Update llama.py - Fix embedding generation error#5

Open

MikeLP wants to merge 32 commits intoinference-sh:mainfrom

Commits on Jul 4, 2025

feat: add streaming tool use

lsorber
authored and
okaris
committed
fix: remove strict=True to support Python 3.9

lsorber
authored and
okaris
committed
feat: improve tool use robustness

lsorber
authored and
okaris
committed
test: skip if insufficient resources on macOS

lsorber
authored and
okaris
committed
fix: apply missing _convert_text_completion_logprobs_to_chat

lsorber
authored and
okaris
committed
feat: Add Gemma3 chat handler (#1976)

kossum
authored and
okaris
committed
resolve the image embedding issue in gemma3

kossum
authored and
okaris
committed
fix: added n_ctx check for prompt requirements when embedding images in Gemma3ChatHandler

kossum
authored and
okaris
committed
fix: modify the gemma3 chat template to be compatible with openai api

kossum
authored and
okaris
committed
fix: add compatibility with v0.3.9 for Gemma3ChatHandler

kossum
authored and
okaris
committed
feat: abstract context creation and expose for recreation
okaris
committed
feat: add usage to streamin response
okaris
committed
switch to llama.cpp fork and llama : expose C API to get layer device type
okaris
committed
chore: empty commit to trigger rebuild downstream
okaris
committed
c definitions
okaris
committed
chore: bump empty commit
okaris
committed
migrate llava to mtmd
okaris
committed
port kv_cache to new memory
okaris
committed
cleanup
okaris
committed
fixes
okaris
committed
migrate clip to mtmd
okaris
committed
migrate clip to mtmd
okaris
committed
add general purpose function calling handler
okaris
committed
add general purpose function calling handler
okaris
committed
add general purpose function calling handler
okaris
committed
fix recreate context
okaris
committed
bump llama.cpp
okaris
committed
fix deprecated
okaris
committed
fixes
okaris
committed
fixes
okaris
committed
fixes
okaris
committed

Commits on Jul 12, 2025

Update llama.py - Fix embedding generation error
MikeLP
authored