Allow for user specified embedding pooling type #5849

iamlemec · 2024-03-03T05:13:51Z

This replaces the bool do_pooling context flag with a full llama_pooling_type pooling_type option. It also adds an additional value LLAMA_POOLING_TYPE_UNSPECIFIED to llama_pooling_type that falls back to the model default. Pooling type is optional in GGUF files so that generative LLMs or embedding models with no default pooling type (such as WhereIsAI/UAE-Large-V1) will work smoothly.

This is backwards compatible with existing GGUF files (checked with a couple models including nomic). It should be independent of but complementary to what is afoot in #5796.

ggerganov · 2024-03-03T09:01:52Z

llama.h

@@ -258,7 +259,7 @@ extern "C" {
        bool logits_all;  // the llama_decode() call computes all logits, not just the last one (DEPRECATED - set llama_batch.logits instead)
        bool embedding;   // embedding mode only
        bool offload_kqv; // whether to offload the KQV ops (including the KV cache) to GPU
-        bool do_pooling;  // whether to pool (sum) embedding results by sequence id (ignored if no pooling layer)
+        enum llama_pooling_type pooling_type;  // whether to pool (sum) embedding results by sequence id (ignored if no pooling layer)


Move this next to rope_scaling_type above and pass as int32_t

Should I have it be an int32_t in llama_cparams and llm_build_context as well? Otherwise will need a static_cast<llama_pooling_type> in llama_new_context_with_model.

cebtenzzre · 2024-03-03T16:36:52Z

llama.h

+        enum llama_rope_scaling_type rope_scaling_type; // RoPE scaling type, from `enum llama_rope_scaling_type`
+        enum llama_pooling_type      pooling_type;      // whether to pool (sum) embedding results by sequence id


Unless we use C23/C++11 syntax to specify the underlying type of the enum, this defeats the purpose of using fixed-size types like int32_t in the API. In C this will be an int; in C++, there are few guarantees:

Declares an unscoped enumeration type whose underlying type is not fixed (in this case, the underlying type is an implementation-defined integral type that can represent all enumerator values; this type is not larger than int unless the value of an enumerator cannot fit in an int or unsigned int. [...])

Ok, I'll go back to int32_t within #5796

Edit: On second thought, using enums is OK. I doubt this would ever cause problems

* allow for user specified pooling type * llama : use enum types over int --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

@JohannesGaessler

* fix mul_mat fault in cpy_f32_f16 * rm unused function * add wait() for memcpy * restore ci/run.sh, rename struct defination, fix bug in ggml_sycl_op_mul_mat_sycl * fix format issue * llama : fix segfault from unknown model arch name (#5820) * llama : fix segfault from unknown model arch name * llama : make all LLM maps const This also requires using `std::map::at` instead of its `operator[]` which does not exist for const maps. * llama : name LLM_ARCH_UNKNOWN to "(unknown)" This avoids errors from `std::map::at` when getting the general name of the model architecture. Using "(unknown)" instead of an empty string as per suggestion #5820 (comment) * llama : remove redundant inner const for LLM_TENSOR_NAMES The extra const won't do anything here as const maps return const references to values. Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * llama : remove redundant nullptr check in llm_arch_from_string Since LLM_ARCH_NAMES is a const map, no spurious elements with a NULL name are inserted anymore, so this check is dead code. --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * llama : refactor internal quantization functions (#5830) * scripts : add pod-llama.sh * ggml : IQ3_S improvements (#5829) * iq3_s: somewhat faster AVX2 dot product On Ryzen a 7950X TG-128 increases to 16 t/s from 15.5 t/s using 16 threads. For 8 threads it is 13.85 t/s vs 11.75 t/s. PP-512 increases to 28.5 t/s from 23.8 t/s. * iq3_s: somewhat faster ARM_NEON dot product Still dog slow - 10.7 t/s up from 9.9 t/s. * iq3_s: another small ARM_NEON improvement 10.7 -> 11.0 t/s. Using vmulq_s8 is faster than the xor - sub trick that works best on AVX2. * iq3_s: minor improvement on Metal 49.4 t/s -> 50.3 t/s * iq3_s: PPL improvement E.g., for a context of 4096 LLaMA-v2-7B goes to 5.1340 from 5.1653. * iq3_s: use new grid everywhere * Fix ARM_NEON --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> * convert-hf : make model class definitions self-contained (#5825) * convert : automatically fall back to HfVocab if tokenizer.model doesn't exist (#5821) * ggml : fix IQ3_S AVX implementation (#5834) ggml-ci * llama : add abort_callback to interrupt computation (#5409) * using abort_callback from ggml to stop llama computation * format fix * a brief explaining comment --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * server: tests: passkey challenge / self-extend with context shift demo (#5832) * 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 * flake.lock: Update (#5842) Flake lock file updates: • Updated input 'flake-parts': 'github:hercules-ci/flake-parts/b253292d9c0a5ead9bc98c4e9a26c6312e27d69f' (2024-02-01) → 'github:hercules-ci/flake-parts/f7b3c975cf067e56e7cda6cb098ebe3fb4d74ca2' (2024-03-01) • Updated input 'flake-parts/nixpkgs-lib': 'github:NixOS/nixpkgs/97b17f32362e475016f942bbdfda4a4a72a8a652?dir=lib' (2024-01-29) → 'github:NixOS/nixpkgs/1536926ef5621b09bba54035ae2bb6d806d72ac8?dir=lib' (2024-02-29) • Updated input 'nixpkgs': 'github:NixOS/nixpkgs/cbc4211f0afffe6dfd2478a62615dd5175a13f9a' (2024-02-23) → 'github:NixOS/nixpkgs/1536926ef5621b09bba54035ae2bb6d806d72ac8' (2024-02-29) Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com> * server : init http requests thread pool with --parallel if set (#5836) * ci : schedule slow server tests only on Release or on demand (#5839) * llama : fix llama_copy_state_data with fragmented KV cache (#5840) The row size of the saved states was based on kv_self.head while it should be based on llama_kv_cache_cell_max. Existing session files should still work. * llama : fix llama_kv_cache_cell_max inability to return 1 I've also changed its return type to uint32_t, because this function is always used to set the value of uint32_t variables, and because the index already has this type. * llama : fix state size calculation Some bytes in the state were unaccounted for in llama_get_state_size. Since the logits reserve so much space, it did not cause problems. * gguf-dump : support i-quants (#5841) Co-authored-by: Black_Fox <radekliska@gmail.com> * llama : allow for user specified embedding pooling type (#5849) * allow for user specified pooling type * llama : use enum types over int --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * readme : add API changes section * cuda : fix data race in soft max (#5853) * main : support special tokens as reverse/anti prompt (#5847) * Support special tokens as reverse/anti prompt. * Tokenize antiprompts only once. * main : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * common : use LLAMA_DEFAULT_SEED (#5855) * add some new ops, fix some operators and add batch operations to certain operators. (ggml/747) * cuda: fix group_norm * cuda: add batch inference support for ggml_pad/ggml_upscale * add ggml_arrange * add ggml_timestep_embedding * update ggml_arange/ggml_timestep_embedding tests * cuda: fix im2col * add ggml_arange/ggml_timestep_embbeding support for metal backend * fix some bugs * fix some bugs * Update ggml.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update ggml-cuda.cu Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update ggml-metal.metal Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * modify according to the review comments * ggml : fix compile warnings + code style * ggml : normalize compute_forward calls + fix seg fault in debug * minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com> * sync : ggml * add alias for chat template (#5858) * speculative : implement stochastic speculative sampling (#5625) * (WIP) Implement stochastic speculative decoding * sample from residual distribution on draft accept failure * fix #5657: force greedy sampling with probs when temp is 0 * remove p_accept parameter * fix style * remove unused variables * add srand() in speculative.cpp * replace use of rand() with mt19937 sampling * fixes based on review (@JohannesGaessler) * fix r random generation * randomly select next sequence to verify + fix bug in memory freeing * fix bug in active_seqs sync * fix uniform int distribution initialization * remove warnings from comparison between int and size_t * check grammar in `llama_sample_probability_distribution_impl` * remove malloc code by utilizing vectors * add PR link to README * cmake : handle cases where git index is not found in .git (#5844) * Update CMakeLists.txt * Update CMakeLists.txt * ggml : introduce ggml_status (ggml/750) * using enum as an exit code instead of macros * update return type from enum to unsigned int * indentation fix * compound update ggml_compute_exit_code -> ggml_status changed ggml_status from a bit-field type to simple codes ggml_status to string cast * ggml_status to string cast * GGML_CALL was removed Co-authored-by: slaren <slarengh@gmail.com> --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * sync : ggml ggml-ci * ggml : fix unknown status (#0) * flake : fix * llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list * nix: static build (#5814) * fix speculative decoding build on windows (#5874) * rebase and rm tailing space --------- Co-authored-by: LiangtaoJin <liang-tao.jin@intel.com> Co-authored-by: compilade <113953597+compilade@users.noreply.github.com> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Kawrakow <48489457+ikawrakow@users.noreply.github.com> Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Michael Podvitskiy <podvitskiymichael@gmail.com> Co-authored-by: Pierrick Hymbert <pierrick.hymbert@gmail.com> Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com> Co-authored-by: Nindaleth <Nindaleth@users.noreply.github.com> Co-authored-by: Black_Fox <radekliska@gmail.com> Co-authored-by: Douglas Hanley <thesecretaryofwar@gmail.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: DAN™ <dranger003@gmail.com> Co-authored-by: leejet <leejet714@gmail.com> Co-authored-by: Minsoo Cheong <54794500+mscheong01@users.noreply.github.com> Co-authored-by: Dane Madsen <dane_madsen@hotmail.com> Co-authored-by: hutli <6594598+hutli@users.noreply.github.com> Co-authored-by: Jeffrey Quesnelle <emozilla@nousresearch.com>

