* server tests : remove seemingly redundant newlines in print()
* server tests : use built-in subprocess features, not os.kill and psutil
* server tests : do not catch e.g. SystemExit; use print_exc
* server tests: handle TimeoutExpired exception
* server tests: fix connect on dual-stack systems
* server: tests: add new tokens regex on windows generated following new repeat penalties default changed in (#6127)
* server: tests: remove the hack on windows since now we get the good socket family
* server: tests: add new tokens regex following new repeat penalties default changed in (#6127)
* server: tests: add new tokens regex following new repeat penalties default changed in (#6127)
---------
Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
* llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs
ggml-ci
* server : add -ub, --ubatch-size parameter
* fix server embedding test
* llama : fix Mamba inference for pipeline parallelism
Tested to work correctly with both `main` and `parallel` examples.
* llama : limit max batch size to n_batch
* add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism
default increase to 4 (from 2)
changing this value may improve performance for some systems, but increases memory usage
* fix hip build
* fix sycl build (disable cpy_tensor_async)
* fix hip build
* llama : limit n_batch and n_ubatch to n_ctx during context creation
* llama : fix norm backend
* batched-bench : sync after decode
* swiftui : sync after decode
* ggml : allow ggml_get_rows to use multiple threads if they are available
* check n_ubatch >= n_tokens with non-casual attention
* llama : do not limit n_batch to n_ctx with non-casual attn
* server : construct batch with size of llama_n_batch
* ggml_backend_cpu_graph_compute : fix return value when alloc fails
* llama : better n_batch and n_ubatch comment
* fix merge
* small fix
* reduce default n_batch to 2048
---------
Co-authored-by: Francis Couture-Harpin <git@compilade.net>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* server: format error to json
* server: do not crash on grammar error
* fix api key test case
* revert limit max n_predict
* small fix
* correct coding style
* update completion.js
* launch_slot_with_task
* update docs
* update_slots
* update webui
* update readme
* server: ci: windows build and tests
* server: ci: remove tmp push branch
* server: ci: EOF EOL
* Use builti
Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
* server: tests: server graceful shutdown, then kill, then hard kill
* server: tests: remove python2 unicode string
* server: tests: remove wrong comment on server starting, close_fds is always true
* server: tests: server kill, if pid exists
* server: tests: remove dependency to killall
* server: tests: ci windows: pid exists better handling
---------
Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
* server: bench: Init a bench scenario with K6
See #5827
* server: bench: EOL EOF
* server: bench: PR feedback and improved k6 script configuration
* server: bench: remove llamacpp_completions_tokens_seconds as it include prompt processing time and it's misleading
server: bench: add max_tokens from SERVER_BENCH_MAX_TOKENS
server: bench: increase truncated rate to 80% before failing
* server: bench: fix doc
* server: bench: change gauge custom metrics to trend
* server: bench: change gauge custom metrics to trend
server: bench: add trend custom metrics for total tokens per second average
* server: bench: doc add an option to debug http request
* server: bench: filter dataset too short and too long sequences
* server: bench: allow to filter out conversation in the dataset based on env variable
* server: bench: fix assistant message sent instead of user message
* server: bench: fix assistant message sent instead of user message
* server : add defrag thold parameter
* server: bench: select prompts based on the current iteration id not randomly to make the bench more reproducible
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* add cmake build toggle to enable ssl support in server
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
* add flags for ssl key/cert files and use SSLServer if set
All SSL setup is hidden behind CPPHTTPLIB_OPENSSL_SUPPORT in the same
way that the base httlib hides the SSL support
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
* Update readme for SSL support in server
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
* Add LLAMA_SERVER_SSL variable setup to top-level Makefile
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
---------
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
* mamba : begin working on support for Mamba SSM
* mamba : begin figuring out how to (ab)use the kv cache for Mamba
* mamba : recurrent inference almost works, but incoherent
* mamba : recurrent inference WORKS!!!
* convert : optionally use d_conv and d_state from config.json for Mamba
* mamba : refactor recurrent conv, resulting in 20% perf increase
It's still slower than I'd like, but I did not really optimize `ggml_exp` yet.
I also refactored `ggml_exp` to work with tensors with more than 2 dimensions.
* ggml : parallelize ggml_exp
This results in 8% faster token generation for Mamba-130M.
* mamba : simplify the conv step with a self-overlapping view
Turns out the conv_state can be made smaller by one column.
Note that this breaks existing GGUFs of Mamba,
because the key_value_length field is tied to the conv_state size.
Convolution with a self-overlapping view is cool!
And it's much simpler than what I initially thought would be necessary
to make the convolution step work with more than 1 token at a time.
Next step is to make the SSM step work on batches of tokens too,
and thus I need to figure out a way to make a parallel selective scan
which will keep the ssm_state small and won't make it bigger
by a factor of (n_layer * batch_size).
* llama : fix Mamba KV self size wrongly displaying as f16 instead of f32
Relatedly, I also tried to see if other types than f32 worked for the states,
but they don't, because of the operators used.
It's probably better anyway to keep lots of precision there,
since the states are small anyway.
* mamba : fix self-overlapping view depth stride
* mamba : handle batches of more than 1 token
This means running Mamba no longer crashes when using the default settings!
And probably also slightly faster prompt processing.
Both batched and non-batched processing yield the same output.
Previously, the state was not cleared when starting a sequence.
Next step is to make the KV cache API work as expected for Mamba models.
* ggml: add ggml_ssm_scan to help with parallel selective scan
If the selective scan was implemented without a custom operator,
there would be waaay too many nodes in the graph. For example,
for Mamba-130M, with a batch size of 512 (the default),
a naive selective scan could add at least 24*512=12288 nodes,
which is more than LLAMA_MAX_NODES (8192),
and that's only for the smallest Mamba model.
So it's much cleaner with a custom operator.
Not sure about the name, though.
* ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation
This will help with performance on CPU if ggml_vec_mul_f32
and ggml_vec_add_f32 are ever optimized with SIMD.
* mamba : very basic quantization support
Mostly works, but there is currently no difference
between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same).
Most of the SSM-specific weights can be kept in f32 without affecting
the size that much, since they are relatively small.
(the linear projection weights are responsible for most of Mamba's size)
Too much quantization seems to make the state degrade quite fast, and
the model begins to output gibberish.
It seems to affect bigger models to a lesser extent than small models,
but I'm not sure by how much.
Experimentation will be needed to figure out which weights are more important
for the _M (and _L?) variants of k-quants for Mamba.
* convert : fix wrong name for layer norm weight of offical Mamba models
I was using Q-bert/Mamba-* models before, which have a slighlty different
naming scheme for the weights.
(they start with "model.layers" instead of "backbone.layers")
* mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator
This increases performance on CPU by around 30% for prompt processing,
and by around 20% for text generation.
However, it also makes the ggml_exp and ggml_soft_plus operators unused.
Whether or not they should be kept will be decided later.
* convert : for Mamba, also consider the "MambaLMHeadModel" arch name
It's the name of the class of the official implementation,
though they don't use it (yet) in the "architectures" field of config.json
* mamba : fix vocab size problems with official models
The perplexity was waaaay to high for models with a non-round vocab size.
Not sure why, but it needed to be fixed in the metadata.
Note that this breaks existing GGUF-converted Mamba models,
but **only if** the vocab size was not already rounded.
* ggml : remove ggml_exp and ggml_soft_plus
They did not exist anyway outside of this branch,
and since ggml_ssm_scan fused operations together, they are unused.
It's always possible to bring them back if needed.
* mamba : remove some useless comments
No code change.
* convert : fix flake8 linter errors
* mamba : apply suggestions from code review
* mamba : remove unecessary branch for row-wise ssm_state and C multiplication
It was previously done to avoid permuting when only one token is processed
at a time (like when generating text), but permuting is cheap,
and dynamically changing the compute graph is not future-proof.
* ggml : in ggml_ssm_scan, use more appropriate asserts
* ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32
* mamba : multiple sequences, but one at a time
This is a step towards making this Mamba implementation usable
with the server example (the way the system prompt is kept when clearing
the client slots will need to be changed before this can work, though).
The KV cache size for this kind of model is tied to the maximum number
of sequences kept at any single time.
For now, this number is obtained from n_parallel (plus one,
to have an extra sequence to dedicate to the system prompt),
but there might be a better way to do this which won't also
make the main example use 2 cells even if only 1 is really used.
(for this specific case, --parallel 0 helps)
Simultaneous sequence processing will probably require changes to
ggml_ssm_scan, and possibly a new operator for the conv step.
* mamba : support llama_kv_cache_seq_cp
This (mis)uses the logic around K shifts, because tokens in a state
can't be shifted anyway, and because inp_K_shift has the right shape and type.
Using ggml_get_rows is a nice way to do copies, but copy chains can't work.
Fortunately, copy chains don't really seem to be used in the examples.
Each KV cell is dedicated to the sequence ID corresponding to its own index.
* mamba : use a state mask
It's cleaner than the previous heuristic of
checking for the pos of the first token in the batch.
inp_KQ_mask could not be re-used for this, because it has the wrong shape
and because it seems more suited to the next step of
simultaneous sequence processing (helping with the problem of
remembering which token belongs to which sequence(s)/state(s)).
* llama : replace the usage of n_ctx with kv_self.size in many places
* mamba : use n_tokens directly instead of n_tok
* mamba : in comments, properly refer to KV cells instead of slots
* mamba : reduce memory usage of ggml_ssm_scan
From 290.37 MiB to 140.68 MiB of CPU compute buffer size
with Mamba 3B with a batch size of 512.
The result tensor of ggml_ssm_scan was previously a big part
of the CPU compute buffer size. To make it smaller,
it does not contain the intermediate ssm states anymore.
Both y and the last ssm state are combined in the result tensor,
because it seems only a single tensor can be returned by an operator
with the way the graph is built.
* mamba : simultaneous sequence processing
A batch can now contain tokens from multiple sequences.
This is necessary for at least the parallel example, the server example,
and the HellaSwag test in the perplexity example.
However, for this to be useful, uses of llama_kv_cache_seq_rm/cp
will need to be changed to work on whole sequences.
* ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba
This operator makes it possible to use and update the correct states
for each token of the batch in the same way as ggml_ssm_scan.
Other solutions which use existing operators would need loops which would
add too many nodes to the graph (at least the ones I thought of).
Using this operator further reduces the size of the CPU compute buffer
from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512.
And (at least on CPU), it's a bit faster than before.
Note that "ggml_ssm_conv" is probably not the most appropriate name,
and it could be changed if a better one is found.
* llama : add inp_s_seq as a new input tensor
The most convenient implementation to select the correct state (for Mamba)
for each token is to directly get the correct index from a tensor.
This is why inp_s_seq is storing int32_t and not floats.
The other, less convenient way to select the correct state would be
to have inp_KQ_mask contain 1.0f for each state used by a token
and 0.0f otherwise. This complicates quickly fetching the first used
state of a token, and is also less efficient because a whole row
of the mask would always need to be read for each token.
Using indexes makes it easy to stop searching when there are
no more sequences for a token, and the first sequence assigned
is always very quickly available (it's the first element of each row).
* mamba : support llama_kv_cache_seq_cp copy chains
* mamba : support shifting and dividing the kv cache pos
* mamba : make the server and parallel examples work with whole sequences
A seq_id is dedicated to the system prompt in both cases.
* llama : make llama_kv_cache_seq_rm return whether it succeeded or not
* mamba : dedicate an input tensor for state copy indices
This is cleaner and makes it easier to adapt when/if token positions
(and by extension, inp_K_shift) are no longer integers.
* mamba : adapt perplexity, batched, and batched-bench examples
* perplexity : limit the max number of sequences
This adapts to what the loaded model can provide.
* llama : add llama_n_max_seq to get the upper limit for seq_ids
Used by the perplexity example.
* batched : pass n_parallel to the model's context params
This should have been there already, but it wasn't.
* batched-bench : reserve sequences to support Mamba
* batched-bench : fix tokens being put in wrong sequences
Generation quality isn't what's measured in there anyway,
but at least using the correct sequences avoids using non-consecutive
token positions.
* mamba : stop abusing attention metadata
This breaks existing converted-to-GGUF Mamba models,
but will allow supporting mixed architectures like MambaFormer
without needing to break Mamba models.
This will also allow changing the size of Mamba's states
without having to reconvert models in the future.
(e.g. using something else than d_conv - 1 columns for the conv_states
will not require breaking existing converted Mamba models again)
* gguf-py : add new KV metadata key-value pairs for Mamba
* llama : add new metadata key-value pairs for Mamba
* llama : guard against divisions by zero when n_head is 0
* mamba : rename "unlimited" KV cache property to "recurrent"
* mamba : more correctly update the "used" field of the KV cache
* ggml : in ggml_ssm_scan, use a threshold for soft_plus
This is how the official Mamba implementation does it,
and it's also what torch.nn.Softplus does.
* convert : for Mamba, fallback to internal NeoX tokenizer
The resulting models are exactly the same
as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there.
* mamba : support state saving and restoring
* ggml : implicitly pass src tensors through dst for Mamba-related ops
* mamba : clarify some comments
* server : fix cache_tokens not getting correctly resized
Otherwise, when the "we have to evaluate at least 1 token" special case
was triggered, an extra token was kept in cache_tokens even if it was
removed from the KV cache.
For Mamba, this caused useless prompt reprocessing when the previous
request triggered the above case.
* convert-hf : support new metadata keys for Mamba
For the models available at
https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406
* mamba : rename metadata to be more similar to transformers library
This breaks existing converted-to-GGUF models,
but the metadata names are more "standard".
* mamba : support mamba-*-hf models
These models share their token_embd.weight with their output.weight
* mamba : add missing spaces
This is purely a formatting change.
* convert-hf : omit output.weight when identical with token_embd.weight
Only for Mamba for now, but it might be relevant for other models eventually.
Most Mamba models actually share these two tensors, albeit implicitly.
* readme : add Mamba to supported models, and add recent API changes
* mamba : move state_seq and state_mask views outside layer loop
A few tensors were also missing `struct` in front of `ggml_tensor`.
* server : refactoring (wip)
* server : remove llava/clip objects from build
* server : fix empty prompt handling + all slots idle logic
* server : normalize id vars
* server : code style
* server : simplify model chat template validation
* server : code style
* server : minor
* llama : llama_chat_apply_template support null buf
* server : do not process embedding requests when disabled
* server : reorganize structs and enums + naming fixes
* server : merge oai.hpp in utils.hpp
* server : refactor system prompt update at start
* server : disable cached prompts with self-extend
* server : do not process more than n_batch tokens per iter
* server: tests: embeddings use a real embeddings model (#5908)
* server, tests : bump batch to fit 1 embedding prompt
* server: tests: embeddings fix build type Debug is randomly failing (#5911)
* server: tests: embeddings, use different KV Cache size
* server: tests: embeddings, fixed prompt do not exceed n_batch, increase embedding timeout, reduce number of concurrent embeddings
* server: tests: embeddings, no need to wait for server idle as it can timout
* server: refactor: clean up http code (#5912)
* server : avoid n_available var
ggml-ci
* server: refactor: better http codes
* server : simplify json parsing + add comment about t_last
* server : rename server structs
* server : allow to override FQDN in tests
ggml-ci
* server : add comments
---------
Co-authored-by: Pierrick Hymbert <pierrick.hymbert@gmail.com>
* server: tests: add models endpoint scenario
* server: /v1/models add some metadata
* server: tests: add debug field in context before scenario
* server: tests: download model from HF, add batch size
* server: tests: add passkey test
* server: tests: add group attention params
* server: do not truncate prompt tokens if self-extend through group attention is enabled
* server: logs: do not truncate log values
* server: tests - passkey - first good working value of nga
* server: tests: fix server timeout
* server: tests: fix passkey, add doc, fix regex content matching, fix timeout
* server: tests: fix regex content matching
* server: tests: schedule slow tests on master
* server: metrics: fix when no prompt processed
* server: tests: self-extend add llama-2-7B and Mixtral-8x7B-v0.1
* server: tests: increase timeout for completion
* server: tests: keep only the PHI-2 test
* server: tests: passkey add a negative test
* Add "/chat/completions" as alias for "/v1/chat/completions"
* merge to upstream master
* minor : fix trailing whitespace
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* server: docs - refresh and tease a little bit more the http server
* Rephrase README.md server doc
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Update examples/server/README.md
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Update examples/server/README.md
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Update README.md
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
The system prompt is now decoded in batches.
* server : fix off-by-one n_past when start of prompt matches whole cache
The tokens right after the matching part would otherwise skip a pos value.
