mirror of
https://github.com/MindWorkAI/AI-Studio.git
synced 2026-10-18 07:58:00 +00:00
Port the Llama and Qwen open weights
This commit is contained in:
7 files changed
+338
-13
No files matched your search
@@ -8,11 +8,12 @@ namespace AIStudio.Models.Alibaba;
|
||||
/// The Qwen models Alibaba Cloud Model Studio serves.
|
||||
/// </summary>
|
||||
/// <remarks>
|
||||
/// Everything in this folder is bound to Alibaba Cloud, and that is the point of it. Model Studio
|
||||
/// sells commercial models -- qwen-max, qwen3.7-max, qwq-plus -- which carry the family names of
|
||||
/// the open weights without being them, and it answers differently for several names the open
|
||||
/// weights share with it. The old rules kept the two apart by having two functions; here they are
|
||||
/// kept apart by saying which provider a rule speaks for.
|
||||
/// Everything called Model Studio here is bound to Alibaba Cloud, and that is the point of it.
|
||||
/// Model Studio sells commercial models -- qwen-max, qwen3.7-max, qwq-plus -- which carry the
|
||||
/// family names of the open weights without being them, and it answers differently for several
|
||||
/// names the open weights share with it. The old rules kept the two apart by having two functions;
|
||||
/// here they are kept apart by saying which provider a rule speaks for. The unbound families next
|
||||
/// to these are the open weights, which answer everywhere else.
|
||||
///
|
||||
/// The first rule is the catalog's own fallback, and it is written as a plain substring on purpose:
|
||||
/// a substring is the weakest thing a rule can be, so every other rule here beats it without anyone
|
||||
|
||||
Reference in new issue
Block a user