github/tinygrad
1
1
Fork 0
mirror of https://github.com/tinygrad/tinygrad.git synced 2026-01-08 22:48:25 -05:00
Code Issues Packages Projects Releases Wiki Activity

TRAIN=0 to only eval llama (#13804)

Browse Source
This commit is contained in:
chenyu
2025-12-22 11:55:46 -05:00
committed by GitHub
parent 39d962106f
commit 27d899ce97
1 changed files with 22 additions and 21 deletions
Download Patch File Download Diff File Expand all files Collapse all files

43
examples/mlperf/model_train.py
View File

@@ -1438,33 +1438,34 @@ def train_llama3():
iter = get_train_iter() iter = get_train_iter()
i, sequences_seen = resume_ckpt, 0 i, sequences_seen = resume_ckpt, 0
for tokens in tqdm(iter, total=SAMPLES//GBS): for tokens in tqdm(iter, total=SAMPLES//GBS):
t = time.perf_counter()
GlobalCounters.reset() GlobalCounters.reset()
loss, lr = train_step(model, tokens) if getenv("TRAIN", 1):
loss = loss.float().item() t = time.perf_counter()
lr = lr.item() loss, lr = train_step(model, tokens)
loss = loss.float().item()
lr = lr.item()
i += 1 i += 1
sequences_seen += tokens.shape[0] sequences_seen += tokens.shape[0]
sec = time.perf_counter()-t sec = time.perf_counter()-t
tqdm.write( tqdm.write(
f"{i:5} {sec:.2f} s run, {loss:.4f} loss, {lr:.12f} LR, {GlobalCounters.mem_used / 1e9:.2f} GB used, " f"{i:5} {sec:.2f} s run, {loss:.4f} loss, {lr:.12f} LR, {GlobalCounters.mem_used / 1e9:.2f} GB used, "
f"{GlobalCounters.global_ops * 1e-9 / sec:9.2f} GFLOPS") f"{GlobalCounters.global_ops * 1e-9 / sec:9.2f} GFLOPS")
if (fname:=getenv("LOSS_FILE", "")): if (fname:=getenv("LOSS_FILE", "")):
with open(fname, "a") as f: with open(fname, "a") as f:
f.write(f"{i} {loss:.4f} {lr.item():.12f} {GlobalCounters.mem_used / 1e9:.2f}\n") f.write(f"{i} {loss:.4f} {lr.item():.12f} {GlobalCounters.mem_used / 1e9:.2f}\n")
if (ckpt_freq := getenv("CKPT")) and (i % ckpt_freq == 0 and (i != 1 or ckpt_freq == 1)): if (ckpt_freq := getenv("CKPT")) and (i % ckpt_freq == 0 and (i != 1 or ckpt_freq == 1)):
tqdm.write("saving checkpoint") tqdm.write("saving checkpoint")
if not os.path.exists(ckpt_dir := "./ckpts"): os.mkdir(ckpt_dir) if not os.path.exists(ckpt_dir := "./ckpts"): os.mkdir(ckpt_dir)
fn = f"{ckpt_dir}/llama3_{i}.safe" fn = f"{ckpt_dir}/llama3_{i}.safe"
safe_save(get_state_dict(model), fn) safe_save(get_state_dict(model), fn)
tqdm.write("saving optim checkpoint") tqdm.write("saving optim checkpoint")
fn = f"{ckpt_dir}/llama3_{i}_optim.safe" fn = f"{ckpt_dir}/llama3_{i}_optim.safe"
safe_save(get_state_dict(scheduler), fn) safe_save(get_state_dict(scheduler), fn)
if sequences_seen % EVAL_FREQ == 0 and (i != 1 or EVAL_FREQ == 1): if sequences_seen % EVAL_FREQ == 0 and (i != 1 or EVAL_FREQ == 1):
tqdm.write(f"evaluating after {sequences_seen} sequences") tqdm.write(f"evaluating after {sequences_seen} sequences")