* allow for user specified pooling type * llama : use enum types over int --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

@JohannesGaessler

* fix mul_mat fault in cpy_f32_f16 * rm unused function * add wait() for memcpy * restore ci/run.sh, rename struct defination, fix bug in ggml_sycl_op_mul_mat_sycl * fix format issue * llama : fix segfault from unknown model arch name (ggerganov#5820) * llama : fix segfault from unknown model arch name * llama : make all LLM maps const This also requires using `std::map::at` instead of its `operator[]` which does not exist for const maps. * llama : name LLM_ARCH_UNKNOWN to "(unknown)" This avoids errors from `std::map::at` when getting the general name of the model architecture. Using "(unknown)" instead of an empty string as per suggestion ggerganov#5820 (comment) * llama : remove redundant inner const for LLM_TENSOR_NAMES The extra const won't do anything here as const maps return const references to values. Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * llama : remove redundant nullptr check in llm_arch_from_string Since LLM_ARCH_NAMES is a const map, no spurious elements with a NULL name are inserted anymore, so this check is dead code. --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * llama : refactor internal quantization functions (ggerganov#5830) * scripts : add pod-llama.sh * ggml : IQ3_S improvements (ggerganov#5829) * iq3_s: somewhat faster AVX2 dot product On Ryzen a 7950X TG-128 increases to 16 t/s from 15.5 t/s using 16 threads. For 8 threads it is 13.85 t/s vs 11.75 t/s. PP-512 increases to 28.5 t/s from 23.8 t/s. * iq3_s: somewhat faster ARM_NEON dot product Still dog slow - 10.7 t/s up from 9.9 t/s. * iq3_s: another small ARM_NEON improvement 10.7 -> 11.0 t/s. Using vmulq_s8 is faster than the xor - sub trick that works best on AVX2. * iq3_s: minor improvement on Metal 49.4 t/s -> 50.3 t/s * iq3_s: PPL improvement E.g., for a context of 4096 LLaMA-v2-7B goes to 5.1340 from 5.1653. * iq3_s: use new grid everywhere * Fix ARM_NEON --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> * convert-hf : make model class definitions self-contained (ggerganov#5825) * convert : automatically fall back to HfVocab if tokenizer.model doesn't exist (ggerganov#5821) * ggml : fix IQ3_S AVX implementation (ggerganov#5834) ggml-ci * llama : add abort_callback to interrupt computation (ggerganov#5409) * using abort_callback from ggml to stop llama computation * format fix * a brief explaining comment --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * server: tests: passkey challenge / self-extend with context shift demo (ggerganov#5832) * 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 * flake.lock: Update (ggerganov#5842) Flake lock file updates: • Updated input 'flake-parts': 'github:hercules-ci/flake-parts/b253292d9c0a5ead9bc98c4e9a26c6312e27d69f' (2024-02-01) → 'github:hercules-ci/flake-parts/f7b3c975cf067e56e7cda6cb098ebe3fb4d74ca2' (2024-03-01) • Updated input 'flake-parts/nixpkgs-lib': 'github:NixOS/nixpkgs/97b17f32362e475016f942bbdfda4a4a72a8a652?dir=lib' (2024-01-29) → 'github:NixOS/nixpkgs/1536926ef5621b09bba54035ae2bb6d806d72ac8?dir=lib' (2024-02-29) • Updated input 'nixpkgs': 'github:NixOS/nixpkgs/cbc4211f0afffe6dfd2478a62615dd5175a13f9a' (2024-02-23) → 'github:NixOS/nixpkgs/1536926ef5621b09bba54035ae2bb6d806d72ac8' (2024-02-29) Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com> * server : init http requests thread pool with --parallel if set (ggerganov#5836) * ci : schedule slow server tests only on Release or on demand (ggerganov#5839) * llama : fix llama_copy_state_data with fragmented KV cache (ggerganov#5840) The row size of the saved states was based on kv_self.head while it should be based on llama_kv_cache_cell_max. Existing session files should still work. * llama : fix llama_kv_cache_cell_max inability to return 1 I've also changed its return type to uint32_t, because this function is always used to set the value of uint32_t variables, and because the index already has this type. * llama : fix state size calculation Some bytes in the state were unaccounted for in llama_get_state_size. Since the logits reserve so much space, it did not cause problems. * gguf-dump : support i-quants (ggerganov#5841) Co-authored-by: Black_Fox <radekliska@gmail.com> * llama : allow for user specified embedding pooling type (ggerganov#5849) * allow for user specified pooling type * llama : use enum types over int --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * readme : add API changes section * cuda : fix data race in soft max (ggerganov#5853) * main : support special tokens as reverse/anti prompt (ggerganov#5847) * Support special tokens as reverse/anti prompt. * Tokenize antiprompts only once. * main : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * common : use LLAMA_DEFAULT_SEED (ggerganov#5855) * add some new ops, fix some operators and add batch operations to certain operators. (ggml/747) * cuda: fix group_norm * cuda: add batch inference support for ggml_pad/ggml_upscale * add ggml_arrange * add ggml_timestep_embedding * update ggml_arange/ggml_timestep_embedding tests * cuda: fix im2col * add ggml_arange/ggml_timestep_embbeding support for metal backend * fix some bugs * fix some bugs * Update ggml.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update ggml-cuda.cu Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update ggml-metal.metal Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * modify according to the review comments * ggml : fix compile warnings + code style * ggml : normalize compute_forward calls + fix seg fault in debug * minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com> * sync : ggml * add alias for chat template (ggerganov#5858) * speculative : implement stochastic speculative sampling (ggerganov#5625) * (WIP) Implement stochastic speculative decoding * sample from residual distribution on draft accept failure * fix ggerganov#5657: force greedy sampling with probs when temp is 0 * remove p_accept parameter * fix style * remove unused variables * add srand() in speculative.cpp * replace use of rand() with mt19937 sampling * fixes based on review (@JohannesGaessler) * fix r random generation * randomly select next sequence to verify + fix bug in memory freeing * fix bug in active_seqs sync * fix uniform int distribution initialization * remove warnings from comparison between int and size_t * check grammar in `llama_sample_probability_distribution_impl` * remove malloc code by utilizing vectors * add PR link to README * cmake : handle cases where git index is not found in .git (ggerganov#5844) * Update CMakeLists.txt * Update CMakeLists.txt * ggml : introduce ggml_status (ggml/750) * using enum as an exit code instead of macros * update return type from enum to unsigned int * indentation fix * compound update ggml_compute_exit_code -> ggml_status changed ggml_status from a bit-field type to simple codes ggml_status to string cast * ggml_status to string cast * GGML_CALL was removed Co-authored-by: slaren <slarengh@gmail.com> --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * sync : ggml ggml-ci * ggml : fix unknown status (#0) * flake : fix * llama : fix embeddings (ggerganov#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list * nix: static build (ggerganov#5814) * fix speculative decoding build on windows (ggerganov#5874) * rebase and rm tailing space --------- Co-authored-by: LiangtaoJin <liang-tao.jin@intel.com> Co-authored-by: compilade <113953597+compilade@users.noreply.github.com> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Kawrakow <48489457+ikawrakow@users.noreply.github.com> Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Michael Podvitskiy <podvitskiymichael@gmail.com> Co-authored-by: Pierrick Hymbert <pierrick.hymbert@gmail.com> Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com> Co-authored-by: Nindaleth <Nindaleth@users.noreply.github.com> Co-authored-by: Black_Fox <radekliska@gmail.com> Co-authored-by: Douglas Hanley <thesecretaryofwar@gmail.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: DAN™ <dranger003@gmail.com> Co-authored-by: leejet <leejet714@gmail.com> Co-authored-by: Minsoo Cheong <54794500+mscheong01@users.noreply.github.com> Co-authored-by: Dane Madsen <dane_madsen@hotmail.com> Co-authored-by: hutli <6594598+hutli@users.noreply.github.com> Co-authored-by: Jeffrey Quesnelle <emozilla@nousresearch.com>

