Commit Graph

3096 Commits

Author SHA1 Message Date
Georgi Gerganov
08737ef720 cuda : fix convert function (#1412) 2023-05-13 17:40:58 +03:00
Georgi Gerganov
bda4d7c215 make : fix PERF build with cuBLAS 2023-05-13 17:25:09 +03:00
Georgi Gerganov
5a5aeb1e91
llama : fix unused warning 2023-05-13 16:55:14 +03:00
Georgi Gerganov
66841fdb0e
ggml : multi-thread mul and diag_mask ops (#1428) 2023-05-13 16:48:03 +03:00
Johannes Gäßler
905d87b70a
ggml : GPU-accelerated token generation (#1412)
* CUDA kernel for q4_0 dequant. + mat. vec. mult.

* Added q4_1 via template

* Added missing __syncthreads();

* --gpu_layers -> --gpu-layers

* Shorter dequantize_mul_mat_vec line

* q5_0 dequantize_mul_mat kernel

* More readable dequantize_mul_mat_vec logic

* dequantize_mul_mat_vec kernels for q5_1, q8_0, f16

* llama : offload "output" tensor to GPU too + coding style fixes

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-05-13 16:38:36 +03:00
xaedes
f954edda93
ggml : implement backward pass for llama + small training-llama-from-scratch example (#1360)
* implement 8 of 14 missing backward pass operations used by llama

- GGML_OP_ADD_AT
- GGML_OP_CPY
- GGML_OP_MUL_MAT (src0.grad)
- GGML_OP_PERMUTE
- GGML_OP_RESHAPE
- GGML_OP_SCALE
- GGML_OP_TRANSPOSE
- GGML_OP_VIEW

implement additional ggml operation GGML_OP_ADD_AT, which is necessary for backward pass of GGML_OP_VIEW.

this operation adds src1 to src0 with data offset, i.e. to view(src0, ..., offset).
the values are return in a tensor size of src0. values outside of [data+offset:data+offset+nbytes(src1)] are just the original values from src0.

still missing backward passes for llama:

- GGML_OP_DIAG_MASK_INF
- GGML_OP_GET_ROWS
- GGML_OP_RMS_NORM
- GGML_OP_ROPE
- GGML_OP_SILU
- GGML_OP_SOFT_MAX

* implement 5 of 6 missing backward pass operations used by llama

- GGML_OP_DIAG_MASK_INF
- GGML_OP_GET_ROWS
- GGML_OP_RMS_NORM
- GGML_OP_SILU
- GGML_OP_SOFT_MAX

add necessary ggml operations GGML_OP_ADD1, GGML_OP_SILU_BACK, GGML_OP_RMS_NORM_BACK, GGML_OP_DIAG_MASK_ZERO, and GGML_OP_ROPE_BACK

GGML_OP_ADD1 is necessary to add a scalar value in the backward pass of GGML_OP_SOFT_MAX
GGML_OP_ADD1 could also be replaced by using GGML_OP_ADD and GGML_OP_REPEAT, but the performance would be worse. additionally GGML_OP_REPEAT will return unexpected value when the the input to GGML_OP_SOFT_MAX contains only a single scalar. in this case GGML_OP_REPEAT will not return the value that should be repeated (src1) but the value which shape the result should take (src0). So in this case it can not replace GGML_OP_ADD1.

GGML_OP_SILU_BACK, GGML_OP_RMS_NORM_BACK and GGML_OP_ROPE_BACK are necessary for backward pass of GGML_OP_SILU, GGML_OP_RMS_NORM and GGML_OP_ROPE. The backward pass for these functions cannot be easily composed of existing operations. Since the backward pass builds a computation graph we need operations forward pass implementations of the the required backward passes. Sounds a bit confusing at first, I know...

GGML_OP_DIAG_MASK_ZERO is necessary for backward pass of GGML_OP_DIAG_MASK_INF.

Some operations where previously inplace-only. for backward pass there needs to be non-inplace variants.
staying consistent with other operations that have non-inplace and inplace variants, the operations are changed to non-inplace and
functions with "_inplace" are added which are inplace.
in llama we need to call the inplace variants so that it is implemented as before.
for llama backward pass we need to use the non-inplace variants.

still not completely implemented backward passes for llama:

- GGML_OP_ROPE: needs forward pass for GGML_OP_ROPE_BACK
- GGML_OP_GET_ROWS: only necessary for tokenizer

* norm & rms_norm can not be threaded:

after investigation rms norm for quite some time I come to the conclusion that neither norm, nor rms_norm can be threaded, because we need mean over all items, not just of the slices each thread sees.

