osmarks/nanogpt-experiments
1
0
Fork 0
mirror of https://github.com/osmarks/nanogpt-experiments.git synced 2024-12-18 14:10:28 +00:00
Code Issues Releases Wiki Activity

allow sample.py to init from a pretrained gpt2 checkpoints as well, in similar style to train.py

Browse Source
This commit is contained in:
Andrej Karpathy 2023-01-25 00:55:29 +00:00
parent 6c40a08b41
commit 21675d7755
1 changed files with 19 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
sample.py
View File

@ -9,7 +9,8 @@ import tiktoken
from model import GPTConfig, GPT from model import GPTConfig, GPT
# ----------------------------------------------------------------------------- # -----------------------------------------------------------------------------
out_dir = 'out' init_from = 'resume' # either 'resume' (from an out_dir) or a gpt2 variant (e.g. 'gpt2-xl')
out_dir = 'out' # ignored if init_from is not 'resume'
start = "\n" # or "<|endoftext|>" or whatever you like start = "\n" # or "<|endoftext|>" or whatever you like
num_samples = 10 # number of samples to draw num_samples = 10 # number of samples to draw
max_new_tokens = 500 # number of tokens generated in each sample max_new_tokens = 500 # number of tokens generated in each sample
@ -31,6 +32,8 @@ ptdtype = {'float32': torch.float32, 'bfloat16': torch.bfloat16, 'float16': torc
ctx = nullcontext() if device_type == 'cpu' else torch.amp.autocast(device_type=device_type, dtype=ptdtype) ctx = nullcontext() if device_type == 'cpu' else torch.amp.autocast(device_type=device_type, dtype=ptdtype)
# model # model
if init_from == 'resume':
# init from a model saved in a specific directory
ckpt_path = os.path.join(out_dir, 'ckpt.pt') ckpt_path = os.path.join(out_dir, 'ckpt.pt')
checkpoint = torch.load(ckpt_path, map_location=device) checkpoint = torch.load(ckpt_path, map_location=device)
gptconf = GPTConfig(**checkpoint['model_args']) gptconf = GPTConfig(**checkpoint['model_args'])
@ -41,6 +44,10 @@ for k,v in list(state_dict.items()):
if k.startswith(unwanted_prefix): if k.startswith(unwanted_prefix):
state_dict[k[len(unwanted_prefix):]] = state_dict.pop(k) state_dict[k[len(unwanted_prefix):]] = state_dict.pop(k)
model.load_state_dict(state_dict) model.load_state_dict(state_dict)
elif init_from.startswith('gpt2'):
# init from a given GPT-2 model
model = GPT.from_pretrained(init_from, dict(dropout=0.0))
model.eval() model.eval()
model.to(device) model.to(device)
if compile: if compile:
@ -48,7 +55,7 @@ if compile:
# look for the meta pickle in case it is available in the dataset folder # look for the meta pickle in case it is available in the dataset folder
load_meta = False load_meta = False
if 'config' in checkpoint and 'dataset' in checkpoint['config']: # older checkpoints might not have these... if init_from == 'resume' and 'config' in checkpoint and 'dataset' in checkpoint['config']: # older checkpoints might not have these...
meta_path = os.path.join('data', checkpoint['config']['dataset'], 'meta.pkl') meta_path = os.path.join('data', checkpoint['config']['dataset'], 'meta.pkl')
load_meta = os.path.exists(meta_path) load_meta = os.path.exists(meta_path)
if load_meta: if load_meta: