GitLab

* Upgrading our deps.

* fixup.

* Fixup.

* Upgrade outlines to 0.1.1

* Update for new API

* Check if allowed tokens is None

---------
Co-authored-by: Nicolas Patry <patry.nicolas@protonmail.com>

compressed-tensors is a safetensors extension for sparse, quantized
tensors. The format is more powerful than earlier AWQ/GPTQ/FP8
quantization, because

- Different quantizer configurations can be used for different targets.
- The format can specify input/output quantizers in addition to weight
  quantizers.
- Configurable exclusions for quantization.

This change adds a dependency on the `compressed-tensors` package for
its configuration parsing and layer matching functionality.

The following types of quantization are supported in this PR:

- W8A16 and W4A16 INT using GPTQ-Marlin kernels.
- W8A8 and W8A16 FP using FP8-Marlin and cutlass kernels.

Support for other quantization types will be added in subsequent PRs.

* Add support for FP8 KV cache scales

Since FP8 only has limited dynamic range, we can scale keys/values
before storing them into the cache (and unscale them in attention). To
avoid rescaling the cache as the absmax values change, good scales are
usually determined per layer using calibration calibration data and stored
in the checkpoint.

This change adds support for for using key-value scales and loading them
from checkpoints in the two most common formats:

- Separate per-layer `k_scale` and `v_scale` scalars.
- Per-layer `kv_scale` scalar (older format).

Currently, scales are only used with an `float8_e4m3fn` cache.

Besides adding support for key/value scales, the `fp8_quantize` function
is also extended to support quantization with a kernel vendored from
vLLM. This is slightly faster than the PyTorch implementation, but also
scales in FP32, potentially improving accuracy.

* Update FP8 KV cache test to use checkpoint with scales

* `can_scale`: check that the attention is flashinfer

* wip

* rollback

* refactor to use prefix/postfix namming + fix all_input_ids_tensor

* maybe patching vlms?

* fix filter and concat

* wip, no filter, no concat

* current

* add prepare_for_prefill

* working

* load tested

* re-create slots

* re-create slots

* fix slot_filtering_indices

* feedback loop

* remove log

* fix benchmarker

* fix vlm and seq2seq

* rename to cache and input lengths

* fix prefill logprobs

* fix launcher

* fix logprobs?

* idk at this point

* max input length

* omfg

* remove debugging lines

* fix tests

* fix mllama

* fix cargo tests

* remove support chunking for paged

* Fixing non blocked attentions

* Fixing dtype + AMD, Ipex targets.

* lint fix.

* rename

* Fix prefix_caching variable, remove defaults in server (confusing a lot
of the times).

* Add simple resolution when user specifies ATTENTION=paged.

* Put back non default simple tests.

* Fix env name

---------
Co-authored-by: Nicolas Patry <patry.nicolas@protonmail.com>

* (feat) fp8 fnuz support for rocm

* (review comments) Fix compression_config load, type hints

* (bug) update all has_tensor

* (review_comments) fix typo and added comments

* (nit) improved comment

XPU backend is available natively (without IPEX) in pytorch starting
from pytorch 2.4. This commit extends TGI to cover the case when user
has XPU support thru pytorch 2.4, but does not have IPEX installed.
Models which don't require attention can work. For attention required
models more work is needed to provide attention implementation.

Tested with the following models:
* teknium/OpenHermes-2.5-Mistral-7B
* bigscience/bloom-560m
* google/gemma-7b
* google/flan-t5-xxl
Signed-off-by: Dmitry Rogozhkin <dmitry.v.rogozhkin@intel.com>

* Move to moe-kernels package and switch to common MoE layer

This change introduces the new `moe-kernels` package:

- Add `moe-kernels` as a dependency.
- Introduce a `SparseMoELayer` module that can be used by MoE
  models.
- Port over Mixtral and Deepseek.

