GitLab

* fix crash in multi-modal
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>

* update according to review comment
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>

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

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

* chore: update to torch 2.4

* remove un-necessary patch

* fix

* Add support for Llama 3 rotary embeddings

* Update transformers to 4.43

* Preparing for release.

* Updating docs.

* Fixing token within the docker image for the launcher.

* Support passing head_dim through config

* Using `head_dim` as a fallback is necessary since it's a non standard
key in mistralConfig (as defined in transformers).

* Shorter diff.

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

* 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(l4): fix fp8 logic on l4

* also quant weights with single scale

* use marlin even on 89

* Softcapping for gemma2.

* Less clutter.

* No access to transformers config, only config_dict here.

* 0.0 is the null value in the C++ API.

* fix(server): fix fp8 weight loading

* fixed scales loading

* update snap

* revert default dtype

* test new instances
Signed-off-by: Adrien <adrien@huggingface.co>

* improve build ci
Signed-off-by: Adrien <adrien@huggingface.co>

---------
Signed-off-by: Adrien <adrien@huggingface.co>

Update README.md

point to huggingface_hub inference clients instead

* Update idefics_causal_lm.py

Fix syntax issues

* fix dbrx & opt model prefix bug

* Hotfix: fix of use of unquantized weights in Mixtral GQA loading

* 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

* re-push to internal registry
Signed-off-by: Adrien <adrien@huggingface.co>

* fix name
Signed-off-by: Adrien <adrien@huggingface.co>

* debug
Signed-off-by: Adrien <adrien@huggingface.co>

* debug
Signed-off-by: Adrien <adrien@huggingface.co>

* wip
Signed-off-by: Adrien <adrien@huggingface.co>

* wip
Signed-off-by: Adrien <adrien@huggingface.co>

* wip debug
Signed-off-by: Adrien <adrien@huggingface.co>

* add debug
Signed-off-by: Adrien <adrien@huggingface.co>

* should
Signed-off-by: Adrien <adrien@huggingface.co>

* wip
Signed-off-by: Adrien <adrien@huggingface.co>

* ww
Signed-off-by: Adrien <adrien@huggingface.co>

* wip
Signed-off-by: Adrien <adrien@huggingface.co>

* wip
Signed-off-by: Adrien <adrien@huggingface.co>

* ww
Signed-off-by: Adrien <adrien@huggingface.co>

* wip
Signed-off-by: Adrien <adrien@huggingface.co>

* wip
Signed-off-by: Adrien <adrien@huggingface.co>

* debug
Signed-off-by: Adrien <adrien@huggingface.co>

* w
Signed-off-by: Adrien <adrien@huggingface.co>

* revert tests
Signed-off-by: Adrien <adrien@huggingface.co>

* last reverts
Signed-off-by: Adrien <adrien@huggingface.co>

* another one
Signed-off-by: Adrien <adrien@huggingface.co>

---------
Signed-off-by: Adrien <adrien@huggingface.co>

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.

* fix: adjust default tool choice

* feat: improve tool choice syntax and response parsing/errors

* fix: remove dev tests

* feat: add ToolChoice to docs

quick fix

* draft of usage stats

* fix wrong link

* launcher doesn't need sysinfo dep

* only tokenizer class instead of hole struct

* unused import

* fix clippy errors

* update openAPI doc

* cargo fmt

* fix error in passing flags to router

* try again to update docs

* run pre-commit locally

* Update router/src/main.rs
Co-authored-by: Hugo Larcher <hugo.larcher@huggingface.co>

* Update router/src/main.rs
Co-authored-by: Hugo Larcher <hugo.larcher@huggingface.co>

* on crash use anonymous error event

* delete json_output and ngrok

* more robust way of checking if is in container

* more robust nvidia smi

* parse xpu more robustly

* fix errors

* add nvidia-smi details in docs

* cargo fmt

* fix clippy

* should make docs check pass

* Update router/src/usage_stats.rs
Co-authored-by: Hugo Larcher <hugo.larcher@huggingface.co>

* error reason can't be in nested json

* cargo fmt

---------
Co-authored-by: Hugo Larcher <hugo.larcher@huggingface.co>
Co-authored-by: Erik Kaunismäki <erikkaum@Eriks-MacBook-Pro.local>

* 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.

Remove bitsandbytes installation when running cpu-only install

* fix to not ignore HUGGINGFACE_HUB_CACHE in cache

* delete printlns

* delete newlines

* maybe fix trailing whitespace

* feat: simple mistral lora integration tests

* fix: include args in docker launcher

* fix: disable cuda graphs with lora and warn

* fix: adjust docs and precommit issues

* fix: re update docs

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`.

* fix: append DONE message to chat stream

* fix: update completions endpoint

Use FP8 GPTQ-Marlin kernels to enable FP8 support on CUDA GPUs
with compute capability >=8.0 and <8.9.
Co-authored-by: Florian Zimmermeister <flozi00.fz@gmail.com>