* remove already resolved TODO

* implement backward pass of ggml_rope and ggml_rope_back

* implement backward pass for ggml_get_rows and for new operation ggml_get_rows_back

* add test-grad0.c

* use GGML_PRINT_DEBUG for debug messages which will otherwise flood the console

* test both gradients of mul_mat

* disable graph dot export as it floods console

* bug fixes for silu_back

* successfully test silu backward

* bug fix for scale backward pass

use sum instead of mean for gradient of scalar scale parameter

* successfully test scale backward

* improve performance of sum backward pass

use add1(x,y) instead of add(x,repeat(y,x))

* improve performance of sqr backward pass

use scale(x,y) instead of mul(x,repeat(y,x))

* successfully test rope backward

* bug fix for cpy backward pass

* successfully test cpy backward

* bug fix for reshape backward pass

* successfully test reshape backward

* add test-opt.c

this uses ggml_opt to train a,b for minimal e=sum(sqr(c - a*b)) for random initial a,b,c

* correctly implement softmax backward pass using new operation ggml_diag

ggml_diag constructs diagonal matrices with entries.
ggml_diag(shape[a,1,c,d]) -> shape[a,a,c,d]

* successfully test soft_max backward

* align shape annotations

* add shape annotations for llama

* de-duplicate ggml_forward_dup code taking care of contiguous tensors of same type.

with this we can duplicate tensor of any typ as long as they are contiguous.

* fix ggml_compute_forward_dup_same_cont for when nelements < nthreads

when more threads are used than elements exist ie1 was less than ie0, resulting in invalid negative byte count argument in memcpy

* bug fix for add_at forward

required for view backward pass

src0 values must be copied to dst, because during addition we don't touch all dst elements in contrast to the normal add function.

* successfully test view backward

* minor code format improvement

* fix ggml_forward_add functions to work correctly with transposed tensors

uses the same logic as in ggml_compute_forward_add_q_f32, but make it consistent across all ggml_compute_forward_add_... functions.
this also slightly changes the mem access pattern of the different threads to works as in ggml_compute_forward_add_q_f32.

* fix ggml_forward_add1 functions to work correctly with transposed tensors

uses the same logic as in ggml_compute_forward_add1_q_f32, but make it consistent across all ggml_compute_forward_add1_... functions.
this also slightly changes the mem access pattern of the different threads to works as in ggml_compute_forward_add1_q_f32.

* test-grad0.c : add print_elements to help with debugging

* successfully test permute backward

* some minor test-grad0 fixes

* fix sub, mul and div functions to work correctly with transposed tensors

uses the same logic as in add

* implement ggml_cont backward pass

* successfully test transpose backward and permute for all permutations

also test sub, mul and div up to max n_dims

* test-grad0.c add TODO for view_2d and view_3d

add_at (required for view backward pass) is a bit tricky for n_dims > 1.

* fix comments

* successfully test diag_mask_inf and diag_mask_zero backward

* test-grad0 : fix test for div

nargs and ndims was swapped, corrupting the stack

* fix diag_mask to work with non-inplace input

* move dup call into the actual add_at functions

* fix get rows backward pass

* successfully test get_rows backward

* fix view backward pass

add nb parameters to add_at like in view.
together with offset they define how to view dst and src0 during the add_at operation.

* successfully test backward pass of view_1d, view_2d and view_3d

* fix backward pass for rms_norm

I would have used formulas from other frameworks, but they differed so I could not decide which is correct.
Instead it was derived here in comment using manual forward-backward automatic differention of rms_norm and simplification.

* successfully test backward pass of rms_norm

some tests may fail when gradients are large.
could not find a satisfying configuration to check for abs error and relative error that passes all tests while still actually testing the results with tight enough error bounds.
when looking at the values the "failed" tests look actually ok. for example:

rms_norm: ndims=2, i=0, k=2, x0=0.000153, xm=0.000053, xp=0.000253, f0=0.278594, f1=0.086213, g0=961.905457, g1=966.064941, eps=0.000100, error_abs=4.159485, error_rel=0.004324

it is due to the test logic in check_gradients that they fail.

* add todos for llama backward pass

- implementation for ADD1 backward pass should probably use sum instead of mean (but this backward pass is not required)
- repeat is not yet tested and looks like it only works for single element src0 inputs.

* add operation ggml_sum_rows

ggml_sum_rows(shape[a,b,c,d]) -> shape[1,b,c,d]

* add missing GGML_OP_SUM_ROWS

* fix backward pass for repeat

requires ggml_sum_rows

* successfully test backward pass of repeat

* update quantization types in switch-case of add_at and add1

* add baby-llama example training a very small llama model from scratch to output a sinusoidal wave.

had to increase maximum number of optimization parameters to train from scratch.

