Update app.py

update and simplify page to support OpenELM-270M

app.py

@@ -12,50 +12,46 @@ DEFAULT_MAX_NEW_TOKENS = 256
 MAX_INPUT_TOKEN_LENGTH = 512
 
 DESCRIPTION = """\
-# OpenELM-3B-Instruct
+# OpenELM-270M-Instruct -- Running on CPU
 
-This Space demonstrates [OpenELM-3B-Instruct](https://huggingface.co/apple/OpenELM-3B-Instruct) by Apple. Please, check the original model card for details.
-You can see the other models of the OpenELM family [here](https://huggingface.co/apple/OpenELM)
-The following Colab notebooks are available:
-* [OpenELM-3B-Instruct (GPU)](https://gist.github.com/Norod/4f11bb36bea5c548d18f10f9d7ec09b0)
-* [OpenELM-270M (CPU)](https://gist.github.com/Norod/5a311a8e0a774b5c35919913545b7af4)
+This Space demonstrates [apple/OpenELM-270M-Instruct](https://huggingface.co/apple/OpenELM-270M-Instruct) by Apple. Please, check the original model card for details.
 
-You might also be interested in checking out Apple's [CoreNet Github page](https://github.com/apple/corenet?tab=readme-ov-file).
+For detail on the OpenELM model refer to the Paper page [here](https://huggingface.co/papers/2404.14619)
 
-If you duplicate this space, make sure you have access to [meta-llama/Llama-2-7b-hf](https://huggingface.co/meta-llama/Llama-2-7b-hf)
-because this model uses it as a tokenizer.
-
-# Note: Use this model for only for completing sentences and instruction following.
+For detail on the pre-training, instruct tuning, and parameter-efficient finetuning process refer to the [OpenELM page in the CoreNet GitHub repository](https://github.com/apple/corenet/tree/main/projects/openelm)
 """
 
 LICENSE = """
 <p/>
 
 ---
-As a derivative work of [OpenELM-3B-Instruct](https://huggingface.co/apple/OpenELM-3B-Instruct) by Apple,
-this demo is governed by the original [license](https://huggingface.co/apple/OpenELM-3B-Instruct/blob/main/LICENSE).
+As a derivative work of [apple/OpenELM-270M-Instruct](https://huggingface.co/apple/OpenELM-270M-Instruct) by Apple,
+this demo is governed by the original [license](https://huggingface.co/apple/OpenELM-270M-Instruct/blob/main/LICENSE).
+---
+based on the [Norod78/OpenELM_3B_Demo](https://huggingface.co/spaces/Norod78/OpenELM_3B_Demo) space - I encourage you to like his space as well.  I have a lot of respect for how he promoted and shared information about this unique model.
 """
 
-if not torch.cuda.is_available():
-    DESCRIPTION += "\n<p>Running on CPU 🥶 This demo does not work on CPU.</p>"
 
+model = AutoModelForCausalLM.from_pretrained(
+    "apple/OpenELM-270M-Instruct",
+    trust_remote_code=True,
+)
+tokenizer = AutoTokenizer.from_pretrained(
+    "NousResearch/Llama-2-7b-hf",
+    trust_remote_code=True,
+    tokenizer_class=LlamaTokenizer,
+)
 
-if torch.cuda.is_available():
-    model_id = "apple/OpenELM-3B-Instruct"    
-    model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto", trust_remote_code=True, low_cpu_mem_usage=True)
-    tokenizer_id = "meta-llama/Llama-2-7b-hf"
-    tokenizer = AutoTokenizer.from_pretrained(tokenizer_id)
-    if tokenizer.pad_token == None:
-        tokenizer.pad_token = tokenizer.eos_token
-        tokenizer.pad_token_id = tokenizer.eos_token_id
-        model.config.pad_token_id = tokenizer.eos_token_id
+if tokenizer.pad_token == None:
+    tokenizer.pad_token = tokenizer.eos_token
+    tokenizer.pad_token_id = tokenizer.eos_token_id
+    model.config.pad_token_id = tokenizer.eos_token_id
 
-@spaces.GPU
 def generate(
     message: str,
     chat_history: list[tuple[str, str]],
     max_new_tokens: int = 1024,
-    temperature: float = 0.6,
+    temperature: float = 0.1,
     top_p: float = 0.9,
     top_k: int = 50,
     repetition_penalty: float = 1.4,
@@ -139,17 +135,16 @@ chat_interface = gr.ChatInterface(
     ],
     stop_btn=None,
     examples=[
-        ["A recipe for a chocolate cake:"],
-        ["Can you explain briefly to me what is the Python programming language?"],
-        ["Explain the plot of Cinderella in a sentence."],
-        ["Question: What is the capital of France?\nAnswer:"],
-        ["Question: I am very tired, what should I do?\nAnswer:"],
+        ["Tell me a joke about a sandwich:"],
+        ["What would a polite pirate say?"],
+        ["Explain quantum physics in 5 words or less:"],
+        ["Question: Why don't scientists trust atoms?\nAnswer:"],
+        ["Question: What do you call a bear with no teeth?\nAnswer:"],
     ],
 )
 
 with gr.Blocks(css="style.css") as demo:
     gr.Markdown(DESCRIPTION)
-    gr.DuplicateButton(value="Duplicate Space for private use", elem_id="duplicate-button")
     chat_interface.render()
     gr.Markdown(LICENSE)