* allow for user specified pooling type * llama : use enum types over int --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

@JohannesGaessler

* fix mul_mat fault in cpy_f32_f16 * rm unused function * add wait() for memcpy * restore ci/run.sh, rename struct defination, fix bug in ggml_sycl_op_mul_mat_sycl * fix format issue * llama : fix segfault from unknown model arch name (ggerganov#5820) * llama : fix segfault from unknown model arch name * llama : make all LLM maps const This also requires using `std::map::at` instead of its `operator[]` which does not exist for const maps. * llama : name LLM_ARCH_UNKNOWN to "(unknown)" This avoids errors from `std::map::at` when getting the general name of the model architecture. Using "(unknown)" instead of an empty string as per suggestion ggerganov#5820 (comment) * llama : remove redundant inner const for LLM_TENSOR_NAMES The extra const won't do anything here as const maps return const references to values. Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * llama : remove redundant nullptr check in llm_arch_from_string Since LLM_ARCH_NAMES is a const map, no spurious elements with a NULL name are inserted anymore, so this check is dead code. --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * llama : refactor internal quantization functions (ggerganov#5830) * scripts : add pod-llama.sh * ggml : IQ3_S improvements (ggerganov#5829) * iq3_s: somewhat faster AVX2 dot product On Ryzen a 7950X TG-128 increases to 16 t/s from 15.5 t/s using 16 threads. For 8 threads it is 13.85 t/s vs 11.75 t/s. PP-512 increases to 28.5 t/s from 23.8 t/s. * iq3_s: somewhat faster ARM_NEON dot product Still dog slow - 10.7 t/s up from 9.9 t/s. * iq3_s: another small ARM_NEON improvement 10.7 -> 11.0 t/s. Using vmulq_s8 is faster than the xor - sub trick that works best on AVX2. * iq3_s: minor improvement on Metal 49.4 t/s -> 50.3 t/s * iq3_s: PPL improvement E.g., for a context of 4096 LLaMA-v2-7B goes to 5.1340 from 5.1653. * iq3_s: use new grid everywhere * Fix ARM_NEON --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> * convert-hf : make model class definitions self-contained (ggerganov#5825) * convert : automatically fall back to HfVocab if tokenizer.model doesn't exist (ggerganov#5821) * ggml : fix IQ3_S AVX implementation (ggerganov#5834) ggml-ci * llama : add abort_callback to interrupt computation (ggerganov#5409) * using abort_callback from ggml to stop llama computation * format fix * a brief explaining comment --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * server: tests: passkey challenge / self-extend with context shift demo (ggerganov#5832) * 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 * flake.lock: Update (ggerganov#5842) Flake lock file updates: • Updated input 'flake-parts': 'github:hercules-ci/flake-parts/b253292d9c0a5ead9bc98c4e9a26c6312e27d69f' (2024-02-01) → 'github:hercules-ci/flake-parts/f7b3c975cf067e56e7cda6cb098ebe3fb4d74ca2' (2024-03-01) • Updated input 'flake-parts/nixpkgs-lib': 'github:NixOS/nixpkgs/97b17f32362e475016f942bbdfda4a4a72a8a652?dir=lib' (2024-01-29) → 'github:NixOS/nixpkgs/1536926ef5621b09bba54035ae2bb6d806d72ac8?dir=lib' (2024-02-29) • Updated input 'nixpkgs': 'github:NixOS/nixpkgs/cbc4211f0afffe6dfd2478a62615dd5175a13f9a' (2024-02-23) → 'github:NixOS/nixpkgs/1536926ef5621b09bba54035ae2bb6d806d72ac8' (2024-02-29) Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com> * server : init http requests thread pool with --parallel if set (ggerganov#5836) * ci : schedule slow server tests only on Release or on demand (ggerganov#5839) * llama : fix llama_copy_state_data with fragmented KV cache (ggerganov#5840) The row size of the saved states was based on kv_self.head while it should be based on llama_kv_cache_cell_max. Existing session files should still work. * llama : fix llama_kv_cache_cell_max inability to return 1 I've also changed its return type to uint32_t, because this function is always used to set the value of uint32_t variables, and because the index already has this type. * llama : fix state size calculation Some bytes in the state were unaccounted for in llama_get_state_size. Since the logits reserve so much space, it did not cause problems. * gguf-dump : support i-quants (ggerganov#5841) Co-authored-by: Black_Fox <radekliska@gmail.com> * llama : allow for user specified embedding pooling type (ggerganov#5849) * allow for user specified pooling type * llama : use enum types over int --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * readme : add API changes section * cuda : fix data race in soft max (ggerganov#5853) * main : support special tokens as reverse/anti prompt (ggerganov#5847) * Support special tokens as reverse/anti prompt. * Tokenize antiprompts only once. * main : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * common : use LLAMA_DEFAULT_SEED (ggerganov#5855) * add some new ops, fix some operators and add batch operations to certain operators. (ggml/747) * cuda: fix group_norm * cuda: add batch inference support for ggml_pad/ggml_upscale * add ggml_arrange * add ggml_timestep_embedding * update ggml_arange/ggml_timestep_embedding tests * cuda: fix im2col * add ggml_arange/ggml_timestep_embbeding support for metal backend * fix some bugs * fix some bugs * Update ggml.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update ggml-cuda.cu Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update ggml-metal.metal Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * modify according to the review comments * ggml : fix compile warnings + code style * ggml : normalize compute_forward calls + fix seg fault in debug * minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com> * sync : ggml * add alias for chat template (ggerganov#5858) * speculative : implement stochastic speculative sampling (ggerganov#5625) * (WIP) Implement stochastic speculative decoding * sample from residual distribution on draft accept failure * fix ggerganov#5657: force greedy sampling with probs when temp is 0 * remove p_accept parameter * fix style * remove unused variables * add srand() in speculative.cpp * replace use of rand() with mt19937 sampling * fixes based on review (@JohannesGaessler) * fix r random generation * randomly select next sequence to verify + fix bug in memory freeing * fix bug in active_seqs sync * fix uniform int distribution initialization * remove warnings from comparison between int and size_t * check grammar in `llama_sample_probability_distribution_impl` * remove malloc code by utilizing vectors * add PR link to README * cmake : handle cases where git index is not found in .git (ggerganov#5844) * Update CMakeLists.txt * Update CMakeLists.txt * ggml : introduce ggml_status (ggml/750) * using enum as an exit code instead of macros * update return type from enum to unsigned int * indentation fix * compound update ggml_compute_exit_code -> ggml_status changed ggml_status from a bit-field type to simple codes ggml_status to string cast * ggml_status to string cast * GGML_CALL was removed Co-authored-by: slaren <slarengh@gmail.com> --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * sync : ggml ggml-ci * ggml : fix unknown status (#0) * flake : fix * llama : fix embeddings (ggerganov#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list * nix: static build (ggerganov#5814) * fix speculative decoding build on windows (ggerganov#5874) * rebase and rm tailing space --------- Co-authored-by: LiangtaoJin <liang-tao.jin@intel.com> Co-authored-by: compilade <113953597+compilade@users.noreply.github.com> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Kawrakow <48489457+ikawrakow@users.noreply.github.com> Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Michael Podvitskiy <podvitskiymichael@gmail.com> Co-authored-by: Pierrick Hymbert <pierrick.hymbert@gmail.com> Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com> Co-authored-by: Nindaleth <Nindaleth@users.noreply.github.com> Co-authored-by: Black_Fox <radekliska@gmail.com> Co-authored-by: Douglas Hanley <thesecretaryofwar@gmail.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: DAN™ <dranger003@gmail.com> Co-authored-by: leejet <leejet714@gmail.com> Co-authored-by: Minsoo Cheong <54794500+mscheong01@users.noreply.github.com> Co-authored-by: Dane Madsen <dane_madsen@hotmail.com> Co-authored-by: hutli <6594598+hutli@users.noreply.github.com> Co-authored-by: Jeffrey Quesnelle <emozilla@nousresearch.com>

