Asuncom/Llama-3.1-8B-bnb-4bit-shiji

Uploaded model

• Developed by: Asuncom
• License: apache-2.0
• Finetuned from model : unsloth/Meta-Llama-3.1-8B-Instruct-bnb-4bit

This llama model was trained 2x faster with Unsloth and Huggingface's TRL library.

!pip install "unsloth[colab-new] @ git+https://github.com/unslothai/unsloth.git"

!pip install --upgrade pip

!pip install --no-deps "xformers<0.0.26" "trl<0.9.0" peft accelerate bitsandbytes

from unsloth import FastLanguageModel
import torch
max_seq_length = 2048 # Choose any! We auto support RoPE Scaling internally!
dtype = None # None for auto detection. Float16 for Tesla T4, V100, Bfloat16 for Ampere+
load_in_4bit = True # Use 4bit quantization to reduce memory usage. Can be False.

# 4bit pre quantized models we support for 4x faster downloading + no OOMs.
fourbit_models = [
    "unsloth/Meta-Llama-3.1-8B-bnb-4bit",      # Llama-3.1 15 trillion tokens model 2x faster!
    "unsloth/Meta-Llama-3.1-8B-Instruct-bnb-4bit",
    "unsloth/Meta-Llama-3.1-70B-bnb-4bit",
    "unsloth/Meta-Llama-3.1-405B-bnb-4bit",    # We also uploaded 4bit for 405b!
    "unsloth/Mistral-Nemo-Base-2407-bnb-4bit", # New Mistral 12b 2x faster!
    "unsloth/Mistral-Nemo-Instruct-2407-bnb-4bit",
    "unsloth/mistral-7b-v0.3-bnb-4bit",        # Mistral v3 2x faster!
    "unsloth/mistral-7b-instruct-v0.3-bnb-4bit",
    "unsloth/Phi-3-mini-4k-instruct",          # Phi-3 2x faster!d
    "unsloth/Phi-3-medium-4k-instruct",
    "unsloth/gemma-2-9b-bnb-4bit",
    "unsloth/gemma-2-27b-bnb-4bit",            # Gemma 2x faster!
] # More models at https://huggingface.co/unsloth

model, tokenizer = FastLanguageModel.from_pretrained(
    model_name = "unsloth/Meta-Llama-3.1-8B-Instruct-bnb-4bit",
    max_seq_length = max_seq_length,
    dtype = dtype,
    load_in_4bit = load_in_4bit,
    # token = "hf_...", # use one if using gated models like meta-llama/Llama-2-7b-hf
)

# ========================================================
# Test before training
# ========================================================
alpaca_prompt = """Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.

### Instruction:
{}

### Input:
{}

### Response:
{}"""
FastLanguageModel.for_inference(model) # Enable native 2x faster inference
inputs = tokenizer(
[
    alpaca_prompt.format(
        "请把现代汉语翻译成古文", # instruction
        "其品行廉正，所以至死也不放松对自己的要求。", # input
        "", # output - leave this blank for generation!
    )
], return_tensors = "pt").to("cuda")

from transformers import TextStreamer
text_streamer = TextStreamer(tokenizer)
_ = model.generate(**inputs, streamer = text_streamer, max_new_tokens = 128)

model = FastLanguageModel.get_peft_model(
    model,
    r = 16, # Choose any number > 0 ! Suggested 8, 16, 32, 64, 128
    target_modules = ["q_proj", "k_proj", "v_proj", "o_proj",
                      "gate_proj", "up_proj", "down_proj",],
    lora_alpha = 16,
    lora_dropout = 0, # Supports any, but = 0 is optimized
    bias = "none",    # Supports any, but = "none" is optimized
    # [NEW] "unsloth" uses 30% less VRAM, fits 2x larger batch sizes!
    use_gradient_checkpointing = "unsloth", # True or "unsloth" for very long context
    random_state = 3407,
    use_rslora = False,  # We support rank stabilized LoRA
    loftq_config = None, # And LoftQ
)

alpaca_prompt = """Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.

