From 9ecfa16b12c13ac6ed136929258829208ed8afc5 Mon Sep 17 00:00:00 2001 From: Nicolas Patry Date: Mon, 11 Dec 2023 12:46:30 +0100 Subject: [PATCH] Speculative (#1308) --- Cargo.lock | 66 +-- docs/source/basic_tutorials/launcher.md | 8 + .../test_flash_medusa_all_params.json | 98 +++++ .../test_flash_medusa_load.json | 414 ++++++++++++++++++ .../test_flash_medusa_simple.json | 103 +++++ integration-tests/models/test_flash_medusa.py | 59 +++ .../models/test_flash_mistral.py | 4 +- launcher/src/main.rs | 15 + load_tests/common.js | 21 +- load_tests/tgi.js | 4 +- proto/generate.proto | 39 +- router/client/src/client.rs | 4 +- router/client/src/lib.rs | 8 +- router/src/infer.rs | 94 ++-- router/src/queue.rs | 61 ++- router/src/server.rs | 1 + server/Makefile-vllm | 33 +- server/tests/models/test_bloom.py | 4 +- server/tests/models/test_causal_lm.py | 4 +- server/tests/models/test_seq2seq_lm.py | 4 +- server/text_generation_server/cli.py | 28 +- .../text_generation_server/models/__init__.py | 36 +- .../models/causal_lm.py | 19 +- .../models/flash_causal_lm.py | 188 +++++--- .../models/flash_llama.py | 13 + .../models/flash_mistral.py | 63 ++- .../models/idefics_causal_lm.py | 16 +- server/text_generation_server/models/model.py | 7 + .../models/seq2seq_lm.py | 18 +- server/text_generation_server/models/types.py | 40 +- server/text_generation_server/server.py | 6 +- server/text_generation_server/utils/medusa.py | 51 +++ .../text_generation_server/utils/speculate.py | 12 + server/text_generation_server/utils/tokens.py | 92 +++- 34 files changed, 1350 insertions(+), 283 deletions(-) create mode 100644 integration-tests/models/__snapshots__/test_flash_medusa/test_flash_medusa_all_params.json create mode 100644 integration-tests/models/__snapshots__/test_flash_medusa/test_flash_medusa_load.json create mode 100644 integration-tests/models/__snapshots__/test_flash_medusa/test_flash_medusa_simple.json create mode 100644 integration-tests/models/test_flash_medusa.py create mode 100644 server/text_generation_server/utils/medusa.py create mode 100644 server/text_generation_server/utils/speculate.py diff --git a/Cargo.lock b/Cargo.lock index fa1a45a7..93e13056 100644 --- a/Cargo.lock +++ b/Cargo.lock @@ -60,30 +60,30 @@ checksum = "7079075b41f533b8c61d2a4d073c4676e1f8b249ff94a393b0595db304e0dd87" [[package]] name = "anstyle-parse" -version = "0.2.2" +version = "0.2.3" source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "317b9a89c1868f5ea6ff1d9539a69f45dffc21ce321ac1fd1160dfa48c8e2140" +checksum = "c75ac65da39e5fe5ab759307499ddad880d724eed2f6ce5b5e8a26f4f387928c" dependencies = [ "utf8parse", ] [[package]] name = "anstyle-query" -version = "1.0.0" +version = "1.0.1" source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "5ca11d4be1bab0c8bc8734a9aa7bf4ee8316d462a08c6ac5052f888fef5b494b" +checksum = "a3a318f1f38d2418400f8209655bfd825785afd25aa30bb7ba6cc792e4596748" dependencies = [ - "windows-sys 0.48.0", + "windows-sys 0.52.0", ] [[package]] name = "anstyle-wincon" -version = "3.0.1" +version = "3.0.2" source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "f0699d10d2f4d628a98ee7b57b289abbc98ff3bad977cb3152709d4bf2330628" +checksum = "1cd54b81ec8d6180e24654d0b371ad22fc3dd083b6ff8ba325b72e00c87660a7" dependencies = [ "anstyle", - "windows-sys 0.48.0", + "windows-sys 0.52.0", ] [[package]] @@ -333,9 +333,9 @@ checksum = "baf1de4339761588bc0619e3cbc0120ee582ebb74b53b4efbf79117bd2da40fd" [[package]] name = "clap" -version = "4.4.10" +version = "4.4.11" source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "41fffed7514f420abec6d183b1d3acfd9099c79c3a10a06ade4f8203f1411272" +checksum = "bfaff671f6b22ca62406885ece523383b9b64022e341e53e009a62ebc47a45f2" dependencies = [ "clap_builder", "clap_derive", @@ -343,9 +343,9 @@ dependencies = [ [[package]] name = "clap_builder" -version = "4.4.9" +version = "4.4.11" source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "63361bae7eef3771745f02d8d892bec2fee5f6e34af316ba556e7f97a7069ff1" +checksum = "a216b506622bb1d316cd51328dce24e07bdff4a6128a47c7e7fad11878d5adbb" dependencies = [ "anstream", "anstyle", @@ -392,9 +392,9 @@ dependencies = [ [[package]] name = "core-foundation" -version = "0.9.3" +version = "0.9.4" source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "194a7a9e6de53fa55116934067c844d9d749312f75c6f6d0980e8c252f8c2146" +checksum = "91e195e091a93c46f7102ec7818a2aa394e1e1771c3ab4825963fa03e45afb8f" dependencies = [ "core-foundation-sys", "libc", @@ -402,9 +402,9 @@ dependencies = [ [[package]] name = "core-foundation-sys" -version = "0.8.4" +version = "0.8.6" source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "e496a50fda8aacccc86d7529e2c1e0892dbd0f898a6b5645b5561b89c3210efa" +checksum = "06ea2b9bc92be3c2baa9334a323ebca2d6f074ff852cd1d7b11064035cd3868f" [[package]] name = "cpufeatures" @@ -549,9 +549,9 @@ dependencies = [ [[package]] name = "deranged" -version = "0.3.9" +version = "0.3.10" source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "0f32d04922c60427da6f9fef14d042d9edddef64cb9d4ce0d64d0685fbeb1fd3" +checksum = "8eb30d70a07a3b04884d2677f06bec33509dc67ca60d92949e5535352d3191dc" dependencies = [ "powerfmt", ] @@ -1218,9 +1218,9 @@ dependencies = [ [[package]] name = "linux-raw-sys" -version = "0.4.11" +version = "0.4.12" source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "969488b55f8ac402214f3f5fd243ebb7206cf82de60d3172994707a4bcc2b829" +checksum = "c4cd1a83af159aa67994778be9070f0ae1bd732942279cabb14f86f986a21456" [[package]] name = "lock_api" @@ -1612,9 +1612,9 @@ dependencies = [ [[package]] name = "openssl" -version = "0.10.60" +version = "0.10.61" source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "79a4c6c3a2b158f7f8f2a2fc5a969fa3a068df6fc9dbb4a43845436e3af7c800" +checksum = "6b8419dc8cc6d866deb801274bba2e6f8f6108c1bb7fcc10ee5ab864931dbb45" dependencies = [ "bitflags 2.4.1", "cfg-if", @@ -1644,9 +1644,9 @@ checksum = "ff011a302c396a5197692431fc1948019154afc178baf7d8e37367442a4601cf" [[package]] name = "openssl-sys" -version = "0.9.96" +version = "0.9.97" source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "3812c071ba60da8b5677cc12bcb1d42989a65553772897a7e0355545a819838f" +checksum = "c3eaad34cdd97d81de97964fc7f29e2d104f483840d906ef56daa1912338460b" dependencies = [ "cc", "libc", @@ -2309,15 +2309,15 @@ dependencies = [ [[package]] name = "rustix" -version = "0.38.25" +version = "0.38.26" source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "dc99bc2d4f1fed22595588a013687477aedf3cdcfb26558c559edb67b4d9b22e" +checksum = "9470c4bf8246c8daf25f9598dca807fb6510347b1e1cfa55749113850c79d88a" dependencies = [ "bitflags 2.4.1", "errno", "libc", "linux-raw-sys", - "windows-sys 0.48.0", + "windows-sys 0.52.0", ] [[package]] @@ -2567,9 +2567,9 @@ dependencies = [ [[package]] name = "slotmap" -version = "1.0.6" +version = "1.0.7" source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "e1e08e261d0e8f5c43123b7adf3e4ca1690d655377ac93a03b2c9d3e98de1342" +checksum = "dbff4acf519f630b3a3ddcfaea6c06b42174d9a44bc70c620e9ed1649d58b82a" dependencies = [ "version_check", ] @@ -3835,18 +3835,18 @@ dependencies = [ [[package]] name = "zerocopy" -version = "0.7.26" +version = "0.7.28" source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "e97e415490559a91254a2979b4829267a57d2fcd741a98eee8b722fb57289aa0" +checksum = "7d6f15f7ade05d2a4935e34a457b936c23dc70a05cc1d97133dc99e7a3fe0f0e" dependencies = [ "zerocopy-derive", ] [[package]] name = "zerocopy-derive" -version = "0.7.26" +version = "0.7.28" source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "dd7e48ccf166952882ca8bd778a43502c64f33bf94c12ebe2a7f08e5a0f6689f" +checksum = "dbbad221e3f78500350ecbd7dfa4e63ef945c05f4c61cb7f4d3f84cd0bba649b" dependencies = [ "proc-macro2", "quote", diff --git a/docs/source/basic_tutorials/launcher.md b/docs/source/basic_tutorials/launcher.md index 62abe8c6..9590e463 100644 --- a/docs/source/basic_tutorials/launcher.md +++ b/docs/source/basic_tutorials/launcher.md @@ -67,6 +67,14 @@ Options: - bitsandbytes-nf4: Bitsandbytes 4bit. Can be applied on any model, will cut the memory requirement by 4x, but it is known that the model will be much slower to run than the native f16 - bitsandbytes-fp4: Bitsandbytes 4bit. nf4 should be preferred in most cases but maybe this one has better perplexity performance for you model +``` +## SPECULATE +```shell + --speculate + The number of input_ids to speculate on If using a medusa model, the heads will be picked up automatically Other wise, it will use n-gram speculation which is relatively free in terms of compute, but the speedup heavily depends on the task + + [env: SPECULATE=] + ``` ## DTYPE ```shell diff --git a/integration-tests/models/__snapshots__/test_flash_medusa/test_flash_medusa_all_params.json b/integration-tests/models/__snapshots__/test_flash_medusa/test_flash_medusa_all_params.json new file mode 100644 index 00000000..d8a298eb --- /dev/null +++ b/integration-tests/models/__snapshots__/test_flash_medusa/test_flash_medusa_all_params.json @@ -0,0 +1,98 @@ +{ + "details": { + "best_of_sequences": null, + "finish_reason": "length", + "generated_tokens": 10, + "prefill": [ + { + "id": 1, + "logprob": null, + "text": "" + }, + { + "id": 338, + "logprob": -10.0078125, + "text": "is" + }, + { + "id": 21784, + "logprob": -15.515625, + "text": "Deep" + }, + { + "id": 29257, + "logprob": -2.8847656, + "text": "Learning" + }, + { + "id": 29973, + "logprob": -4.140625, + "text": "?" + } + ], + "seed": 0, + "tokens": [ + { + "id": 13, + "logprob": -1.1582031, + "special": false, + "text": "\n" + }, + { + "id": 2772, + "logprob": -0.23083496, + "special": false, + "text": "De" + }, + { + "id": 1022, + "logprob": 0.0, + "special": false, + "text": "ep" + }, + { + "id": 6509, + "logprob": 0.0, + "special": false, + "text": " learning" + }, + { + "id": 29892, + "logprob": -0.61816406, + "special": false, + "text": "," + }, + { + "id": 607, + "logprob": -0.7089844, + "special": false, + "text": " which" + }, + { + "id": 508, + "logprob": -1.7724609, + "special": false, + "text": " can" + }, + { + "id": 367, + "logprob": 0.0, + "special": false, + "text": " be" + }, + { + "id": 5545, + "logprob": 0.0, + "special": false, + "text": " considered" + }, + { + "id": 408, + "logprob": -0.3869629, + "special": false, + "text": " as" + } + ] + }, + "generated_text": "What is Deep Learning?