Commit Graph

8 Commits

Author SHA1 Message Date
Yingda Chen
2b1c839918 Format llm pipeline (#1094)
* format llm pipeline

Co-authored-by: Yingda Chen <yingda.chen@alibaba-inc.com>
2024-11-22 20:04:59 +08:00
Firmament-cyou
8cdbeec3f6 Support swift model inference for llm pipeline (#880)
* Support swift model inference for llm pipeline

* fix bug

* Add 'swfit' llm_framework and fix stream bug

* For pass the unittest run action
2024-06-30 19:43:13 +08:00
Firmament-cyou
a0fe5947b7 Support stream_generate for LLMPipeline (#768)
* support streaming output for llm_pipeline
* add qwen2 format_messages
2024-03-05 10:05:18 +08:00
Firmament-cyou
8f0f9d4a33 Register llm format map (#659)
* add LLMAdapterRegistry

* fix bug

* replace traceback with cache
2023-12-06 16:22:50 +08:00
Firmament-cyou
fafb0fe013 llm pipeline support chatglm3 (#618) 2023-11-04 20:44:55 +08:00
hemu.zp
4cad376298 Add llm_first parameter for pipeline
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/14264249

* support llm_first parameter

* register_module(Tasks.text_generation)

* fix bug

* update format & fix out_base64 for int4

* pre-commit
2023-10-13 14:04:04 +08:00
hemu.zp
d64cfa48bc Support int4 model for llm_pipeline
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/14214673

* use ms qwen update

* support baichuan2 int4

* update qwen vl unittest and fix bug

* init from AutoModelForCausalLM

* add todo for AutoModelForCausalLM
2023-10-09 16:16:33 +08:00
狄咖
74abc2e63f [feat]chat pipeline 2023-09-28 00:54:41 +08:00