pos

* llama : adapt new models to F16 KQ_mask

* ggml : fix F16 store (ARM NEON)

* llama : fix type of KQ_mask and KQ_pos

* ggml : fix CPU soft_max

* tests : add hs=256

* cuda : fix build

* metal : improve perf via smaller int registers

* cuda : adapt soft_max to F16 mask and pos

* CUDA: faster FlashAttention, kernel for bs == 1

* 16 cols for Phi-2

* no vec for hs, no hs==256 ncols==32 for Volta

* adjust kernel selection logic

* 4 warps, 256 stride for all D

* no ncols == 64

* Multiple parallel blocks for batch size 1

* fix compile warnings

* fix excessive KQ_b loads

* fix cmake build

* fix KV cache padding, NaN from INFINITY (llama/6438)

* llama : flash_attn cparam + fix defrag

* server: support flash_attn param

* server: bench: enable flash_attn param

* CUDA: refactor host code, dyn. par. blocks

* fix flash_attn_vec_f16 race condition

* flush softmax exp below threshold to 0

* store temp KQ in registers

* Calculate KQ as FP32 if KQV has GGML_PREC_F32

* Add __hgt2_mask implementation for CUDA 11

* fix KQ FP32 precision fpr parallel_blocks > 1

* llama-bench : add -fa,--flash-attn arg

* metal : add BS=1 kernel for flash attention (llama/6508)

* metal : add BS=1 kernel for flash attention (wip)

* metal : support more than 1 warps

* metal : opts

* metal : opt

* metal : switch to parallel reduce

* metal : reduce registers

* metal : simplify

* metal : initial FA vec kernel

* metal : use F32 attention accumulators

* batched-bench : add fattn arg

* llama : simplify llama_build_kv_store

ggml-ci

* llama : adapt build_olmo to changes

* ggml : fix arm fp16 store on windows

* metal : clean-up

* metal : clean-up kernel code

* metal : minor

* tests : remove benchmarks

ggml-ci

* ggml : fix avx512 const correctness

ggml-ci

* ggml : fix soft_max with bias on CPU

ggml-ci

* common : print --flash-attn in help

* ggml : fix num dimensions in ggml_flash_attn_ext

* llama : force disable flash attention for incompatible models

* ggml : ggml_soft_max support F16/F32 mask/pos

ggml-ci

* cuda : uint -> uint32_t

* cuda : "constexpr dim3" -> "const dim3"

ggml-ci

* cuda : try to fix __hgt2_mask

ggml-ci

* ggml : add TODO's for F16/F32 mask/pos support in other backends

* llama : replace bool need_kq_pos with use_alibi

* llama : prep ALiBi support for BERT models

ggml-ci

* llama : fix n_batch requirements

ggml-ci

* cont

* server : add help for --flash-attn arg

* llama : disable FA for AMD

* tests : remove TMP_ATTN_BENCH

ggml-ci

* llama : support save/load state with FA enabled

ggml-ci

* ci : add CUDA save-load-state tests

ggml-ci

* llama : llama_kv_cache_clear zeroes data + fix save-load seq

ggml-ci

* llama : fix copy-paste errors, add TODO

* llama : disallow incompatible states

* llama : update llama_state_get_size after v_trans field

* metal : remove tmp log

* llama : add static reminder for llama_state_get_size

* metal : fix max nsg

ggml-ci

* ci : fix arg order

ggml-ci

---------

Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-05-13 11:02:26 +03:00
slaren
c96b0a938e ggml : group all experts in a single ggml_mul_mat_id (llama/6505)
* ggml : group all experts in a single ggml_mul_mat_id
cuda : improve mmid row copy

* cuda : fix bin bcast with non-cont src0

* test-backend-ops : only run all mul mat tests for base types

* llama : disable moe offloading with SYCL

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-05-13 11:02:26 +03:00
Shijie
fdb2c87350 llama : add qwen2moe (llama/6074)
* support qwen2moe

* fix-review

* metal : support unary ops for nelements % 4 != 0

* metal : require contiguousness for float4 unary kernels

* metal : require contiguousness for float4 unary kernels (cont)

* fix-review

* names : for brevity "SHARED_EXP" -> "SHEXP"

* llama : reuse build_moe_ffn()

* llama : add model type name

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-05-13 11:02:26 +03:00
Dave
9d6d50d933 Added support for GGML_OP_CLAMP in Metal (llama/6662)
* Added support for GGML_OP_CLAMP in Metal

* Corrected size

---------

Co-authored-by: dave-fl <dave@Davids-MacBook-Pro.local>
2024-05-13 11:02:26 +03:00
slaren
00a0947c65 metal : unify mul_mv_id kernels (llama/6556) 2024-05-13 11:02:26 +03:00
Justina Cho
37e6757453 feat: implemented sigmoid function (ggml/806)
* added sigmoid function

* implemented metal kernel for sigmoid

* implemented cuda kernel for sigmoid

* added sigmoid unary op and incremented count
2024-05-13 11:02:26 +03:00
slaren
1dce94cf26
ggml : mul_mat_id use the same tensor for all the experts (llama/6387)
* ggml : update mul_mat_id to use the same tensor for all the experts

* update cuda

* minor

* update metal

* update test-backend-ops

* fix cuda

* Update ggml-metal.m

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

* update convert.py

* update convert-hf-to-gguf.py

* update convert.py for mixtral hf models

* Update convert-hf-to-gguf.py

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

* cuda : support non-pow-2 number of experts

* allow quantize to work for split and merged experts models in the same way

* cleanup + disable mmap automatically with split tensors models

* update imatrix

* test-backend-ops : test qwen argsort

* update grok model loading

* llama : add merged experts tensors to the grok tensor map

* minor

* gguf : bump version

* fix quantizing of merged experts

* convert-hf-to-gguf.py : update grok (untested)

* make linter happy

* cuda/argsort : use shared memory instead of pool memory

* convert : fix grok tensor names

* metal : add support for non-pow-2 argsort

* llama : more loader cleanup, better error checking

* cuda : fix warning

* llama : still use mmap for loading old models, but copy the data to a host buffer

* add review note

* llama : remove ffn tensor counting + add sanity check

ggml-ci

* convert : fix handling of n_experts == None

ggml-ci

* imatrix : fix ncall counters

* llama : produce error if imatrix size does not match

* quantize : terminate on errors + trace logs

ggml-ci

* metal : pad shared memory to 16 bytes

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-07 16:15:57 +03:00
Georgi Gerganov
2948c740a2
sync : ggml (#2001)
* sync : update scripts

* sync : ggml

* talk-llama : sync llama.cpp

* make : WHISPER_CUBLAS -> WHISPER_CUDA

* ci : try to fix sycl build

* talk-llama : fix make build
2024-03-27 18:55:10 +02:00
Georgi Gerganov
c8709d4604
metal : build metallib + fix embed path (llama/6015)
* metal : build metallib + fix embed path

ggml-ci

* metal : fix embed build + update library load logic

ggml-ci

* metal : fix embeded library build

ggml-ci

* ci : fix iOS builds to use embedded library
2024-03-15 14:01:14 +02:00
Georgi Gerganov
46e3c3f112
ggml : reuse quantum structs across backends (llama/5943)
* ggml : reuse quant blocks across backends

ggml-ci

* ggml : define helper constants only for CUDA and SYCL

ggml-ci

* ggml : define helper quantum constants for SYCL

ggml-ci
2024-03-15 14:01:13 +02:00
Kawrakow
9dc60fc02d
1.5 bit: we can do even better (llama/5999)
* iq1_s: we can do even better

Spent one of the 4 scale bits on a signs of a 0.125 shift.
I.e., quants are now -1 + delta, delta, 1 + delta, where delta
is +/- 0.125.

CUDA works, same performance as before.
PPL(LLaMA-v2-7B) is now 11.85!

* iq1_s: make scalar and AVX2 work with the new version

* iq1_s: make Neon work with new version.

~10% drop in performance, so will need some more work.