* Make `cargo check` pass

* Update runner

fix: pass missing revision arg for lora adapter when loading multiple adapters

* feat: support lora revisions and qkv_proj weights

* fix: add qkv_proj weights to weight test

* fix: allocate tmp based on sgmv kernel if available

* fix: re add copy build artifacts step for punica kernels

* hotfix: fix xpu crash brought by code refine. torch.xpu rely on import ipex
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>

* reable gemma2 in xpu
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>

* fix in regression in ipex flashattention
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>

---------
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>
Co-authored-by: Wang, Yi A <yi.a.wang@intel.com>

The `GPTWeightLoader` was structured like this in pseudocode:

if marlin:
  Set up tensors in a way that GPTQ-Marlin expects
else:
  Set up tensors in a way that ExLlama/GPTQ/AWQ expect

However, the GPT-Marlin implementation details should really be in the
`marlin` module. So move the former part out to a separate
`GPTQMarlinWeightsLoader`.

* quick fix

* allow silent failure

* explicit todo that this is only short term

* feat: add ruff and resolve issue

* fix: update client exports and adjust after rebase

* fix: adjust syntax to avoid circular import

* fix: adjust client ruff settings

* fix: lint and refactor import check and avoid model enum as global names

* fix: improve fbgemm_gpu check and lints

* fix: update lints

* fix: prefer comparing model enum over str

* fix: adjust lints and ignore specific rules

* fix: avoid unneeded quantize check

* fix: refactor adapter weight loading and mapping

* feat: enable lora load from directory

* fix: adjust launcher for local lora adapters

* feat: improve weight loading and add tests

* fix: improve logging and rebase syntax issue

* fix: impove adapter merge comments and remove unused conditional

* fix: improve get_model_with_lora_adapters naming

* fix: comment typo

* Add support for repacking AWQ weights for GPTQ-Marlin

So far we couldn't support AWQ because virtually all AWQ models use
symmetric quantization, which GPTQ-Marlin did not suppors. GPTQ-Marlin
has recently added support AWQ repacking and AWQ asymmetric quantization
(zero_point=True).

This change updates all GPTQ-Marlin kernels from upstream and wires up
AWQ support. For now enabling AWQ using Marlin requires running TGI with
`--quantize gptq`.

* Enable Marlin for supported AWQ configurations by default

This makes the AWQ -> GPTQ repack test redundant, since we are now
testing this with the regular AWQ test.

* fix(server): fix fp8 weight loading

* fixed scales loading

* update snap

* revert default dtype

* feat(fp8): add support for fbgemm

* allow loading fp8 weights directly

* update outlines

* fix makefile

* build fbgemm

* avoid circular import and fix dockerfile

* add default dtype

* refactored weights loader

* fix auto conversion

* fix quantization config parsing

* force new nccl on install

* missing get_weights implementation

* increase timeout

Deepseek V2 is a MoE model from Deepseek. Relevant variations
compared to other models:

- Grouped top-K in expert selection.
- mscale in yarn is calculated using the `mscale` and `mscale_all_dim`
  configuration options.
- `mscale_all_dim` is also used in scaling attention softmax.
- Permuting of the query/key representations before applying rotary
  embeddings.
- Some projections cannot be sharded (`q_a_proj`, `kv_a_proj_with_mqa`).
  So, we need weight loads that supports quantized weights. To this
  end `{Weights,WeightLoader}.get_weight` was added.
- The query/key head dimensionality differs from that of the value,
  so we need to pad during attention.
- Heads with size 192, needs an extension to our paged attention
  fork and we need to ensure that the KV cache is allocated with the
  correct size.
- Shared experts.

* Improve the handling of quantized weights

Handling of quantized weights was split between two mechanisms:

- For quantized checkpoints, we used the new weight loader
  infrastructure.
- For quantization while loading (EETQ, FP8, bitsandbytes) we
  instead relied on conditional in `get_linear`.

