rixty
/
KoboldAI-Client
mirror of https://github.com/KoboldAI/KoboldAI-Client.git
1
1
Fork 0
Code Issues Projects Releases Wiki Activity

Fix a strange bug where max tokens was capped at 1024 

This seems to be related to the model config files, because only certain
models have this problem, and replacing ALL configuration files of a
"bad" model with those of a "good" model of the same type would fix the
problem.

Shouldn't be required anymore.
This commit is contained in:
Gnome Ann 2021-12-15 00:45:41 -05:00
parent 1551c45ba4
commit 2687135e05
1 changed files with 9 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

18
aiserver.py
View File

@ -2000,17 +2000,17 @@ def calcsubmitbudget(actionlen, winfo, mem, anotetxt, actions):
lnanote = 0 # Placeholder for Author's Note length
# Calculate token budget
prompttkns = tokenizer.encode(vars.comregex_ai.sub('', vars.prompt))
prompttkns = tokenizer.encode(vars.comregex_ai.sub('', vars.prompt), max_length=1+int(vars.max_length), truncation=True)
lnprompt = len(prompttkns)
memtokens = tokenizer.encode(mem)
memtokens = tokenizer.encode(mem, max_length=1+int(vars.max_length), truncation=True)
lnmem = len(memtokens)
witokens = tokenizer.encode(winfo)
witokens = tokenizer.encode(winfo, max_length=1+int(vars.max_length), truncation=True)
lnwi = len(witokens)
if(anotetxt != ""):
anotetkns = tokenizer.encode(anotetxt)
anotetkns = tokenizer.encode(anotetxt, max_length=1+int(vars.max_length), truncation=True)
lnanote = len(anotetkns)
lnsp = vars.sp.shape[0] if vars.sp is not None else 0
@ -2039,7 +2039,7 @@ def calcsubmitbudget(actionlen, winfo, mem, anotetxt, actions):
if(budget <= 0):
break
acttkns = tokenizer.encode(chunk)
acttkns = tokenizer.encode(chunk, max_length=int(vars.max_length), truncation=True)
tknlen = len(acttkns)
if(tknlen < budget):
tokens = acttkns + tokens
@ -2173,7 +2173,7 @@ def calcsubmit(txt):
#==================================================================#
def _generate(txt, minimum, maximum, found_entries):
gen_in = tokenizer.encode(txt, return_tensors="pt", truncation=True).long()
gen_in = tokenizer.encode(txt, return_tensors="pt", max_length=int(vars.max_length), truncation=True).long()
if(vars.sp is not None):
soft_tokens = torch.arange(
model.config.vocab_size,
@ -2226,7 +2226,7 @@ def _generate(txt, minimum, maximum, found_entries):
winfo, mem, anotetxt, _found_entries = calcsubmitbudgetheader(txt, force_use_txt=True)
found_entries[i].update(_found_entries)
txt, _, _ = calcsubmitbudget(len(actions), winfo, mem, anotetxt, actions)
encoded.append(tokenizer.encode(txt, return_tensors="pt", truncation=True)[0].long().to(genout.device))
encoded.append(tokenizer.encode(txt, return_tensors="pt", max_length=int(vars.max_length), truncation=True)[0].long().to(genout.device))
max_length = len(max(encoded, key=len))
encoded = torch.stack(tuple(torch.nn.functional.pad(e, (max_length - len(e), 0), value=model.config.pad_token_id or model.config.eos_token_id) for e in encoded))
genout = torch.cat(
@ -2532,8 +2532,8 @@ def getnewcontent(txt):
return txt
# Tokenize the last context and the generated content
ctxtokens = tokenizer.encode(vars.lastctx)
txttokens = tokenizer.encode(txt)
ctxtokens = tokenizer.encode(vars.lastctx, max_length=1+int(vars.max_length), truncation=True)
txttokens = tokenizer.encode(txt, max_length=1+int(vars.max_length), truncation=True)
dif = (len(txttokens) - len(ctxtokens)) * -1
# Remove the context from the returned text