VoucherVision / vouchervision /LLM_local_custom_fine_tune.py
phyloforfun's picture
July 18 update
c5e57d6
raw
history blame
15 kB
import os, re, json, yaml, torch
from peft import AutoPeftModelForCausalLM
from transformers import AutoTokenizer
import json, torch, transformers, gc
from transformers import BitsAndBytesConfig
from langchain.output_parsers.retry import RetryOutputParser
from langchain.prompts import PromptTemplate
from langchain_core.output_parsers import JsonOutputParser
from huggingface_hub import hf_hub_download
from langchain_community.llms.huggingface_pipeline import HuggingFacePipeline
from vouchervision.utils_LLM import SystemLoadMonitor, run_tools, count_tokens, save_individual_prompt, sanitize_prompt
from vouchervision.utils_LLM_JSON_validation import validate_and_align_JSON_keys_with_template
# MODEL_NAME = "unsloth/mistral-7b-instruct-v0.2-bnb-4bit"
# sltp_version = 'HLT_MICH_Angiospermae_SLTPvA_v1-0_medium__OCR-C25-L25-E50-R05'
# LORA = "phyloforfun/mistral-7b-instruct-v2-bnb-4bit__HLT_MICH_Angiospermae_SLTPvC_v1-0_medium_OCR-C25-L25-E50-R05"
TEXT = "HERBARIUM OF MARCUS W. LYON , JR . Tracaulon sagittatum Indiana : Porter Co. Mincral Springs edge wet subdural woods 1927 TX 11 Flowers pink UNIVERSIT HERBARIUM MICHIGAN MICH University of Michigan Herbarium 1439649 copyright reserved PERSICARIA FEB 26 1965 cm "
PARENT_MODEL = "unsloth/mistral-7b-instruct-v0.2-bnb-4bit"
class LocalFineTuneHandler:
RETRY_DELAY = 2 # Wait 2 seconds before retrying
MAX_RETRIES = 5 # Maximum number of retries
STARTING_TEMP = 0.001
TOKENIZER_NAME = None
VENDOR = 'mistral'
MAX_GPU_MONITORING_INTERVAL = 2 # seconds
def __init__(self, cfg, logger, model_name, JSON_dict_structure, config_vals_for_permutation=None):
# self.model_id = f"phyloforfun/{self.model_name}"
# model_name = LORA #######################################################
# self.JSON_dict_structure = JSON_dict_structure
# self.JSON_dict_structure_str = json.dumps(self.JSON_dict_structure, sort_keys=False, indent=4)
self.JSON_dict_structure_str = """{"catalogNumber": "", "scientificName": "", "genus": "", "specificEpithet": "", "scientificNameAuthorship": "", "collector": "", "recordNumber": "", "identifiedBy": "", "verbatimCollectionDate": "", "collectionDate": "", "occurrenceRemarks": "", "habitat": "", "locality": "", "country": "", "stateProvince": "", "county": "", "municipality": "", "verbatimCoordinates": "", "decimalLatitude": "", "decimalLongitude": "", "minimumElevationInMeters": "", "maximumElevationInMeters": ""}"""
self.cfg = cfg
self.print_output = True
self.tool_WFO = self.cfg['leafmachine']['project']['tool_WFO']
self.tool_GEO = self.cfg['leafmachine']['project']['tool_GEO']
self.tool_wikipedia = self.cfg['leafmachine']['project']['tool_wikipedia']
self.logger = logger
self.has_GPU = torch.cuda.is_available()
if self.has_GPU:
self.device = "cuda"
else:
self.device = "cpu"
self.monitor = SystemLoadMonitor(logger)
self.model_name = model_name.split("/")[1]
self.model_id = model_name
# self.model_path = hf_hub_download(repo_id=self.model_id, repo_type="model",filename="config.json")
self.starting_temp = float(self.STARTING_TEMP)
self.temp_increment = float(0.2)
self.adjust_temp = self.starting_temp
self.load_in_4bit = False
self.parser = JsonOutputParser()
self._load_model()
self._create_prompt()
self._set_config()
self._build_model_chain_parser()
def _set_config(self):
# self._clear_VRAM()
self.config = {'max_new_tokens': 1024,
'temperature': self.starting_temp,
'seed': 2023,
'top_p': 1,
# 'top_k': 1,
# 'top_k': 40,
'do_sample': False,
'n_ctx':4096,
# Activate 4-bit precision base model loading
# 'use_4bit': True,
# # Compute dtype for 4-bit base models
# 'bnb_4bit_compute_dtype': "float16",
# # Quantization type (fp4 or nf4)
# 'bnb_4bit_quant_type': "nf4",
# # Activate nested quantization for 4-bit base models (double quantization)
# 'use_nested_quant': False,
}
def _adjust_config(self):
new_temp = self.adjust_temp + self.temp_increment
if self.json_report:
self.json_report.set_text(text_main=f'Incrementing temperature from {self.adjust_temp} to {new_temp}')
self.logger.info(f'Incrementing temperature from {self.adjust_temp} to {new_temp}')