\nDeep learning, which can be considered as" +} diff --git a/integration-tests/models/__snapshots__/test_flash_medusa/test_flash_medusa_load.json b/integration-tests/models/__snapshots__/test_flash_medusa/test_flash_medusa_load.json new file mode 100644 index 00000000..413af1d7 --- /dev/null +++ b/integration-tests/models/__snapshots__/test_flash_medusa/test_flash_medusa_load.json @@ -0,0 +1,414 @@ +[ + { + "details": { + "best_of_sequences": null, + "finish_reason": "length", + "generated_tokens": 10, + "prefill": [ + { + "id": 1, + "logprob": null, + "text": "" + }, + { + "id": 1724, + "logprob": -10.734375, + "text": "What" + }, + { + "id": 338, + "logprob": -1.5488281, + "text": "is" + }, + { + "id": 21784, + "logprob": -9.2890625, + "text": "Deep" + }, + { + "id": 29257, + "logprob": -1.2753906, + "text": "Learning" + }, + { + "id": 29973, + "logprob": -0.48046875, + "text": "?" + } + ], + "seed": null, + "tokens": [ + { + "id": 13, + "logprob": -1.1845703, + "special": false, + "text": "\n" + }, + { + "id": 2772, + "logprob": -0.5727539, + "special": false, + "text": "De" + }, + { + "id": 1022, + "logprob": -0.00010967255, + "special": false, + "text": "ep" + }, + { + "id": 6509, + "logprob": -0.1239624, + "special": false, + "text": " learning" + }, + { + "id": 338, + "logprob": -0.04510498, + "special": false, + "text": " is" + }, + { + "id": 263, + "logprob": -0.018295288, + "special": false, + "text": " a" + }, + { + "id": 11306, + "logprob": -0.45922852, + "special": false, + "text": " subset" + }, + { + "id": 310, + "logprob": -0.00020992756, + "special": false, + "text": " of" + }, + { + "id": 4933, + "logprob": -0.0046539307, + "special": false, + "text": " machine" + }, + { + "id": 6509, + "logprob": -0.00025844574, + "special": false, + "text": " learning" + } + ] + }, + "generated_text": "\nDeep learning is a subset of machine learning" + }, + { + "details": { + "best_of_sequences": null, + "finish_reason": "length", + "generated_tokens": 10, + "prefill": [ + { + "id": 1, + "logprob": null, + "text": "" + }, + { + "id": 1724, + "logprob": -10.734375, + "text": "What" + }, + { + "id": 338, + "logprob": -1.5488281, + "text": "is" + }, + { + "id": 21784, + "logprob": -9.2890625, + "text": "Deep" + }, + { + "id": 29257, + "logprob": -1.2724609, + "text": "Learning" + }, + { + "id": 29973, + "logprob": -0.47729492, + "text": "?" + } + ], + "seed": null, + "tokens": [ + { + "id": 13, + "logprob": -1.1826172, + "special": false, + "text": "\n" + }, + { + "id": 2772, + "logprob": -0.56689453, + "special": false, + "text": "De" + }, + { + "id": 1022, + "logprob": -0.000108003616, + "special": false, + "text": "ep" + }, + { + "id": 6509, + "logprob": -0.1239624, + "special": false, + "text": " learning" + }, + { + "id": 338, + "logprob": -0.044433594, + "special": false, + "text": " is" + }, + { + "id": 263, + "logprob": -0.018295288, + "special": false, + "text": " a" + }, + { + "id": 11306, + "logprob": -0.45922852, + "special": false, + "text": " subset" + }, + { + "id": 310, + "logprob": -0.0002104044, + "special": false, + "text": " of" + }, + { + "id": 4933, + "logprob": -0.004711151, + "special": false, + "text": " machine" + }, + { + "id": 6509, + "logprob": -0.00025892258, + "special": false, + "text": " learning" + } + ] + }, + "generated_text": "\nDeep learning is a subset of machine learning" + }, + { + "details": { + "best_of_sequences": null, + "finish_reason": "length", + "generated_tokens": 10, + "prefill": [ + { + "id": 1, + "logprob": null, + "text": "" + }, + { + "id": 1724, + "logprob": -10.734375, + "text": "What" + }, + { + "id": 338, + "logprob": -1.5488281, + "text": "is" + }, + { + "id": 21784, + "logprob": -9.2890625, + "text": "Deep" + }, + { + "id": 29257, + "logprob": -1.2724609, + "text": "Learning" + }, + { + "id": 29973, + "logprob": -0.47729492, + "text": "?" + } + ], + "seed": null, + "tokens": [ + { + "id": 13, + "logprob": -1.1826172, + "special": false, + "text": "\n" + }, + { + "id": 2772, + "logprob": -0.56689453, + "special": false, + "text": "De" + }, + { + "id": 1022, + "logprob": -0.000108003616, + "special": false, + "text": "ep" + }, + { + "id": 6509, + "logprob": -0.1239624, + "special": false, + "text": " learning" + }, + { + "id": 338, + "logprob": -0.044433594, + "special": false, + "text": " is" + }, + { + "id": 263, + "logprob": -0.018295288, + "special": false, + "text": " a" + }, + { + "id": 11306, + "logprob": -0.45922852, + "special": false, + "text": " subset" + }, + { + "id": 310, + "logprob": -0.0002104044, + "special": false, + "text": " of" + }, + { + "id": 4933, + "logprob": -0.004711151, + "special": false, + "text": " machine" + }, + { + "id": 6509, + "logprob": -0.00025892258, + "special": false, + "text": " learning" + } + ] + }, + "generated_text": "\nDeep learning is a subset of machine learning" + }, + { + "details": { + "best_of_sequences": null, + "finish_reason": "length", + "generated_tokens": 10, + "prefill": [ + { + "id": 1, + "logprob": null, + "text": "" + }, + { + "id": 1724, + "logprob": -10.734375, + "text": "What" + }, + { + "id": 338, + "logprob": -1.5488281, + "text": "is" + }, + { + "id": 21784, + "logprob": -9.2890625, + "text": "Deep" + }, + { + "id": 29257, + "logprob": -1.2724609, + "text": "Learning" + }, + { + "id": 29973, + "logprob": -0.47729492, + "text": "?" + } + ], + "seed": null, + "tokens": [ + { + "id": 13, + "logprob": -1.1826172, + "special": false, + "text": "\n" + }, + { + "id": 2772, + "logprob": -0.56689453, + "special": false, + "text": "De" + }, + { + "id": 1022, + "logprob": -0.000108003616, + "special": false, + "text": "ep" + }, + { + "id": 6509, + "logprob": -0.1239624, + "special": false, + "text": " learning" + }, + { + "id": 338, + "logprob": -0.044433594, + "special": false, + "text": " is" + }, + { + "id": 263, + "logprob": -0.018295288, + "special": false, + "text": " a" + }, + { + "id": 11306, + "logprob": -0.45922852, + "special": false, + "text": " subset" + }, + { + "id": 310, + "logprob": -0.0002104044, + "special": false, + "text": " of" + }, + { + "id": 4933, + "logprob": -0.004711151, + "special": false, + "text": " machine" + }, + { + "id": 6509, + "logprob": -0.00025892258, + "special": false, + "text": " learning" + } + ] + }, + "generated_text": "\nDeep learning is a subset of machine learning" + } +] diff --git a/integration-tests/models/__snapshots__/test_flash_medusa/test_flash_medusa_simple.json b/integration-tests/models/__snapshots__/test_flash_medusa/test_flash_medusa_simple.json new file mode 100644 index 00000000..15754b14 --- /dev/null +++ b/integration-tests/models/__snapshots__/test_flash_medusa/test_flash_medusa_simple.json @@ -0,0 +1,103 @@ +{ + "details": { + "best_of_sequences": null, + "finish_reason": "length", + "generated_tokens": 10, + "prefill": [ + { + "id": 1, + "logprob": null, + "text": "" + }, + { + "id": 1724, + "logprob": -10.734375, + "text": "What" + }, + { + "id": 338, + "logprob": -1.5488281, + "text": "is" + }, + { + "id": 21784, + "logprob": -9.2890625, + "text": "Deep" + }, + { + "id": 29257, + "logprob": -1.2753906, + "text": "Learning" + }, + { + "id": 29973, + "logprob": -0.48046875, + "text": "?" + } + ], + "seed": null, + "tokens": [ + { + "id": 13, + "logprob": -1.1845703, + "special": false, + "text": "\n" + }, + { + "id": 2772, + "logprob": -0.5727539, + "special": false, + "text": "De" + }, + { + "id": 1022, + "logprob": -0.000108122826, + "special": false, + "text": "ep" + }, + { + "id": 6509, + "logprob": -0.1239624, + "special": false, + "text": " learning" + }, + { + "id": 338, + "logprob": -0.044433594, + "special": false, + "text": " is" + }, + { + "id": 263, + "logprob": -0.01852417, + "special": false, + "text": " a" + }, + { + "id": 11306, + "logprob": -0.45922852, + "special": false, + "text": " subset" + }, + { + "id": 310, + "logprob": -0.0002104044, + "special": false, + "text": " of" + }, + { + "id": 4933, + "logprob": -0.004787445, + "special": false, + "text": " machine" + }, + { + "id": 6509, + "logprob": -0.00026226044, + "special": false, + "text": " learning" + } + ] + }, + "generated_text": "\nDeep learning is a subset of machine learning" +} diff --git a/integration-tests/models/test_flash_medusa.py b/integration-tests/models/test_flash_medusa.py new file mode 100644 index 00000000..003409b0 --- /dev/null +++ b/integration-tests/models/test_flash_medusa.py @@ -0,0 +1,59 @@ +import pytest + + +@pytest.fixture(scope="module") +def flash_medusa_handle(launcher): + with launcher("FasterDecoding/medusa-vicuna-7b-v1.3", num_shard=2) as handle: + yield handle + + +@pytest.fixture(scope="module") +async def flash_medusa(flash_medusa_handle): + await flash_medusa_handle.health(300) + return flash_medusa_handle.client + + +@pytest.mark.asyncio +@pytest.mark.private +async def test_flash_medusa_simple(flash_medusa, response_snapshot): + response = await flash_medusa.generate( + "What is Deep Learning?", max_new_tokens=10, decoder_input_details=True + ) + + assert response.details.generated_tokens == 10 + assert response == response_snapshot + + +@pytest.mark.asyncio +@pytest.mark.private +async def test_flash_medusa_all_params(flash_medusa, response_snapshot): + response = await flash_medusa.generate( + "What is Deep Learning?", + max_new_tokens=10, + repetition_penalty=1.2, + return_full_text=True, + stop_sequences=["test"], + temperature=0.5, + top_p=0.9, + top_k=10, + truncate=5, + typical_p=0.9, + watermark=True, + decoder_input_details=True, + seed=0, + ) + + assert