* iq1_s: make Metal work with new version

* iq1_s: very slightly faster dequantize on Metal

* iq1_s: fix dequantize on the CPU

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-03-15 14:01:13 +02:00
Kawrakow
f79d0d4f74
Better 1.5 bit quantization (llama/5971)
* Trying blocvks of 16 for IQ1_S - seems slightly better

* iq1s_blocks16: Adjust scale fudge factor to 1.125

* iq1s_blocks16: going to blocks of 32

with 2048 lattice points, so same bpw.
This is even better than blocks of 16.
Should I try blocks of 64? But to keep the same
bpw, when I go to 4096 lattice points, I need to
remove blocks alltogether and just have superblocks of
256 weights.

* iq1s_blocks16: Use 2*<x^2> as sigma2 in weight adjustment

* iq1s_blocks16: scalar and AVX2 dot products

* iq1s_blocks16: CUDA dot product

* iq1s_blocks16: Metal works, Neon does not

Metal works but TG is dog slow (35 t/s). PP is OKish (493 t/s).
Not seeing the bug in the Neon implementation for now.

* iq1s_blocks16: fixed Neon

* iq1s_blocks16: very slightly faster TG on Metal

Still pathetic at 37 t/s

* iq1s_blocks16: speedup Metal by packing codebook into uint32_t's

* Formatting

* iq1s_blocks16: uint32_t codebook is also better in CUDA

TG-128 is now 204 t/s up from 194 t/s.
PP-512 is 5890 t/s, so significantly better than other quants

* iq1s_blocks16: slightly faster Neon dot product

* iq1s_blocks16: faster AVX2 dot product

* iq1s_blocks16: adjust to ggml-common.h

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-03-15 14:01:12 +02:00
Georgi Gerganov
7bdb1de9ec
metal : move mm_id indices to shared mem (llama/5982) 2024-03-15 14:01:12 +02:00
Georgi Gerganov
24eba5a2ff
ggml : add ggml-common.h to deduplicate shared code (llama/5940)
* ggml : add ggml-common.h to shared code

ggml-ci

* scripts : update sync scripts

* sycl : reuse quantum tables

ggml-ci

* ggml : minor

* ggml : minor

* sycl : try to fix build
2024-03-15 14:01:12 +02:00
Kawrakow
21d295180d
ggml : IQ3_S improvements (llama/5829)
* iq3_s: somewhat faster AVX2 dot product

On Ryzen a 7950X TG-128 increases to 16 t/s from 15.5 t/s using
16 threads. For 8 threads it is 13.85 t/s vs 11.75 t/s.
PP-512 increases to 28.5 t/s from 23.8 t/s.

* iq3_s: somewhat faster ARM_NEON dot product

Still dog slow - 10.7 t/s up from 9.9 t/s.

* iq3_s: another small ARM_NEON improvement

10.7 -> 11.0 t/s. Using vmulq_s8 is faster than the xor - sub trick
that works best on AVX2.

* iq3_s: minor improvement on Metal

49.4 t/s -> 50.3 t/s

* iq3_s: PPL improvement

E.g., for a context of 4096 LLaMA-v2-7B goes to 5.1340 from 5.1653.

* iq3_s: use new grid everywhere

* Fix ARM_NEON

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-03-08 11:38:32 +02:00
leejet
8cee7c08b6
add some new ops, fix some operators and add batch operations to certain operators. (ggml/747)
* cuda: fix group_norm

* cuda: add batch inference support for ggml_pad/ggml_upscale

* add ggml_arrange

* add ggml_timestep_embedding

* update ggml_arange/ggml_timestep_embedding tests

* cuda: fix im2col

* add ggml_arange/ggml_timestep_embbeding support for metal backend

* fix some bugs

* fix some bugs

* Update ggml.h

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

* Update ggml-cuda.cu

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

* Update ggml-metal.m

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

* Update ggml-metal.m

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

* Update ggml-metal.metal

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

* modify according to the review comments

* ggml : fix compile warnings + code style

* ggml : normalize compute_forward calls + fix seg fault in debug

* minor

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
Co-authored-by: slaren <slarengh@gmail.com>
2024-03-08 11:38:31 +02:00
Kawrakow
2533305596
ggml : make i-quants work with super-blocks of 64 (CPU,Metal) (llama/5760)
* WIP: make i-quants work for QK_K = 64

* iq2_xs: attempt to fix AVX dot product for QK_K = 64

Tests pass, but I get gibberish.

