-
Notifications
You must be signed in to change notification settings - Fork 189
Pull requests: google-ai-edge/litert-torch
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
Replace dictionary key access on
conversation.send_message(...) with
#1245
opened Sep 18, 2026 by
copybara-service
Bot
Loading…
Provide a default text-only implementation for
#1244
opened Sep 17, 2026 by
copybara-service
Bot
Loading…
Move RenderSingleTurnTemplate from ModelDataProcessor to prompt_utils:
#1243
opened Sep 17, 2026 by
copybara-service
Bot
Loading…
Remove unused capability fields (supports_tools, supports_tool_calls,
#1242
opened Sep 17, 2026 by
copybara-service
Bot
Loading…
Fix the github action for generative api examples
#1236
opened Sep 15, 2026 by
copybara-service
Bot
Loading…
Enable the transposed KV cache layout for Qwen3.
#1229
opened Sep 14, 2026 by
copybara-service
Bot
Loading…
npu_export: document the Qualcomm HTP prefill-mask size limit and warn when cache_length crosses it
#1226
opened Sep 11, 2026 by
john-rocky
Contributor
•
Draft
Introduce LiteRT Torch Converter V2 for decoupled weight and MLIR bytecode export.
#1218
opened Sep 3, 2026 by
copybara-service
Bot
Loading…
Derive prompt_templates.model.prefix from the generation prompt
#1215
opened Sep 3, 2026 by
john-rocky
Contributor
Loading…
Restructure Gated Delta Net custom op to gated_delta_update and optimize recurrent execution.
#1208
opened Sep 1, 2026 by
copybara-service
Bot
Loading…
tokenizer_to_sentencepiece: preserve byte-level BPE byte semantics
#1207
opened Sep 1, 2026 by
john-rocky
Contributor
Loading…
Point export_hf int4 docs/examples at the blockwise dynamic_wi4b32_afp32 recipe
CLA:yes
#1202
opened Aug 27, 2026 by
john-rocky
Contributor
Loading…
Fix TFLite FlatBuffer SignatureDef indexing and low-bit constant export
#1199
opened Aug 27, 2026 by
copybara-service
Bot
Loading…
Write start_token only when the model input actually starts with the BOS
CLA:yes
#1198
opened Aug 27, 2026 by
john-rocky
Contributor
Loading…
Probe the chat template with a sentinel instead of searching the rendered output for the prompt
#1197
opened Aug 27, 2026 by
john-rocky
Contributor
Loading…
Preserve single-element structured outputs in the JAX bridge
#1195
opened Aug 26, 2026 by
cdeil
Loading…
Add optional per-signature decode cache sizes for split-context export
#1189
opened Aug 24, 2026 by
anikulk
Loading…
Fix unaligned weight access in XNNPACK depthwise conv packing.
#1187
opened Aug 24, 2026 by
copybara-service
Bot
Loading…
Calibrate with the model's own chat template instead of Gemma's turn markers
#1182
opened Aug 22, 2026 by
john-rocky
Contributor
Loading…
Apply full torch_tfl decomposition table on default lowering path (M1)
#1174
opened Aug 17, 2026 by
copybara-service
Bot
Loading…
Add litert_lm_system_metadata to export config to support custom model metadata.
#1167
opened Aug 11, 2026 by
copybara-service
Bot
Loading…
Add model_type override to allow for tiny_gemma model type population
#1165
opened Aug 6, 2026 by
copybara-service
Bot
Loading…
Previous Next
ProTip!
Updated in the last three days: updated:>2026-09-15.