response.details.generated_tokens == 10 + assert response == response_snapshot + + +@pytest.mark.asyncio +@pytest.mark.private +async def test_flash_medusa_load(flash_medusa, generate_load, response_snapshot): + responses = await generate_load(flash_medusa, "What is Deep Learning?", max_new_tokens=10, n=4) + + assert len(responses) == 4 + assert all([r.generated_text == responses[0].generated_text for r in responses]), f"{[r.generated_text for r in responses]}" + assert responses[0].generated_text == '\nDeep learning is a subset of machine learning' + + assert responses == response_snapshot diff --git a/integration-tests/models/test_flash_mistral.py b/integration-tests/models/test_flash_mistral.py index 63cb09b5..7d21afd9 100644 --- a/integration-tests/models/test_flash_mistral.py +++ b/integration-tests/models/test_flash_mistral.py @@ -21,6 +21,7 @@ async def test_flash_mistral(flash_mistral, response_snapshot): ) assert response.details.generated_tokens == 10 + assert response.generated_text == ": Let n = 10 - 1" assert response == response_snapshot @@ -55,6 +56,7 @@ async def test_flash_mistral_load(flash_mistral, generate_load, response_snapsho ) assert len(responses) == 4 - assert all([r.generated_text == responses[0].generated_text for r in responses]) + assert all([r.generated_text == responses[0].generated_text for r in responses]), f"{[r.generated_text for r in responses]}" + assert responses[0].generated_text == ": Let n = 10 - 1" assert responses == response_snapshot diff --git a/launcher/src/main.rs b/launcher/src/main.rs index b4fc86b7..4e230205 100644 --- a/launcher/src/main.rs +++ b/launcher/src/main.rs @@ -155,6 +155,13 @@ struct Args { #[clap(long, env, value_enum)] quantize: Option, + /// The number of input_ids to speculate on + /// If using a medusa model, the heads will be picked up automatically + /// Other wise, it will use n-gram speculation which is relatively free + /// in terms of compute, but the speedup heavily depends on the task. + #[clap(long, env)] + speculate: Option, + /// The dtype to be forced upon the model. This option cannot be used with `--quantize`. #[clap(long, env, value_enum)] dtype: Option, @@ -375,6 +382,7 @@ fn shard_manager( model_id: String, revision: Option, quantize: Option, + speculate: Option, dtype: Option, trust_remote_code: bool, uds_path: String, @@ -432,6 +440,11 @@ fn shard_manager( shard_args.push(quantize.to_string()) } + if let Some(speculate) = speculate { + shard_args.push("--speculate".to_string()); + shard_args.push(speculate.to_string()) + } + if let Some(dtype) = dtype { shard_args.push("--dtype".to_string()); shard_args.push(dtype.to_string()) @@ -882,6 +895,7 @@ fn spawn_shards( let shutdown_sender = shutdown_sender.clone(); let otlp_endpoint = args.otlp_endpoint.clone(); let quantize = args.quantize; + let speculate = args.speculate; let dtype = args.dtype; let trust_remote_code = args.trust_remote_code; let master_port = args.master_port; @@ -896,6 +910,7 @@ fn spawn_shards( model_id, revision, quantize, + speculate, dtype, trust_remote_code, uds_path, diff --git a/load_tests/common.js b/load_tests/common.js index be812e9b..5d71abea 100644 --- a/load_tests/common.js +++ b/load_tests/common.js @@ -7,7 +7,9 @@ const seed = 0; const host = __ENV.HOST || '127.0.0.1:8000'; const timePerToken = new Trend('time_per_token', true); -const throughput = new Counter('tokens_per_s'); +const tokens = new Counter('tokens'); +const new_tokens = new Counter('new_tokens'); +const input_tokens = new Counter('input_tokens'); randomSeed(seed); // const shareGPT = JSON.parse(open("ShareGPT_V3_unfiltered_cleaned_split.json")) @@ -19,7 +21,7 @@ export function get_options(reference_latency_ms){ thresholds: { http_req_failed: ['rate==0'], time_per_token: [{ - threshold: `p(50)<${3 * reference_latency_ms}`, + threshold: `p(50)<${5 * reference_latency_ms}`, abortOnFail: true, delayAbortEval: '10s' }], @@ -28,7 +30,7 @@ export function get_options(reference_latency_ms){ load_test: { executor: 'constant-arrival-rate', duration: '60s', - preAllocatedVUs: 100, + preAllocatedVUs: 10, rate: 10, timeUnit: '1s', }, @@ -48,17 +50,22 @@ export function run(host, generate_payload, max_new_tokens) { return; } + check(res, { 'Post status is 200': (r) => res.status === 200, }); - const n_tokens = max_new_tokens; - const timings = res.timings.duration; + const duration = res.timings.duration; if (res.status === 200) { - const latency_ms_per_token = timings / n_tokens; + const body = res.json(); + const n_tokens = body.details.tokens.length; + const latency_ms_per_token = duration / n_tokens; timePerToken.add(latency_ms_per_token); const latency_in_s = latency_ms_per_token / 1000; const individual_throughput = 1 / latency_in_s; - throughput.add(individual_throughput); + const _input_tokens = body.details.prefill.length; + tokens.add(n_tokens + _input_tokens); + input_tokens.add(_input_tokens); + new_tokens.add(n_tokens); } } diff --git a/load_tests/tgi.js b/load_tests/tgi.js index 93a0e278..1db4ab6f 100644 --- a/load_tests/tgi.js +++ b/load_tests/tgi.js @@ -1,13 +1,13 @@ import { get_options, run } from "./common.js"; -const reference_latency_ms = 30; +const reference_latency_ms = 70; const host = __ENV.HOST || '127.0.0.1:8000'; const max_new_tokens = 50; function generate_payload(gpt){ const input = gpt["conversations"][0]["value"]; - return {"inputs": input, "parameters": {"max_new_tokens": max_new_tokens, "temperature" : 0.5}} + return {"inputs": input, "parameters": {"max_new_tokens": max_new_tokens, "decoder_input_details": true}} } export const options = get_options(reference_latency_ms); diff --git a/proto/generate.proto b/proto/generate.proto index c873e661..19ec059b 100644 --- a/proto/generate.proto +++ b/proto/generate.proto @@ -1,6 +1,6 @@ syntax = "proto3"; -package generate.v1; +package generate.v2; service TextGenerationService { /// Model Info @@ -32,6 +32,7 @@ message InfoResponse { string dtype = 2; string device_type = 3; optional uint32 window_size = 4; + uint32 speculate = 5; } /// Empty request @@ -135,43 +136,27 @@ message GeneratedText { optional uint64 seed = 4; } -message PrefillTokens { - /// Prefill Token IDs +message Tokens { + /// Token IDs repeated uint32 ids = 1; - /// Prefill Logprobs + /// Logprobs repeated float logprobs = 2; - /// Prefill tokens + /// tokens repeated string texts = 3; -} - -message TopTokens { - /// Top Token IDs - repeated uint32 ids = 1; - /// Top Logprobs - repeated float logprobs = 2; - /// Top Token Texts - repeated string texts = 3; - /// If the tokens are special - repeated bool is_special = 6; + /// special + repeated bool is_special = 4; } message Generation { /// Request ID uint64 request_id = 1; /// Prefill tokens (optional) - PrefillTokens prefill_tokens = 2; - /// Token ID - uint32 token_id = 3; - /// Logprob - float token_logprob = 4; - /// Text - string token_text = 5; - /// Is it a special token - bool token_is_special = 6; + Tokens prefill_tokens = 2; + Tokens tokens = 3; /// Complete generated text - optional GeneratedText generated_text = 7; + optional GeneratedText generated_text = 4; /// Top tokens - TopTokens top_tokens = 8; + repeated Tokens top_tokens = 5; } message FilterBatchRequest { diff --git a/router/client/src/client.rs b/router/client/src/client.rs index 341e70fd..1560f19c 100644 --- a/router/client/src/client.rs +++ b/router/client/src/client.rs @@ -1,6 +1,6 @@ /// Single shard Client -use crate::pb::generate::v1::text_generation_service_client::TextGenerationServiceClient; -use crate::pb::generate::v1::*; +use crate::pb::generate::v2::text_generation_service_client::TextGenerationServiceClient; +use crate::pb::generate::v2::*; use crate::Result; use grpc_metadata::InjectTelemetryContext; use std::cmp::min; diff --git a/router/client/src/lib.rs b/router/client/src/lib.rs index f334be21..c38b931b 100644 --- a/router/client/src/lib.rs +++ b/router/client/src/lib.rs @@ -6,11 +6,11 @@ mod pb; mod sharded_client; pub use client::Client; -pub use pb::generate::v1::HealthResponse; -pub use pb::generate::v1::InfoResponse as ShardInfo; -pub use pb::generate::v1::{ +pub use pb::generate::v2::HealthResponse; +pub use pb::generate::v2::InfoResponse as ShardInfo; +pub use pb::generate::v2::{ Batch, CachedBatch, FinishReason, GeneratedText, Generation, NextTokenChooserParameters, - PrefillTokens, Request, StoppingCriteriaParameters, + Request, StoppingCriteriaParameters, Tokens, }; pub use sharded_client::ShardedClient; use thiserror::Error; diff --git a/router/src/infer.rs b/router/src/infer.rs index aa6dc664..2e199ce2 100644 --- a/router/src/infer.rs +++ b/router/src/infer.rs @@ -9,7 +9,7 @@ use std::sync::{ Arc, }; use text_generation_client::{ - Batch, CachedBatch, ClientError, GeneratedText, Generation, PrefillTokens, ShardedClient, + Batch, CachedBatch, ClientError, GeneratedText, Generation, ShardedClient, Tokens, }; use thiserror::Error; use tokio::sync::mpsc::error::SendError; @@ -50,10 +50,11 @@ impl Infer { max_concurrent_requests: usize, requires_padding: bool, window_size: Option, + speculate: u32, generation_health: Arc, ) -> Self { // Infer shared state - let queue = Queue::new(requires_padding, 16, window_size); + let queue = Queue::new(requires_padding, 16, window_size, speculate); let shared = Arc::new(Shared { batching_task: Notify::new(), }); @@ -523,50 +524,63 @@ fn send_responses( } // Create last Token - let token = Token { - id: generation.token_id, - text: generation.token_text, - logprob: generation.token_logprob, - special: generation.token_is_special, - }; - - // generation.top_tokens - - let mut top_tokens = Vec::new(); - if let Some(top_tokens_) = generation.top_tokens { - top_tokens.extend( + let tokens_ = generation.tokens.expect("Non empty tokens in generation"); + let n = tokens_.ids.len(); + metrics::histogram!