* QK_K = 64 tests pass on ARM_NEON and Metal

Sadly, that does not mean it actually works.

* Make CUDA compile with QK_K = 64

Tests don't pass, plus we get misaligned access

* Q2_K: fixed bug in imatrix quantization for QK_K = 64

* iq1_s: turn off SIMD implementation for QK_K = 64 (it does not work)

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-28 13:00:30 +02:00
Kawrakow
013e394a4b
IQ4_XS: a 4.25 bpw quantization (llama/5747)
* Try IQ4_NL with blocks of 64 - does not look good

* iq4_xs: go to super-blocks of 256 and 6-bit scales for blocks of 32

* iq4_xs: CUDA works - 133.2 t/s

* iq4_xs: AVX2 dot product

* iq4_xs: ARM_NEON dot product

* iq4_nl: Metal implementation

As usual, Metal / Apple Silicon don't like my quants.

* iq3_xs: minor fix

* iq4_xs: shrink by using IQ3_S for attn_k and attn_q

* iq4_xs: revert using IQ3_S for attn_k and attn_v

PPL vs size is good, but CPU performance suffers: on M2 Max
TG-128 drops to 21.7 t/s from 28.8, and on a Ryzen-7950X
to 14.5 t/s from 15.8 t/s. On CUDA we have 135 t/s when
using IQ3_S vs 133 t/s with pure IQ4_XS.

* Fix CI

* iq4_xs: Added forgotten check for 256 divisibility

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-28 13:00:29 +02:00
Kawrakow
7b1d8ea7e0
Adding IQ2_S and IQ2_M to complete coverage of the 2-3 bit quantization range (llama/5721)
* Adding IQ2_S and IQ2_M as a single cumulative commit

* Update examples/quantize/quantize.cpp

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

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-28 13:00:29 +02:00
Kawrakow
9e7c5212a1
IQ3_S: a much better alternative to Q3_K (llama/5676)
* iq4_nl: squash commits for easier rebase

* Basics (quantize, dequantize)
* CUDA dequantize and dot product
* Slightly faster CUDA dot product (120 t/s)
* Switch to 6-bit scales
* Scalar dot product
* AVX2 dot product
* ARM_NEON dot product
* Works on metal, but still slow
* Slightly better Metal dot product
* Another small Metal improvement
* Metal dot product is getting there
* Faster CUDA dot product
* Add 1/8 ffn_down layers as Q5_K when no imatrix has been provided
* Report the actual bpw
* Add _xs mix that is 4.05 bpw for non-MoE models
* Remove IQ4_XS for now, slightly adjust kvalues_iq4nl
* AVX2 dot product uses Q8_0 instead of Q8_K
* Add to test-backend-ops
* Minor fix
* Also use use Q5_K for attn_output in MoE models
* Fixes after merging latest master
* Switching to blocks of 32
* AVX2 for blocks of 32
* Scaler dot product for blocks of 32
* ARM_NEON dot product for blocks of 32
* Metal kernels for blocks of 32
* Slightly faster Metal kernels

* Resurrecting iq3_xs

After all the experimentation, nothing was better than this.