* allow for user specified pooling type * llama : use enum types over int --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

@JohannesGaessler

* fix mul_mat fault in cpy_f32_f16 * rm unused function * add wait() for memcpy * restore ci/run.sh, rename struct defination, fix bug in ggml_sycl_op_mul_mat_sycl * fix format issue * llama : fix segfault from unknown model arch name (ggerganov#5820) * llama : fix segfault from unknown model arch name * llama : make all LLM maps const This also requires using `std::map::at` instead of its `operator[]` which does not exist for const maps. * llama : name LLM_ARCH_UNKNOWN to "(unknown)" This avoids errors from `std::map::at` when getting the general name of the model architecture. Using "(unknown)" instead of an empty string as per suggestion ggerganov#5820 (comment) * llama : remove redundant inner const for LLM_TENSOR_NAMES The extra const won't do anything here as const maps return const references to values. Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * llama : remove redundant nullptr check in llm_arch_from_string Since LLM_ARCH_NAMES is a const map, no spurious elements with a NULL name are inserted anymore, so this check is dead code. --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * llama : refactor internal quantization functions (ggerganov#5830) * scripts : add pod-llama.sh * ggml : IQ3_S improvements (ggerganov#5829) * iq3_s: somewhat faster AVX2 dot product On Ryzen a 7950X TG-128 increases to 16 t/s from 15.5 t/s using 16 threads. For 8 threads it is 13.85 t/s vs 11.75 t/s. PP-512 increases to 28.5 t/s from 23.8 t/s. * iq3_s: somewhat faster ARM_NEON dot product Still dog slow - 10.7 t/s up from 9.9 t/s. * iq3_s: another small ARM_NEON improvement 10.7 -> 11.0 t/s. Using vmulq_s8 is faster than the xor - sub trick that works best on AVX2. * iq3_s: minor improvement on Metal 49.4 t/s -> 50.3 t/s * iq3_s: PPL improvement E.g., for a context of 4096 LLaMA-v2-7B goes to 5.1340 from 5.1653. * iq3_s: use new grid everywhere * Fix ARM_NEON --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> * convert-hf : make model class definitions self-contained (ggerganov#5825) * convert : automatically fall back to HfVocab if tokenizer.model doesn't exist (ggerganov#5821) * ggml : fix IQ3_S AVX implementation (ggerganov#5834) ggml-ci * llama : add abort_callback to interrupt computation (ggerganov#5409) * using abort_callback from ggml to stop llama computation * format fix * a brief explaining comment --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * server: tests: passkey challenge / self-extend with context shift demo (ggerganov#5832) * 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 * flake.lock: Update (ggerganov#5842) Flake lock file updates: • Updated input 'flake-parts': 'github:hercules-ci/flake-parts/b253292d9c0a5ead9bc98c4e9a26c6312e27d69f' (2024-02-01) → 'github:hercules-ci/flake-parts/f7b3c975cf067e56e7cda6cb098ebe3fb4d74ca2' (2024-03-01) • Updated input 'flake-parts/nixpkgs-lib': 'github:NixOS/nixpkgs/97b17f32362e475016f942bbdfda4a4a72a8a652?dir=lib' (2024-01-29) → 'github:NixOS/nixpkgs/1536926ef5621b09bba54035ae2bb6d806d72ac8?dir=lib' (2024-02-29) • Updated input 'nixpkgs': 'github:NixOS/nixpkgs/cbc4211f0afffe6dfd2478a62615dd5175a13f9a' (2024-02-23) → 'github:NixOS/nixpkgs/1536926ef5621b09bba54035ae2bb6d806d72ac8' (2024-02-29) Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com> * server : init http requests thread pool with --parallel if set (ggerganov#5836) * ci : schedule slow server tests only on Release or on demand (ggerganov#5839) * llama : fix llama_copy_state_data with fragmented KV cache (ggerganov#5840) The row size of the saved states was based on kv_self.head while it should be based on llama_kv_cache_cell_max. Existing session files should still work. * llama : fix llama_kv_cache_cell_max inability to return 1 I've also changed its return type to uint32_t, because this function is always used to set the value of uint32_t variables, and because the index already has this type. * llama : fix state size calculation Some bytes in the state were unaccounted for in llama_get_state_size. Since the logits reserve so much space, it did not cause problems. * gguf-dump : support i-quants (ggerganov#5841) Co-authored-by: Black_Fox <radekliska@gmail.com> * llama : allow for user specified embedding pooling type (ggerganov#5849) * allow for user specified pooling type * llama : use enum types over int --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * readme : add API changes section * cuda : fix data race in soft max (ggerganov#5853) * main : support special tokens as reverse/anti prompt (ggerganov#5847) * Support special tokens as reverse/anti prompt. * Tokenize antiprompts only once. * main : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * common : use LLAMA_DEFAULT_SEED (ggerganov#5855) * add some new ops, fix some operators and add batch operations to certain operators. (ggml/747) * cuda: fix group_norm * cuda: add batch inference support for ggml_pad/ggml_upscale * add ggml_arrange * add ggml_timestep_embedding * update ggml_arange/ggml_timestep_embedding tests * cuda: fix im2col * add ggml_arange/ggml_timestep_embbeding support for metal backend * fix some bugs * fix some bugs * Update ggml.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update ggml-cuda.cu Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update ggml-metal.metal Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * modify according to the review comments * ggml : fix compile warnings + code style * ggml : normalize compute_forward calls + fix seg fault in debug * minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com> * sync : ggml * add alias for chat template (ggerganov#5858) * speculative : implement stochastic speculative sampling (ggerganov#5625) * (WIP) Implement stochastic speculative decoding * sample from residual distribution on draft accept failure * fix ggerganov#5657: force greedy sampling with probs when temp is 0 * remove p_accept parameter * fix style * remove unused variables * add srand() in speculative.cpp * replace use of rand() with mt19937 sampling * fixes based on review (@JohannesGaessler) * fix r random generation * randomly select next sequence to verify + fix bug in memory freeing * fix bug in active_seqs sync * fix uniform int distribution initialization * remove warnings from comparison between int and size_t * check grammar in `llama_sample_probability_distribution_impl` * remove malloc code by utilizing vectors * add PR link to README * cmake : handle cases where git index is not found in .git (ggerganov#5844) * Update CMakeLists.txt * Update CMakeLists.txt * ggml : introduce ggml_status (ggml/750) * using enum as an exit code instead of macros * update return type from enum to unsigned int * indentation fix * compound update ggml_compute_exit_code -> ggml_status changed ggml_status from a bit-field type to simple codes ggml_status to string cast * ggml_status to string cast * GGML_CALL was removed Co-authored-by: slaren <slarengh@gmail.com> --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * sync : ggml ggml-ci * ggml : fix unknown status (#0) * flake : fix * llama : fix embeddings (ggerganov#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list * nix: static build (ggerganov#5814) * fix speculative decoding build on windows (ggerganov#5874) * rebase and rm tailing space --------- Co-authored-by: LiangtaoJin <liang-tao.jin@intel.com> Co-authored-by: compilade <113953597+compilade@users.noreply.github.com> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Kawrakow <48489457+ikawrakow@users.noreply.github.com> Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Michael Podvitskiy <podvitskiymichael@gmail.com> Co-authored-by: Pierrick Hymbert <pierrick.hymbert@gmail.com> Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com> Co-authored-by: Nindaleth <Nindaleth@users.noreply.github.com> Co-authored-by: Black_Fox <radekliska@gmail.com> Co-authored-by: Douglas Hanley <thesecretaryofwar@gmail.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: DAN™ <dranger003@gmail.com> Co-authored-by: leejet <leejet714@gmail.com> Co-authored-by: Minsoo Cheong <54794500+mscheong01@users.noreply.github.com> Co-authored-by: Dane Madsen <dane_madsen@hotmail.com> Co-authored-by: hutli <6594598+hutli@users.noreply.github.com> Co-authored-by: Jeffrey Quesnelle <emozilla@nousresearch.com>