("tgi_request_skipped_tokens", (n - 1) as f64); + let mut iterator = tokens_ + .ids + .into_iter() + .zip(tokens_.logprobs.into_iter()) + .zip(tokens_.texts.into_iter()) + .zip(tokens_.is_special.into_iter()) + .enumerate() + .peekable(); + while let Some((i, (((id, logprob), text), special))) = iterator.next() { + let token = Token { + id, + text, + logprob, + special, + }; + let top_tokens = if let Some(top_tokens_) = generation.top_tokens.get(i) { top_tokens_ .ids - .into_iter() - .zip(top_tokens_.logprobs.into_iter()) - .zip(top_tokens_.texts.into_iter()) - .zip(top_tokens_.is_special.into_iter()) - .map(|(((id, logprob), text), special)| Token { + .iter() + .zip(top_tokens_.logprobs.iter()) + .zip(top_tokens_.texts.iter()) + .zip(top_tokens_.is_special.iter()) + .map(|(((&id, &logprob), text), &special)| Token { id, - text, + text: text.to_string(), logprob, special, - }), - ) + }) + .collect() + } else { + vec![] + }; + match (&generation.generated_text, iterator.peek()) { + (Some(generated_text), None) => { + // Generation has ended + stopped = true; + // Send message + entry.response_tx.send(Ok(InferStreamResponse::End { + token, + top_tokens, + generated_text: generated_text.clone(), + queued: entry.queue_time, + start: entry.batch_time.unwrap(), + }))?; + } + _ => { + // Send message + entry + .response_tx + .send(Ok(InferStreamResponse::Intermediate { token, top_tokens }))?; + } + } } - if let Some(generated_text) = generation.generated_text { - // Generation has ended - stopped = true; - // Send message - entry.response_tx.send(Ok(InferStreamResponse::End { - token, - top_tokens, - generated_text, - queued: entry.queue_time, - start: entry.batch_time.unwrap(), - }))?; - } else { - // Send message - entry - .response_tx - .send(Ok(InferStreamResponse::Intermediate { token, top_tokens }))?; - } Ok(stopped) } @@ -591,7 +605,7 @@ fn send_errors(error: ClientError, entries: &mut IntMap) { #[derive(Debug)] pub(crate) enum InferStreamResponse { // Optional first message - Prefill(PrefillTokens), + Prefill(Tokens), // Intermediate messages Intermediate { token: Token, diff --git a/router/src/queue.rs b/router/src/queue.rs index bbb8db0e..106cacc4 100644 --- a/router/src/queue.rs +++ b/router/src/queue.rs @@ -34,7 +34,12 @@ pub(crate) struct Queue { } impl Queue { - pub(crate) fn new(requires_padding: bool, block_size: u32, window_size: Option) -> Self { + pub(crate) fn new( + requires_padding: bool, + block_size: u32, + window_size: Option, + speculate: u32, + ) -> Self { // Create channel let (queue_sender, queue_receiver) = mpsc::unbounded_channel(); @@ -43,6 +48,7 @@ impl Queue { requires_padding, block_size, window_size, + speculate, queue_receiver, )); @@ -91,9 +97,10 @@ async fn queue_task( requires_padding: bool, block_size: u32, window_size: Option, + speculate: u32, mut receiver: mpsc::UnboundedReceiver, ) { - let mut state = State::new(requires_padding, block_size, window_size); + let mut state = State::new(requires_padding, block_size, window_size, speculate); while let Some(cmd) = receiver.recv().await { match cmd { @@ -136,10 +143,18 @@ struct State { /// Sliding window window_size: Option, + + /// Speculation amount + speculate: u32, } impl State { - fn new(requires_padding: bool, block_size: u32, window_size: Option) -> Self { + fn new( + requires_padding: bool, + block_size: u32, + window_size: Option, + speculate: u32, + ) -> Self { Self { entries: VecDeque::with_capacity(128), next_id: 0, @@ -147,6 +162,7 @@ impl State { requires_padding, block_size, window_size, + speculate, } } @@ -229,7 +245,7 @@ impl State { } if prefill_tokens > prefill_token_budget - || (prefill_tokens + decode_tokens) > token_budget + || (prefill_tokens + decode_tokens + self.speculate) > token_budget { // Entry is over budget // Add it back to the front @@ -359,7 +375,7 @@ mod tests { #[test] fn test_append() { - let mut state = State::new(false, 1, None); + let mut state = State::new(false, 1, None, 0); let (entry, _guard) = default_entry(); assert_eq!(state.next_id, 0); @@ -375,7 +391,7 @@ mod tests { #[test] fn test_next_batch_empty() { - let mut state = State::new(false, 1, None); + let mut state = State::new(false, 1, None, 0); assert!(state.next_batch(None, 1, 1).is_none()); assert!(state.next_batch(Some(1), 1, 1).is_none()); @@ -383,7 +399,7 @@ mod tests { #[test] fn test_next_batch_min_size() { - let mut state = State::new(false, 1, None); + let mut state = State::new(false, 1, None, 0); let (entry1, _guard1) = default_entry(); let (entry2, _guard2) = default_entry(); state.append(entry1); @@ -415,7 +431,7 @@ mod tests { #[test] fn test_next_batch_token_budget() { - let mut state = State::new(false, 1, None); + let mut state = State::new(false, 1, None, 0); let (entry1, _guard1) = default_entry(); let (entry2, _guard2) = default_entry(); state.append(entry1); @@ -448,14 +464,14 @@ mod tests { #[tokio::test] async fn test_queue_append() { - let queue = Queue::new(false, 1, None); + let queue = Queue::new(false, 1, None, 0); let (entry, _guard) = default_entry(); queue.append(entry); } #[tokio::test] async fn test_queue_next_batch_empty() { - let queue = Queue::new(false, 1, None); + let queue = Queue::new(false, 1, None, 0); assert!(queue.next_batch(None, 1, 1).await.is_none()); assert!(queue.next_batch(Some(1), 1, 1).await.is_none()); @@ -463,7 +479,7 @@ mod tests { #[tokio::test] async fn test_queue_next_batch_min_size() { - let queue = Queue::new(false, 1, None); + let queue = Queue::new(false, 1, None, 0); let (entry1, _guard1) = default_entry(); let (entry2, _guard2) = default_entry(); queue.append(entry1); @@ -496,7 +512,7 @@ mod tests { #[tokio::test] async fn test_queue_next_batch_token_budget() { - let queue = Queue::new(false, 1, None); + let queue = Queue::new(false, 1, None, 0); let (entry1, _guard1) = default_entry(); let (entry2, _guard2) = default_entry(); queue.append(entry1); @@ -519,9 +535,28 @@ mod tests { assert_eq!(batch.size, 2); } + #[tokio::test] + async fn test_queue_next_batch_token_speculate() { + let queue = Queue::new(false, 1, None, 2); + let (entry1, _guard1) = default_entry(); + let (entry2, _guard2) = default_entry(); + queue.append(entry1); + queue.append(entry2); + + // Budget of 1 is not enough + assert!(queue.next_batch(None, 1, 1).await.is_none()); + + let (entries, batch, _) = queue.next_batch(None, 6, 6).await.unwrap(); + assert_eq!(entries.len(), 2); + assert!(entries.contains_key(&0)); + assert!(entries.contains_key(&1)); + assert_eq!(batch.id, 0); + assert_eq!(batch.size, 2); + } + #[tokio::test] async fn test_queue_next_batch_dropped_receiver() { - let queue = Queue::new(false, 1, None); + let queue = Queue::new(false, 1, None, 0); let (entry, _) = default_entry(); queue.append(entry); diff --git a/router/src/server.rs b/router/src/server.rs index f254afd8..5f41fd5e 100644 --- a/router/src/server.rs +++ b/router/src/server.rs @@ -596,6 +596,7 @@ pub async fn run( max_concurrent_requests, shard_info.requires_padding, shard_info.window_size, + shard_info.speculate, generation_health, ); diff --git a/server/Makefile-vllm b/server/Makefile-vllm index ddb648ea..c9c1d520 100644 --- a/server/Makefile-vllm +++ b/server/Makefile-vllm @@ -1,22 +1,25 @@ -build-vllm-cuda: REPOSITORY=https://github.com/vllm-project/vllm.git -build-vllm-cuda: VLLM_COMMIT=f8a1e39fae05ca610be8d5a78be9d40f5274e5fc -build-vllm-cuda: BRANCH=main -build-vllm-cuda: build-vllm - -build-vllm-rocm: REPOSITORY=https://github.com/fxmarty/vllm-public.git -build-vllm-rocm: VLLM_COMMIT=ad9b7c4095ef54419a0533d254f2ad84bd2dfcae -build-vllm-rocm: BRANCH=rotary-no-positions-split-cos-sin -build-vllm-rocm: build-vllm - -vllm: +vllm-cuda: # Clone vllm pip install -U ninja packaging --no-cache-dir - git clone --single-branch --branch $(BRANCH) $(REPOSITORY) vllm + git clone https://github.com/vllm-project/vllm.git vllm -build-vllm: vllm - cd vllm && git fetch && git checkout $(VLLM_COMMIT) +build-vllm-cuda: vllm-cuda + cd vllm && git fetch && git checkout f8a1e39fae05ca610be8d5a78be9d40f5274e5fc cd vllm && python setup.py build -install-vllm: build-vllm +install-vllm-cuda: build-vllm-cuda + pip uninstall vllm -y || true + cd vllm && python setup.py install + +vllm-rocm: + # Clone vllm + pip install -U ninja packaging --no-cache-dir + git clone https://github.com/fxmarty/vllm-public.git vllm + +build-vllm-rocm: vllm-rocm + cd vllm && git fetch && git checkout ad9b7c4095ef54419a0533d254f2ad84bd2dfcae + cd vllm && python setup.py build + +install-vllm-rocm: build-vllm-rocm pip uninstall vllm -y || true cd vllm && python setup.py install diff --git a/server/tests/models/test_bloom.py b/server/tests/models/test_bloom.py index 71013cb6..1990ef8b 100644 --- a/server/tests/models/test_bloom.py +++ b/server/tests/models/test_bloom.py @@ -133,8 +133,8 @@ def test_causal_lm_generate_token(default_bloom, default_bloom_batch): ) assert all([generation.generated_text is None for generation in generations]) assert all([len(generation.prefill_tokens) == 1 for generation in generations]) - assert all([generation.token_id.item() == 10264 for generation in generations]) - assert all([generation.token_text == "Test" for generation in generations]) + assert all([token_id.item() == 10264 for generation in generations for token_id in generation.tokens.token_ids]) + assert all([token_text == "Test" for generation in generations for token_text in generation.tokens.texts]) assert generations[0].request_id == 0 diff --git a/server/tests/models/test_causal_lm.py b/server/tests/models/test_causal_lm.py index 0f9dab2c..f105ce6f 100644 --- a/server/tests/models/test_causal_lm.py +++ b/server/tests/models/test_causal_lm.py @@ -129,8 +129,8 @@ def test_causal_lm_generate_token(default_causal_lm, default_causal_lm_batch): ) assert all([generation.generated_text is None for generation in generations]) assert