huggingface/text-generation-inference

Pull Requests Commits

add 'mamba' as model config

ErikKaum committed 1 year ago

060b2db0

Unify attention output handling (#2343)

danieldk committed 1 year ago

Verified 47447ef0

Fix cache block size for flash decoding (#2351)

danieldk committed 1 year ago

Verified 22fb1be5

enable HuggingFaceM4/idefics-9b in intel gpu (#2338)

sywangyi committed 1 year ago

Verified 9ab99374

refactor usage stats (#2339)

ErikKaum committed 1 year ago

Verified 7451041e

Pr 2290 ci run (#2329)

drbh committed 1 year ago

Verified f7f61876

Handle GPTQ-Marlin loading in `GPTQMarlinWeightLoader` (#2300)

danieldk committed 1 year ago

Verified 34f7dcfd

Rebase TRT-llm (#2331)

Narsil committed 1 year ago

Verified 2b19d671

server quantize: store quantizer config in standard format (#2299)

danieldk committed 1 year ago

Verified 53aec273

fix: adjust test snapshots and small refactors (#2323)

drbh committed 1 year ago

Verified 0b95693f

patch-error-on-invalid-grammar (#2282)

ErikKaum committed 1 year ago

Verified 3d7f4f41

fix: reject grammars without properties (#2309)

drbh committed 1 year ago

Verified f15e808d

Install Marlin from standalone package (#2320)

danieldk committed 1 year ago

Verified 922732b2

Run ci api key (#2315)

ErikKaum committed 1 year ago

Verified 583d37a2

fix: fix buildkit config in ci

XciD committed 1 year ago

Verified fd2e0631

feat: add ruff and resolve issue (#2262)

drbh committed 1 year ago

Verified bab02ff2

Support tied embeddings in 0.5B and 1.5B Qwen2 models (#2313)

danieldk committed 1 year ago

Verified 4b49c50f

Fix registry name (#2307)

XciD committed 1 year ago

Verified 3905f854

Fixing idefics on g6 tests. (#2306)

Narsil committed 1 year ago

Verified 17ed42be

Some small fixes for the Torch 2.4.0 update (#2304)

danieldk committed 1 year ago

Verified 9256d7c3

Using g6 instead of g5. (#2281)

Narsil committed 1 year ago

Verified 26614057

fix: refactor adapter weight loading and mapping (#2193)

drbh committed 1 year ago

Verified 5d85a958

Split up `layers.marlin` into several files (#2292)

danieldk committed 1 year ago

Verified 93d2b9fe

fix of use of unquantized weights in cohere GQA loading, also enable … (#2291)

sywangyi committed 1 year ago

Verified 86422506

fix crash in multi-modal (#2245)

sywangyi committed 1 year ago

Verified 5ad39dd3

hotfix: update nccl

OlivierDehaene committed 1 year ago

Verified a8950294

chore: update to torch 2.4 (#2259)

OlivierDehaene committed 1 year ago

Verified e7e3aa6c

hotfix: pin numpy (#2289)

danieldk committed 1 year ago

Verified bc9593a5

Add support for Llama 3 rotary embeddings (#2286)

danieldk committed 1 year ago

Verified 4ab41737

Preparing for release. (#2285)

Narsil committed 1 year ago

Verified 5d121a97

Older