### Instruction:
{}

### Input:
{}

### Response:
{}"""

EOS_TOKEN = tokenizer.eos_token # Must add EOS_TOKEN
def formatting_prompts_func(examples):
    instructions = examples["instruction"]
    inputs       = examples["input"]
    outputs      = examples["output"]
    texts = []
    for instruction, input, output in zip(instructions, inputs, outputs):
        # Must add EOS_TOKEN, otherwise your generation will go on forever!
        text = alpaca_prompt.format(instruction, input, output) + EOS_TOKEN
        texts.append(text)
    return { "text" : texts, }
pass

from datasets import load_dataset
dataset = load_dataset("Asuncom/shiji-qishiliezhuan", split = "train")
dataset = dataset.map(formatting_prompts_func, batched = True,)

from trl import SFTTrainer
from transformers import TrainingArguments
from unsloth import is_bfloat16_supported

trainer = SFTTrainer(
    model = model,
    tokenizer = tokenizer,
    train_dataset = dataset,
    dataset_text_field = "text",
    max_seq_length = max_seq_length,
    dataset_num_proc = 2,
    packing = False, # Can make training 5x faster for short sequences.
    args = TrainingArguments(
        per_device_train_batch_size = 2,
        gradient_accumulation_steps = 4,
        warmup_steps = 5,
        # num_train_epochs = 1, # Set this for 1 full training run.
        max_steps = 100,
        learning_rate = 2e-4,
        fp16 = not is_bfloat16_supported(),
        bf16 = is_bfloat16_supported(),
        logging_steps = 1,
        optim = "adamw_8bit",
        weight_decay = 0.01,
        lr_scheduler_type = "linear",
        seed = 3407,
        output_dir = "outputs",
    ),
)

#@title Show current memory stats
gpu_stats = torch.cuda.get_device_properties(0)
start_gpu_memory = round(torch.cuda.max_memory_reserved() / 1024 / 1024 / 1024, 3)
max_memory = round(gpu_stats.total_memory / 1024 / 1024 / 1024, 3)
print(f"GPU = {gpu_stats.name}. Max memory = {max_memory} GB.")
print(f"{start_gpu_memory} GB of memory reserved.")

import wandb

# 初始化一个离线模式的W&B运行
wandb.init(mode="offline", project="asuncom", entity="asuncom")

trainer_stats = trainer.train()

#@title Show final memory and time stats
used_memory = round(torch.cuda.max_memory_reserved() / 1024 / 1024 / 1024, 3)
used_memory_for_lora = round(used_memory - start_gpu_memory, 3)
used_percentage = round(used_memory         /max_memory*100, 3)
lora_percentage = round(used_memory_for_lora/max_memory*100, 3)
print(f"{trainer_stats.metrics['train_runtime']} seconds used for training.")
print(f"{round(trainer_stats.metrics['train_runtime']/60, 2)} minutes used for training.")
print(f"Peak reserved memory = {used_memory} GB.")
print(f"Peak reserved memory for training = {used_memory_for_lora} GB.")
print(f"Peak reserved memory % of max memory = {used_percentage} %.")
print(f"Peak reserved memory for training % of max memory = {lora_percentage} %.")

# alpaca_prompt = Copied from above
FastLanguageModel.for_inference(model) # Enable native 2x faster inference
inputs = tokenizer(
[
    alpaca_prompt.format(
        "请把现代汉语翻译成古文", # instruction
        "其品行廉正，所以至死也不放松对自己的要求。", # input
        "", # output - leave this blank for generation!
    )
], return_tensors = "pt").to("cuda")

from transformers import TextStreamer
text_streamer = TextStreamer(tokenizer)
_ = model.generate(**inputs, streamer = text_streamer, max_new_tokens = 128)

