v2.0.1

What's Changed

• feat: improve tools to include name and add tests by @drbh in #1693
• Update response type for /v1/chat/completions and /v1/completions by @Wauplin in #1747
• accept list as prompt for OpenAI API by @drbh in #1702
• fix ROCm docker image

Full Changelog: v2.0.0...v2.0.1

v2.0.0

TGI is back to Apache 2.0!

Highlights

• License was reverted to Apache 2.0
• Cuda graphs are now used by default. They improve latency substancially on high end nodes.
• Llava-next was added. It is the second multimodal model available on TGI after Idefics.
• Cohere Command R+ support. TGI is the fastest open source backend for Command R+
• FP8 support.
• We now share the vocabulary for all medusa heads, greatly improving latency and memory use.

Try out Command R+ with Medusa heads on 4xA100s with:

model=text-generation-inference/commandrplus-medusa
volume=$PWD/data # share a volume with the Docker container to avoid downloading weights every run

docker run --gpus all --shm-size 1g -p 8080:80 -v $volume:/data ghcr.io/huggingface/text-generation-inference:2.0 --model-id $model --speculate 3 --num-shard 4

What's Changed

• Add cuda graphs sizes and make it default. by @Narsil in #1703
• Pickle conversion now requires --trust-remote-code. by @Narsil in #1704
• Push users to streaming in the readme. by @Narsil in #1698
• Fixing cohere tokenizer. by @Narsil in #1697
• Force weights_only (before fully breaking pickle files anyway). by @Narsil in #1710
• Regenerate ld.so.cache by @oOraph in #1708
• Revert license to Apache 2.0 by @OlivierDehaene in #1714
• Automatic quantization config. by @Narsil in #1719
• Adding Llava-Next (Llava 1.6) with full support. by @Narsil in #1709
• fix: fix CohereForAI/c4ai-command-r-plus by @OlivierDehaene in #1707
• Update libraries by @abhishekkrthakur in #1713
• Dev/mask ldconfig output v2 by @oOraph in #1716
• Fp8 Support by @Narsil in #1726
• Upgrade EETQ (Fixes the cuda graphs). by @Narsil in #1729
• fix(router): fix a possible deadlock in next_batch by @OlivierDehaene in #1731
• chore(cargo-toml): apply lto fat and codegen-units of one by @somehowchris in #1651
• Improve the defaults for the launcher by @Narsil in #1727
• feat: medusa shared by @OlivierDehaene in #1734
• Fix typo in guidance.md by @eltociear in #1735

New Contributors

• @somehowchris made their first contribution in #1651

Full Changelog: v1.4.5...v2.0.0

v.1.4.5

Highlights

• DBRX support #1685. See #1679 on how to prompt the model.

What's Changed

• fix: adjust logprob response logic by @drbh in #1682
• fix: handle batches with and without grammars by @drbh in #1676
• feat: Add dbrx support by @OlivierDehaene in #1685

Full Changelog: v1.4.4...v1.4.5

v.1.4.4

Highlights

• CohereForAI/c4ai-command-r-v01 model support

What's Changed

• Handle concurrent grammar requests by @drbh in #1610
• Fix idefics default. by @Narsil in #1614
• Fix async client timeout by @hugoabonizio in #1617
• accept legacy request format and response by @drbh in #1527
• add missing stop parameter for chat request by @drbh in #1619
• correctly index into mask when applying grammar by @drbh in #1618
• Use a better model for the quick tour by @lewtun in #1639
• Upgrade nix version from 0.27.1 to 0.28.0 by @yuanwu2017 in #1638
• Update peft + transformers + accelerate + bnb + safetensors by @abhishekkrthakur in #1646
• Fix index in ChatCompletionChunk by @Wauplin in #1648
• Fixing minor typo in documentation: supported hardware section by @SachinVarghese in #1632
• bump minijina and add test for core templates by @drbh in #1626
• support force downcast after FastRMSNorm multiply for Gemma by @drbh in #1658
• prefer spaces url over temp url by @drbh in #1662
• improve tool type, bump pydantic and outlines by @drbh in #1650
• Remove unecessary cuda graph. by @Narsil in #1664
• Repair idefics integration tests. by @Narsil in #1663
• fix: LlamaTokenizerFast to AutoTokenizer at flash_mistral.py by @SeongBeomLEE in #1637
• Inline images for multimodal models. by @Narsil in #1666

New Contributors

• @hugoabonizio made their first contribution in #1617
• @yuanwu2017 made their first contribution in #1638
• @abhishekkrthakur made their first contribution in #1646
• @Wauplin made their first contribution in #1648
• @SachinVarghese made their first contribution in #1632
• @SeongBeomLEE made their first contribution in #1637

Full Changelog: v1.4.3...v1.4.4

v1.4.3

Highlights

• Add support for Starcoder 2
• Add support for Qwen2

What's Changed

• fix openapi schema by @OlivierDehaene in #1586
• avoid default message by @drbh in #1579
• Revamp medusa implementation so that every model can benefit. by @Narsil in #1588
• Support tools by @drbh in #1587
• Fixing x-compute-time. by @Narsil in #1606
• Fixing guidance docs. by @Narsil in #1607
• starcoder2 by @OlivierDehaene in #1605
• Qwen2 by @Jason-CKY in #1608

Full Changelog: v1.4.2...v1.4.3

v1.4.2

Highlights

• Add support for Google Gemma models

What's Changed

• Fix mistral with length > window_size for long prefills (rotary doesn't create long enough cos, sin). by @Narsil in #1571
• improve endpoint support by @drbh in #1577
• refactor syntax to correctly include structs by @drbh in #1580
• fix openapi and add jsonschema validation by @OlivierDehaene in #1578
• add support for Gemma by @OlivierDehaene in #1583

Full Changelog: v1.4.1...v1.4.2

v1.4.1

Highlights

• Mamba support by @drbh in #1480 and by @Narsil in #1552
• Experimental support for cuda graphs by @OlivierDehaene in #1428
• Outlines guided generation by @drbh in #1539
• Added name field to OpenAI compatible API Messages by @amihalik in #1563

What's Changed

• Fixing top_n_tokens. by @Narsil in #1497
• Sending compute type from the environment instead of hardcoded string by @Narsil in #1504
• Create the compute type at launch time (if not provided in the env). by @Narsil in #1505
• Modify default for max_new_tokens in python client by @freitng in #1336
• feat: eetq gemv optimization when batch_size <= 4 by @dtlzhuangz in #1502
• fix: improve messages api docs content and formatting by @drbh in #1506
• GPTNeoX: Use static rotary embedding by @dwyatte in #1498
• Hotfix the / health - route. by @Narsil in #1515
• fix: tokenizer config should use local model path when possible by @drbh in #1518
• Updating tokenizers. by @Narsil in #1517
• [docs] Fix link to Install CLI by @pcuenca in #1526
• feat: add ie update to message docs by @drbh in #1523
• feat: use existing add_generation_prompt variable from config in temp… by @drbh in #1533
• Update to peft 0.8.2 by @Stillerman in #1537
• feat(server): add frequency penalty by @OlivierDehaene in #1541
• chore: bump ci rust version by @drbh in #1543
• ROCm AWQ support by @IlyasMoutawwakil in #1514
• feat(router): add max_batch_size by @OlivierDehaene in #1542
• feat: add deserialize_with that handles strings or objects with content by @drbh in #1550
• Fixing glibc version in the runtime. by @Narsil in #1556
• Upgrade intermediary layer for nvidia too. by @Narsil in #1557
• Improving mamba runtime by using updates by @Narsil in #1552
• Small cleanup. by @Narsil in #1560
• Bugfix: eos and bos tokens positions are inconsistent by @amihalik in #1567
• chore: add pre-commit by @OlivierDehaene in #1569
• feat: add chat template struct to avoid tuple ordering errors by @OlivierDehaene in #1570
• v1.4.1 by @OlivierDehaene in #1568

New Contributors

• @freitng made their first contribution in #1336
• @dtlzhuangz made their first contribution in #1502
• @dwyatte made their first contribution in #1498
• @pcuenca made their first contribution in #1526
• @Stillerman made their first contribution in #1537
• @IlyasMoutawwakil made their first contribution in #1514
• @amihalik made their first contribution in #1563

Full Changelog: v1.4.0...v1.4.1

v1.4.0

Highlights

• OpenAI compatible API #1427
• exllama v2 Tensor Parallel #1490
• GPTQ support for AMD GPUs #1489
• Phi support #1442

What's Changed

• fix: fix local loading for .bin models by @OlivierDehaene in #1419
• Fix missing make target platform for local install: 'install-flash-attention-v2' by @deepily in #1414
• fix: follow base model for tokenizer in router by @OlivierDehaene in #1424
• Fix local load for Medusa by @PYNing in #1420
• Return prompt vs generated tokens. by @Narsil in #1436
• feat: supports openai chat completions API by @drbh in #1427
• feat: support raise_exception, bos and eos tokens by @drbh in #1450
• chore: bump rust version and annotate/fix all clippy warnings by @drbh in #1455
• feat: conditionally toggle chat on invocations route by @drbh in #1454
• Disable decoder_input_details on OpenAI-compatible chat streaming, pass temp and top-k from API by @EndlessReform in #1470
• Fixing non divisible embeddings. by @Narsil in #1476
• Add messages api compatibility docs by @drbh in #1478
• Add a new /tokenize route to get the tokenized input by @Narsil in #1471
• feat: adds phi model by @drbh in #1442
• fix: read stderr in download by @OlivierDehaene in #1486
• fix: show warning with tokenizer config parsing error by @drbh in #1488
• fix: launcher doc typos by @Narsil in #1473
• Reinstate exl2 with tp by @Narsil in #1490
• Add sealion mpt support by @Narsil in #1477
• Trying to fix that flaky test. by @Narsil in #1491
• fix: launcher doc typos by @thelinuxkid in #1462
• Update the docs to include newer models. by @Narsil in #1492
• GPTQ support on ROCm by @fxmarty in #1489
• feat: add tokenizer-config-path to launcher args by @drbh in #1495

New Contributors

• @deepily made their first contribution in #1414
• @PYNing made their first contribution in #1420
• @drbh made their first contribution in #1427
• @EndlessReform made their first contribution in #1470
• @thelinuxkid made their first contribution in #1462

Full Changelog: v1.3.4...v1.4.0

v1.3.4

What's Changed

• feat: relax mistral requirements by @OlivierDehaene in #1351
• fix: fix logic if sliding window key is not present in config by @OlivierDehaene in #1352
• fix: fix offline (#1341) by @OlivierDehaene in #1347
• fix: fix gpt-q with groupsize = -1 by @OlivierDehaene in #1358
• Peft safetensors. by @Narsil in #1364
• Change URL for Habana Gaudi support in doc by @regisss in #1343
• feat: update exllamav2 kernels by @OlivierDehaene in #1370
• Fix local load for peft by @Narsil in #1373

Full Changelog: v1.3.3...v1.3.4

v1.3.3

What's Changed

• fix gptq params loading
• improve decode latency for long sequences two fold
• feat: add more latency metrics in forward by @OlivierDehaene in #1346
• fix: max_past default value must be -1, not 0 by @OlivierDehaene in #1348

Full Changelog: v1.3.2...v1.3.3