* fix softmax in baby-llama example

* switching from training with adam to lbfgs produces much better results in the baby-llama example

* train with two examples, creating new tensors each time..

* fix bug when using ggml_opt to optimize params in one context and use a renewable context for eval and opt

when not keeping gradients of model parameters they are overwritten by tensors created by opt, which may be invalid after opt context is renewed.
so we need to keep the original gradients and make dups for opt

* train on multiple examples, generate & print tokens with trained model afterwards

ctx0 for evaluation and optimization is renewed for each sample

* add ggml_reshape_1d, ggml_reshape_4d and ggml_view_4d

* fix soft_max backward pass for input->ne[1] != 1

* add ggml_log operation necessary for cross entropy loss

* add test for ggml_log gradients

* implement backward pass for ggml_sum_rows, necessary for cross entropy loss

* implement ggml_repeat support for rank > 2 tensors

* add test for ggml_sum_rows gradients

* fix training get_example_targets

predict the next token, not the current token!

* add square_error_loss and cross_entropy_loss functions

* optimize loss over multiple samples

this increases computation graph, need parallel batched forward for more efficiency.

* fix backward pass for add_at and change arguments to have same order as in view

* add ggml_set(ctx, a, b) to set b in view of a and return modified a

necessary to set values into kv_self cache and properly propagate the gradients

* fix kv_self gradients for training

use ggml_set instead of ggml_cpy to set kv_self cache with properly propagating gradients

* replace inplace operations for training with copying operations to allow gradient propagation

* add GGML_ASSERT to catch ggml_rope and back value errors

* add trainable lora-only model with all big matrices C split into A,B with A*B=C

this is not a lora-finetune, but the whole model changed to have only low-rank "lora" matrices.

training this instead of the normal model resulted in much worse results though...

* vastly improve training results

instead of logit targets 0 and 1 use -1 and +1.

* shorten code using a variable

* change name of GGML_OP_ADD_AT to GGML_OP_ACC

* smaller default values for baby llama model parameters

* update static assert of GGML_OP_COUNT

* remove shape annotations in llama_eval_internal

* revert disabling of threading for rms_norm and norm

* rename print functions in baby-llama example

* fix call to ggml_set_name

* add missing include for strcmp, etc

* remove trailing whitespace

* reduce number of test-grad0 iterations

avoid exceeding timeout of automated tests

* remove busy loop that was used as sleep for slower sinus wave generation

* disable slow tests grad0 and opt to avoid exceeding timeouts

* c++ in baby-llama example

use c++ includes instead of c includes
use std::min, std::max instead of MIN, MAX macros

* c++ in baby-llama example

use c++ includes instead of c includes
use std::min, std::max instead of MIN, MAX macros

* ggml : fix compiler warnings + cosmetic changes

* ggml : fix nullptr derefs in GGML_OP_CONT and GGML_OP_RESHAPE back

* swap arguments to vDSP_vdiv call

documentation for vDSP_vdiv states: "Note that B comes before A!"

* swap arguments to vDSP_vdiv call

documentation for vDSP_vdiv states: "Note that B comes before A!"

* ggml : swap vDSP_vsub args as per documentation

* add parallel batched forward function for baby-llama training

* cleanup code for batched training

* remove trailing whitespace

* minor : fix compiler warnings + indentation style

* ggml : fix null ptr deref in backward pass

* ggml : remove Q4_2 remnants

* ggml : fix clang-tidy warnings

* baby-llama : couple of clang-tidy warnings

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-05-13 15:56:40 +03:00
Georgi Gerganov
f048af0230
ggml : sync alibi fix from ggml repo 2023-05-13 11:54:33 +03:00
3ooabkhxtn
ac0cd259d5
Adding SSE instructions to ggml_vec_dot_q4_0_q8_0 (#1413) 2023-05-13 08:43:33 +00:00
Georgi Gerganov
0cd22e190a
llama : fix various warnings 2023-05-13 11:23:15 +03:00
Rinne
6456a4eb9f
embedding : remove unused code (#1426) 2023-05-13 10:24:20 +03:00
Georgi Gerganov
cdd5350892
readme : update Q4_0 perplexities
I think these were affected by the removal of the `round` during quantization
2023-05-13 09:12:44 +03:00
Georgi Gerganov
738ace394a
llama : free ggml context in set / copy state data (close #1425) 2023-05-13 09:08:52 +03:00
Henri Vasserman
699b1ad7fe
opencl : fix kernels for the new formats (#1422)
* Fix OpenCL kernels for the new formats