* server: logs - always use JSON logger, add add thread_id in message, log task_id and slot_id
* server : skip GH copilot requests from logging
* server : change message format of server_log()
* server : no need to repeat log in comment
* server : log style consistency
* server : fix compile warning
* server : fix tests regex patterns on M2 Ultra
* server: logs: PR feedback on log level
* server: logs: allow to choose log format in json or plain text
* server: tests: output server logs in text
* server: logs switch init logs to server logs macro
* server: logs ensure value json value does not raised error
* server: logs reduce level VERBOSE to VERB to max 4 chars
* server: logs lower case as other log messages
* server: logs avoid static in general
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* server: logs PR feedback: change text log format to: LEVEL [function_name] message | additional=data
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* server: monitoring - add /metrics prometheus compatible endpoint
* server: concurrency issue, when 2 task are waiting for results, only one call thread is notified
* server: metrics - move to a dedicated struct
* server: tests: init scenarios
- health and slots endpoints
- completion endpoint
- OAI compatible chat completion requests w/ and without streaming
- completion multi users scenario
- multi users scenario on OAI compatible endpoint with streaming
- multi users with total number of tokens to predict exceeds the KV Cache size
- server wrong usage scenario, like in Infinite loop of "context shift" #3969
- slots shifting
- continuous batching
- embeddings endpoint
- multi users embedding endpoint: Segmentation fault #5655
- OpenAI-compatible embeddings API
- tokenize endpoint
- CORS and api key scenario
* server: CI GitHub workflow
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* server: fallback to chatml
* add new chat template
* server: add AlphaMonarch to test chat template
* server: only check model template if there is no custom tmpl
* remove TODO
* server: health: fix race condition on slots data using tasks queue
* server: health:
* include_slots only if slots_endpoint
* fix compile warning task.target_id not initialized.
* server: enrich health endpoint with available slots, return 503 if not slots are available
* server: document new status no slot available in the README.md
* Added numa options to allow finer grained control as well as plumbing for a new mirror mode that will require numa.h
* Reverted Makefile
* Fixed include
* Removed sched.h from ggml.h, moved ggml_get_numa_affinity into ggml.c, removed trailing whitespace and fixed up a few inconsistent variables
* removed trailing whitespace
* Added numa options to allow finer grained control as well as plumbing for a new mirror mode that will require numa.h
* Reverting Makefile
* Fixed a number of issues with the move from BOOL to ggml_numa_strategies. Added a note about mirror mode note being implemented yet
* Removing MIRROR_MODE code for this PR
* Removing last bit of MIRROR_MODE code for this PR
* Removing unneeded branch in server.cpp example and moving get_numa_affinity and making it static
* Fixed lingering init_llama_backend() bool calls in tests and examples
* Remote enum llama_numa_strategies
* Revert bad merge with dynatemp flags
* add missing enum ggml_numa_strategies declaration and revert sync problem with master
* add missing enum ggml_numa_strategies declaration
* fixed ggml_init_numa variable
* Update ggml.h
Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
* Update READMEs with info about numa flags, change INTERLEAVE strategy name to DISTRIBUTE everywhere, implement the improved distribution strategy from @rankaiyx, fix a spelling mistake and un-merge some bad merges
* split numa init out from llama_backend_init and created llama_numa_init. Updated all code paths and samples
* Fix up some boolean vs enum comparisons
* Added #ifdefs for non-Linux OS that don't have cpu_set_t datatype
* Update ggml.h
Align enum values
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Update ggml.c
Remove whitespace
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Update ggml.c
align paremeters
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Update examples/server/server.cpp
remove whitespace and align brace
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Update common/common.cpp
Remove whitespace and align brace
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* unified ggml_numa_strategy enum and fixed text alignment in server.cpp example
* Update ggml.c
simplified return for platforms without NUMA support
Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
* removed redundant else from cli argument processing of --numa
* whitespace
---------
Co-authored-by: root <root@nenya.lothlorien.ca>
Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
Co-authored-by: Jared Van Bortel <jared@nomic.ai>
* Fix memory management in llava and server code
Fixes this error:
llama_new_context_with_model: graph splits (measure): 3
Available slots:
-> Slot 0 - max context: 6000
{"timestamp":1707926446,"level":"INFO","function":"main","line":2623,"message":"model loaded"}
all slots are idle and system prompt is empty, clear the KV cache
slot 0 - loaded image
slot 0 is processing [task id: 0]
slot 0 : kv cache rm - [0, end)
slot 0 - encoding image [id: 1]
munmap_chunk(): invalid pointer
Aborted
* Make it cleaner by checking size in batch free wrapper
* Create llava-survery-v2.py
* Update convert-image-encoder-to-gguf.py
* Update convert-image-encoder-to-gguf.py
* Rename llava-survery-v2.py to llava-surgery-v2.py
* Update convert-image-encoder-to-gguf.py
will now search for projector
* Update convert-image-encoder-to-gguf.py
whoops
* Update llava-surgery-v2.py
* Clip: Bugfix for normalization (it did not loat the 3 std and mean values)
Clip: bicubic resize function
Clip: added save-to-bmp/pil for debugging and conversion from/to 32/8 images
Clip: added normalization with FP16 precision simulation (image tensors match HF implementation, can be switched off, only used for llava-1.6)
Clip: added newline tensor, mergetype kv, image-grid kv, new resize-pad function with resolution from gridpoints
Clip: clip_image_preprocess now returns a float * vector instead of float, this way llava 1.5 and 1.6 is supported
llava: added ggml cpu graph for embedding patching, added spatial_unpad preliminary support, added a lot of comments that need to be cleaned when all is final
convert-image-encoder: fixed image-grid flattening
* whitespace corrections
* ws
* Tensors are now properly permuted.
Before the embeddings were inserted 1:1, now they are split into the 24x24 patches as in reference.
* ws
* added verbose_prompt support into cli
added stopwords for llava-1.6 into cli
* moved llava functions to llava.cpp, made clip.h C compatible API, replaced vector style functions with pointers, added a debug define to remove functions from compilation while not needed
* ws
* convert : skip unknown tensors (need for LLaVA)
* llava : update readme
* llava : fix compile warnings
* llava : style
* convert : add --skip-unknown CLI arg
* server : remove clip structs
* bugfix for non llava-1.6
It should now work with llava-1.5 as well
* clip : minor code rearrange
* llava : update readme a bit
---------
Co-authored-by: John <cmt-nct@users.noreply.github.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* server: allow to specify tokens as strings in logit_bias
* Apply suggestions from code review
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* include total "num_slots" in default_generation_settings_for_props
* cleanup total_slots return value in /props endpoint
* update /props endpoint docs with total_slots
* remove num_slots from default_generation_settings_for_props
* update /props endpoint section
server : fix deadlock when prompt array contains strings and numbers
server : removed an unnecessary generation when generating multi-prompts
server : removed an unnecessary assert
* server : fix context shift + simplify self-extend
* server : take system_tokens into account
* server : more n_past fixes
* server : rever n_past_se changes
* Changed ugly xxd to literals.
HPP files are much more readable as multiline literals rather than hex arrays.
* Dashes in literal variable names.
Replace . and - with _ in file names -> variable names.
* Comment on removing xxd.
XXD-> string literals
* XXD to string literals.
Replaced these unreadable headers with string literal versions using new deps.sh.
* first update for migration
* update init_cublas
* add debug functio, commit all help code
* step 1
* step 2
* step3 add fp16, slower 31->28
* add GGML_LIST_DEVICE function
* step 5 format device and print
* step6, enhance error check, remove CUDA macro, enhance device id to fix none-zero id issue
* support main device is non-zero
* step7 add debug for code path, rm log
* step 8, rename all macro & func from cuda by sycl
* fix error of select non-zero device, format device list
* ren ggml-sycl.hpp -> ggml-sycl.h
* clear CMAKE to rm unused lib and options
* correct queue: rm dtct:get_queue
* add print tensor function to debug
* fix error: wrong result in 658746bb26702e50f2c59c0e4ada8e9da6010481
* summary dpct definition in one header file to replace folder:dpct
* refactor device log
* mv dpct definition from folder dpct to ggml-sycl.h
* update readme, refactor build script
* fix build with sycl
* set nthread=1 when sycl, increase performance
* add run script, comment debug code
* add ls-sycl-device tool
* add ls-sycl-device, rm unused files
* rm rear space
* dos2unix
* Update README_sycl.md
* fix return type
* remove sycl version from include path
* restore rm code to fix hang issue
* add syc and link for sycl readme
* rm original sycl code before refactor
* fix code err
* add know issue for pvc hang issue
* enable SYCL_F16 support
* align pr4766
* check for sycl blas, better performance
* cleanup 1
* remove extra endif
* add build&run script, clean CMakefile, update guide by review comments
* rename macro to intel hardware
* editor config format
* format fixes
* format fixes
* editor format fix
* Remove unused headers
* skip build sycl tool for other code path
* replace tab by space
* fix blas matmul function
* fix mac build
* restore hip dependency
* fix conflict
* ren as review comments
* mv internal function to .cpp file
* export funciton print_sycl_devices(), mv class dpct definition to source file
* update CI/action for sycl code, fix CI error of repeat/dup
* fix action ID format issue
* rm unused strategy
* enable llama_f16 in ci
* fix conflict
* fix build break on MacOS, due to CI of MacOS depend on external ggml, instead of internal ggml
* fix ci cases for unsupported data type
* revert unrelated changed in cuda cmake
remove useless nommq
fix typo of GGML_USE_CLBLAS_SYCL
* revert hip cmake changes
* fix indent
* add prefix in func name
* revert no mmq
* rm cpu blas duplicate
* fix no_new_line
* fix src1->type==F16 bug.
* pass batch offset for F16 src1
* fix batch error
* fix wrong code
* revert sycl checking in test-sampling
* pass void as arguments of ggml_backend_sycl_print_sycl_devices
* remove extra blank line in test-sampling
* revert setting n_threads in sycl
* implement std::isinf for icpx with fast math.
* Update ci/run.sh
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Update examples/sycl/run-llama2.sh
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Update examples/sycl/run-llama2.sh
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Update CMakeLists.txt
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Update CMakeLists.txt
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Update CMakeLists.txt
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Update CMakeLists.txt
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* add copyright and MIT license declare
* update the cmd example
---------
Co-authored-by: jianyuzh <jianyu.zhang@intel.com>
Co-authored-by: luoyu-intel <yu.luo@intel.com>
Co-authored-by: Meng, Hengyu <hengyu.meng@intel.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* feat: add Dockerfiles for each platform that user ./server instead of ./main
* feat: update .github/workflows/docker.yml to build server-first docker containers
* doc: add information about running the server with Docker to README.md
* doc: add information about running with docker to the server README
* doc: update n-gpu-layers to show correct GPU usage
* fix(doc): update container tag from `server` to `server-cuda` for README example on running server container with CUDA
* server: add llama_server_queue struct
* server: add llama_server_response_event
* server: add comments
* server: move all mutexes away from server.cpp
* server: correct multitask response
* server: only add back deferred tasks when one slot is available
* server: fix a race condition cause by "request_completion"
* llama : ggml-backend integration
* ggml-backend : add names to buffers
* fix unmap after loading
* batched-bench : add tensor_split param
* llama : check for null tensor_split
* ggml-backend : increase GGML_MAX_BACKENDS
* improve graph splitting, partial fix for --no-kv-offload
* cuda : add ggml-backend split buffer support
* cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available)
* ggml : fix null backend dereference (#4807)
* ggml : fix null backend dereference
* ggml : also check ggml_backend_is_cpu
* test-backend-ops : check buffer allocation failures
* llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row)
* ggml : fix mul_mat_id work size
* llama : rewrite session kv load/set without graphs
* minor
* llama : only initialize used backends, free backends on context free
* llama : abort ctx if cuda backend init fails
* llama : rewrite lora with ggml-backend and compute on CPU
ggml-ci
* llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer
* opencl : add ggml-backend buffer type
* cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf)
* llama : on Metal, by default offload the full model
ggml-ci
* metal : page align the data ptr (#4854)
* Apply suggestions from code review
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
* cuda : fix split buffer free
* address review comments
* llama-bench : add split-mode parameter
* fix whitespace
* opencl : fix double initialization
* server : add --split-mode parameter
* use async copy and compute to improve multi-gpu performance
ggml-ci
* use async memcpys to copy the graph outputs to the CPU
* fix opencl
* use a host buffer for the cpu compute buffer for faster copies to the gpu
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
* server: added support for multiple api keys, added loading api keys from file
* minor: fix whitespace
* added file error handling to --api-key-file, changed code to better
reflect current style
* server: update README.md for --api-key-file
---------
Co-authored-by: Michael Coppola <info@michaeljcoppola.com>
* added /health endpoint to the server
* added comments on the additional /health endpoint
* Better handling of server state
When the model is being loaded, the server state is `LOADING_MODEL`. If model-loading fails, the server state becomes `ERROR`, otherwise it becomes `READY`. The `/health` endpoint provides more granular messages now according to the server_state value.
* initialized server_state
* fixed a typo
* starting http server before initializing the model
* Update server.cpp
* Update server.cpp
* fixes
* fixes
* fixes
* made ServerState atomic and turned two-line spaces into one-line
* updated `server` readme to document the `/health` endpoint too
* used LOG_INFO after successful model loading
* added /health endpoint to the server
* added comments on the additional /health endpoint
* Better handling of server state
When the model is being loaded, the server state is `LOADING_MODEL`. If model-loading fails, the server state becomes `ERROR`, otherwise it becomes `READY`. The `/health` endpoint provides more granular messages now according to the server_state value.
* initialized server_state
* fixed a typo
* starting http server before initializing the model
* Update server.cpp
* Update server.cpp
* fixes
* fixes
* fixes
* made ServerState atomic and turned two-line spaces into one-line
* updated `server` readme to document the `/health` endpoint too
* added /health endpoint to the server
* added comments on the additional /health endpoint
* Better handling of server state
When the model is being loaded, the server state is `LOADING_MODEL`. If model-loading fails, the server state becomes `ERROR`, otherwise it becomes `READY`. The `/health` endpoint provides more granular messages now according to the server_state value.
* initialized server_state
* fixed a typo
* starting http server before initializing the model
* Update server.cpp
* Update server.cpp
* fixes
* fixes
* fixes
* made ServerState atomic and turned two-line spaces into one-line
* updated server readme to reflect the gg/server-token-probs-4088 commit
added explanation for the API's completion result which now includes `completion_probabilities`. Also added a JSON schema that shows the type/structure of `completion_probabilities`.
* simplified the `completion_probabilities` JSON schema
It's now easier to understand what the structure of `completion_probabilities` looks like.
* minor : fix trailing whitespace
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Changes to server to allow metadata override
* documentation
* flake.nix: expose full scope in legacyPackages
* flake.nix: rocm not yet supported on aarch64, so hide the output
* flake.nix: expose checks
* workflows: nix-ci: init; build flake outputs
* workflows: nix-ci: add a job for eval
* workflows: weekly `nix flake update`
* workflows: nix-flakestry: drop tag filters
...and add a job for flakehub.com
* workflows: nix-ci: add a qemu job for jetsons
* flake.nix: suggest the binary caches
* flake.lock: update
to a commit recently cached by nixpkgs-cuda-ci
---------
Co-authored-by: John <john@jLap.lan>
Co-authored-by: Someone Serge <sergei.kozlukov@aalto.fi>
The server currently schedules tasks using a sleep(5ms) busy loop. This
adds unnecessary latency since most sleep implementations do a round up
to the system scheduling quantum (usually 10ms). Other libc sleep impls
spin for smaller time intervals which results in the server's busy loop
consuming all available cpu. Having the explicit notify() / wait() code
also helps aid in the readability of the server code.
See mozilla-Ocho/llamafile@711344b
The default values for tfs_z and typical_p were being set to zero, which
caused the token candidates array to get shrunk down to one element thus
preventing any sampling. Note this only applies to OpenAI API compatible
HTTP server requests.
The solution is to use the default values that OpenAI documents, as well
as ensuring we use the llama.cpp defaults for the rest. I've tested this
change still ensures deterministic output by default. If a "temperature"
greater than 0 is explicitly passed, then output is unique each time. If
"seed" is specified in addition to "temperature" then the output becomes
deterministic once more.
See mozilla-Ocho/llamafile#117
See mozilla-Ocho/llamafile@9e4bf29
* Add API key authentication for enhanced server-client security
* server : to snake_case
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* * add multiprompt support
* * cleanup
* * more cleanup
* * remove atomicity of id_gen, and change lock_guard to unique_lock on completion requests
* * remove all references to mutex_multitasks
* Update examples/server/server.cpp
Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
* Update examples/server/server.cpp
Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
* Update examples/server/server.cpp
Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
* Update examples/server/server.cpp
Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
* * change to set
---------
Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
* fix oai proxy
fix generation not stoped while bot stop talking in chat mode
fix possible `slot_id` not exist
response for cors (and pre flight)
* oai proxy: workaround for some client (such as Chatbox)
* use stop as separator to replace hardcoded `\n`
* Add openai-compatible POST /v1/chat/completions API endpoint to server example
* fix code style
* Update server README.md
* Improve server README.md
* Fix server.cpp code style according to review
* server : some style changes
* server : indentation
* server : enable special tokens during tokenization by default
* server : minor code style
* server : change random string generator
* straightforward /v1/models endpoint
---------
Co-authored-by: kir-gadjello <111190790+kir-gadjello@users.noreply.github.com>
Co-authored-by: Tobi Lütke <tobi@Tobis-MacBook-Pro.local>
* gguf-py: gguf-dump: Respect --no-tensor flag in JSON mode.
* Respect add_bos_token GGUF metadata value
* gguf-py: Try to fix SpecialVocab giving up too easily for the Nth time
* Update server.cpp with min_p after it was introduced in https://github.com/ggerganov/llama.cpp/pull/3841
* Use spaces instead of tabs
* Update index.html.hpp after running deps.sh
* Fix test - fix line ending
* wip llava python bindings compatibility
* add external llava API
* add base64 in-prompt image support
* wip refactor image loading
* refactor image load out of llava init
* cleanup
* further cleanup; move llava-cli into its own file and rename
* move base64.hpp into common/
* collapse clip and llava libraries
* move llava into its own subdir
* wip
* fix bug where base64 string was not removed from the prompt
* get libllava to output in the right place
* expose llava methods in libllama.dylib
* cleanup memory usage around clip_image_*
* cleanup and refactor *again*
* update headerdoc
* build with cmake, not tested (WIP)
* Editorconfig
* Editorconfig
* Build with make
* Build with make
* Fix cyclical depts on Windows
* attempt to fix build on Windows
* attempt to fix build on Windows
* Upd TODOs
* attempt to fix build on Windows+CUDA
* Revert changes in cmake
* Fix according to review comments
* Support building as a shared library
* address review comments
---------
Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com>
Co-authored-by: Jared Van Bortel <jared@nomic.ai>
* cmake : fix build when .git does not exist
* cmake : simplify BUILD_INFO target
* cmake : add missing dependencies on BUILD_INFO
* build : link against build info instead of compiling against it
* zig : make build info a .cpp source instead of a header
Co-authored-by: Matheus C. França <matheus-catarino@hotmail.com>
* cmake : revert change to CMP0115
---------
Co-authored-by: Matheus C. França <matheus-catarino@hotmail.com>
* Extend llama_kv_cache_seq_rm to allow matichng any sequence
* Replace llama_kv_cache_tokens_rm with llama_kv_cache_clear
Use llama_kv_cache_clear for cache clearing
Change calls to llama_kv_cache_tokens_rm that want to delete by position to use llama_kv_cache_seq_rm functionality
* added `llama_model_token_*` variants to all the `llama_token_*` functions.
* added `LLAMA_API`
* formatting
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* removed old `llama_token` functions
* changed 3 more functions to take in model
- `llama_token_get_text`
- `llama_token_get_score`
- `llama_token_get_type`
* added back docs
* fixed main.cpp
* changed token functions to use new model variants
* changed token functions to use new model variants
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* implementing parallel decoding in server example
* crash fixed
* save dev progress
* refactored sampling function
* completion endpoint working
* multiple client support
* grammar + no stream completion
* cached prompt support
* chat.mjs support cached prompt + some fixes
* server ui now support multiple clients
* unused change reverted
* fixed timings per slot
* add context swap
* add changes to README.md
* llava multimodal integration
* fixed tokens probs
* add multimodal input - alfa
* refactor code + remove unused comments + improved README.md
* fix compilation errors with llvm
* notify the user from server ui that multimodality is unavialable
* some ci fixes
* fix ci make build undefined ref errors
* fix long prompt than ctx proposed in #3639
* fixed premature end due stop word
* context shift fixed
* fix llava implementation
* sync README.md changes
* readme change
* update api like OpenAI
* multimodal support enabled by default
* fix make bui;d errors
* fix multiple clients
* fix zig build
* new sampling API
* latest changes of sampling API
* server : coding-style normalization
* server : coding-style normalization (part 2)
* server : remove beam-search functionality
* server : bug fix in ingest_images
n_tokens is incremented internally by llama_batch_add
* server : use refs + use llama_batch_clear()
* server : snake case
* server : minor sync
* added thread safe pipeline
* server : bach has to be allocated for n_parallel sequences
* server : no need for atomic int - already using mutex
* server : logs + minor code style
* server : fix multibyte handle in partial response (#3706)
* fix image load + view image in chat
* make : silence stb warnings
* clip : link to ggml, not to llama
* server : fix switch fallthrough
* server : fix crash in Debug on macOS (I have no idea why this fixes it!?)
* server : refactor ctx_sampling init + n_ctx + names
* server : bug fix for prompt caching
* Do not save/load image_data to localStorage
* editorconfig : new line in index.html
* server : completion requests remember slot_id
* Update readme to document multimodal in server
* server : minor style
* Update readme to document multimodal in server
* server : hide ctx_sampling->prev behind API (#3696)
* server : apply fix from #3722
* server : fix slot reuse
* server : add comment about changing slot_state to bool
---------
Co-authored-by: FSSRepo <go778sgt@gmail.com>
Co-authored-by: Damian Stewart <d@damianstewart.com>
Co-authored-by: Steward Garcia <57494570+FSSRepo@users.noreply.github.com>
Co-authored-by: Jhen-Jie Hong <iainst0409@gmail.com>
Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com>
* Added documentation of JSON return value of /completion endpoint
* Update examples/server/README.md
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Fix mirostat state when using multiple sequences
* Fix mirostat by completely refactoring sampling!
* Try to fix zig build.
* Export function to fetch/create default sampler states
Code formatting cleanups and add some comments
Silence a warning about id not being used when logging is disabled
* Apply some renaming suggestions.
Fix comments that were out of sync with the pull.
* Use more consistant naming convention for sampling contexts
* infill tokens correction
* serverinfill tokens correction
* removing any leading whitespace from infill suffix and removing leeading space token from suffix when params.escape
* removing any leading whitespace from infill suffix and removing leeading space token from suffix when params.escape
* only rm when params.escape, rm space if possible which is added back or rm added space token
* only rm when params.escape, rm space if possible which is added back or rm added space token
* Revert "only rm when params.escape, rm space if possible which is added back or rm added space token"
This reverts commit 63ba0b621f.
* fix interactive prompt escaping and fix server infill leading space handling
* rm unnecessary bool check
* vvhg-code-infill (#1)
* infill in separate example (#2)
* reverted changes to main and added infill example
* cleanup
* naming improvement
* make : add missing blank line
* fix missing semicolon
* brought infill up to current main code
* cleanup
---------
Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com>
* llama.cpp : split llama_context_params into model and context params
ggml-ci
* fix metal build
* fix freq_base/scale default to model value
* llama-bench : keep the same model between tests when possible
* move n_threads to llama_context_params, add n_threads_batch
* fix mpi build
* remove kv_size(), cuda scratch fixes
* remove low-vram option
* add n_threads_batch to system info, refactor to get_system_info()
* add documentation about --threads-batch to the READMEs
* llama-bench fix
* main : fix rope freq/scale warning
* llama.cpp : add llama_get_model
common : add llama_tokenize from model
* remove duplicated ctx/model functions
ggml-ci
* cuda : print total VRAM used
* fix track_max_mem in forward_batch_wo_cache_flash_attn_train
* remove unnecessary Adam(W) optimizer tensors.
reduces optimizer memory overhead from 7*modelsize to 2*modelsize.
additionally allows to optimize models with more than 2^31 parameters by replacing int with int64_t.
bumps training checkpoint file version, but old checkpoints can still be read.
new version with less tensors is saved.
* add gradient clipping to AdamW
* Fix reset of unused g->nodes and g->grads to NULL
* implement gradient checkpointing for training
reduces memory overhead from O(n_layer) to O(sqrt(n_layer))
as explained in readme of https://github.com/cybertronai/gradient-checkpointing
* remove unused compute buffer 3
* add and use function ggml_build_backward_expand to avoid stack overflows with large maximum number of nodes
GGML_API void ggml_build_backward_expand(struct ggml_context * ctx, struct ggml_cgraph * gf, struct ggml_cgraph * gb, bool keep);
* change AdamW decay parameter to work like the torch AdamW decay parameter
It is now relative to Adam learning rate `alpha*sched`.
Before that it was relative to `sched` only.
`alpha` being the maximum learning rate and `sched` being a scaling parameter in [0..1]
* change default AdamW weight decay parameter used in training to 0.1 as used in nanoGPT
* change default AdamW weight decay parameter defined in ggml to 0.0, making Adam default instead of AdamW
btw: the default weight decay parameter for torch.optim.AdamW is 0.01
* bug fixes for cross entropy loss
ggml_cross_entropy_loss: sums where not correctly added in workload of each thread
ggml_cross_entropy_loss_back: simplify backward process, reducing numerical issues
guard usage of exp f16 lookup in cross entropy by #define GGML_CROSS_ENTROPY_EXP_FP16
cross entropy loss is only used once during training, but it is quite sensitive to numerical errors introduced by exp-f16-lookup.
so exp-f16-lookup for cross entropy loss is disabled by default, trading better gradients for very slightly worse runtime performance.
* fix test-grad0 for cross_entropy_loss
the second argument to cross_entropy_loss must sum up to 1 for each row
* fix test-grad0 for soft_max
dont use only sum as aggregation, because sum of softmax is always 1 -> finite differences should not work
instead use sum(log(soft_max()*(1-eps)+eps)); use eps to avoid log(0)
* improve finite differences of test-grad0 by using double instead of float
* change cross_entropy_loss to output average over all rows
this helps keeping the loss and gradients in a sane range
* improve gradient checkpointing
sqrt(n_layers) is only the best checkpoint step when mem size of checkpoints and mem size of layers are equal.
since layers require more memory than the single-tensor-checkpoint we use, the optimal values are compute different:
```
given: n, u, v
objective: minimize(a*u+b*v) where a*b=n, a>0, b>0
b=n/a
minimize(a*u+v*n/a)
diff(a*u+v*n/a, a) = u - (v*n/a)/a
diff(a*u+v*n/a, a) == 0
u - (v*n/a)/a == 0
u == v*n/(a*a)
u*a*a = v*n
a*a = v*n/u
a = sqrt(n*v/u)
```
this change results in more checkpoints, requiring less layers to store between checkpoints, overall improving memory usage.
