Duplicate from stabilityai/stablelm-tuned-alpha-chat

Co-authored-by: reshinth.adith <reshinthadith@users.noreply.huggingface.co>

.gitattributes

@@ -0,0 +1,34 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,13 @@
+---
+title: Stablelm Tuned Alpha Chat
+emoji: 👀
+colorFrom: purple
+colorTo: pink
+sdk: gradio
+sdk_version: 3.27.0
+app_file: app.py
+pinned: false
+duplicated_from: stabilityai/stablelm-tuned-alpha-chat
+---
+
+Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -0,0 +1,111 @@
+import gradio as gr
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline, StoppingCriteria, StoppingCriteriaList
+import time
+import numpy as np
+from torch.nn import functional as F
+import os
+# auth_key = os.environ["HF_ACCESS_TOKEN"]
+print(f"Starting to load the model to memory")
+m = AutoModelForCausalLM.from_pretrained(
+    "stabilityai/stablelm-tuned-alpha-7b", torch_dtype=torch.float16).cuda()
+tok = AutoTokenizer.from_pretrained("stabilityai/stablelm-tuned-alpha-7b")
+generator = pipeline('text-generation', model=m, tokenizer=tok, device=0)
+print(f"Sucessfully loaded the model to the memory")
+
+start_message = """<|SYSTEM|># StableAssistant
+- StableAssistant is A helpful and harmless Open Source AI Language Model developed by Stability and CarperAI.
+- StableAssistant is excited to be able to help the user, but will refuse to do anything that could be considered harmful to the user.
+- StableAssistant is more than just an information source, StableAssistant is also able to write poetry, short stories, and make jokes.
+- StableAssistant will refuse to participate in anything that could harm a human."""
+
+
+class StopOnTokens(StoppingCriteria):
+    def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor, **kwargs) -> bool:
+        stop_ids = [50278, 50279, 50277, 1, 0]
+        for stop_id in stop_ids:
+            if input_ids[0][-1] == stop_id:
+                return True
+        return False
+
+
+def contrastive_generate(text, bad_text):
+    with torch.no_grad():
+        tokens = tok(text, return_tensors="pt")[
+            'input_ids'].cuda()[:, :4096-1024]
+        bad_tokens = tok(bad_text, return_tensors="pt")[
+            'input_ids'].cuda()[:, :4096-1024]
+        history = None
+        bad_history = None
+        curr_output = list()
+        for i in range(1024):
+            out = m(tokens, past_key_values=history, use_cache=True)
+            logits = out.logits
+            history = out.past_key_values
+            bad_out = m(bad_tokens, past_key_values=bad_history,
+                        use_cache=True)
+            bad_logits = bad_out.logits
+            bad_history = bad_out.past_key_values
+            probs = F.softmax(logits.float(), dim=-1)[0][-1].cpu()
+            bad_probs = F.softmax(bad_logits.float(), dim=-1)[0][-1].cpu()
+            logits = torch.log(probs)
+            bad_logits = torch.log(bad_probs)
+            logits[probs > 0.1] = logits[probs > 0.1] - bad_logits[probs > 0.1]
+            probs = F.softmax(logits)
+            out = int(torch.multinomial(probs, 1))
+            if out in [50278, 50279, 50277, 1, 0]:
+                break
+            else:
+                curr_output.append(out)
+            out = np.array([out])
+            tokens = torch.from_numpy(np.array([out])).to(
+                tokens.device)
+            bad_tokens = torch.from_numpy(np.array([out])).to(
+                tokens.device)
+        return tok.decode(curr_output)
+
+
+def generate(text, bad_text=None):
+    stop = StopOnTokens()
+    result = generator(text, max_new_tokens=1024, num_return_sequences=1, num_beams=1, do_sample=True,
+                       temperature=1.0, top_p=0.95, top_k=1000, stopping_criteria=StoppingCriteriaList([stop]))
+    return result[0]["generated_text"].replace(text, "")
+
+
+def user(user_message, history):
+    history = history + [[user_message, ""]]
+    return "", history, history
+
+
+def bot(history, curr_system_message):
+    messages = curr_system_message + \
+        "".join(["".join(["<|USER|>"+item[0], "<|ASSISTANT|>"+item[1]])
+                for item in history])
+    output = generate(messages)
+    history[-1][1] = output
+    time.sleep(1)
+    return history, history
+
+
+with gr.Blocks() as demo:
+    history = gr.State([])
+    gr.Markdown("## StableLM-Tuned-Alpha-7b Chat")
+    gr.HTML('''<center><a href="https://huggingface.co/spaces/stabilityai/stablelm-tuned-alpha-chat?duplicate=true"><img src="https://bit.ly/3gLdBN6" alt="Duplicate Space"></a>Duplicate the Space to skip the queue and run in a private space</center>''')
+    chatbot = gr.Chatbot().style(height=500)
+    with gr.Row():
+        with gr.Column(scale=0.70):
+            msg = gr.Textbox(label="Chat Message Box", placeholder="Chat Message Box", show_label=False).style(container=False)
+        with gr.Column(scale=0.30):
+          with gr.Row():
+              submit = gr.Button("Submit")
+              clear = gr.Button("Clear")    
+    system_msg = gr.Textbox(
+        start_message, label="System Message", interactive=False, visible=False)
+
+    msg.submit(fn=user, inputs=[msg, history], outputs=[msg, chatbot, history], queue=False).then(
+        fn=bot, inputs=[chatbot, system_msg], outputs=[chatbot, history], queue=True)
+    submit.click(fn=user, inputs=[msg, history], outputs=[msg, chatbot, history], queue=False).then(
+        fn=bot, inputs=[chatbot, system_msg], outputs=[chatbot, history], queue=True)
+    clear.click(lambda: [None, []], None, [chatbot, history], queue=False)
+demo.queue(concurrency_count=5)
+demo.launch()

requirements.txt

@@ -0,0 +1,4 @@
+gradio
+torch
+transformers
+numpy