* Fix Q5_0 alignment issues.
2023-05-13 09:01:15 +03:00
Georgi Gerganov
fb62f92433
llama : fix --mtest option (close #1414) 2023-05-12 21:44:20 +03:00
Johannes Gäßler
773ee249fb
CLI args use - instead of _, backwards compatible (#1416) 2023-05-12 14:34:55 +00:00
slaren
553fd4d4b5
Add clang-tidy reviews to CI (#1407) 2023-05-12 15:40:53 +02:00
Rinne
089b1c93ba
readme : add C#/.NET bindings repo (#1409) 2023-05-12 08:39:40 +03:00
Georgi Gerganov
b9fd7eee57
ggml : remove bit shuffling (#1405)
* ggml : remove Q4_0 bit shufling (ARM NEON)

* ggml : remove Q4_1 bit shuffling (ARM NEON + reference)

* ggml : nibbles_from_floats() + bytes_from_nibbles() (ARM NEON)

* ggml : remove Q4_2 bit shuffling (WIP, BROKEN)

* ggml : remove Q5_0 bit shuffling (ARM NEON)

* ggml : 2x faster scalar implementations

* ggml : remove Q5_1 bit shuffling (ARM NEON + scalar)

* ggml : simplify scalar dot

* ggml : remove WASM SIMD bit shuffling + remove vzip for ARM 32-bit

* ggml : fix Q4_1 quantization

* ggml : update cuBLAS + normalize variable names

* ggml : remove Q4_2 mode

* ggml : minor formatting

* ggml : fix Q5_0 quantization

* scripts : add script for measuring the time per token

* AVX implementations (#1370)

* ggml : uniform 5th bit extraction

* llama : produce error upon loading old model files

* llama : fix model magic/version write

* ggml : speed-up Q5_0 + Q5_1 at 4 threads

* ggml : preserve old Q4 and Q5 formats

* ggml : simplify Q8_1 - no need for low / high sums anymore

* ggml : fix Q8_0 and Q8_1 rounding

* Revert "AVX implementations (#1370)"

This reverts commit 948d124837.

* ggml : fix AVX2 implementation

* sha : update hashes for 7B and 13B

* readme : update timings + remove warning banner

* llama : update v2 PR number to 1405

* ggml : fix WASM comments

* ggml : back to original bit order

* readme : add note that Q4 and Q5 have been changed

* llama : fix return for unknown version

---------

Co-authored-by: Stephan Walter <stephan@walter.name>
2023-05-12 00:23:08 +03:00
CRD716
b608b55a3e
prompts : model agnostic DAN (#1304)
* add model-agnostic dan prompt

* quick readme update

* save a token

* Revert "quick readme update"

This reverts commit 8dc342c069.
2023-05-11 18:10:19 +03:00
Evan Jones
cf348a60e0
main : add option to save full output to session (#1338)
* main : add option to save full output to session

* split behavior into --session and --prompt-cache

* restore original implementation with new names

* PR comments

* move the check for incompatible parameters to gpt_params_parse

* Fix whitespace

Co-authored-by: DannyDaemonic <DannyDaemonic@gmail.com>

---------

Co-authored-by: DannyDaemonic <DannyDaemonic@gmail.com>
2023-05-10 11:37:14 -04:00
DannyDaemonic
e6a46b0ed1
Locale fix for Windows (#1379) 2023-05-09 19:53:28 +02:00
Sami Farin
9f8dbc4787
use pause asm insn in busyloop to run the CPU (13600K) 10 °C cooler (#1314)
* use pause asm insn in busyloop to run the CPU (13600K) 10 °C cooler

Tested with a 13B model.

* use _mm_pause() in busyloop

* use _mm_pause() in busyloop on x86_64 to reduce power consumption
2023-05-09 14:29:20 +02:00
DannyDaemonic
41654efea8
Interface improvements and --multiline-input (previously --author-mode) (#1040)
* Interface improvements
* Multiline input
* Track character width
* Works with all characters and control codes + Windows console fixes
2023-05-08 19:45:48 -07:00
Georgi Gerganov
56551bc11f
readme : add notice about upcoming breaking change 2023-05-08 22:52:18 +03:00
AlpinDale
fe60904eef
readme : add TOC and Pygmalion instructions (#1359) 2023-05-08 19:33:30 +03:00
Pavol Rusnak
003ba2fb43
llama : fix hparams shadow (#1367)
fixes #1363
2023-05-08 17:48:21 +03:00
Georgi Gerganov
f9a6364912
llama : require first token to be BOS (#1303)
* llama : require first token to be BOS