* disable gradient checkpointing debug output
* llama : fix rope usage in train-text-from-scratch after ChatGLM change
* add more training parameters:
--enable-restart N Only for Adam optimizer. Enable restarts of cos-decay
--disable-restart N Only for Adam optimizer. Disable restarts of cos-decay
--opt-past N Number of optimization iterations to track for delta convergence test. Disabled when zero.
--opt-delta N Maximum delta for delta convergence test. Disabled when <= zero.
--opt-max-no-improvement N Maximum number of optimization iterations with no improvement. Disabled when <= zero.
--adam-epsf N AdamW epsilon for convergence test. Disabled when <= zero.
--adam-min-alpha N Adam minimum learning rate alpha, usually 0.1 * alpha
* replace memcpy with reshape operation so that the graph is not cut at the input
this makes it possible to store other values into the input tensor and then simply recompute the graph without rebuilding it
* remove unused function argument from get_example_targets_batch
* measure and print total training time
* add optimization callback to ggml_opt_resume_g
this callback is called before each iteration with custom data and pointer to learning schedule parameter (only used in Adam(W)).
can be used for dynamic learning schedule and setting input data for batches before each iteration
* use optimization callback in training
allows dynamic learning schedule and different batch data for each iteration without relying on low n_iter and high n_examples parameters
reduces runtime by avoiding restart of optimization function and improves training convergence by providing a different batch for each iteration
* add minimum number of tensor dimensions to apply weight decay (default 2)
this allows to not apply weight decay to bias parameters
* rename training parameter cos-decay-alpha to cos-decay-min and clarify that adam-min-alpha also applies to warmup
* fix increase of model.train_samples and model.train_tokens
now that each optimizer iteration gets its own batch we need to multiply by number of opt iterations
* change sampling parameters for prediction after training to defaults of common.h
and clarify what is context for prediction and what are generated tokens
* tighten abs error bounds for cross_entropy_loss in test-grad0
* add conditional compilation of using F16 exp in flash attention
uncomment `// #define GGML_FLASH_ATTN_EXP_FP16` to enable usage of f16 exp in flash attention
* tighten abs error bounds for flash_attn in test-grad0
* tighten abs error bounds for sqrt in test-grad0
* remove out-commented vectorized code of opt_adam
the vectorized code might be bit faster for low number of parameters, but it had a big memory usage overhead
* ggml : update ggml_rms_norm_back with configurable eps
* llama training : fix ggml_rms_norm_back calls to pass configurable eps
* remove trailing whitespace
* add train function using automatic gradient checkpointing backward pass and allocator
* in train function replace add_inplace by regular add
because using add_inplace seems to result in different gradients
* don't use allocate hash_map on context
because the context has no_alloc=True when using memory allocator resulting in NULL data pointers
* correctly clone reshape and permute operations by also cloning tensor->nb values
* fix variable name and add missing type cast
* terminate recursive tensor cloning when reaching tensor without src tensors
* correctly clone view tensors by setting data pointers
without this the checkpointing would only work when being used together with memory allocator
* fix variable names
* swap arguments to commutative ops to be the same as in `forward_batch_wo_cache_flash_attn`
* add input tensors as checkpoints
so that recursive tensor cloning of gradient checkpointing terminates on input tensors
* fix variable name and add missing boolean negation
* make sure some tensors are not reallocated by inserting new temporary nodes depending on them:
output and parameter gradient tensors need to be available at the end of the graph execution
parameter gradient tensors also need to be available before the graph execution because they are set to zero before each optimizer iteration
checkpoint tensors are allocated all together to reduce memory allocator fragmentation
afterwards, in addition to the temporary nodes, we also need to reset the temporary leafs
* fix ASSERT to work with zero layers
* add training options whether to use allocator and/or unified training function
* integrate unified training function which may use memory allocator
the unified training function also supports arguments whether to use flash attention and/or gradient checkpointing
* format name of cloned tensors with " (clone)" suffix
* set names for tensors in unified train function for easier debugging
* allocate graph on context using ggml_new_graph
* remove handwritten training functions
* remove unused training parameters "use_scratch" and "use_unified"
* remove trailing whitespace
* remove unused train params: mem_compute1_gb & mem_compute2_gb
mem_compute_gb is used for compute when automatic memory allocator is not enabled, otherwise it can be very small to only hold the tensor definitions
mem_compute0_gb is used for automatic memory allocator (as long as measurement of max required size is not implemented)
* remove unused forward_batch function
* add debug asserts in ggml_allocr_alloc to some common pitfalls when using this function directly
* only use ggml_allocr_alloc when tensor has NULL data and is no view
* fix test when to create temporary backward graph
temporary backward graph is only necessary when using checkpointing
* fix memory "leak" in optimizers
each iteration a new cplan with new memory for work data was allocated.
now cplan creation only happens at the start of optimization, with each iteration reusing the cplan and its work data.
* reverse order of for loop in ggml_build_backward_expand to save memory when using gradient checkpointing and allocator
with this loop order gradient checkpointing with allocator on 16 layer model saves 13% memory; 2 layer memory it saves 2% memory.
the computation results are the same
* add API functions to access llama model tensors
* add stub example for finetuning, based on train-text-from-scratch
* move and remove code
* add API functions to access remaining model parameters:
mult, head and rot
* first draft for LORA finetune training
* remove const model and layer arguments in API functions for accessing model tensors
* bug fixes to make finetune compile
automatic allocator does not work yet
* add debug prints for training memory improvements
* fix names of lora tensors
* avoid stack overflow resulting from big ggml_cgraph
replace stack allocation and ggml_build_forward by ggml_new_graph in combination with ggml_build_forward_expand
* replace llama API functions to get model tensors by one function to get model tensor by name
LLAMA_API struct ggml_tensor * llama_get_model_tensor(struct llama_model * model, const char * name);
* remove unused call to not existing llama_get_layer_from_model
* implement ggml_compute_forward_out_prod_q_f32
* remove trailing whitespace
* add lora finetune support on quantized base model tensors
* add ggml_add_cast API function
this function works like ggml_add, but accepts a data type for the resulting tensor.
only supported for quantized src0 input.
* use ggml_add_cast in finetuning
lora-applied weights will now have data type F32, which improves gradients when finetuning quantized base models
* bug fix: actually use result type passed to ggml_add_cast
* make sure base model tensors data cannot be used in viewable operations
memory allocator would try to make lora application inplace on base model tensors.
since those are memory mapped this will result in memory access violations
* fix bug in ggml_out_prod which resulted in wrong n_dims of result tensors
* avoid keeping in memory ALL of the gradients
The problem here stems from ggml_graph_reset. This function is called in the optimization function, before each graph computation, to reset the gradients to zero. This required a unique memory slot for each gradient: allocating memory from a previosly freed memory location might lead to non-zero input gradients.
During ggml_compute_backward the gradients are build stepwise by adding or substracting new values, starting from a OP_NONE tensor which needs to contain zero-values. This requires the graph reset.
To avoid this I now remember in ggml_build_backward_expand the original OP_NONE gradient tensors in a hash table, which is passed to ggml_compute_backward. There instead of using add (or sub or similar) I test whether the existing gradient to be changed is a zero-valued-tensor by looking up its existence in the hash table. When it is such a zero-tensor it will not be modified, but replaced by the value to be added, otherwise the regular add (not inplace, allocator will take care of this) will be used. This way none of those zero-tensor values will be necessary in the final backward graph and more importantly they won't need a unique memory slot, just to make them zero.
* remove trailing whitespace
* remove debug prints and function to compute tensor data hash
* improve optimization iteration prints
* adjust maximal values to support finetuning 3B models
* change default finetune params lora_r and lora_alpha to match the n_rank parameters of 4
* bug fix: make sure finetune input gradient is allocated at begin and kept until end
* remove unnecessary src tensor from ggml_get_rows_back
we don't need data of src[2] for computation, only to setup the correct output shape.
remove dependency on src[2], so that allocator can work more freely.
the computational graph is still completely determined, because the output shape is naturally included.
this is similar to how ggml_reshape does it.
* remove unnecessary src tensor from ggml_repeat & ggml_repeat_back
we don't need data of src[1] for computation, only to setup the correct output shape.
remove dependency on src[1], so that allocator can work more freely.
the computational graph is still completely determined, because the output shape is naturally included
* resolve todo
allocator will only make it inplace when they are of the same type
* mixing multiple LORA adapters is now possible
pass more than one '--lora FNAME' argument to apply more than one LORA.
use '--lora-scaled FNAME S' when you want to specify a user-defined scale for an adapter.