self.adjust_temp += self.temp_increment
def _reset_config(self):
if self.json_report:
self.json_report.set_text(text_main=f'Resetting temperature from {self.adjust_temp} to {self.starting_temp}')
self.logger.info(f'Resetting temperature from {self.adjust_temp} to {self.starting_temp}')
self.adjust_temp = self.starting_temp
def _load_model(self):
self.model = AutoPeftModelForCausalLM.from_pretrained(
pretrained_model_name_or_path=self.model_id, # YOUR MODEL YOU USED FOR TRAINING
load_in_4bit = self.load_in_4bit,
low_cpu_mem_usage=True,
).to(self.device)
self.tokenizer = AutoTokenizer.from_pretrained(PARENT_MODEL)
self.eos_token_id = self.tokenizer.eos_token_id
# def _build_model_chain_parser(self):
# self.local_model_pipeline = transformers.pipeline("text-generation",
# model=self.model_id,
# max_new_tokens=self.config.get('max_new_tokens'),
# # top_k=self.config.get('top_k'),
# top_p=self.config.get('top_p'),
# do_sample=self.config.get('do_sample'),
# model_kwargs={"load_in_4bit": self.load_in_4bit})
# self.local_model = HuggingFacePipeline(pipeline=self.local_model_pipeline)
# # Set up the retry parser with the runnable
# # self.retry_parser = RetryWithErrorOutputParser.from_llm(parser=self.parser, llm=self.local_model, max_retries=self.MAX_RETRIES)
# self.retry_parser = RetryOutputParser(parser=self.parser, llm=self.local_model, max_retries=self.MAX_RETRIES)
# # Create an llm chain with LLM and prompt
# self.chain = self.prompt | self.local_model # LCEL
def _build_model_chain_parser(self):
self.local_model_pipeline = transformers.pipeline(
"text-generation",
model=self.model_id,
max_new_tokens=self.config.get('max_new_tokens'),
top_k=self.config.get('top_k', None),
top_p=self.config.get('top_p'),
do_sample=self.config.get('do_sample'),
model_kwargs={"load_in_4bit": self.load_in_4bit},
)
self.local_model = HuggingFacePipeline(pipeline=self.local_model_pipeline)
self.retry_parser = RetryOutputParser(parser=self.parser, llm=self.local_model, max_retries=self.MAX_RETRIES)
def _create_prompt(self):
self.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:
{}"""
self.template = """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:
{}""".format("{instructions}", "{OCR_text}", "{empty}")
self.instructions_text = """Refactor the unstructured text into a valid JSON dictionary. The key names follow the Darwin Core Archive Standard. If a key lacks content, then insert an empty string. Fill in the following JSON structure as required: """
self.instructions_json = self.JSON_dict_structure_str.replace("\n ", " ").strip().replace("\n", " ")
self.instructions = ''.join([self.instructions_text, self.instructions_json])
# Create a prompt from the template so we can use it with Langchain
self.prompt = PromptTemplate(template=self.template, input_variables=["instructions", "OCR_text", "empty"])
# Set up a parser
self.parser = JsonOutputParser()
def extract_json(self, response_text):
# Assuming the response is a list with a single string entry
# response_text = response[0]
response_pattern = re.compile(r'### Response:(.*)', re.DOTALL)
response_match = response_pattern.search(response_text)
if not response_match:
raise ValueError("No '### Response:' section found in the provided text")
response_text = response_match.group(1)
# Use a regular expression to find JSON objects in the response text
json_objects = re.findall(r'\{.*?\}', response_text, re.DOTALL)
if json_objects:
# Assuming you want the first JSON object if there are multiple
json_str = json_objects[0]
# Convert the JSON string to a Python dictionary
json_dict = json.loads(json_str)
return json_str, json_dict
else:
raise ValueError("No JSON object found in the '### Response:' section")
def call_llm_local_custom_fine_tune(self, OCR_text, json_report, paths):
_____, ____, _, __, ___, json_file_path_wiki, txt_file_path_ind_prompt = paths
self.json_report = json_report
if self.json_report:
self.json_report.set_text(text_main=f'Sending request to {self.model_name}')
self.monitor.start_monitoring_usage()
nt_in = 0
nt_out = 0
self.inputs = self.tokenizer(
[
self.alpaca_prompt.format(
self.instructions, # instruction
OCR_text, # input
"", # output - leave this blank for generation!
)
], return_tensors = "pt").to(self.device)
ind = 0
while ind < self.MAX_RETRIES:
ind += 1
try:
# Fancy
# Dynamically set the temperature for this specific request
model_kwargs = {"temperature": self.adjust_temp}
# Invoke the chain to generate prompt text
# results = self.chain.invoke({"instructions": self.instructions, "OCR_text": OCR_text, "empty": "", "model_kwargs": model_kwargs})
# Use retry_parser to parse the response with retry logic
# output = self.retry_parser.parse_with_prompt(results, prompt_value=OCR_text)
results = self.local_model.invoke(OCR_text)
output = self.retry_parser.parse_with_prompt(results, prompt_value=OCR_text)
# Should work:
# output = self.model.generate(**self.inputs, eos_token_id=self.eos_token_id, max_new_tokens=512) # Adjust max_length as needed
# Decode the generated text
# generated_text = self.tokenizer.decode(output[0], skip_special_tokens=True)
# json_str, json_dict = self.extract_json(generated_text)
if self.print_output:
# print("\nJSON String:")
# print(json_str)
print("\nJSON Dictionary:")
print(output)
if output is None:
self.logger.error(f'Failed to extract JSON from:\n{results}')
self._adjust_config()
del results
else:
nt_in = count_tokens(self.instructions+OCR_text, self.VENDOR, self.TOKENIZER_NAME)
nt_out = count_tokens(results, self.VENDOR, self.TOKENIZER_NAME)
output = validate_and_align_JSON_keys_with_template(output, json.loads(self.JSON_dict_structure_str))
if output is None:
self.logger.error(f'[Attempt {ind}] Failed to extract JSON from:\n{results}')
self._adjust_config()
else:
self.monitor.stop_inference_timer() # Starts tool timer too
if self.json_report:
self.json_report.set_text(text_main=f'Working on WFO, Geolocation, Links')
output_WFO, WFO_record, output_GEO, GEO_record = run_tools(output, self.tool_WFO, self.tool_GEO, self.tool_wikipedia, json_file_path_wiki)
save_individual_prompt(sanitize_prompt(self.instructions+OCR_text), txt_file_path_ind_prompt)
self.logger.info(f"Formatted JSON:\n{json.dumps(output,indent=4)}")
usage_report = self.monitor.stop_monitoring_report_usage()
if self.adjust_temp != self.starting_temp:
self._reset_config()
if self.json_report:
self.json_report.set_text(text_main=f'LLM call successful')
del results
return output, nt_in, nt_out, WFO_record, GEO_record, usage_report
except Exception as e:
self.logger.error(f'{e}')
self.logger.info(f"Failed to extract valid JSON after [{ind}] attempts")
if self.json_report:
self.json_report.set_text(text_main=f'Failed to extract valid JSON after [{ind}] attempts')
self.monitor.stop_inference_timer() # Starts tool timer too
usage_report = self.monitor.stop_monitoring_report_usage()
if self.json_report:
self.json_report.set_text(text_main=f'LLM call failed')
return None, nt_in, nt_out, None, None, usage_report
# # Create a prompt from the template so we can use it with Langchain
# self.prompt = PromptTemplate(template=template, input_variables=["query"])
# # Set up a parser
# self.parser = JsonOutputParser()
model_name = "unsloth/mistral-7b-instruct-v0.2-bnb-4bit"
sltp_version = 'HLT_MICH_Angiospermae_SLTPvA_v1-0_medium__OCR-C25-L25-E50-R05'
lora_name = "phyloforfun/mistral-7b-instruct-v2-bnb-4bit__HLT_MICH_Angiospermae_SLTPvA_v1-0_medium__OCR-C25-L25-E50-R05"
OCR_test = "HERBARIUM OF MARCUS W. LYON , JR . Tracaulon sagittatum Indiana : Porter Co. Mincral Springs edge wet subdural woods 1927 TX 11 Flowers pink UNIVERSIT HERBARIUM MICHIGAN MICH University of Michigan Herbarium 1439649 copyright reserved PERSICARIA FEB 26 1965 cm "
# model.merge_and_unload()
# Generate the output