Weight loaders support context managers to selectively load
particular layers with different weight loaders, which is useful
for models like Idefics2 AWQ, which uses a quantized text model,
but unquantized vision and connector models. However, the context
manager would be overrided by `get_linear`, which string-checks
`quantizer`. Also, the context manager would not work with
EETQ, FP8, and bitsandbytes.

This change migrates all quantizers to the weight loader infrastructure.
This has several benefits:

- We can use context managers with all quantizers.
- All the implementation details move down to the quantizer layers,
  `get_linear` does not need to know how to handle quantizer linear
  layers.
- All quantizer weights are strongly typed, we don't pass around
  raw tensors.
- We don't have to pass around the `quantizer` string everywhere.

* Exclude non-MLP layers when using FP8 quantization with Llama

Fixes #2236.

Fixes #2036.

Packing of asymmetric quantization is broken, all (q)zeros values
of `0` get reset to `1`, resulting in a loss of accuracy. So instead
use symmetric quantization. To be able to distinguish models with
symmetric and asymmetric quantization, a new config tensor `gptq_sym` is
added. If this tensor is not present, we assume `sym=False`.

Quantized weights were loaded in the `Weights` class, but this was
getting quite unwieldy, where every higher level method to load weights
was a long conditional to cover all the different quantizers.

This change moves loading of quantized weights out of the `Weights`
class. This is done by defining a simple `WeightsLoader` interface
that is implemented by `Exl2WeightsLoader`, `GPTQWeightsLoader`,
and `MarlinWeightsLoader`. These implementations are in the quantizers'
respective modules. The `Weights` class provides the low-level load
operations (such as loading tensors or sharded tensors), but delegates
loads that need quantizer-specific weight processing to a loader. The
loaders still use the low-level functionality provided by `Weights`.

I initially tried making a hierarchy where a class like `GPTQWeights`
would inherit from `Weights`. But it is not very flexible (e.g. does
not work well with the new weight storage mock used in tests) and
the implicit indirections made the code harder to follow.

* refine get xpu free memory
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>

* enable qwen2 in xpu
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>

* enable gemma/gemma2/phi in intel platform
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>

---------
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>

GPTQ-Marlin is currently the best-performing kernel for GPTQ models. So
let's use it by default if the kernels are installed, the GPU supports
it, and the kernels support the configuration.

For models generated by `text-generation-server quantize`, use
`sym=False`. This subcommand symmetric quantization since the beginning
and incorrectly reporting the model to be symmetric will use
GPTQ-Marlin (which does not support asymmetric quantization).

This change adds support for 2:4 sparsity when using Marlin
quantization. The 2:4 kernel is used when:

* The quantizer is `marlin`;
* the quantizer checkpoint format is `marlin_24`.

Fixes #2098.

* feat: first draft load multiple lora

* feat: load weights within layer and refactor lora pass

* fix: refactor and reduce lora math

* feat: baseline impl single request multi lora support

* feat: prefer lorax implementation and port loading logic

* fix: prefer adapter_data and refactors

* feat: perfer loraxs custom punica kernels and add mlp loras

* fix: adjust batch for bgmv

* fix: adjust adapter_segments logic when in batch

* fix: refactor and move changes to v3 proto

* fix: pass model_id for all flash causal lms

* fix: pass model_id for all causal and seq2seq lms

* fix: add model_id to model test

* feat: add lora support to mistral and refactors

* feat: prefer model id in request

* fix: include rust code for adapter id

* feat: bump launcher and add new lora docs

* feat: support base model generation and refactors

* fix: rename doc to retry ci build

* feat: support if vlm models

* fix: add adapter_data param and avoid missing layers

* fix: add adapter_data param to phi and neox

* fix: update all models forwards to include adapter_data

* fix: add model_id to IdeficsCausalLM

* Update lora.md

Fixed a typo

* Update lora.md

Fixing spam image

* fix: add lora kernel to dockerfile, support running without kernels and refactors

* fix: avoid dockerfile conflict

* fix: refactors and adjust flash llama lora logic

* fix: skip llama test due to CI issue (temp)

* fix: skip llama test CI (temp) 2

* fix: revert skips and prefer updated ci token for tests

* fix: refactors and helpful comments

* fix: add noop in TensorParallelAdapterRowLinear too

* fix: refactor and move shard_lora_weights logic

* fix: exit early if no adapter_data

---------
Co-authored-by: Derek <datavistics@gmail.com>

* Removing IPEX_AVAIL.