model.save_pretrained("lora_model") # Local saving
tokenizer.save_pretrained("lora_model")
model.push_to_hub("Asuncom/Llama-3.1-8B-bnb-4bit-shiji", token = "hf_huggingface的密钥NeKb") # Online saving
tokenizer.push_to_hub("Asuncom/Llama-3.1-8B-bnb-4bit-shiji", token = "hf_huggingface的密钥saving

if False:
    from unsloth import FastLanguageModel
    model, tokenizer = FastLanguageModel.from_pretrained(
        model_name = "lora_model", # YOUR MODEL YOU USED FOR TRAINING
        max_seq_length = max_seq_length,
        dtype = dtype,
        load_in_4bit = load_in_4bit,
    )
    FastLanguageModel.for_inference(model) # Enable native 2x faster inference

# alpaca_prompt = You MUST copy from above!

inputs = tokenizer(
[
    alpaca_prompt.format(
        "What is a famous tall tower in Paris?", # instruction
        "", # input
        "", # output - leave this blank for generation!
    )
], return_tensors = "pt").to("cuda")

from transformers import TextStreamer
text_streamer = TextStreamer(tokenizer)
_ = model.generate(**inputs, streamer = text_streamer, max_new_tokens = 128)

# Merge to 16bit
if False: model.save_pretrained_merged("model", tokenizer, save_method = "merged_16bit",)
if False: model.push_to_hub_merged("Asuncom/Llama-3.1-8B-bnb-4bit-shiji", tokenizer, save_method = "merged_16bit", token = "hf_huggingface的密钥NeKb")

# Merge to 4bit
if False: model.save_pretrained_merged("model", tokenizer, save_method = "merged_4bit",)
if False: model.push_to_hub_merged("Asuncom/Llama-3.1-8B-bnb-4bit-shiji", tokenizer, save_method = "merged_4bit", token = "hf_huggingface的密钥oRA adapters
if False: model.save_pretrained_merged("model", tokenizer, save_method = "lora",)
if False: model.push_to_hub_merged("Asuncom/Llama-3.1-8B-bnb-4bit-shiji", tokenizer, save_method = "lora", token = "hf_huggingface的密钥

# Save to 8bit Q8_0
if False: model.save_pretrained_gguf("model", tokenizer,)
# Remember to go to https://huggingface.co/settings/tokens for a token!
# And change hf to your username!
if False: model.push_to_hub_gguf("Asuncom/Llama-3.1-8B-bnb-4bit-shiji", tokenizer, token = "")

# Save to 16bit GGUF
if False: model.save_pretrained_gguf("model", tokenizer, quantization_method = "f16")
if False: model.push_to_hub_gguf("Asuncom/Llama-3.1-8B-bnb-4bit-shiji", tokenizer, quantization_method = "f16", token = "")

# Save to q4_k_m GGUF
if False: model.save_pretrained_gguf("model", tokenizer, quantization_method = "q4_k_m")
if True: model.push_to_hub_gguf("Asuncom/Llama-3.1-8B-bnb-4bit-shiji", tokenizer, quantization_method = "q4_k_m", token = "hf_xxxxx")

# Save to multiple GGUF options - much faster if you want multiple!
if False:
    model.push_to_hub_gguf(
        "Asuncom/Llama-3.1-8B-bnb-4bit-shiji", # Change hf to your username!
        tokenizer,
        quantization_method = ["q4_k_m", "q8_0", "q5_k_m",],
        token = "hf_huggingface的密钥NeKb", # Get a token at https://huggingface.co/settings/tokens
    )

model.push_to_hub_gguf(
        "Asuncom/Llama-3.1-8B-bnb-4bit-shiji", # Change hf to your username!
        tokenizer,
        quantization_method = ["q4_k_m", "q8_0", "q5_k_m",],
        token = "hf_huggingface的密钥NeKb", # Get a token at https://huggingface.co/settings/tokens
    )

