Upload folder using huggingface_hub

Dockerfile

@@ -1,12 +1,39 @@
-# Use the base image from your Docker container
-FROM trcoot/cpu-casuallm:latest
+# Use the full Python 3.9 image (if you need specific modules)
+FROM python:3.9.19
+
+RUN useradd -m -u 1000 user
+USER user
+ENV PATH="/home/user/.local/bin:$PATH"
+
+# Working Directory
+WORKDIR /app
+
+COPY --chown=user ./models/ models/
+COPY --chown=user ./app.py app.py
+
+RUN pip install --no-cache-dir torch==2.2.2
+RUN pip install --no-cache-dir packaging
+
+# Copy Dependencies (if you have any)
+COPY --chown=user ./requirements.txt requirements.txt
+
+# Install Dependencies (if you have any)
+RUN pip install --no-cache-dir -r requirements.txt
+RUN pip install -U git+https://github.com/sustcsonglin/flash-linear-attention
+
+# Copy Custom Modules (Adjust paths if needed)
+COPY --chown=user  ./causal-conv1d/ causal-conv1d/
+RUN cd /app/causal-conv1d && python setup.py install --user
+
+COPY --chown=user ./mamba/ mamba/
+RUN cd /app/mamba && python setup.py install --user
 
 # Set the environment variable for the Hugging Face cache directory
 ENV HF_HOME=/app/.cache 
 
 # Create the cache directory and give the appropriate permissions
 RUN mkdir -p /app/.cache && chmod 777 /app/.cache
-RUN mkdir -p /app/gradio_flagged && chmod 777 /app/gradio_flagged
 
-# Start the application (modify this based on your app's entry point)
-CMD ["python", "app.py"] 
+# Print Messages
+# CMD ["bash"]
+CMD ["python", "app.py"]

README.md

@@ -1,11 +1,6 @@
 ---
-title: Cpu Casuallm
-emoji: 🌍
-colorFrom: pink
-colorTo: purple
-sdk: docker
-pinned: false
-license: apache-2.0
+title: cpu-casuallm
+app_file: app.py
+sdk: gradio
+sdk_version: 4.42.0
 ---
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -2,6 +2,7 @@ import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer
 import time
 import gradio as gr
+from gradio import deploy
 
 def generate_prompt(instruction, input=""):
     instruction = instruction.strip().replace('\r\n','\n').replace('\n\n','\n')
@@ -47,7 +48,7 @@ def generate_text(input_text):
     input_ids = tokenizer(prompt, return_tensors="pt").input_ids
 
     generated_text = ""
-    stop_sequence_found = False
+
     for i in range(333):
         output = model.generate(input_ids, max_new_tokens=1, do_sample=True, temperature=1.0, top_p=0.3, top_k=0)
         new_word = tokenizer.decode(output[0][-1:], skip_special_tokens=True)
@@ -55,15 +56,9 @@ def generate_text(input_text):
         print(new_word, end="", flush=True)
         generated_text += new_word
 
-        if new_word == '\n' or new_word == '.':
-            stop_sequence_found = True
-            break
 
         input_ids = output
 
-    if stop_sequence_found:
-        print("\n(Stop sequence found)")
-    print()
     return generated_text
 
 # Create the Gradio interface
@@ -73,11 +68,13 @@ iface = gr.Interface(
     outputs="text",
     title="RWKV Chatbot",
     description="Enter your prompt below:",
-    flagging_dir="/app/gradio_flagged" 
+    # flagging_callback=None  
+    flagging_dir="gradio_flagged/" 
 )
 
 # For local testing:
-# iface.launch()
+# iface.launch(share=True)
+deploy()
 
 
 # Hugging Face Spaces will automatically launch the interface.

app.py_09_23_24

@@ -0,0 +1,62 @@
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer
+import time
+
+def generate_prompt(instruction, input=""):
+    instruction = instruction.strip().replace('\r\n','\n').replace('\n\n','\n')
+    input = input.strip().replace('\r\n','\n').replace('\n\n','\n')
+    if input:
+        return f"""Instruction: {instruction}
+
+Input: {input}
+
+Response:"""
+    else:
+        return f"""User: hi
+
+Lover: Hi. I am your assistant and I will provide expert full response in full details. Please feel free to ask any question and I will always answer it.
+
+User: {instruction}
+
+Lover:"""
+
+model_path = "models/rwkv-6-world-1b6/" # Path to your local model directory
+
+model = AutoModelForCausalLM.from_pretrained(
+    model_path, 
+    trust_remote_code=True, 
+    use_flash_attention_2=False  # Explicitly disable Flash Attention
+).to(torch.float32)
+
+
+tokenizer = AutoTokenizer.from_pretrained(
+    model_path, 
+    bos_token="</s>",
+    eos_token="</ s>",
+    unk_token="<unk>",
+    pad_token="<pad>",
+    trust_remote_code=True, 
+    padding_side='left', 
+    clean_up_tokenization_spaces=False  # Or set to True if you prefer
+)
+
+print(tokenizer.special_tokens_map)
+
+text = "Hi"
+
+prompt = generate_prompt(text)
+
+input_ids = tokenizer(prompt, return_tensors="pt").input_ids
+
+# Generate text word by word with stop sequence
+generated_text = ""
+for i in range(333):  # Generate up to 333 tokens
+    output = model.generate(input_ids, max_new_tokens=1, do_sample=True, temperature=1.0, top_p=0.3, top_k=0)
+    new_word = tokenizer.decode(output[0][-1:], skip_special_tokens=True)
+
+    print(new_word, end="", flush=True)  # Print word-by-word
+    generated_text += new_word
+
+    input_ids = output  # Update input_ids for next iteration
+
+print()  # Add a newline at the end