* Minor PPL improvement via a block scale fudge factor

* Minor improvement via 3 neighbours

* iq3_xs: working scalar and AVX2 dot products

* iq3_xs: ARM_NEON dot product - works but extremely slow (10 t/s)

* iq3_xs: working Metal implementation

* Adding IQ3_M - IQ3_XS mix with mostly Q4_K

* iiq3_xs: a 3.4375 bpw variant

* iq3_xs: make CUDA work for new version

* iq3_xs: make scalar and AVX2 work for new version

* iq3_s: make ARM_NEON work with new version

* iq3_xs: make new version work on metal

Performance is very similar to Q3_K_S

* iq3_xs: tiny Metal speed improvement

* iq3_xs: tiny Metal speed improvement

* Fix stupid warning

* Q3_K_XS now uses a mix of IQ3_XS and IQ3_XXS

* iq3_xs: rename to iq3_s

* iq3_s: make tests pass

* Move Q3_K_XS mix to 3.25 bpw

* Attempt to fix failing tests

* Another attempt to fix the Windows builds

* Attempt to fix ROCm

* ROCm again

* iq3_s: partial fix for QK_K = 64

* iq3_s: make it work on metal for QK_K = 64

Pleasent surprise: the coding was super-block size independent,
so all it took was to delete some QK_K == 256 guards.

* Will this fix ROCm?

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-25 19:58:46 +02:00
Georgi Gerganov
ce411498f6
sync : llama.cpp (ggml/0)
ggml-ci
2024-02-22 15:12:36 +02:00
slaren
5ec1e0edfa
cuda, metal : fix nans in soft_max (llama/5574)
* cuda : fix nans in soft_max

* metal : fix nans in soft_max

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-19 15:53:24 +02:00
Georgi Gerganov
0775374750
metal : fix unused warnings (llama/0) 2024-02-19 15:53:24 +02:00
Kawrakow
2c1ad21ba8
1.5 bit quantization (llama/5453)
* iq1_s: WIP basics

* iq1_s: CUDA is working

* iq1_s: scalar CPU dot product

* iq1_s: WIP AVX2 dot product - something is not right

* Fix tests

* Fix shadow warnings

* Fix after merge with latest master

* iq1_s: AVX2 finally works

* iq1_s: ARM_NEON dot product. Works, but not very fast

* iq1_s: better grid

* iq1_s: use IQ2_XXS for attn_output

At a cost of 0.04 extra bpw this gives a big improvement in PPL.

* iq1_s: Metal basics

Dequantize works, but not dot product

* iq1_s: Metal works, but quite slow

As usual, Apple Silicon does not like the code I write.

* iq1_s: Tests

* iq1_s: slightly faster dot product

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-19 15:53:23 +02:00
Georgi Gerganov
eca5ff9868
ggml : add ALiBi support for ggml_soft_max_ext (llama/5488) 2024-02-19 15:53:23 +02:00
Georgi Gerganov
9b735cea77
metal : add im2col F32 dst support (llama/5132) 2024-02-10 09:55:46 +02:00
Kawrakow
361ecebe90
ggml : fix IQ3_XXS on Metal (llama/5219)
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-30 21:28:00 +02:00
Kawrakow
6adb969b09
SOTA 3-bit quants (llama/5196)
* iq3_xxs: quantize/dequantize

RMSE seems a bit high-ish at about half-way between q2_K and
q3_K, so need to check more.

* iq3_xxs: CUDA dequantize works

* iq2_xxs: tuning quantization

* iq3_xxs: starting to look better

PPL on wiki.test.raw
LLaMA-v1-7B: 6.4218
LLaMA-v2-7B: 6.3560
Mistral-7B : 6.0717

This is better than Q3_K_XS, with a 5% reduction in quantized model
size.

* iq3_xxs: CUDA dot product

We have
PP-512: 5891 t/s
TG-128: 143.9 t/s

* iq3_xxs: scalar and AVX2 dot products

* iq3_xxs: ARM_NEON and Metal

Metal performance is decent, ARM_NEON is pathetic

* iq3_xxs: slightly better grid points

* Faster iq3_xxs and iq2_xs dot products on CUDA

* iq3_xxs: add some quant mix

* iq3_xxs: fix failing quantization test

Dot product still fails. Is this real?

* iq3_xxs: hopefully fix ROCm

* iq3_xxs: failing tests

This time the dot product accuracy did find an actual bug
in the AVX2 implementation.