all([len(generation.prefill_tokens) == 1 for generation in generations]) - assert all([generation.token_id.item() == 13 for generation in generations]) - assert all([generation.token_text == "." for generation in generations]) + assert all([token_id.item() == 13 for generation in generations for token_id in generation.tokens.token_ids]) + assert all([token_text == "." for generation in generations for token_text in generation.tokens.texts]) assert generations[0].request_id == 0 diff --git a/server/tests/models/test_seq2seq_lm.py b/server/tests/models/test_seq2seq_lm.py index 299340f8..d553067e 100644 --- a/server/tests/models/test_seq2seq_lm.py +++ b/server/tests/models/test_seq2seq_lm.py @@ -151,8 +151,8 @@ def test_seq2seq_lm_generate_token(default_seq2seq_lm, default_seq2seq_lm_batch) ) assert all([generation.generated_text is None for generation in generations]) assert all([len(generation.prefill_tokens) == 1 for generation in generations]) - assert all([generation.token_id.item() == 259 for generation in generations]) - assert all([generation.token_text == " " for generation in generations]) + assert all([token_id.item() == 259 for generation in generations for token_id in generation.tokens.token_ids]) + assert all([token_text == " " for generation in generations for token_text in generation.tokens.texts]) assert generations[0].request_id == 0 diff --git a/server/text_generation_server/cli.py b/server/text_generation_server/cli.py index 3abe86af..cb151173 100644 --- a/server/text_generation_server/cli.py +++ b/server/text_generation_server/cli.py @@ -32,6 +32,7 @@ def serve( revision: Optional[str] = None, sharded: bool = False, quantize: Optional[Quantization] = None, + speculate: Optional[int] = None, dtype: Optional[Dtype] = None, trust_remote_code: bool = False, uds_path: Path = "/tmp/text-generation-server", @@ -81,7 +82,7 @@ def serve( "Only 1 can be set between `dtype` and `quantize`, as they both decide how goes the final model." ) server.serve( - model_id, revision, sharded, quantize, dtype, trust_remote_code, uds_path + model_id, revision, sharded, quantize, speculate, dtype, trust_remote_code, uds_path ) @@ -116,7 +117,7 @@ def download_weights( logger.info("Files are already present on the host. " "Skipping download.") return # Local files not found - except (utils.LocalEntryNotFoundError, FileNotFoundError): + except (utils.LocalEntryNotFoundError, FileNotFoundError, utils.EntryNotFoundError): pass is_local_model = (Path(model_id).exists() and Path(model_id).is_dir()) or os.getenv( @@ -137,6 +138,29 @@ def download_weights( except (utils.LocalEntryNotFoundError, utils.EntryNotFoundError): pass + try: + import json + medusa_head = hf_hub_download(model_id, revision=revision, filename="medusa_lm_head.pt") + if auto_convert: + medusa_sf = Path(medusa_head[:-len(".pt")] + ".safetensors") + if not medusa_sf.exists(): + utils.convert_files([Path(medusa_head)], [medusa_sf], []) + medusa_config = hf_hub_download(model_id, revision=revision, filename="config.json") + with open(medusa_config, "r") as f: + config = json.load(f) + + model_id = config["base_model_name_or_path"] + revision = "main" + try: + utils.weight_files(model_id, revision, extension) + logger.info(f"Files for parent {model_id} are already present on the host. " "Skipping download.") + return + # Local files not found + except (utils.LocalEntryNotFoundError, FileNotFoundError, utils.EntryNotFoundError): + pass + except (utils.LocalEntryNotFoundError, utils.EntryNotFoundError): + pass + # Try to download weights from the hub try: filenames = utils.weight_hub_files(model_id, revision, extension) diff --git a/server/text_generation_server/models/__init__.py b/server/text_generation_server/models/__init__.py index ab3b25b7..27e3897d 100644 --- a/server/text_generation_server/models/__init__.py +++ b/server/text_generation_server/models/__init__.py @@ -6,6 +6,7 @@ from transformers.configuration_utils import PretrainedConfig from transformers.models.auto import modeling_auto from typing import Optional +from text_generation_server.utils.speculate import get_speculate, set_speculate from text_generation_server.models.model import Model from text_generation_server.models.causal_lm import CausalLM from text_generation_server.models.flash_causal_lm import FlashCausalLM @@ -77,12 +78,12 @@ except ImportError as e: if MISTRAL: __all__.append(FlashMistral) - def get_model( model_id: str, revision: Optional[str], sharded: bool, quantize: Optional[str], + speculate: Optional[int], dtype: Optional[str], trust_remote_code: bool, ) -> Model: @@ -97,6 +98,11 @@ def get_model( else: raise RuntimeError(f"Unknown dtype {dtype}") + if speculate is not None: + set_speculate(speculate) + else: + set_speculate(0) + if "facebook/galactica" in model_id: return GalacticaSharded( model_id, @@ -131,6 +137,33 @@ def get_model( config_dict, _ = PretrainedConfig.get_config_dict( model_id, revision=revision, trust_remote_code=trust_remote_code ) + + use_medusa = None + if "medusa_num_heads" in config_dict: + use_medusa = model_id + medusa_config = config_dict + model_id = config_dict["base_model_name_or_path"] + revision = "main" + speculate_medusa = config_dict["medusa_num_heads"] + if speculate is not None: + if speculate > speculate_medusa: + raise RuntimeError("Speculate is set to `{speculate}` but this medusa models only has `{speculate_medusa}` heads, please make them match") + else: + set_speculate(speculate) + else: + set_speculate(speculate_medusa) + + config_dict, _ = PretrainedConfig.get_config_dict( + model_id, revision=revision, trust_remote_code=trust_remote_code + ) + method = "medusa" + else: + method = "n-gram" + + speculate = get_speculate() + if speculate > 0: + logger.info(f"Using speculation {method} with {speculate} input ids.") + model_type = config_dict["model_type"] if model_type == "gpt_bigcode": @@ -206,6 +239,7 @@ def get_model( quantize=quantize, dtype=dtype, trust_remote_code=trust_remote_code, + use_medusa=use_medusa ) elif sharded: raise NotImplementedError(FLASH_ATT_ERROR_MESSAGE.format("Sharded Llama")) diff --git a/server/text_generation_server/models/causal_lm.py b/server/text_generation_server/models/causal_lm.py index 8056a8ec..c571a022 100644 --- a/server/text_generation_server/models/causal_lm.py +++ b/server/text_generation_server/models/causal_lm.py @@ -10,10 +10,9 @@ from typing import Optional, Tuple, List, Type, Dict from text_generation_server.models import Model from text_generation_server.models.types import ( Batch, - PrefillTokens, + Tokens, Generation, GeneratedText, - TopTokens, ) from text_generation_server.pb import generate_pb2 from text_generation_server.utils import NextTokenChooser, StoppingCriteria, Sampling @@ -676,8 +675,8 @@ class CausalLM(Model): clean_up_tokenization_spaces=False, skip_special_tokens=False, ) - prefill_tokens = PrefillTokens( - prefill_token_ids, prefill_logprobs, prefill_texts + prefill_tokens = Tokens( + prefill_token_ids, prefill_logprobs, prefill_texts, is_special=[] ) else: prefill_tokens = None @@ -691,7 +690,7 @@ class CausalLM(Model): special_toptokens = [ token_id in self.all_special_ids for token_id in top_token_ids ] - top_tokens = TopTokens( + top_tokens = Tokens( top_token_ids, top_token_logprobs, toptoken_texts, @@ -703,10 +702,12 @@ class CausalLM(Model): generation = Generation( request.id, prefill_tokens, - next_token_id_squeezed, - next_token_logprob, - next_token_text, - next_token_id_squeezed.item() in self.all_special_ids, + Tokens( + [next_token_id_squeezed], + [next_token_logprob], + [next_token_text], + [next_token_id_squeezed.item() in self.all_special_ids], + ), generated_text, top_tokens, ) diff --git a/server/text_generation_server/models/flash_causal_lm.py b/server/text_generation_server/models/flash_causal_lm.py index f1a4854f..79344ea1 100644 --- a/server/text_generation_server/models/flash_causal_lm.py +++ b/server/text_generation_server/models/flash_causal_lm.py @@ -11,13 +11,13 @@ from opentelemetry import trace from transformers import PreTrainedTokenizerBase from typing import Optional, Tuple, List, Type, Union, Dict -from text_generation_server.models import Model +from text_generation_server.models import Model +from text_generation_server.utils.speculate import get_speculate from text_generation_server.models.types import ( Batch, - PrefillTokens, + Tokens, Generation, GeneratedText, - TopTokens, ) from text_generation_server.models.cache_manager import ( get_cache_manager, @@ -41,6 +41,7 @@ class FlashCausalLMBatch(Batch): # Decoder values input_ids: torch.Tensor position_ids: torch.Tensor + speculative_ids: torch.Tensor # Flash Attention values @@ -120,6 +121,7 @@ class FlashCausalLMBatch(Batch): )["input_ids"] position_ids = [] + speculative_ids = [] cu_seqlen_prefill = [0] needed_blocks_slots = [] start_slots = [] @@ -163,6 +165,8 @@ class FlashCausalLMBatch(Batch): input_length = len(tokenized_input) input_lengths.append(input_length) + + prefix_offsets.append(input_length - 5) read_offsets.append(input_length) @@ -186,7 +190,8 @@ class FlashCausalLMBatch(Batch): # Paged attention # Remove one as the first token des not have a past - total_tokens = input_length + max_new_tokens - 1 + speculative_length = get_speculate() + total_tokens = input_length + max_new_tokens - 1 + speculative_length needed_blocks = math.ceil(total_tokens / BLOCK_SIZE) blocks += needed_blocks needed_blocks_slots.append((needed_blocks, total_tokens)) @@ -224,7 +229,7 @@ class FlashCausalLMBatch(Batch): cumulative_max_length += total_tokens max_seqlen = max(max_seqlen, input_length) max_blocks = max(max_blocks, needed_blocks) - max_length = max(max_length, input_length + max_new_tokens) + max_length = max(max_length, input_length + max_new_tokens + speculative_length) next_token_chooser = HeterogeneousNextTokenChooser.from_pb( next_token_chooser_parameters, dtype, device @@ -255,7 +260,6 @@ class FlashCausalLMBatch(Batch): cu_seqlen_prefill = torch.tensor( cu_seqlen_prefill, device=device, dtype=torch.int32 ) - position_ids = position_ids.to(device) slot_indices = slot_indices.to(device) input_ids = torch.tensor(input_ids, dtype=torch.int64, device=device) @@ -309,6 +313,7 @@ class FlashCausalLMBatch(Batch): top_n_tokens_tensor=top_n_tokens_tensor, blocks=blocks, max_blocks=max_blocks, + speculative_ids=None, ) @tracer.start_as_current_span("filter") @@ -419,6 +424,7 @@ class FlashCausalLMBatch(Batch): slots = self.slots[slot_filtering_indices] next_token_chooser = self.next_token_chooser.filter(indices) top_n_tokens_tensor = self.top_n_tokens_tensor[indices] + speculative_ids = self.speculative_ids[indices] if self.speculative_ids is not None else None start_slots = torch.tensor(start_slots, dtype=torch.int64) @@ -454,6 +460,7 @@ class FlashCausalLMBatch(Batch): top_n_tokens_tensor=top_n_tokens_tensor, blocks=blocks, max_blocks=max_blocks, + speculative_ids=speculative_ids, ) @classmethod @@ -473,6 +480,7 @@ class FlashCausalLMBatch(Batch): total_batch_size += len(b) total_slots += len(b.slots) blocks += b.blocks + speculative_length = b.speculative_ids.shape[1] if b.speculative_ids is not None else 0 max_blocks = max(max_blocks, b.max_blocks) max_seqlen = max(max_seqlen, b.max_seqlen) max_length = max( @@ -480,6 +488,7 @@ class FlashCausalLMBatch(Batch): max( input_length + stopping_criteria.max_new_tokens + + speculative_length - stopping_criteria.current_tokens for input_length, stopping_criteria in zip( b.input_lengths, b.stopping_criterias @@ -577,6 +586,8 @@ class FlashCausalLMBatch(Batch): device=batches[0].next_token_chooser.device, ) + speculative_ids = torch.cat([b.speculative_ids for b in batches], dim=0) if batches[0].speculative_ids is not None else None + # Needed to avoid dropping blocks when the batches will go out of scope for b in batches: b.block_tables = None @@ -611,6 +622,7 @@ class FlashCausalLMBatch(Batch): top_n_tokens_tensor=top_n_tokens_tensor, blocks=blocks, max_blocks=max_blocks, + speculative_ids=speculative_ids ) def __del__(self): @@ -714,16 +726,55 @@ class FlashCausalLM(Model): def forward(self, batch: FlashCausalLMBatch) -> Tuple[torch.Tensor, torch.Tensor]: # Model Forward + if batch.speculative_ids is not None: + input_ids=batch.input_ids + position_ids=batch.position_ids + cu_seqlen_prefill=batch.cu_seqlen_prefill + kv_cache=get_cache_manager().kv_cache + block_tables=batch.block_tables_tensor + slots=batch.slots[batch.slot_indices] + input_lengths=batch.input_lengths_tensor + max_s=batch.max_seqlen + lm_head_indices=batch.prefill_head_indices + + speculative_ids = batch.speculative_ids + + B, speculative_length = speculative_ids.shape + new_length = speculative_length + 1 + new_input_ids = torch.cat([input_ids.unsqueeze(-1), speculative_ids], dim=1).reshape(-1) + arange = torch.arange(new_length, device=position_ids.device).unsqueeze(0) + arange_int = arange.to(dtype=torch.int32) + new_position_ids = (position_ids.unsqueeze(-1).expand(B, new_length) + arange).view(-1) + slots = (slots.unsqueeze(-1).expand(B, new_length) + arange_int).view(-1) + input_lengths = (input_lengths.unsqueeze(-1).expand(B, new_length) + arange_int).view(-1) + + # Add Copy the block tables for all members + block_tables = block_tables.unsqueeze(1).expand(B, new_length, -1).reshape(B* new_length, -1).contiguous() + max_s = max_s + speculative_length + + input_ids = new_input_ids + position_ids = new_position_ids + else: + input_ids=batch.input_ids + position_ids=batch.position_ids + cu_seqlen_prefill=batch.cu_seqlen_prefill + kv_cache=get_cache_manager().kv_cache + block_tables=batch.block_tables_tensor + slots=batch.slots[batch.slot_indices] + input_lengths=batch.input_lengths_tensor + max_s=batch.max_seqlen + lm_head_indices=batch.prefill_head_indices + return self.model.forward( - input_ids=batch.input_ids, - position_ids=batch.position_ids, - cu_seqlen_prefill=batch.cu_seqlen_prefill, - kv_cache=get_cache_manager().kv_cache, - block_tables=batch.block_tables_tensor, - slots=batch.slots[batch.slot_indices], - input_lengths=batch.input_lengths_tensor, - max_s=batch.max_seqlen, - lm_head_indices=batch.prefill_head_indices, + input_ids=input_ids, + position_ids=position_ids, + cu_seqlen_prefill=cu_seqlen_prefill, + kv_cache=kv_cache, + block_tables=block_tables, + slots=slots, + input_lengths=input_lengths, + max_s=max_s, + lm_head_indices=lm_head_indices, ) @tracer.start_as_current_span("generate_token") @@ -752,21 +803,32 @@ class FlashCausalLM(Model): del batch raise e + if isinstance(out, tuple): + out, speculative_logits = out + else: + speculative_logits = None + + if prefill: next_token_logits = ( out[batch.prefill_next_token_indices] if prefill_logprobs else out ) + if speculative_logits is not None: + speculative_logits = ( + speculative_logits[batch.prefill_next_token_indices] if prefill_logprobs else speculative_logits + ) else: next_token_logits = out - next_input_ids, next_token_logprobs, logprobs = batch.next_token_chooser( - batch.all_input_ids_tensor[:, : batch.max_seqlen], next_token_logits + next_input_ids, next_token_logprobs, logprobs, accepted_ids, speculative_ids = batch.next_token_chooser( + batch.all_input_ids_tensor[:, : batch.max_seqlen], next_token_logits, get_speculate(), batch.speculative_ids, speculative_logits ) batch_top_token_ids, batch_top_token_logprobs = batch_top_tokens( batch.top_n_tokens, batch.top_n_tokens_tensor, logprobs ) + speculative_length = 0 if speculative_ids is None else speculative_ids.shape[1] if prefill: if len(batch) > 1 and prefill_logprobs: # We create the prefill_tokens_indices tensor that will be used to gather prefill logprobs @@ -792,6 +854,7 @@ class FlashCausalLM(Model): iterator = zip( batch.input_lengths, batch.all_input_ids, + accepted_ids ) # We do two for loops as the first one can run completely asynchronously from the GPU while for the second @@ -799,9 +862,11 @@ class FlashCausalLM(Model): # It is faster if we delay this sync for the maximum amount of time # For each member of the batch + index = 0 for i, ( input_length, all_input_ids, + n_accepted_ids ) in enumerate(iterator): # Indexing metadata start_index = cumulative_length @@ -830,15 +895,18 @@ class FlashCausalLM(Model): start_index + 1 : start_index + out_length ] - batch.all_input_ids_tensor[i, input_length] = next_input_ids[i] + for j in range(n_accepted_ids): + batch.all_input_ids_tensor[i, input_length + j] = next_input_ids[index] + index += 1 cumulative_length += input_length - # Set values in batch - batch.input_ids = next_input_ids - batch.position_ids = next_position_ids + 1 - batch.input_lengths_tensor += 1 - batch.slot_indices += 1 + + batch.input_ids = next_input_ids[accepted_ids.cumsum(dim=-1) - 1] + batch.speculative_ids = speculative_ids + batch.position_ids = next_position_ids + accepted_ids + batch.input_lengths_tensor += accepted_ids + batch.slot_indices += accepted_ids if prefill and prefill_logprobs: # Get prefill logprobs @@ -851,7 +919,7 @@ class FlashCausalLM(Model): # GPU <-> CPU sync next_token_logprobs = next_token_logprobs.tolist() - next_token_ids = batch.input_ids.tolist() + next_token_ids = next_input_ids.tolist() # Zipped iterator iterator = zip( @@ -864,13 +932,13 @@ class FlashCausalLM(Model): batch.next_token_chooser.do_sample, batch.next_token_chooser.seeds, batch.top_n_tokens, - next_token_ids, - next_token_logprobs, + accepted_ids, batch_top_token_ids, batch_top_token_logprobs, ) # For each member of the batch + index = 0 for i, ( request, input_length, @@ -881,29 +949,43 @@ class FlashCausalLM(Model): do_sample, seed, top_n_tokens, - next_token_id, - next_token_logprob, + n_accepted_ids, top_token_ids, top_token_logprobs, ) in enumerate(iterator): # Append next token to all tokens - all_input_ids.append(next_token_id) + next_token_texts = [] + left = 0 + before = stopping_criteria.current_tokens - # Generated token - next_token_text, prefix_offset, read_offset = self.decode_token( - all_input_ids, - prefix_offset, - read_offset, - ) + current_stopped = False + for j in range(index, index + n_accepted_ids): + # Generated token + next_token_id = next_token_ids[j] + all_input_ids.append(next_token_id) + next_token_text, prefix_offset, read_offset = self.decode_token( + all_input_ids, + prefix_offset, + read_offset, + ) + next_token_texts.append(next_token_text) - # Evaluate stopping criteria - stop, reason = stopping_criteria( - next_token_id, - next_token_text, - ) + stop, reason = stopping_criteria( + next_token_id, + next_token_text, + ) - if not stop: - stopped = False + if stop: + left = index + n_accepted_ids - j - 1 + current_stopped = True + break + else: + current_stopped = False + stopped = stopped and current_stopped + + _next_token_ids = next_token_ids[index: index+n_accepted_ids - left] + _next_token_logprobs = next_token_logprobs[index: index+n_accepted_ids - left] + index += n_accepted_ids # Shard generations # All generations will be appended in the rust sharded client @@ -943,8 +1025,9 @@ class FlashCausalLM(Model): clean_up_tokenization_spaces=False, skip_special_tokens=False, ) - prefill_tokens = PrefillTokens( - prefill_token_ids, request_prefill_logprobs, prefill_texts + + prefill_tokens = Tokens( + prefill_token_ids, request_prefill_logprobs, prefill_texts, is_special = [] ) else: prefill_tokens = None @@ -958,7 +1041,7 @@ class FlashCausalLM(Model): special_toptokens = [ token_id in self.all_special_ids for token_id in top_token_ids ] - top_tokens = TopTokens( + top_tokens = Tokens( top_token_ids, top_token_logprobs, toptoken_texts, @@ -970,10 +1053,12 @@ class FlashCausalLM(Model): generation = Generation( request.id, prefill_tokens, - next_token_id, - next_token_logprob, - next_token_text, - next_token_id in self.all_special_ids, + Tokens( + _next_token_ids, + _next_token_logprobs, + next_token_texts, + [nid