Skip to content

Releases: Ai00-X/ai00_server

v0.4.2

29 Apr 06:18
47d0c85
Compare
Choose a tag to compare
Bump `web-rwkv` to v0.7.6 (#107)

v0.4.1

21 Apr 15:46
Compare
Choose a tag to compare
Add a `top_p` filter to mirostat sampler.

v0.4.0

18 Apr 05:02
c7fbd27
Compare
Choose a tag to compare

What's Changed

Full Changelog: v0.3.25...v0.4.0

v0.3.25

08 Apr 13:27
Compare
Choose a tag to compare

What's Changed

Full Changelog: v0.3.24...v0.3.25

v0.3.24

28 Mar 10:33
Compare
Choose a tag to compare

What's Changed?

🎈1. Support Save & Load “Pre-quantized Model” with Extension .perfab #83

You can use /api/models/save (post model_path to this api after loading the model with quantization and LoRA merged to save the prefab), and you will get a smaller model file.
And then you can load this file using out favourite /api/models/load. Note that when loading a prefab, parameters like lora and quant are ignored.

👑2. Support BNF #92

Now you could specify bnf_schema in /api/oai/chat/completions.
This will force the model to generate contents that fit the schema (e.g., in JSON format).

🧨 Fixed Many Minor Issues

🥱🥱🥱

Changed list

New Contributors

Full Changelog: v0.3.23...v0.3.24

v0.3.23

21 Mar 15:06
Compare
Choose a tag to compare

What's Changed?

🎈1. Support TLS and http2/http3

You can enable TLS by turning tls = true in assets/configs/Config.toml.
Thanks to salvo!

👑2. Support openAPI and Swagger-UI

Now you can access https://127.0.0.1:65530/swagger-ui/ to use Swagger UI
Thanks to salvo again!

🔔3. Fix something for Loading Lora

Hmm...

ps. 📢📢📢 Something BIG on the way! (⓿_⓿)

v0.3.22

14 Mar 04:39
Compare
Choose a tag to compare
Bump `web-rwkv` to v0.6.25

v0.3.21

11 Mar 13:39
Compare
Choose a tag to compare
Default `penalty_decay` to 0.996.

v0.3.20

28 Feb 17:36
Compare
Choose a tag to compare
Fix token sampling parallelism.

v0.3.19

25 Feb 07:40
Compare
Choose a tag to compare
Update `web-rwkv` to v0.6.21