Upload 2 files

alpacatest.py

@@ -0,0 +1,63 @@
+from datasets import load_dataset
+from transformers import RwkvForCausalLM, GPTNeoXTokenizerFast,GPT2Config,pipeline,GenerationConfig
+import torch
+import numpy as np
+import gradio as gr
+
+if torch.cuda.is_available():
+    device = "cuda"
+else:
+    device = "cpu"
+
+model = RwkvForCausalLM.from_pretrained("rwkv-alpaca",device_map='auto') #仅500MB,自训练，使用alpaca
+
+tokenizer = GPTNeoXTokenizerFast.from_pretrained("rwkv-alpaca", add_special_tokens=True)
+
+
+
+#rwkv with alpaca 
+def generate_prompt(instruction, input=None):
+    
+    return f"""Below is an instruction that describes a task. Write a response that appropriately completes the request.
+
+### Instruction:
+{instruction}
+
+### Response:"""
+
+def evaluate(
+    instruction,
+    temperature=0.1,
+    top_p=0.75,
+    top_k=40,
+    max_new_tokens=128,
+):
+    prompt = generate_prompt(instruction)
+    input_ids = tokenizer.encode(prompt, return_tensors='pt')
+    out = model.generate(input_ids=input_ids,temperature=temperature,top_p=top_p,top_k=top_k,max_new_tokens=max_new_tokens)
+    answer = tokenizer.decode(out[0])
+    return answer.split("### Response:")[1].strip()
+
+
+gr.Interface(
+    fn=evaluate,#接口函数
+    inputs=[
+        gr.components.Textbox(
+            lines=2, label="Instruction", placeholder="Tell me about alpacas."
+        ),
+        gr.components.Slider(minimum=0, maximum=1, value=0.1, label="Temperature"),
+        gr.components.Slider(minimum=0, maximum=1, value=0.75, label="Top p"),
+        gr.components.Slider(minimum=0, maximum=100, step=1, value=40, label="Top k"),
+        gr.components.Slider(
+            minimum=1, maximum=2000, step=1, value=128, label="Max tokens"
+        ),
+    ],
+    outputs=[
+        gr.inputs.Textbox(
+            lines=5,
+            label="Output",
+        )
+    ],
+    title="RWKV-Alpaca",
+    description="RWKV,easy in HF.",
+).launch()

alpacatrain.py

@@ -0,0 +1,59 @@
+from datasets import load_dataset
+from transformers import RwkvForCausalLM, GPTNeoXTokenizerFast, Trainer, TrainingArguments,DataCollatorForLanguageModeling
+
+MICRO_BATCH_SIZE = 8
+BATCH_SIZE = 128
+GRADIENT_ACCUMULATION_STEPS = BATCH_SIZE // MICRO_BATCH_SIZE
+EPOCHS = 100
+LEARNING_RATE = 2e-5
+CUTOFF_LEN = 256
+
+model = RwkvForCausalLM.from_pretrained("rwkv-430M-pile")
+tokenizer = GPTNeoXTokenizerFast.from_pretrained("rwkv-430M-pile", add_special_tokens=True)
+# model = RwkvForCausalLM.from_pretrained("rwkv-7b-pile")
+# tokenizer = GPTNeoXTokenizerFast.from_pretrained("rwkv-7b-pile", add_special_tokens=True)
+tokenizer.pad_token = tokenizer.eos_token
+tokenizer.pad_token_id = tokenizer.eos_token_id
+
+data = load_dataset("json", data_files="test.json")
+
+def generate_prompt(data_point):
+    
+    return f"""Below is an instruction that describes a task. Write a response that appropriately completes the request.
+
+### Instruction:
+{data_point["instruction"]}
+
+### Response:
+{data_point["output"]}"""
+
+
+data = data.shuffle().map(
+    lambda data_point: tokenizer(
+        generate_prompt(data_point),
+        truncation=True,
+        max_length=CUTOFF_LEN,
+        padding="max_length",
+    )
+)
+
+trainer = Trainer(
+    model=model,
+    train_dataset=data["train"],
+    args=TrainingArguments(
+        per_device_train_batch_size=MICRO_BATCH_SIZE,
+        gradient_accumulation_steps=GRADIENT_ACCUMULATION_STEPS,
+        warmup_steps=100,
+        num_train_epochs=EPOCHS,
+        learning_rate=LEARNING_RATE,
+        fp16=True,
+        logging_steps=1,
+        output_dir="rwkv-alpaca",
+        save_total_limit=3,
+    ),
+    data_collator=DataCollatorForLanguageModeling(tokenizer, mlm=False),
+)
+model.config.use_cache = False
+trainer.train(resume_from_checkpoint=False)
+
+model.save_pretrained("rwkv-alpaca")