in self.all_special_ids for nid in _next_token_ids], + ), generated_text, top_tokens, ) @@ -981,7 +1066,9 @@ class FlashCausalLM(Model): generations.append(generation) # Update values - batch.input_lengths[i] = input_length + 1 + batch.input_lengths[i] = input_length + n_accepted_ids.item() + if batch.input_lengths[i] > batch.max_seqlen: + batch.max_seqlen = batch.input_lengths[i] batch.prefix_offsets[i] = prefix_offset batch.read_offsets[i] = read_offset batch.all_input_ids[i] = all_input_ids @@ -994,6 +1081,5 @@ class FlashCausalLM(Model): batch.prefill_cu_outlens = None batch.prefill_head_indices = None batch.prefill_next_token_indices = None - batch.max_seqlen = batch.max_seqlen + 1 return generations, batch diff --git a/server/text_generation_server/models/flash_llama.py b/server/text_generation_server/models/flash_llama.py index d2ed0b15..3a84b1b6 100644 --- a/server/text_generation_server/models/flash_llama.py +++ b/server/text_generation_server/models/flash_llama.py @@ -28,6 +28,7 @@ class FlashLlama(FlashCausalLM): quantize: Optional[str] = None, dtype: Optional[torch.dtype] = None, trust_remote_code: bool = False, + use_medusa: Optional[str] = None, ): self.process_group, rank, world_size = initialize_torch_distributed() if torch.cuda.is_available(): @@ -66,6 +67,18 @@ class FlashLlama(FlashCausalLM): weights._set_gptq_params(model_id) model = FlashLlamaForCausalLM(config, weights) + if use_medusa: + from text_generation_server.utils.medusa import MedusaModel + from huggingface_hub import hf_hub_download + import json + medusa_config = hf_hub_download(use_medusa, revision=revision, filename="config.json") + with open(medusa_config, "r") as f: + config = json.load(f) + medusa_head = hf_hub_download(use_medusa, revision=revision, filename="medusa_lm_head.pt") + medusa_sf = medusa_head[:-len(".pt")] + ".safetensors" + weights = Weights([medusa_sf], device, dtype, process_group=self.process_group) + lm_head = model.lm_head + model.lm_head = MedusaModel(config, weights, lm_head) torch.distributed.barrier(group=self.process_group) super(FlashLlama, self).__init__( diff --git a/server/text_generation_server/models/flash_mistral.py b/server/text_generation_server/models/flash_mistral.py index 919e4625..e103d9fc 100644 --- a/server/text_generation_server/models/flash_mistral.py +++ b/server/text_generation_server/models/flash_mistral.py @@ -21,6 +21,7 @@ from text_generation_server.models.custom_modeling.flash_mistral_modeling import FlashMistralForCausalLM, MistralConfig, ) +from text_generation_server.utils.speculate import get_speculate from text_generation_server.utils import ( initialize_torch_distributed, weight_files, @@ -132,7 +133,8 @@ class FlashMistralBatch(FlashCausalLMBatch): # Paged attention # Remove one as the first token des not have a past - total_tokens = input_length + max_new_tokens - 1 + speculative_length = get_speculate() + total_tokens = input_length + max_new_tokens - 1 + speculative_length # Needed blocks can not go over SLIDING_WINDOW_BLOCKS needed_blocks = min( @@ -183,7 +185,7 @@ class FlashMistralBatch(FlashCausalLMBatch): cumulative_max_length += total_tokens max_seqlen = max(max_seqlen, input_length) max_blocks = max(max_blocks, needed_blocks) - max_length = max(max_length, input_length + max_new_tokens) + max_length = max(max_length, input_length + max_new_tokens + speculative_length) next_token_chooser = HeterogeneousNextTokenChooser.from_pb( next_token_chooser_parameters, dtype, device @@ -272,6 +274,7 @@ class FlashMistralBatch(FlashCausalLMBatch): blocks=blocks, max_blocks=max_blocks, prefill_cache_indices=prefill_cache_indices, + speculative_ids=None ) @@ -340,17 +343,55 @@ class FlashMistral(FlashCausalLM): def forward(self, batch: FlashMistralBatch) -> Tuple[torch.Tensor, torch.Tensor]: # Model Forward + if batch.speculative_ids is not None: + input_ids=batch.input_ids + position_ids=batch.position_ids + cu_seqlen_prefill=batch.cu_seqlen_prefill + kv_cache=get_cache_manager().kv_cache + block_tables=batch.block_tables_tensor + slots=batch.slots[batch.slot_indices] + input_lengths=batch.input_lengths_tensor + max_s=batch.max_seqlen + lm_head_indices=batch.prefill_head_indices + + speculative_ids = batch.speculative_ids + + B, speculative_length = speculative_ids.shape + new_length = speculative_length + 1 + new_input_ids = torch.cat([input_ids.unsqueeze(-1), speculative_ids], dim=1).reshape(-1) + arange = torch.arange(new_length, device=position_ids.device).unsqueeze(0) + arange_int = arange.to(dtype=torch.int32) + new_position_ids = (position_ids.unsqueeze(-1).expand(B, new_length) + arange).view(-1) + slots = (slots.unsqueeze(-1).expand(B, new_length) + arange_int).view(-1) + input_lengths = (input_lengths.unsqueeze(-1).expand(B, new_length) + arange_int).view(-1) + + # Add Copy the block tables for all members + block_tables = block_tables.unsqueeze(1).expand(B, new_length, -1).reshape(B* new_length, -1).contiguous() + max_s = max_s + speculative_length + + input_ids = new_input_ids + position_ids = new_position_ids + else: + input_ids=batch.input_ids + position_ids=batch.position_ids + cu_seqlen_prefill=batch.cu_seqlen_prefill + kv_cache=get_cache_manager().kv_cache + block_tables=batch.block_tables_tensor + slots=batch.slots[batch.slot_indices] + input_lengths=batch.input_lengths_tensor + max_s=batch.max_seqlen + lm_head_indices=batch.prefill_head_indices logits = self.model.forward( - input_ids=batch.input_ids, - position_ids=batch.position_ids, - cu_seqlen_prefill=batch.cu_seqlen_prefill, - kv_cache=get_cache_manager().kv_cache, - block_tables=batch.block_tables_tensor, - slots=batch.slots[batch.slot_indices], - input_lengths=batch.input_lengths_tensor, - max_s=batch.max_seqlen, + input_ids=input_ids, + position_ids=position_ids, + cu_seqlen_prefill=cu_seqlen_prefill, + kv_cache=kv_cache, + block_tables=block_tables, + slots=slots, + input_lengths=input_lengths, + max_s=max_s, prefill_cache_indices=batch.prefill_cache_indices, - lm_head_indices=batch.prefill_head_indices, + lm_head_indices=lm_head_indices, ) if batch.prefill_cache_indices is not None: batch.prefill_cache_indices = None diff --git a/server/text_generation_server/models/idefics_causal_lm.py b/server/text_generation_server/models/idefics_causal_lm.py index dcad1fa9..2f4bb139 100644 --- a/server/text_generation_server/models/idefics_causal_lm.py +++ b/server/text_generation_server/models/idefics_causal_lm.py @@ -20,7 +20,7 @@ from typing import Optional, Tuple, List, Type, Dict from text_generation_server.models import Model from text_generation_server.models.types import ( Batch, - PrefillTokens, + Tokens, Generation, GeneratedText, ) @@ -791,8 +791,8 @@ class IdeficsCausalLM(Model): clean_up_tokenization_spaces=False, skip_special_tokens=False, ) - prefill_tokens = PrefillTokens( - prefill_token_ids, prefill_logprobs, prefill_texts + prefill_tokens = Tokens( + prefill_token_ids, prefill_logprobs, prefill_texts, is_special=[] ) else: prefill_tokens = None @@ -802,10 +802,12 @@ class IdeficsCausalLM(Model): generation = Generation( request.id, prefill_tokens, - next_token_id_squeezed, - next_token_logprob, - next_token_text, - next_token_id_squeezed.item() in self.all_special_ids, + Tokens( + [next_token_id_squeezed], + [next_token_logprob], + [next_token_text], + [next_token_id_squeezed.item() in self.all_special_ids], + ), generated_text, top_tokens, ) diff --git a/server/text_generation_server/models/model.py b/server/text_generation_server/models/model.py index 17d2ea9b..8552960d 100644 --- a/server/text_generation_server/models/model.py +++ b/server/text_generation_server/models/model.py @@ -6,6 +6,7 @@ from typing import List, Tuple, Optional, TypeVar, Type from transformers import PreTrainedTokenizerBase, PretrainedConfig from text_generation_server.models.types import Batch, Generation +from text_generation_server.utils.speculate import get_speculate from text_generation_server.pb.generate_pb2 import InfoResponse B = TypeVar("B", bound=Batch) @@ -22,6 +23,7 @@ class Model(ABC): rank: int = 0, world_size: int = 1, sliding_window: Optional[int] = None, + speculate: Optional[int] = None, ): self.model = model.eval() self.tokenizer = tokenizer @@ -33,6 +35,10 @@ class Model(ABC): self.world_size = world_size self.sliding_window = sliding_window + if speculate is None: + speculate = get_speculate() + self.speculate = speculate + self.has_position_ids = ( inspect.signature(model.forward).parameters.get("position_ids", None) is not None @@ -50,6 +56,7 @@ class Model(ABC): dtype=str(self.dtype), device_type=self.device.type, window_size=self.sliding_window, + speculate=self.speculate ) @property diff --git a/server/text_generation_server/models/seq2seq_lm.py b/server/text_generation_server/models/seq2seq_lm.py index d4d3cd19..279b5505 100644 --- a/server/text_generation_server/models/seq2seq_lm.py +++ b/server/text_generation_server/models/seq2seq_lm.py @@ -11,8 +11,7 @@ from text_generation_server.models.types import ( GeneratedText, Batch, Generation, - PrefillTokens, - TopTokens, + Tokens, ) from text_generation_server.pb import generate_pb2 from text_generation_server.utils import NextTokenChooser, StoppingCriteria, Sampling @@ -733,10 +732,11 @@ class Seq2SeqLM(Model): # Prefill if stopping_criteria.current_tokens == 1 and request.prefill_logprobs: - prefill_tokens = PrefillTokens( + prefill_tokens = Tokens( [self.tokenizer.bos_token_id], [float("nan")], [self.tokenizer.bos_token], + [False] ) else: prefill_tokens = None @@ -750,7 +750,7 @@ class Seq2SeqLM(Model): special_toptokens = [ token_id in self.all_special_ids for token_id in top_token_ids ] - top_tokens = TopTokens( + top_tokens = Tokens( top_token_ids, top_token_logprobs, toptoken_texts, @@ -762,10 +762,12 @@ class Seq2SeqLM(Model): generation = Generation( request.id, prefill_tokens, - next_token_id_squeezed, - next_token_logprob, - next_token_text, - next_token_id_squeezed.item() in