Phi-3/compile/export_onnx.py

#!/usr/bin/env python3
# ==============================================================================
#
# Copyright (C) 2023 Sophgo Technologies Inc. All rights reserved.
#
# TPU-MLIR is licensed under the 2-Clause BSD License except for the
# third-party components.
#
# ==============================================================================

import os
import torch
import argparse
from tqdm import tqdm
from transformers import AutoTokenizer, AutoModelForCausalLM
torch.set_grad_enabled(False)

parser = argparse.ArgumentParser(description='export onnx')
parser.add_argument('--model_path', type=str, default ="./Phi-3-mini/" ,help='path to the torch model')
parser.add_argument('--seq_length', type=int, default=512, help="sequence length")

args = parser.parse_args()

model_path = args.model_path
folder = f"./tmp/onnx"

origin_model = AutoModelForCausalLM.from_pretrained(
 model_path, trust_remote_code=True).eval()

for param in origin_model.parameters():
 param.requires_grad = False

config = origin_model.config
transformer = origin_model.model
layers = transformer.layers
SEQ_LENGTH = args.seq_length
NUM_LAYERS = config.num_hidden_layers
HIDDEN_SIZE = config.hidden_size
NUM_ATTENTION_HEADS = config.num_attention_heads
HEAD_DIM = HIDDEN_SIZE // NUM_ATTENTION_HEADS
VOCAB_SIZE = config.vocab_size
print(f'Layers: {NUM_LAYERS}\nHidden size: {HIDDEN_SIZE}\n')

tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)

class Embedding(torch.nn.Module):

 def __init__(self):
 super().__init__()

 def forward(self, input_ids):
 hidden_states = transformer.embed_tokens(input_ids)
 return hidden_states


class Block(torch.nn.Module):

 def __init__(self, layer_id):
 super().__init__()
 self.layer_id = layer_id
 self.layer = layers[layer_id]

 def forward(self, hidden_states, position_ids, attention_mask):
 # breakpoint()
 hidden_states, past_kv = self.layer(hidden_states,
 attention_mask,
 position_ids,
 use_cache=True)
 present_k, present_v = past_kv
 return hidden_states, present_k, present_v


class BlockCache(torch.nn.Module):

 def __init__(self, layer_id):
 super().__init__()
 self.layer_id = layer_id
 self.layer = layers[layer_id]

 def forward(self, hidden_states, position_ids, attention_mask, past_k,
 past_v):
 # breakpoint()
 hidden_states, past_kv = self.layer(hidden_states,
 attention_mask,
 position_ids=position_ids,
 past_key_value=(past_k, past_v),
 use_cache=True)
 present_k, present_v = past_kv
 return hidden_states, present_k, present_v


class LmHead(torch.nn.Module):

 def __init__(self):
 super().__init__()

 def forward(self, hidden_states):
 hidden_states = transformer.norm(hidden_states)
 m_logits = origin_model.lm_head(hidden_states)
 return m_logits


class GreedyHead(torch.nn.Module):

 def __init__(self):
 super().__init__()

 def forward(self, m_logits):
 _, token = torch.topk(m_logits.float(), 1)
 return token

 
# refs:https://github.com/huggingface/transformers/blob/main/src/transformers/generation/logits_process.py
class PenaltySampleHead(torch.nn.Module):

 def __init__(self, top_k = 50, min_tokens_to_keep = 5):
 super().__init__()
 self.top_k = top_k
 self.min_tokens_to_keep = min_tokens_to_keep
 self.keep_matrix = torch.zeros((1, self.top_k), dtype=torch.bool)
 self.keep_matrix[0, :self.min_tokens_to_keep] = True

 def forward(self, m_logits, input_ids, top_p, temperature, penalty):
 # repeat penalty
 logits = torch.gather(m_logits, 1, input_ids)
 logits = torch.where(logits < 0, logits * penalty, logits / penalty)
 m_logits.scatter_(1, input_ids, logits)

