rixty/KoboldAI-Client
1
1
Fork 0
mirror of https://github.com/KoboldAI/KoboldAI-Client.git synced 2025-02-17 12:10:49 +01:00
Code Issues Projects Releases Wiki Activity

Merge pull request #177 from ebolam/united

Browse Source 
Fix for secondary model loads leaking settings into secondary model's…
This commit is contained in:
henk717 2022-08-05 14:20:25 +02:00 committed by GitHub
commit ed9391431b
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 37 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

25
aiserver.py
View File

@ -1637,6 +1637,30 @@ def patch_transformers():
return stopping_criteria
transformers.generation_utils.GenerationMixin._get_stopping_criteria = new_get_stopping_criteria
def reset_model_settings():
vars.socketio = socketio
vars.max_length = 2048 # Maximum number of tokens to submit per action
vars.ikmax = 3000 # Maximum number of characters to submit to InferKit
vars.genamt = 80 # Amount of text for each action to generate
vars.ikgen = 200 # Number of characters for InferKit to generate
vars.rep_pen = 1.1 # Default generator repetition_penalty
vars.rep_pen_slope = 0.7 # Default generator repetition penalty slope
vars.rep_pen_range = 1024 # Default generator repetition penalty range
vars.temp = 0.5 # Default generator temperature
vars.top_p = 0.9 # Default generator top_p
vars.top_k = 0 # Default generator top_k
vars.top_a = 0.0 # Default generator top-a
vars.tfs = 1.0 # Default generator tfs (tail-free sampling)
vars.typical = 1.0 # Default generator typical sampling threshold
vars.numseqs = 1 # Number of sequences to ask the generator to create
vars.generated_tkns = 0 # If using a backend that supports Lua generation modifiers, how many tokens have already been generated, otherwise 0
vars.badwordsids = []
vars.fp32_model = False # Whether or not the most recently loaded HF model was in fp32 format
vars.modeldim = -1 # Embedding dimension of your model (e.g. it's 4096 for GPT-J-6B and 2560 for GPT-Neo-2.7B)
vars.sampler_order = [0, 1, 2, 3, 4, 5]
vars.newlinemode = "n"
vars.revision = None
def load_model(use_gpu=True, gpu_layers=None, disk_layers=None, initial_load=False, online_model=""):
global model
global generator
@ -1644,6 +1668,7 @@ def load_model(use_gpu=True, gpu_layers=None, disk_layers=None, initial_load=Fal
global model_config
global GPT2TokenizerFast
global tokenizer
reset_model_settings()
if not utils.HAS_ACCELERATE:
disk_layers = None
vars.noai = False

22
static/application.js
View File

@ -1048,6 +1048,18 @@ function buildLoadModelList(ar, menu, breadcrumbs, showdelete) {
if (breadcrumbs.length > 0) {
$("#loadmodellistbreadcrumbs").append("<hr size='1'>")
}
//If we're in the custom load menu (we need to send the path data back in that case)
if(['NeoCustom', 'GPT2Custom'].includes(menu)) {
$("#loadmodel"+i).off("click").on("click", (function () {
return function () {
socket.send({'cmd': 'selectmodel', 'data': $(this).attr("name"), 'path': $(this).attr("pretty_name")});
highlightLoadLine($(this));
}
})(i));
$("#custommodelname").removeClass("hidden");
$("#custommodelname")[0].setAttribute("menu", menu);
}
for(i=0; i<ar.length; i++) {
if (Array.isArray(ar[i][0])) {
full_path = ar[i][0][0];
@ -1090,16 +1102,6 @@ function buildLoadModelList(ar, menu, breadcrumbs, showdelete) {
disableButtons([load_model_accept]);
}
})(i));
//If we're in the custom load menu (we need to send the path data back in that case)
} else if(['NeoCustom', 'GPT2Custom'].includes(menu)) {
$("#loadmodel"+i).off("click").on("click", (function () {
return function () {
socket.send({'cmd': 'selectmodel', 'data': $(this).attr("name"), 'path': $(this).attr("pretty_name")});
highlightLoadLine($(this));
}
})(i));
$("#custommodelname").removeClass("hidden");
$("#custommodelname")[0].setAttribute("menu", menu);
//Normal load
} else {
$("#loadmodel"+i).off("click").on("click", (function () {