[ 279/ 292]            blk.30.attn_output.weight - [ 4096,  4096,     1,     1], type =    f16, converting to q5_K .. size =    32.00 MiB ->    11.00 MiB
[ 280/ 292]                 blk.30.attn_q.weight - [ 4096,  4096,     1,     1], type =    f16, converting to q5_K .. size =    32.00 MiB ->    11.00 MiB
[ 281/ 292]                 blk.30.attn_v.weight - [ 4096,  1024,     1,     1], type =    f16, converting to q6_K .. size =     8.00 MiB ->     3.28 MiB
[ 282/ 292]               blk.31.ffn_gate.weight - [ 4096, 14336,     1,     1], type =    f16, converting to q5_K .. size =   112.00 MiB ->    38.50 MiB
[ 283/ 292]                 blk.31.ffn_up.weight - [ 4096, 14336,     1,     1], type =    f16, converting to q5_K .. size =   112.00 MiB ->    38.50 MiB
[ 284/ 292]                 blk.31.attn_k.weight - [ 4096,  1024,     1,     1], type =    f16, converting to q5_K .. size =     8.00 MiB ->     2.75 MiB
[ 285/ 292]            blk.31.attn_output.weight - [ 4096,  4096,     1,     1], type =    f16, converting to q5_K .. size =    32.00 MiB ->    11.00 MiB
[ 286/ 292]                 blk.31.attn_q.weight - [ 4096,  4096,     1,     1], type =    f16, converting to q5_K .. size =    32.00 MiB ->    11.00 MiB
[ 287/ 292]                 blk.31.attn_v.weight - [ 4096,  1024,     1,     1], type =    f16, converting to q6_K .. size =     8.00 MiB ->     3.28 MiB
[ 288/ 292]                        output.weight - [ 4096, 128256,     1,     1], type =    f16, converting to q6_K .. size =  1002.00 MiB ->   410.98 MiB
[ 289/ 292]              blk.31.attn_norm.weight - [ 4096,     1,     1,     1], type =    f32, size =    0.016 MB
[ 290/ 292]               blk.31.ffn_down.weight - [14336,  4096,     1,     1], type =    f16, converting to q6_K .. size =   112.00 MiB ->    45.94 MiB
[ 291/ 292]               blk.31.ffn_norm.weight - [ 4096,     1,     1,     1], type =    f32, size =    0.016 MB
[ 292/ 292]                   output_norm.weight - [ 4096,     1,     1,     1], type =    f32, size =    0.016 MB
llama_model_quantize_internal: model size  = 15317.02 MB
llama_model_quantize_internal: quant size  =  5459.93 MB

main: quantize time = 147401.53 ms
main:    total time = 147401.53 ms
Unsloth: Conversion completed! Output location: ./Asuncom/Llama-3.1-8B-bnb-4bit-shiji/unsloth.Q5_K_M.gguf
Unsloth: Uploading GGUF to Huggingface Hub...


unsloth.F16.gguf: 100%|██████████| 16.1G/16.1G [26:20<00:00, 10.2MB/s]   


Saved GGUF to https://huggingface.co/Asuncom/Llama-3.1-8B-bnb-4bit-shiji
Unsloth: Uploading GGUF to Huggingface Hub...


unsloth.Q4_K_M.gguf: 100%|██████████| 4.92G/4.92G [08:05<00:00, 10.1MB/s]


Saved GGUF to https://huggingface.co/Asuncom/Llama-3.1-8B-bnb-4bit-shiji
Unsloth: Uploading GGUF to Huggingface Hub...


unsloth.Q8_0.gguf: 100%|██████████| 8.54G/8.54G [13:48<00:00, 10.3MB/s]


Saved GGUF to https://huggingface.co/Asuncom/Llama-3.1-8B-bnb-4bit-shiji
Unsloth: Uploading GGUF to Huggingface Hub...


unsloth.Q5_K_M.gguf: 100%|██████████| 5.73G/5.73G [09:24<00:00, 10.2MB/s] 


Saved GGUF to https://huggingface.co/Asuncom/Llama-3.1-8B-bnb-4bit-shipython