V2D-1268 Use smaller model in demo and reference larger (#5119)

doc/source/servers/huggingface.md

@@ -78,29 +78,22 @@ You can deploy a custom HuggingFace model by providing the location of the model
 apiVersion: machinelearning.seldon.io/v1alpha2
 kind: SeldonDeployment
 metadata:
-  name: custom-gpt2-hf-model
+  name: custom-tiny-stories-model
 spec:
   protocol: v2
   predictors:
   - graph:
       name: transformer
       implementation: HUGGINGFACE_SERVER
-      modelUri: gs://seldon-models/v1.18.0-dev/huggingface/custom-text-generation
+      modelUri: gs://seldon-models/v1.18.0-dev/huggingface/text-gen-custom-tiny-stories
       parameters:
       - name: task
         type: STRING
         value: text-generation
-    componentSpecs:
-      - spec:
-          containers:
-            - name: transformer
-              resources:
-                limits:
-                  cpu: 1
-                  memory: 4Gi
-                requests:
-                  cpu: 100m
-                  memory: 3Gi
     name: default
     replicas: 1
 ```
+
+````{note}
+As a next step, why not try running a larger-scale model? You can find one in gs://seldon-models/v1.18.0-dev/huggingface/text-gen-custom-gpt2. However, you may need to request more memory!
+````

servers/Makefile

@@ -11,4 +11,6 @@ sklearnserver-models:
 
 
 huggingface-models:
-	make -C huggingface/models/text-generation
+	make -C huggingface/models/text-gen-gpt2
+	make -C huggingface/models/text-gen-tiny-stories
+

servers/huggingface/models/text-generation/Makefile → servers/huggingface/models/text-gen-gpt2/Makefile

@@ -13,4 +13,4 @@ train:
 	.env/bin/python train.py
 
 upload:
-	gsutil cp -r text-generation-model-artefacts/* gs://seldon-models/v${VERSION}/huggingface/custom-text-generation/
+	gsutil cp -r text-generation-model-artefacts/* gs://seldon-models/v${VERSION}/huggingface/text-gen-custom-gpt2/

servers/huggingface/models/text-generation/train.py → servers/huggingface/models/text-gen-gpt2/train.py

@@ -15,5 +15,5 @@ def main() -> None:
 
 
 if __name__ == "__main__":
-    print("Building a custom HuggingFace model...")
+    print("Building a custom GPT2 HuggingFace model...")
     main()

servers/huggingface/models/text-gen-tiny-stories/.gitignore

@@ -0,0 +1,2 @@
+.env/
+text-generation-model-artefacts/

servers/huggingface/models/text-gen-tiny-stories/Makefile

@@ -0,0 +1,16 @@
+VERSION := $(shell cat ../../../../version.txt)
+
+all: env train upload
+
+model: env train
+
+env:
+	python3 -m venv .env
+	./.env/bin/pip install --upgrade pip setuptools
+	./.env/bin/pip install -r requirements.txt
+
+train:
+	.env/bin/python train.py
+
+upload:
+	gsutil cp -r text-generation-model-artefacts/* gs://seldon-models/v${VERSION}/huggingface/text-gen-custom-tiny-stories/

servers/huggingface/models/text-gen-tiny-stories/requirements.txt

@@ -0,0 +1,3 @@
+seldon_core
+transformers >=4.30,<4.32
+torch==2.0.0

servers/huggingface/models/text-gen-tiny-stories/train.py

@@ -0,0 +1,19 @@
+from transformers import (
+    AutoModelForCausalLM,
+    AutoTokenizer,
+    pipeline,
+)
+
+
+def main() -> None:
+    tokenizer = AutoTokenizer.from_pretrained("EleutherAI/gpt-neo-125M")
+    model = AutoModelForCausalLM.from_pretrained('roneneldan/TinyStories-1M')
+
+    p = pipeline(task="text-generation", model=model, tokenizer=tokenizer)
+
+    p.save_pretrained("text-generation-model-artefacts")
+
+
+if __name__ == "__main__":
+    print("Building a custom Tiny Stories HuggingFace model...")
+    main()