* add option to save finetune output every N iterations
* also save latest finetune output with ITERATION="LATEST" and print where files are saved
saving with LATEST makes it easier to resume training from the latest checkpoint
the string "LATEST" can be configured with command line option "--fn-latest STR"
* update checkpoint train stats before saving via "--save-every"
* add command line option `--rank-wo N` for rank of wo tensor
* update finetune README
* fix dump_non_result_info_yaml to output multiple lora adapters
* bug fix: replace GGML_TYPE_SIZE[t] by ggml_type_size(t)
* replace llama_n_mult by llama_n_ff
* finetune bug fixes to compile with merged in code from master
* remove prediction related code to reduce duplicated code with main
use main instead
* reduce large memory overhead in train-text-from-scratch
all gradients had to be pinned so that graph_reset works correctly.
this is no longer necessary with the changes to ggml_compute_backward introduced in this PR.
* add comment explaining why finetune checkpoints are allocated in one block
* make default value of float member a float literal
* handle rms_norm and rope parameters the same as in train-text-from-scratch
* remove unused code
* remove vocab related code as it is unnecessary
* add LLM_KV_TRAINING_TYPE to train-text-from-scratch checkpoints
so that they can be differentiated from lora finetune checkpoints
* add gguf constants and load/save functions from train-text-from-scratch
* add load & save lora finetune checkpoints via gguf
* add python script to convert old finetune checkpoint files to gguf
* remove old checkpoint save & load code
* remove code to print data checksums which was used to verify correctness of new gguf code
* omit tokenization when training is disabled, only save llama lora adapter
training can be disabled by passing '-n 0' to finetune
* remove trailing whitespace
* update README.md
* implement ggml_compute_forward_repeat_f16
* avoid stack overflow of large cgraphs in test-grad0
* add ggml API functions ggml_unravel_index, ggml_get_i32_nd and its analogs for set and for f32
ggml_get_i32_1d, ggml_set_i32_1d, ggml_get_f32_1d, ggml_set_f32_1d now support non-contiguous tensors.
in case of non-contiguous tensor, the 1d index is unraveled into a multi index using ggml_unravel_index to be passed to '_nd' function equivalent.
this fixes a bug in test-grad0 which happens due to ggml_build_backward not building purely contiguous tensors anymore
* increase test-grad0 context mem size to accommodate for bigger cgraph
* add sanity check to ggml_compute_backward, asserting the correct shape of gradients
* fix ggml_acc_or_set to return tensor of correct shape
* remove unused 'inplace' argument from ggml_compute_backward function
inplace operations to add gradients are no longer created by ggml_compute_backward
use allocator to automatically make inplace operations
* add missing argument 'int i0' to ggml_get_i32_nd & ggml_set_i32_nd header declarations
* fix error message in ggml_allocr_alloc to display actual max_avail
* fix check_gradient
ggml_build_backward_expand was previously replaced by ggml_build_backward, but the assignment of forward graph to backward graph missing
* use tensor->view_src instead of ggml_is_view and get_view_source
* move gradient checkpointing code into ggml, new API function:
// build gradient checkpointing backward graph gb for gf using provided checkpoints
// gb_tmp will contain original backward graph with rewritten backward process nodes,
// but without the second forward pass nodes.
GGML_API void ggml_build_backward_gradient_checkpointing(
struct ggml_context * ctx,
struct ggml_cgraph * gf,
struct ggml_cgraph * gb,
struct ggml_cgraph * gb_tmp,
struct ggml_tensor * * checkpoints,
int n_checkpoints);
* replace custom data getters and setters by ggml functions
* train-text-from-scratch can train (full finetune) gguf models
just pass the gguf model via `--checkpoint-in FN`.
after this, to continue training, pass the generated checkpoint instead of the original gguf model.
tested with smaller models, bigger models may exceed available memory.
use (LORA) finetune for those.
* remove trailing whitespace
* add option to save train-text-from-scratch output every N iterations
* update README.md
* fix warnings
* fix warnings
* remove finetune option to disable allocator
the allocator should always be used.
by making sure that it is always used it gets easier to implement automatic memory requirements computation
* add tensor checkpoints only when gradient checkpointing is enabled
* initialize opt ggml context if none was provided
* add ggml-alloc API function 'ggml_allocr_max_size' to get max size of alloc
GGML_API size_t ggml_allocr_max_size(struct ggml_allocr * alloc);
* finetune: automatically allocate all memory and changes to command line options
remove '--n_examples N' parameter, as it no longer makes sense to call optimization process multiple times in a loop.
add '--only_write_lora' command line option: will skip tokenization and training, to only write a llama.cpp comptabile LORA adapter.
remove memory buffer related command line options.
improve iteration console output.
* add finetune to Makefile
* update README.md
* print time per iteration and estimate remaining time
* increase measured alloc size by tensor_alignment
ggml_allocr_reset will reduce the given size by up to tensor_alignment-1
* fix README.md
* add some more allocator debug prints
* bug fix, probably solves the 'ggml_allocr_alloc: not enough space in the buffer' issue
* revert last commit
"bug fix, probably solves the 'ggml_allocr_alloc: not enough space in the buffer' issue"
"alloc was freeing an externally allocated tensor, because it calculated the end of allocator memory as alloc->data + alloc->max_size instead of alloc->data + alloc->size."
This is intentional to reduce the risk of freeing external tensors when measuring. Unless max_size is not properly calculated, I don't see why this is an issue.
* remove unnecessary "0x" before "%p" output
* move measurement memory segment to upper region of the address space
* update README.md
* fix printf format warnings
* add missing gguf_free in load_checkpoint_lora_file
* load default rms_norm and rope parameters from base model
* add gradient accumulation
specify number accumulation steps with '--grad-acc N'.
this will simulate a bigger batch size of grad_acc*batch.
* fix tracking of train_samples and train_tokens
* build : fix compile warnings
* ggml : fix L-BFGS linesearch loop
* improve finetune time measurement
fix printf warnings on system where int64_t is (long int).
change time datatypes to double because values get big with long training times.
exclude file saving from time measurement.
converge faster to actual time per iteration by removing very small first duration before first iteration was performed.
fix bug in output of total training time, the reported value was 1000 times to small.
* specify default lora rank with '--lora-r N'
'--lora-r N' will specify default rank for all tensors
'--rank-wq N', etc. will override this default rank for specific tensor types.
* fix gradient accumulation bug where the same batch was used for each microstep
* fix gradient accumulation bug where the same batch was used for each microstep
* support grouped-query-attention in ggml_flash_attn and ggml_flash_attn_back
k and v can now be repeated in q along ne[2]
in forward pass just use modulo to compute k and v indices, like ik2 = iq2 % nek2.
in backard pass this won't work as easy, because multiple threads will compete to accumulate to the same k->grad[:,ik1,ik2,ik3] and v->grad[:,iv1,iv2,iv3].
so we change the parallelization over q rows to be over k rows. this ensures non-overlapping (ik2,ik3) across threads.
in each thread we then iterate over the number of repetitions of k/v in q to compute iq2 as iq2 = ik2 + irep*nek2.
since ne2 is not the same for q,k and v we also change how the gradients are concatenated into the result tensor.
additionally the offsets of gradq, gradk and gradv in the result tensor are now memory aligned.
we also simplify the compute_backward part of flash_attn to use ggml_reshape instead of switching over the number of dimensions.
this needs a small change to ggml_reshape, removing the assertion of second argument to be contiguous.
since only the shape (ne) of the second reshape argument is of relevance, its memory layout (nb) is irrelevant -> it can very well be non-contiguous.
change test-grad0 to also test for repeated k/v in q.
this changes the rng and now results in small gradient differences in softmax. these solely come from using f16 exp table lookup in forward softmax: when temporarily changing softmax to use actual exp function, the reported gradient differences go away. gradient differences coming solely from f16 table lookup are acceptable.
added a note to explain this.
* add llama API functions to get grouped-query-attention n_head parameter 'n_head_kv'.
* fix finetune to support grouped-query-attention (using flash-attention)
note: ggml changes to ggml_out_prod are necessary to support grouped-query-attention without flash-attention.
* support broadcastable a in out_prod(a, b) and backward pass of broadcasting mul_mat(a, b)
* test broadcasting mul_mat backward pass
* decouple random number generator of each operation test
when changing one test the rng of others tests is not influenced anymore
* add comment briefly describing what ggml_repeat_back does
* simplify broadcasting mul_mat backward using ggml_repeat_back
* add cgraph evaluation order member and corresponding enum type
this controls in which order ggml_build_forward visits source nodes.
by default the nodes are visited left to right, i.e. src[0] first.
in some cases it is beneficial for ggml-alloc to visit in a different order.
two possible orders are supported: left-to-right (src[0] first) and right-to-left (src[0] last).
* measure max compute size for each cgraph eval order and use best order
this can bring huge memory savings:
e.g. codellama-34b with n_ctx=64, n_batch=1 goes from 92927.8mb down to 4627.6 MB
* remove unused command line options
* add sample start patterns and options to force new or by default resume last shuffling
* update shuffle rng state on reshuffle
* exclude known zero values from computations in flash_attn_f32 & flash_attn_back_f32
* remove probably unnecessary exception type flags from stringstream
* pass correct max number of tokens to llama_tokenize
* account for possible leading whitespace that will be added by tokenizer
e.g. '\t' will be tokenized by llama spm tokenizer to [29871, 12]
* use unrolled vec_mad in out_prod
y is vec_mad result vec.
x is vec_mad input vec.
v is vec_mad input scalar.
ggml_vec_mad_f32_unroll will internally loop over x and v with same y.
GGML_VEC_MAD_UNROLL is by default defined to 32.
This value is empirical optimized using performance test runs of out-prod in openllama-3b finetune with 256 context length and batch size 1. It gives 23% performance boost for out_prod.
Full measurements of out-prod runtime in ms:
unroll_xv unroll_yv
1 67014.643 87826.469
2 77117.552 89077.656
4 72091.311 109121.657
8 61077.543 88678.334
16 56914.67 79514.947
24 59024.595 84350.254
28 55952.446 83368.73
32 51476.658 85177.745
36 55973.792 84659.92
40 55139.616 93844.738
48 60736.392 93330.267
64 99856.878 116994.99
Second column is when unrollying yv instead of xv
* set lora_alpha to value of lora_r if it is not set via command line
otherwise only changing lora_r will change scaling of lora adapter used in prediction
* reshuffle original sample order instead of the previous shuffled order
otherwise resumed reshuffle will not result in same sample order
* block tiling for out-prod inspired by mul-mat
block sizes are empirically optimized
roughly doubles the flops of out-prod
* exclude some more known zero values from computations in flash_attn_f32 & flash_attn_back_f32
* add static keywords
* remove outcommented old code
* update train-text-from-scratch with tokenization, sample selection and shuffling from finetune
* remove lbfgs related train parameters
* move common train functions into common/train.[h|cpp]
* move train state into struct train_state
* move train data saving code into callback to unify code of opt_callback
train_params are still different in finetune and train-text-from-scratch, so it can't yet be moved to train.h|cpp
* move common train params into common/train
* move common opt_callback into common/train
* fix consume_common_train_arg
* save and load head_count_kv in lora checkpoints
* increase train_samples by used_samples instead of number of batches
on batch can contain more than one sample when option "fill_with_next_samples" is used
* fix usage of llama_tokenize
* remove static from process_escape since we need it exposed in header
* fix code formating of long function declarations
* fix condition in load_train_state_gguf
* use die("msg") instead of replace GGML_ASSERT(!"msg") or throw std::runtime_error("msg")
* fix saving and loading of training type
* remove terminating '\0' from tokenization
(llama_tokenize is now passed the string length instead of relying on terminating '\0')
* fix compile warnings
* fix compile warnings
* use new/delete for train_state instead of malloc/free
using malloc may result in seg faults when trying to assign string fields
* assert that sample_count > 0, avoiding division by zero
* fix frand to return value in interval [0,1)
* add train option "--sample-random-offsets"
Use samples beginning at random offsets.
The offset is only applied to the first sample in each batch context window.
Together with "--fill-with-next-samples" this may help for training endless text generation.
For example given a dataset containing samples "abcd", "ABCD", "0123".
With context size of 8 and options "--fill-with-next-samples", "--no-separate-with-eos", "--no-separate-with-bos",
the context windows of batches could only be filled with "abcdABCD", "ABCDabcd", "0123abcd", etc.
With "--sample-random-offsets" it can also be filled with "23abcdAB", "bcd0123A", etc.
* deduplicate code into function
* remove n_rot hparam, as it must always be hparam.n_embd_head()
* align code
* assert correct base model tensor shapes
* move some params from lora hparams into model hparams and load model params from gguf
this equalizes the model definition in finetune and text-from-scratch and removes the need for additional llama api functions to get model parameters
* remove now unnecessary llama API functions to get model params that where added by this PR
* train-text-from-scratch: automatically allocate model tensors, remove option '--mem-model N'
* train-text-from-scratch: automatically allocate opt context
* train-text-from-scratch: automatically allocate input tensors
* train-text-from-scratch: automatically allocate compute memory
* remove unused options and equalize train-text-from-scratch with finetune
* initialize opt->loss_after with zero
* add export-lora program
* remove trailing whitespace
* add export-lora build in Makefile
* remove unused struct tensor_info from export-lora
* add export-lora build dependency to llama
because it depends on common, which depends on llama
* update finetune README.md
* cancel optimization when specified number of epochs is completed
* improve handling of export-lora arguments
print errors and warnings when files could not be read or created
* Fix export-lora.cpp "not enough space in the context's memory pool" (#1)
* Fix export-lora.cpp "not enough space in the context's memory pool"
Without this patch, export-lora would sometimes error with "not enough space in the context's memory pool (needed 656784, available 656800)".
* increase required context size by 5*GGML_MEM_ALIGN instead of plain 16
---------
Co-authored-by: xaedes <xaedes@gmail.com>
* improve handling of not yet supported tensor types
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
Co-authored-by: meatbag-18a <145869052+meatbag-18a@users.noreply.github.com>
* tests : write a Python tokenizer test (wip)
* llama : prefix input text for tokenization with whitespace
* llama : distinguish pieces from decoded text + fix detokenization
* common : add comments
* examples : no longer manually add leading space when tokenizing
* tests : use Python to generate tokenizer tests for C++
* tests : add option to tokenize text files
ggml-ci
* tests : add test-tokenizer-1.py
* llama.cpp : fix LF token
* hellaswag : move the concat space for clarity
* tests : add falcon tests (py + cpp, currently do not pass Unicode)
ggml-ci
* common : temporary separate llama_detokenize calls for SPM and BPE
---------
Co-authored-by: klosax <131523366+klosax@users.noreply.github.com>
* Add llama_beam_search().
* Add '// Beam search' heading to llama.{h,cpp} after llama_grammar_accept_token().
* Add space around * pointers and & references.
* Add spaces around comparison and assignment operators.
* Prefer west const.
* Use llama_ prefix for structs in global namespace.
* Delete obsolete comment from an earlier revision.
* Change eos to eob in llama_beam and llama_beam_view structs.
* server : add n_probs param in chat UI
* server : keep message data array & show in probabilites component
* server : add simple popover component
* server : fix completion_probabilities undefined if not set n_probs
* server : implement Probabilites
* server : handle bytes
* server : make n_probs max to 10 for easy scroll
* server : adjust for dark/light mode
* server : Fix regenerated prompt
* server : update index.html.hpp
* server : convert prob to percentage + show original value as div title
* server : fix Probabilites not used if included empty str
* server : skip byte pair in display probabilites
* server : remove array check of completion_probabilities in messages
* skip empty array or byte pair (> 1) in Probabilites
* generate index.html.hpp
* fix incorrect prob convert if the str is already a known token
* use final response to show probabilities on stop
* revert unnecessary change
* correct probabilites usage
* remove unused function
* always send partial response for get correct probs of last to_send
* fix typo
* fix content of format_final_response
* refactor probs render & make pColor transparent if not found
* send empty string when got stop_pos in partial
* avoid unnecessary empty data event & send rest of partial tokens on stop
* use <br /> for new line
* skip -1 tok in loop to avoid send '' on end
* trim last new lines on stop
* revert unnecessary change
* server: allow json array in prompt or content
We accept an array of strings and numbers representing tokens,
in addition to the current string valued prompt or content.
This allows direct token input, so that any special tokens
can be processed and used at the frontend during the construction
of the json data, before sending to the server. And the server
does not need to know or parse special tokens from textual input.
With this, we can use EOS and BOS used in llama-2-chat models.
* server: use tokenizePrompt(json) and default "" if empty prompt
* server: fix prompt check
* server: tokenize endpoint no longer adds BOS
* support for templates in browser LocalStorage
* sync accepted #2409 fix from upstream
* convert autosave invocation to useEffect
* Apply suggestions from code review
Co-authored-by: Jhen-Jie Hong <iainst0409@gmail.com>
* Regen index.html.cpp, suggested from code review
---------
Co-authored-by: Jhen-Jie Hong <iainst0409@gmail.com>
* server : implement json-schema-to-grammar.mjs by follow python impl
* server : add grammar support in chat.mjs
* server : implement grammer param in the UI
* server : generate .hpp
* server : remove trailing whitespaces
* server : generate .hpp
* server : fix sort of prop pairs
* server : optimize regex & iteration
