HF 4.33.1

environments/huggingface.yml

@@ -19,7 +19,6 @@ dependencies:
   - bleach=4.1.0
   - pip
   - git=2.35.1
-  - sentencepiece
   - protobuf
   - marshmallow>=3.13
   - apispec-webframeworks
@@ -32,9 +31,9 @@ dependencies:
     - flask-ngrok
     - flask-cors
     - lupa==1.10
-    - transformers==4.32.1
+    - transformers[sentencepiece]==4.33.1
     - huggingface_hub==0.16.4
-    - optimum==1.12.0
+    - optimum[onnx]==1.12.0
     - safetensors==0.3.3
     - accelerate==0.21.0
     - git+https://github.com/VE-FORBRYDERNE/mkultra

environments/ipex.yml

@@ -15,7 +15,6 @@ dependencies:
   - bleach=4.1.0
   - pip
   - git=2.35.1
-  - sentencepiece
   - protobuf
   - marshmallow>=3.13
   - apispec-webframeworks
@@ -31,9 +30,9 @@ dependencies:
     - flask-ngrok
     - flask-cors
     - lupa==1.10
-    - transformers==4.32.1
+    - transformers[sentencepiece]==4.33.1
     - huggingface_hub==0.16.4
-    - optimum==1.12.0
+    - optimum[onnx]==1.12.0
     - safetensors==0.3.3
     - accelerate==0.20.3
     - git+https://github.com/VE-FORBRYDERNE/mkultra

environments/rocm.yml

@@ -15,7 +15,6 @@ dependencies:
   - bleach=4.1.0
   - pip
   - git=2.35.1
-  - sentencepiece
   - protobuf
   - marshmallow>=3.13
   - apispec-webframeworks
@@ -30,9 +29,9 @@ dependencies:
     - flask-ngrok
     - flask-cors
     - lupa==1.10
-    - transformers==4.32.1
+    - transformers[sentencepiece]==4.33.1
     - huggingface_hub==0.16.4
-    - optimum==1.12.0
+    - optimum[onnx]==1.12.0
     - safetensors==0.3.3
     - accelerate==0.21.0
     - git+https://github.com/VE-FORBRYDERNE/mkultra

modeling/inference_models/hf.py

@@ -230,6 +230,8 @@ class HFInferenceModel(InferenceModel):
     def _post_load(self) -> None:
         self.badwordsids = koboldai_settings.badwordsids_default
         self.model_type = str(self.model_config.model_type)
+        self.model.use_cache = True # Workaround for models that accidentally uploaded with False
+        
         # These are model specific tokenizer overrides if a model has bad defaults
         if self.model_type == "llama":
             # Note: self.tokenizer is a GenericTokenizer, and self.tokenizer.tokenizer is the actual LlamaTokenizer

requirements.txt

@@ -1,6 +1,6 @@
-transformers==4.32.1
+transformers[sentencepiece]==4.33.1
 huggingface_hub==0.16.4
-optimum==1.12.0
+optimum[onnx]==1.12.0
 safetensors==0.3.3
 Flask==2.2.3
 Flask-SocketIO==5.3.2
@@ -15,7 +15,6 @@ dnspython==2.2.1
 lupa==1.10
 markdown
 bleach==4.1.0
-sentencepiece
 protobuf
 accelerate==0.21.0
 flask-session==0.4.0