allow for user specified pooling type

efecd06

ggerganov approved these changes Mar 3, 2024

View reviewed changes

ggerganov requested changes Mar 3, 2024

View reviewed changes

llama : use enum types over int

4661363

ggerganov merged commit 475df1d into ggerganov:master Mar 3, 2024
60 of 61 checks passed

ggerganov added the breaking change Changes that break ABIs, APIs, file formats, or other forms of backwards compatibility. label Mar 3, 2024

SignalRT mentioned this pull request Mar 3, 2024

Llava api SciSharp/LLamaSharp#563

Merged

4 tasks

cebtenzzre reviewed Mar 3, 2024

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Allow for user specified embedding pooling type #5849

Allow for user specified embedding pooling type #5849

iamlemec commented Mar 3, 2024

ggerganov Mar 3, 2024

iamlemec Mar 3, 2024

cebtenzzre Mar 3, 2024

ggerganov Mar 4, 2024 •

edited

Loading

		enum llama_rope_scaling_type rope_scaling_type; // RoPE scaling type, from `enum llama_rope_scaling_type`
		enum llama_pooling_type pooling_type; // whether to pool (sum) embedding results by sequence id

Allow for user specified embedding pooling type #5849

Allow for user specified embedding pooling type #5849

Conversation

iamlemec commented Mar 3, 2024

ggerganov Mar 3, 2024

Choose a reason for hiding this comment

iamlemec Mar 3, 2024

Choose a reason for hiding this comment

cebtenzzre Mar 3, 2024

Choose a reason for hiding this comment

ggerganov Mar 4, 2024 • edited Loading

Choose a reason for hiding this comment

ggerganov Mar 4, 2024 •

edited

Loading