* scripts : add ppl-run-all.sh

* perplexity : add BOS for each chunk

* readme : update perplexity values after BOS fix

* perplexity : add clarifying comments
2023-05-08 17:41:54 +03:00
ubik2
95078cc554
convert: add ability to convert safetensors files (#1276)
* when loading a safetensors file, ignore the metadata header
* check for safetensors files first, and only use PyTorch versions when safetensors aren't available
2023-05-08 13:54:26 +02:00
Johannes Gäßler
1f48b0abcf
Documented CUDA reproducibility, added warning (#1346) 2023-05-08 02:42:01 +02:00
Henri Vasserman
e1295513a4
CI: add Windows CLBlast and OpenBLAS builds (#1277)
* Add OpenCL and CLBlast support

* Add OpenBLAS support

* Remove testing from matrix

* change build name to 'clblast'
2023-05-07 13:20:09 +02:00
swittk
1b0fd45465
ggml : Allow usage of CLBlast alongside Accelerate.framework (#1336)
Minor edit in ggml.c which originally would prevent OpenCL from loading completely if GGML_USE_ACCELERATE was defined.
Minor speedup in prompt eval time.
2023-05-06 23:03:23 -04:00
Jed Fox
3924088512
Remove default arguments from sampling functions (#1343) 2023-05-06 17:01:47 -04:00
DaniAndTheWeb
173d0e6419
makefile: automatic Arch Linux detection (#1332)
This commit is a port of a detection method used in koboldcpp's Makefile in order to automatically set the -lcblas option on Arch Linux
2023-05-05 23:57:14 +02:00
Erik Scholz
a3b85b28da
ci : add cublas to windows release (#1271) 2023-05-05 22:56:09 +02:00
Pavol Rusnak
921dcee00a
readme: add missing info (#1324) 2023-05-05 16:43:36 +02:00
Ionoclast Laboratories
2d13786e91
Fix for OpenCL / clbast builds on macOS. (#1329) 2023-05-05 14:18:21 +02:00
Benjamin Lecaillon
a90e96b266
Convert.py @staticmethod (#1327)
* Line 698 has one #staticmethod and should not

otherwise throw error at unpickle.load() as not callable

* Update convert.py

---------

Co-authored-by: Ivan Stepanov <ivanstepanovftw@gmail.com>
2023-05-05 03:17:07 +03:00
slaren
94c5652fc0
quantize: make output filename optional, default to ggml-model-<ftype>.bin (#1301) 2023-05-05 00:58:56 +02:00
Ivan Stepanov
34d9f22f44
Wrap exceptions in std::exception to verbose output on exception. (#1316) 2023-05-04 18:56:27 +02:00
Ivan Stepanov
d3e8093e9b
convert: support DT_BF16 tensors (#1309)
Co-authored-by: Pavol Rusnak <pavol@rusnak.io>
2023-05-04 18:54:37 +02:00
44670
360cfe5bec
readme : add OpenBuddy link (#1321) 2023-05-04 19:33:31 +03:00
44670
2edbdb0f99
main : add --in-suffix option (#1318)
* adding --in-suffix option

* print input suffix before generation
2023-05-04 18:41:12 +03:00
Ron Jailall
20fbf2a2a0
ggml : change immintrin.h to intrin.h for compatibility (#1307)
* change immintrin.h to intrin.h for compatibility

Building on windows11 arm throws an error on this line. Seems like using intrin.h covers x86 and and arm

* conditional def of intrin.h

* fix typo in ggml.c
2023-05-04 18:05:59 +03:00
DannyDaemonic
db1080876a
Only escape prompts when used with -e (#1311) 2023-05-04 05:08:25 -07:00
DannyDaemonic
c65a7fbfa9
Update main's README.md with new features (#1296) 2023-05-04 03:02:59 -07:00
Tomas
f647ce040f
fix #1224 reverse prompt and multi line (#1297)
* fix reverse prompt and multi line

* Code Formatting

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

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-05-04 03:02:30 -07:00
Georgi Gerganov
799fdc1b5d
ggml : vectorize Q8_0 quantization
https://github.com/ggerganov/ggml/pull/127#issuecomment-1533648531
2023-05-03 23:24:20 +03:00
khimaros
6daa09d879
examples : read chat prompts from a template file (#1196) 2023-05-03 20:58:11 +03:00
Georgi Gerganov
bca9ad938a
minor : fix whitespaces (#1302) 2023-05-03 20:09:42 +03:00
Georgi Gerganov
e2a937ca6a
minor : fix trailing whitespaces 2023-05-03 18:43:23 +03:00