* Add IQ3_XXS to test-backend-ops

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-30 21:27:59 +02:00
Kawrakow
97b12212dd
ggml : SOTA 2-bit quants (add IQ2_XS) (llama/4856)
* iq2_xs: basics

* iq2_xs: this should have been in the basics

* iq2_xs: CUDA and scalar CPU works

* iq2_xs: WIP Metal

* iq2_xs: Metal now works

* iq2_xs: working, but dog slow, ARM_NEON dot product

* iq2_xs: better ARM_NEON dot product

We are now at 19.5 t/s for TG-128 and 61 t/s for PP-512 when
running on the CPU.

* iq2_xs: AVX2 dot product - 19.5 t/s

* iq2_xs: faster AVX2 dit product

21.4 t/s for TG-128, 59.2 t/s for PP-512.
The latter is 2x compared to the previous version.

* iq2_xs: had forgotten to delete iq2-data.h

* Add llama enum for IQ2_XS

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-11 21:50:01 +02:00
Georgi Gerganov
a0a64a19dd
metal : improve dequantize precision to match CPU (llama/4836)
ggml-ci
2024-01-11 21:50:01 +02:00
Kawrakow
10651bddf6
SOTA 2-bit quants (llama/4773)
* iq2_xxs: basics

* iq2_xxs: scalar and AVX2 dot products

Needed to change Q8_K to have quants in the -127...127 range,
else the IQ2_XXS AVX implementation becomes very awkward.
The alternative would have been to use Q8_0 instead. Perhaps
I'll change later, for now this is what we have.

* iq2_xxs: ARM_NEON dot product

Somehow strangely slow (112 ms/token).

* iq2_xxs: WIP Metal

Dequantize works, something is still wrong with the
dot product.

* iq2_xxs: Metal dot product now works

We have
PP-512 = 475 t/s
TG-128 = 47.3 t/s

Not the greatest performance, but not complete garbage either.

* iq2_xxs: slighty faster dot product

TG-128 is now 48.4 t/s

* iq2_xxs: slighty faster dot product

TG-128 is now 50.9 t/s

* iq2_xxs: even faster Metal dot product

TG-128 is now 54.1 t/s.

Strangely enough, putting the signs lookup table
into shared memory has a bigger impact than the
grid values being in shared memory.

* iq2_xxs: dequantize CUDA kernel - fix conflict with master

* iq2_xxs: quantized CUDA dot product (MMVQ)

We get TG-128 = 153.1 t/s

* iq2_xxs: slightly faster CUDA dot product

TG-128 is now at 155.1 t/s.

* iq2_xxs: add to llama ftype enum

* iq2_xxs: fix MoE on Metal

* Fix missing MMQ ops when on hipBLAS

I had put the ggml_supports_mmq call at the wrong place.

* Fix bug in qequantize_row_iq2_xxs

The 0.25f factor was missing.
Great detective work by @ggerganov!

* Fixing tests

* PR suggestion

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-11 21:50:00 +02:00
Georgi Gerganov
4cdd9aad9b metal : add kernel_get_rows_i32
ggml-ci
2024-01-03 14:43:51 +02:00
Georgi Gerganov
f38c057503 metal : optimize ggml_mul_mat_id (faster Mixtral PP) (llama/4725)
* ggml : disable fast-math for Metal (cmake build only)

ggml-ci

* metal : fix Metal API debug warnings

* cmake : add -fno-inline for Metal build (llama/4545)

* metal : fix API debug warnings

* metal : fix compile warnings

* metal : use uint64_t for strides

* cmake : rename option to LLAMA_METAL_SHADER_DEBUG

* metal : fix mat-vec Q8_0 kernel for BS > 1

* metal : normalize mat-vec kernel signatures

* cmake : respect LLAMA_QKK_64 option

* metal : fix mat-vec Q4_K kernel for QK_K == 64

* metal : optimizing ggml_mul_mat_id (wip)

* metal : minor fix

* metal : opt mul_mm_id
2024-01-03 14:43:51 +02:00
Georgi Gerganov
1e5544b39b metal : enable shader debugging (cmake option) (llama/4705)
* ggml : disable fast-math for Metal (cmake build only)

ggml-ci

* metal : fix Metal API debug warnings

* cmake : add -fno-inline for Metal build (llama/4545)