self.all_special_ids, + Tokens( + [next_token_id_squeezed], + [next_token_logprob], + [next_token_text], + [next_token_id_squeezed.item() in self.all_special_ids], + ), generated_text, top_tokens, ) diff --git a/server/text_generation_server/models/types.py b/server/text_generation_server/models/types.py index 0e27680d..87c03d63 100644 --- a/server/text_generation_server/models/types.py +++ b/server/text_generation_server/models/types.py @@ -58,33 +58,15 @@ class GeneratedText: @dataclass -class PrefillTokens: - token_ids: List[int] - logprobs: List[float] - texts: List[str] - - def to_pb(self) -> generate_pb2.PrefillTokens: - return generate_pb2.PrefillTokens( - ids=self.token_ids, logprobs=self.logprobs, texts=self.texts - ) - - def __len__(self): - return len(self.token_ids) - - -@dataclass -class TopTokens: +class Tokens: token_ids: List[int] logprobs: List[float] texts: List[str] is_special: List[bool] - def to_pb(self) -> generate_pb2.TopTokens: - return generate_pb2.TopTokens( - ids=self.token_ids, - logprobs=self.logprobs, - texts=self.texts, - is_special=self.is_special, + def to_pb(self) -> generate_pb2.Tokens: + return generate_pb2.Tokens( + ids=self.token_ids, logprobs=self.logprobs, texts=self.texts, is_special=self.is_special ) def __len__(self): @@ -94,14 +76,11 @@ class TopTokens: @dataclass class Generation: request_id: int - prefill_tokens: Optional[PrefillTokens] - token_id: int - token_logprob: float - token_text: str - token_is_special: bool + prefill_tokens: Optional[Tokens] + tokens: Tokens generated_text: Optional[GeneratedText] # Optional for now, since it's not yet supported for every model. - top_tokens: Optional[TopTokens] + top_tokens: Optional[List[Tokens]] def to_pb(self) -> generate_pb2.Generation: return generate_pb2.Generation( @@ -109,10 +88,7 @@ class Generation: prefill_tokens=self.prefill_tokens.to_pb() if self.prefill_tokens is not None else None, - token_id=self.token_id, - token_logprob=self.token_logprob, - token_text=self.token_text, - token_is_special=self.token_is_special, + tokens=self.tokens.to_pb(), generated_text=self.generated_text.to_pb() if self.generated_text is not None else None, diff --git a/server/text_generation_server/server.py b/server/text_generation_server/server.py index fa831682..ebe066e3 100644 --- a/server/text_generation_server/server.py +++ b/server/text_generation_server/server.py @@ -132,6 +132,7 @@ def serve( revision: Optional[str], sharded: bool, quantize: Optional[str], + speculate: Optional[int], dtype: Optional[str], trust_remote_code: bool, uds_path: Path, @@ -141,6 +142,7 @@ def serve( revision: Optional[str], sharded: bool = False, quantize: Optional[str] = None, + speculate: Optional[int] = None, dtype: Optional[str] = None, trust_remote_code: bool = False, ): @@ -157,7 +159,7 @@ def serve( try: model = get_model( - model_id, revision, sharded, quantize, dtype, trust_remote_code + model_id, revision, sharded, quantize, speculate, dtype, trust_remote_code ) except Exception: logger.exception("Error when initializing model") @@ -205,5 +207,5 @@ def serve( await server.stop(0) asyncio.run( - serve_inner(model_id, revision, sharded, quantize, dtype, trust_remote_code) + serve_inner(model_id, revision, sharded, quantize, speculate, dtype, trust_remote_code) ) diff --git a/server/text_generation_server/utils/medusa.py b/server/text_generation_server/utils/medusa.py new file mode 100644 index 00000000..029de122 --- /dev/null +++ b/server/text_generation_server/utils/medusa.py @@ -0,0 +1,51 @@ +import torch +from dataclasses import dataclass +from text_generation_server.utils.layers import TensorParallelHead, FastLinear + +@dataclass +class Output: + logits: torch.FloatTensor = None + speculative_logits: torch.FloatTensor = None + + +class ResBlock(torch.nn.Module): + def __init__(self, config, prefix, weights): + super().__init__() + self.linear = FastLinear.load(config, prefix=f"{prefix}.linear", weights=weights, bias=True) + self.act = torch.nn.SiLU() + + def forward(self, x): + return x + self.act(self.linear(x)) + + +class MedusaModel(torch.nn.Module): + def __init__( + self, + config, + weights, + lm_head + ): + super().__init__() + self.heads = torch.nn.ModuleList( + [MedusaHead(config, prefix=f"{i}", weights=weights) for i in range(config["medusa_num_heads"])] + ) + self.lm_head = lm_head + + def forward(self, x): + logits = self.lm_head(x) + speculative_logits = torch.stack([head(x) for head in self.heads], dim=1) + return logits, speculative_logits + + +class MedusaHead(torch.nn.Module): + def __init__(self, config, prefix, weights): + super().__init__() + self.blocks = torch.nn.ModuleList([ResBlock(config, prefix=f"{prefix}.{i}", weights=weights) for i in range(config["medusa_num_layers"])]) + n = len(self.blocks) + self.out = FastLinear.load(config, prefix=f"{prefix}.{n}", weights=weights, bias=False) + + def forward(self, x): + for block in self.blocks: + x = block(x) + x = self.out(x) + return x diff --git a/server/text_generation_server/utils/speculate.py b/server/text_generation_server/utils/speculate.py new file mode 100644 index 00000000..38a91972 --- /dev/null +++ b/server/text_generation_server/utils/speculate.py @@ -0,0 +1,12 @@ + +SPECULATE = None + +def get_speculate() -> int: + global SPECULATE + return SPECULATE + +def set_speculate(speculate: int): + global SPECULATE + SPECULATE = speculate + + diff --git a/server/text_generation_server/utils/tokens.py b/server/text_generation_server/utils/tokens.py index 0ff07417..a34c5afc 100644 --- a/server/text_generation_server/utils/tokens.py +++ b/server/text_generation_server/utils/tokens.py @@ -16,7 +16,6 @@ from text_generation_server.utils.logits_process import ( from text_generation_server.utils.watermark import WatermarkLogitsProcessor from transformers import PreTrainedTokenizerBase, RepetitionPenaltyLogitsProcessor - class NextTokenChooser: def __init__( self, @@ -146,6 +145,20 @@ class StoppingCriteria: pb.ignore_eos_token, ) +def create_n_gram_speculation(input_ids: torch.Tensor, next_ids: torch.Tensor, accepted_ids: torch.Tensor, speculate: int, verbose: bool): + # Very trivial approach, find first match in the string. + # This is much less refined than actual n-gram but seems to work + # relatively OK in grounded mode and is by far much faster with + # much less worst case complexity as everything happens on device. + B = accepted_ids.shape[0] + device = input_ids.device + seeds = next_ids[accepted_ids.cumsum(dim=-1) -1 ] + indices = (input_ids == seeds.unsqueeze(-1)).max(dim=1).indices + 1 + all_indices = indices.unsqueeze(-1).expand(B, speculate) + torch.arange(speculate, device=device) + all_indices = torch.clamp(all_indices, max=input_ids.shape[1] - 1) + + speculative_ids = input_ids.gather(dim=-1, index=all_indices) + return speculative_ids class HeterogeneousNextTokenChooser: def __init__( @@ -215,20 +228,79 @@ class HeterogeneousNextTokenChooser: self.dtype = dtype self.device = device - def __call__(self, input_ids: torch.Tensor, scores: torch.Tensor): - if self.watermark_processor is not None: - scores = self.watermark_processor(input_ids, scores) - if self.repetition_processor is not None: - scores = self.repetition_processor(input_ids, scores) + def __call__(self, input_ids: torch.Tensor, scores: torch.Tensor, speculate: int, speculated_ids: Optional[torch.Tensor] = None, speculative_scores: Optional[torch.Tensor] = None, verbose=False): + if speculated_ids is not None: + B = scores.shape[0] // (speculated_ids.shape[1] + 1) + S = speculated_ids.shape[1] + 1 + scores = scores.view(B, S, -1) + else: + B = scores.shape[0] + S = 1 + scores = scores.view(B, S, -1) - for warper in self.warpers: - scores = warper(input_ids, scores) + next_ids = torch.zeros((B, S), device=scores.device, dtype=torch.long) + for j in range(S): + _scores = scores[:, j] + if self.watermark_processor is not None: + _scores = self.watermark_processor(input_ids, _scores) + if self.repetition_processor is not None: + _scores = self.repetition_processor(input_ids, _scores) + + for warper in self.warpers: + _scores = warper(input_ids, _scores) + + + _next_ids = self.choice(_scores) + scores[:, j] = _scores + next_ids[:, j] = _next_ids + next_ids = next_ids.view(B*S) + scores = scores.view( B* S, -1) + + if speculated_ids is not None: + accepted_ids = [] + B = next_ids.shape[0] // (speculated_ids.shape[1] + 1) + S = speculated_ids.shape[1] + 1 + indices = [] + for i in range(B): + _next_ids = next_ids[i*S: (i + 1)*S] + _speculated_ids = speculated_ids[i] + validate_speculative = _next_ids[:-1] == _speculated_ids + index = i * S + accepted = 1 + # First is always valid + indices.append(index) + for valid in validate_speculative.tolist(): + if valid: + index += 1 + accepted += 1 + indices.append(index) + else: + break + accepted_ids.append(accepted) + + accepted_ids = torch.tensor(accepted_ids, device=input_ids.device, dtype=input_ids.dtype) + next_ids = next_ids[indices] + scores = scores[indices] + indices = torch.arange(B, device=input_ids.device) * S + if speculative_scores is not None: + speculative_scores = speculative_scores[indices + accepted_ids - 1] + else: + accepted_ids = torch.ones_like(next_ids) - next_ids = self.choice(scores) logprobs = torch.log_softmax(scores, -1) next_logprobs = torch.gather(logprobs, 1, next_ids.view(-1, 1)).view(-1) - return next_ids, next_logprobs, logprobs + if speculate > 0: + if speculative_scores is not None: + # Medusa provided some scores + speculative_ids = Greedy()(speculative_scores) + else: + # n-gram + speculative_ids = create_n_gram_speculation(input_ids, next_ids, accepted_ids, speculate, verbose) + else: + speculative_ids = None + + return next_ids, next_logprobs, logprobs, accepted_ids, speculative_ids def filter(self, indices): if self.watermark_processor is not None: