rixty/KoboldAI-Client
1
1
Fork 0
mirror of https://github.com/KoboldAI/KoboldAI-Client.git synced 2025-06-05 21:59:24 +02:00
Code Issues Projects Releases Wiki Activity

Potential fix for gpt-neo and gpt-j

Browse Source
This commit is contained in:
ebolam
2023-05-23 19:25:01 -04:00
parent 7a8e4c39da
commit 839d56ebf2
1 changed files with 11 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

21
modeling/inference_models/hf.py
View File

@@ -199,8 +199,9 @@ class HFInferenceModel(InferenceModel):
pass pass
def _post_load(self) -> None: def _post_load(self) -> None:
self.model_type = str(self.model_config.model_type)
# These are model specific tokenizer overrides if a model has bad defaults # These are model specific tokenizer overrides if a model has bad defaults
if utils.koboldai_vars.model_type == "llama": if self.model_type == "llama":
# Note: self.tokenizer is a GenericTokenizer, and self.tokenizer.tokenizer is the actual LlamaTokenizer # Note: self.tokenizer is a GenericTokenizer, and self.tokenizer.tokenizer is the actual LlamaTokenizer
self.tokenizer.add_bos_token = False self.tokenizer.add_bos_token = False
@@ -284,23 +285,23 @@ class HFInferenceModel(InferenceModel):
return result return result
object.__setattr__(self.tokenizer, '__call__', call_wrapper.__get__(self.tokenizer)) object.__setattr__(self.tokenizer, '__call__', call_wrapper.__get__(self.tokenizer))
elif utils.koboldai_vars.model_type == "opt": elif self.model_type == "opt":
self.tokenizer._koboldai_header = self.tokenizer.encode("") self.tokenizer._koboldai_header = self.tokenizer.encode("")
self.tokenizer.add_bos_token = False self.tokenizer.add_bos_token = False
self.tokenizer.add_prefix_space = False self.tokenizer.add_prefix_space = False
# Change newline behavior to match model quirks # Change newline behavior to match model quirks
if utils.koboldai_vars.model_type == "xglm": if self.model_type == "xglm":
# Default to </s> newline mode if using XGLM # Default to </s> newline mode if using XGLM
utils.koboldai_vars.newlinemode = "s" utils.koboldai_vars.newlinemode = "s"
elif utils.koboldai_vars.model_type in ["opt", "bloom"]: elif self.model_type in ["opt", "bloom"]:
# Handle </s> but don't convert newlines if using Fairseq models that have newlines trained in them # Handle </s> but don't convert newlines if using Fairseq models that have newlines trained in them
utils.koboldai_vars.newlinemode = "ns" utils.koboldai_vars.newlinemode = "ns"
# Clean up tokens that cause issues # Clean up tokens that cause issues
if ( if (
utils.koboldai_vars.badwordsids == koboldai_settings.badwordsids_default utils.koboldai_vars.badwordsids == koboldai_settings.badwordsids_default
and utils.koboldai_vars.model_type not in ("gpt2", "gpt_neo", "gptj") and self.model_type not in ("gpt2", "gpt_neo", "gptj")
): ):
utils.koboldai_vars.badwordsids = [ utils.koboldai_vars.badwordsids = [
[v] [v]
@@ -357,15 +358,15 @@ class HFInferenceModel(InferenceModel):
revision=utils.koboldai_vars.revision, revision=utils.koboldai_vars.revision,
cache_dir="cache", cache_dir="cache",
) )
utils.koboldai_vars.model_type = self.model_config.model_type self.model_type = self.model_config.model_type
except ValueError: except ValueError:
utils.koboldai_vars.model_type = { self.model_type = {
"NeoCustom": "gpt_neo", "NeoCustom": "gpt_neo",
"GPT2Custom": "gpt2", "GPT2Custom": "gpt2",
}.get(utils.koboldai_vars.model) }.get(self.model)
if not utils.koboldai_vars.model_type: if not self.model_type:
logger.warning( logger.warning(
"No model type detected, assuming Neo (If this is a GPT2 model use the other menu option or --model GPT2Custom)" "No model type detected, assuming Neo (If this is a GPT2 model use the other menu option or --model GPT2Custom)"
) )
utils.koboldai_vars.model_type = "gpt_neo" self.model_type = "gpt_neo"