* metal : fix API debug warnings

* metal : fix compile warnings

* metal : use uint64_t for strides

* cmake : rename option to LLAMA_METAL_SHADER_DEBUG

* metal : fix mat-vec Q8_0 kernel for BS > 1

* metal : normalize mat-vec kernel signatures

* cmake : respect LLAMA_QKK_64 option

* metal : fix mat-vec Q4_K kernel for QK_K == 64

ggml-ci
2024-01-03 14:43:51 +02:00
Georgi Gerganov
3a5302108d
sync : ggml (ggml_scale, ggml_row_size, etc.) (#1677)
* sync : ggml

* sync : llama.cpp

* talk-llama : fix obsolete param

* ggml-alloc : fix ggml_tallocr_is_own

* talk.wasm : update to new ggml

* ggml : fix type punning in ggml_scale

* ggml : cuda jetson + arm quants warnings
2023-12-22 17:53:39 +02:00
Georgi Gerganov
8171e621fc
sync : ggml (Metal fixes, new ops, tests) (#1633)
* sync : ggml (Metal fixes, new ops, tests)

* cuda : fix bin bcast when src1 and dst have different types
2023-12-13 21:55:03 +02:00
Georgi Gerganov
afce6fa113
sync : ggml (new ops, new backend, etc) (#1602)
* sync : ggml (new ops, new backend, etc)

* whisper : remove obsolete broadcasting code

* ggml : remove backend self-registers + fix ggml_concat + n_task logic

* metal : fix assert

* metal : print resource path

* whisper : fix bug if metal init fails
2023-12-07 22:27:19 +02:00
Georgi Gerganov
b0502836b8
whisper : add full CUDA and Metal offloading (#1472)
* whisper : migrate to ggml-backend

* whisper : fix logit reading

* whisper : fix tensor allocation during load

* whisper : fix beam-search with CUDA

* whisper : free backends + fix compile warning

* whisper : print when CUDA is enabled

* whisper : fix CoreML

* make : clean-up

* talk : fix compile warning

* whisper : support ggml_conv with CUDA and Metal (#1473)

* ggml : add CUDA support for ggml_conv

* whisper : remove ggml_repeat for conv bias + single backend

* cuda : fix im2col kernel

* metal : add im2col support + mul mat-vec f16 x f16

* bench-all : add q4 models

* whisper : clean-up

* quantize-all : fix

* ggml : im2col opts

* whisper : avoid whisper_model_data wrapper

* whisper : add note that ggml_mul_mat_pad does not work with CUDA

* whisper : factor out graph compute in common function

* whisper : fixes

* whisper : fix UB with measure buffers

* whisper : try to fix the parallel whisper_state functionality (#1479)

* whisper : try to fix the parallel whisper_state functionality

* whisper : fix multi-state Metal

* whisper : free backend instances in whisper_state
2023-11-12 15:31:08 +02:00
Georgi Gerganov
f96e1c5b78
sync : ggml (backend v2, k-quants, CUDA opts, Metal opts, etc.) (#1422)
* sync : ggml (backend v2, k-quants, CUDA opts, Metal opts, etc.)

* metal : allow env metal variable to override resource path (#1415)

* Allow env variable to override resource path

* Update ggml-metal.m

---------

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

* sync : restore common / main from `master`

* sync : restore whisper from `master`

* talk-llama : update to latest llama.cpp

* ruby : fix build

* ggml : fix 32-bit ARM build

* ggml : fix MIN / MAX macro collisions + update ios bindings

* ggml : fix ifdefs and MIN / MAX again

* exampels : fix Obj-C and Swift examples

* ggml : fix 32-bit ARM compatibility

* ggml : one more attempt to fix 32-bit ARM compat

* whisper : fix support for larger graphs

---------

Co-authored-by: Chris Raethke <codesoda@users.noreply.github.com>
2023-11-03 21:35:05 +02:00