 # top_k
 logits, token = torch.topk(m_logits.float(), self.top_k)

 # temperature
 logits = logits / temperature

 # top_p
 cumulative_probs = logits.softmax(dim=1).cumsum(dim=1)
 mask = cumulative_probs < top_p
 mask = mask + self.keep_matrix
 filtered_logits = torch.where(mask, logits, torch.FloatTensor([-1000.]))
 probs = filtered_logits.softmax(dim=1)
 return probs, token

def convert_block(layer_id):
 model = Block(layer_id)
 hidden_states = torch.randn((1, SEQ_LENGTH, HIDDEN_SIZE))
 position_ids = torch.tensor([range(SEQ_LENGTH)], dtype=torch.long)
 attention_mask = -1000 * torch.ones((1, 1, SEQ_LENGTH, SEQ_LENGTH), dtype=torch.float32).triu(diagonal=1)
 torch.onnx.export(
 model, (hidden_states, position_ids, attention_mask),
 f'{folder}/block_{layer_id}.onnx',
 verbose=False,
 input_names=['input_states', 'position_ids', 'attention_mask'],
 output_names=['hidden_states', 'past_k', 'past_v'],
 do_constant_folding=True,
 opset_version=15)


def convert_block_cache(layer_id):
 model = BlockCache(layer_id)
 hidden_states = torch.randn((1, 1, HIDDEN_SIZE))
 position_ids = torch.tensor([range(1)], dtype=torch.long)
 attention_mask = -1000 * torch.ones((1, 1, 1, SEQ_LENGTH + 1), dtype=torch.float32).triu(diagonal=1)
 past_k = torch.randn((1, SEQ_LENGTH, config.num_key_value_heads, HEAD_DIM))
 past_v = torch.randn((1, SEQ_LENGTH, config.num_key_value_heads, HEAD_DIM))
 results = model(hidden_states, position_ids, attention_mask, past_k, past_v)
 
 torch.onnx.export(
 model, (hidden_states, position_ids, attention_mask, past_k, past_v),
 f'{folder}/block_cache_{layer_id}.onnx',
 verbose=False,
 input_names=[
 'input_states', 'position_ids', 'attention_mask', 'history_k',
 'history_v'
 ],
 output_names=['hidden_states', 'past_k', 'past_v'],
 do_constant_folding=True,
 opset_version=15)


def convert_embedding():
 model = Embedding()
 input_ids = torch.tensor([range(SEQ_LENGTH)])

 torch.onnx.export(model, (input_ids),
 f'{folder}/embedding.onnx',
 verbose=False,
 input_names=['input_ids'],
 output_names=['input_embed'],
 do_constant_folding=True,
 opset_version=15)


def convert_lm_head():
 model = LmHead()
 input = torch.randn(1, HIDDEN_SIZE)

 torch.onnx.export(model, (input),
 f'{folder}/lm_head.onnx',
 verbose=False,
 input_names=['hidden_states'],
 output_names=['m_logits'],
 do_constant_folding=True,
 opset_version=15)

def convert_greedy_head(): 
 model = GreedyHead()
 m_logits = torch.randn(1, VOCAB_SIZE)

 torch.onnx.export(
 model, (m_logits),
 f'{folder}/greedy_head.onnx',
 verbose=False,
 input_names=['m_logits'],
 output_names=['token'],
 do_constant_folding=True,
 opset_version=15)


def convert_penalty_sample_head(): 
 model = PenaltySampleHead()
 m_logits = torch.randn(1, VOCAB_SIZE)
 input_ids = torch.tensor([range(SEQ_LENGTH)])
 top_p = torch.tensor([0.8])
 temperature = torch.tensor([0.98])
 penalty = torch.tensor([0.98])

 torch.onnx.export(
 model, (m_logits, input_ids, top_p, temperature, penalty),
 f'{folder}/penalty_sample_head.onnx',
 verbose=False,
 input_names=[
 'm_logits', 'input_ids', 'top_p', 'temperature',
 'penalty'
 ],
 output_names=['probs', 'token'],
 do_constant_folding=True,
 opset_version=15)