Chose to unify CPU and XPU under `ipex`. Most code is exactly similar
except for a very few spots.

The biggest number of spots is the kv-cache layout and the flash_xxx.py
files.
Since those files should be removed soon and factored away, we should
not need them.

* Forgot a few places.

* Unrelated change.

* Fixing HF_TOKEN.

* HF_TOKEN

* add CPU tgi support
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>

* ipex distributed ops support
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>

---------
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>
Co-authored-by: Funtowicz Morgan <mfuntowicz@users.noreply.github.com>

* use xpu-smi to dump used memory
xpu use "ZE_AFFINITY_MASK" to control card, usage is like CUDA_VISIBLE_DEVICES
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>

* Update server/text_generation_server/utils/import_utils.py
Co-authored-by: Daniël de Kok <me@github.danieldk.eu>

---------
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>
Co-authored-by: Daniël de Kok <me@github.danieldk.eu>

For Phi-3-Small I need to shard a packed QKV bias tensor, for which
I implemented the `Weights.get_packed_sharded` method. However, this
method can also replace the `Weights._get_qweight` method and the
custom sharding code from `Weights.get_weights_col_packed`.

Add support for GPTQ Marlin kernels

GPTQ Marlin extends the Marlin kernels to support common GPTQ
configurations:

- bits: 4 or 8
- groupsize: -1, 32, 64, or 128
- desc_act: true/false

Using the GPTQ Marlin kernels requires repacking the parameters in the
Marlin quantizer format.

The kernels were contributed by Neural Magic to VLLM. We vendor them
here for convenience.

Add support for Phi-3-medium

The main difference between the medium and mini models is that medium
uses grouped query attention with a packed QKV matrix. This change adds
support for GQA with packed matrixes to `Weights.get_weights_col_packed`
and uses it for Phi-3. This also allows us to remove the custom
implementation of GQA from dbrx attention loading.

The router will now send the input as chunks besides as a single
string. This change modifies the server to process chunked input
rather than strings. This also allows us to remove the image
extraction code from the server.

This change adds support for Marlin-quantized models. Marlin is an
FP16xINT4 matmul kernel, which provides good speedups decoding batches
of 16-32 tokens. It supports quantized models with symmetric
quantization, groupsize -1 or 128, and 4-bit.

Tested with:

- Llama 2
- Llama 3
- Phi 3

# What does this PR do?

The GPTQ code path for column-packed packed tensors assumed that this is
always a QKV matrix. However, models (e.g. Phi-3) can also have
column-packed MLP up/gate matrices.

<!-- Remove if not applicable -->

Fixes # (issue)


## Before submitting
- [ ] This PR fixes a typo or improves the docs (you can dismiss the
other checks if that's the case).
- [x] Did you read the [contributor
guideline](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md#start-contributing-pull-requests),
      Pull Request section?
- [ ] Was this discussed/approved via a Github issue or the
[forum](https://discuss.huggingface.co/)? Please add a link
      to it if that's the case.
- [ ] Did you make sure to update the documentation with your changes?
Here are the
[documentation
guidelines](https://github.com/huggingface/transformers/tree/main/docs),
and
[here are tips on formatting
docstrings](https://github.com/huggingface/transformers/tree/main/docs#writing-source-documentation).
- [ ] Did you write any new necessary tests?


## Who can review?

Anyone in the community is free to review the PR once the tests have
passed. Feel free to tag
members/contributors who may be interested in your PR.

<!-- Your PR will be replied to more quickly if you can figure out the
right person to tag with @


@OlivierDehaene OR @Narsil

 -->