COMMITS
May 20, 2025
W
Fix the crash in default ATTENTION path for Gaudi backend (#3235)
Wang, Yi committed
W
Refine warmup and upgrade to synapse AI 1.21.0 (#3234)
Wang, Yi committed
May 19, 2025
W
Deepseek R1 for Gaudi backend (#3211)
Wang, Yi committed
May 16, 2025
May 15, 2025
Y
Enable Llama4 for Gaudi backend (#3223)
Yuan Wu committed
D
Update to Torch 2.7.0 (#3221)
Daniël de Kok committed
May 12, 2025
K
Adjust the `round_up_seq` logic in Gaudi backend (#3224)
kaixuanliu committed
May 10, 2025
K
Change HPU warmup logic: seq length should be with exponential growth (#3217)
kaixuanliu committed
May 9, 2025
D
Prepare for 3.3.0 (#3220)
Daniël de Kok committed
May 6, 2025
M
Chunked Prefill VLM (#3188)
Mohit Sharma committed
W
forward and tokenize chooser use the same shape (#3196)
Wang, Yi committed
W
IPEX support FP8 kvcache/softcap/slidingwindow (#3144)
Wang, Yi committed
R
Fix `HF_HUB_OFFLINE=1` for Gaudi backend (#3193)
regisss committed
May 1, 2025
J
Update client SDK snippets (#3207)
Julien Chaumond committed
D
fix: bump snaps for mllama (#3202)
drbh committed
D
Pr 2982 ci branch (#3046)
drbh committed
J
doc typo (#3206)
Julien Chaumond committed
A
Skip `{% generation %}` and `{% endgeneration %}` template handling (#3204)
Alvaro Bartolome committed
April 29, 2025
N
Fixing the router + template for Qwen3. (#3200)
Nicolas Patry committed
April 24, 2025
N
Put more wiggle room. (#3189)
Nicolas Patry committed
W
Warmup gaudi backend (#3172)
Wang, Yi committed
April 23, 2025
M
Add option to configure prometheus port (#3187)
Mohit Sharma committed
April 18, 2025
N
Fixing CI (#3184)
Nicolas Patry committed
A
Bump `sccache` to 0.10.0 (#3179)
Alvaro Bartolome committed
H
Get opentelemetry trace id from request headers instead of creating a new trace (#2648)
Hyeongchan Kim committed
April 17, 2025
D
Support flashinfer for Gemma3 prefill (#3167)
Daniël de Kok committed
April 15, 2025
N
Hotfix gaudi2 with newer transformers. (#3176)
Nicolas Patry committed
N
Hotfixing gaudi deps. (#3174)
Nicolas Patry committed
N
Upgrading the dependencies in Gaudi backend. (#3170)
Nicolas Patry committed
W
transformers flash llm/vlm enabling in ipex (#3152)
Wang, Yi committed