# create folder to store onnx
if not os.path.exists(folder):
 os.makedirs(folder)

# export models
print(f'Convert block & block_cache')
for i in tqdm(range(NUM_LAYERS)):
 convert_block_cache(i)
 convert_block(i)

print(f'Convert embedding')
convert_embedding()

print(f'Convert lm_head')
convert_lm_head()
convert_greedy_head()
convert_penalty_sample_head()
print("Done")

# def test_net_with_mask():
# import numpy as np
# num_layers = NUM_LAYERS
# MAX_LEN = SEQ_LENGTH
# embed = Embedding()
# blocks = [Block(i) for i in range(num_layers)]
# block_kvs = [BlockCache(i) for i in range(num_layers)]
# ids = tokenizer.encode('你好')
# query = '你好'
# print(query)
# # promt = tokenizer.build_prompt(query)
# # ids = tokenizer.encode(promt)
# ids = [ 1, 32006, 887, 526, 263, 8444, 13436, 20255, 29889, 3529,
# 3867, 9109, 29892, 11314, 936, 322, 16232, 2472, 304, 278,
# 1404, 29889, 32007, 32010, 11644, 526, 366, 29973, 32007, 32001]
# MAX_LEN = len(ids) + 20
# print("input ids:{}".format(ids))
# token_len = len(ids)
# ids = ids + (MAX_LEN - token_len) * [0]
# input_ids = torch.tensor(ids).view(MAX_LEN)

# out = embed(input_ids).view(1, MAX_LEN, HIDDEN_SIZE)
# # breakpoint()
# position_ids = list(range(token_len)) + (MAX_LEN - token_len) * [0]
# position_ids = torch.tensor([position_ids])
# attention_mask = torch.ones((MAX_LEN, MAX_LEN)) * -10000.0
# for i in range(token_len):
# attention_mask[i,:i+1] = 0
# attention_mask = attention_mask.view((1,1,MAX_LEN,MAX_LEN))
# k_cache = []
# v_cache = []

# for i in tqdm(range(num_layers)):
# out, k, v = blocks[i](out, position_ids, attention_mask)
# k[:,token_len:,:,:] = 0
# v[:,token_len:,:,:] = 0
# k_cache.append(k)
# v_cache.append(v)
# out = out[0, token_len - 1:token_len, :].view(1, HIDDEN_SIZE)
# lm = LmHead()
# m_logits = lm(out)
# greedy_head = GreedyHead()
# token = greedy_head(m_logits)
# out_ids = [int(token)]
# word = tokenizer._convert_id_to_token(int(token[0]))
# print(word, end="")
# for i in tqdm(range(10)):
# token_len += 1
# input_ids = torch.tensor([token])
# out = embed(input_ids).view(1, 1, HIDDEN_SIZE)
# position_ids = torch.tensor([[token_len - 1]])
# attention_mask = torch.ones((1, 1, 1, MAX_LEN + 1)) * -10000.0
# attention_mask[:, :, :, :token_len] = 0
# attention_mask[:, :, :, -1] = 0

# for i in range(num_layers):
# out, k_cache_present, v_cache_present = block_kvs[i](out, position_ids,
# attention_mask,
# k_cache[i], v_cache[i])
# k_cache[i][:,token_len-1:token_len,:,:] = k_cache_present
# v_cache[i][:,token_len-1:token_len,:,:] = v_cache_present
# m_logits = lm(out)
# token = greedy_head(m_logits)
# out_ids.append(int(token))
# word = tokenizer._convert_id_to_token(int(token[0]))
# print(word, end="")
# print("\noutput_ids:{}".format(out_ids))

# test_net_with_mask()

#My: [306, 626, 385, 319, 29902, 13436, 20255, 8688, 304, 3867, 8444]
# [306, 626, 385, 319, 29902, 13436, 20255, 8688, 304, 3867, 8444, 29892, 9109, 29892, 322, 16232, 2472, 304, 4160, 29889, 1619, 6437, 338, 304, 6985, 366, 411, 263, 9377, 3464]