rixty/KoboldAI-Client
1
1
Fork 0
mirror of https://github.com/KoboldAI/KoboldAI-Client.git synced 2025-04-16 19:27:18 +02:00
Code Issues Projects Releases Wiki Activity

Path Fixes

Browse Source 
Fixes the tokenizer cache being hit when we already have a local model
This commit is contained in:
henk717 2021-12-27 01:56:59 +01:00
parent 1a64f8bdc4
commit 6d1bf76ef1
1 changed files with 4 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
aiserver.py
View File

@ -902,9 +902,9 @@ if(not vars.model in ["InferKit", "Colab", "OAI", "ReadOnly", "TPUMeshTransforme
elif(os.path.isdir(vars.model.replace('/', '_'))): elif(os.path.isdir(vars.model.replace('/', '_'))):
with(maybe_use_float16()): with(maybe_use_float16()):
try: try:
tokenizer = AutoTokenizer.from_pretrained(vars.custmodpth, cache_dir="cache/") tokenizer = AutoTokenizer.from_pretrained(vars.model.replace('/', '_'), cache_dir="cache/")
except ValueError as e: except ValueError as e:
tokenizer = GPT2TokenizerFast.from_pretrained(vars.custmodpth, cache_dir="cache/") tokenizer = GPT2TokenizerFast.from_pretrained(vars.model.replace('/', '_'), cache_dir="cache/")
try: try:
model = AutoModelForCausalLM.from_pretrained(vars.model.replace('/', '_'), cache_dir="cache/", **lowmem) model = AutoModelForCausalLM.from_pretrained(vars.model.replace('/', '_'), cache_dir="cache/", **lowmem)
except ValueError as e: except ValueError as e:
@ -912,9 +912,9 @@ if(not vars.model in ["InferKit", "Colab", "OAI", "ReadOnly", "TPUMeshTransforme
else: else:
print("Model does not exist locally, attempting to download from Huggingface...") print("Model does not exist locally, attempting to download from Huggingface...")
try: try:
tokenizer = AutoTokenizer.from_pretrained(vars.custmodpth, cache_dir="cache/") tokenizer = AutoTokenizer.from_pretrained(vars.model, cache_dir="cache/")
except ValueError as e: except ValueError as e:
tokenizer = GPT2TokenizerFast.from_pretrained(vars.custmodpth, cache_dir="cache/") tokenizer = GPT2TokenizerFast.from_pretrained(vars.model, cache_dir="cache/")
with(maybe_use_float16()): with(maybe_use_float16()):
try: try:
model = AutoModelForCausalLM.from_pretrained(vars.model, cache_dir="cache/", **lowmem) model = AutoModelForCausalLM.from_pretrained(vars.model, cache_dir="cache/", **lowmem)