052953d9-0495-43db-9d15-ae98fff06f18.txt

import os
import sys
with open(sys.argv[0]) as f:
    code = f.read() # read the code of this file ASAP, for logging
import uuid
import time
import contextlib
from dataclasses import dataclass
from pathlib import Path

import torch
from torch import nn
import torch.nn.functional as F
import torch.distributed as dist
import torch._inductor.config as config
from torch.nn.parallel import DistributedDataParallel as DDP
from torch.nn.attention.flex_attention import BlockMask, flex_attention #KoszarskyB

# -----------------------------------------------------------------------------
# Muon optimizer

@torch.compile
def zeropower_via_newtonschulz5(G, steps=10, eps=1e-7):
    """
    Newton-Schulz iteration to compute the zeroth power / orthogonalization of G. We opt to use a
    quintic iteration whose coefficients are selected to maximize the slope at zero. For the purpose
    of minimizing steps, it turns out to be empirically effective to keep increasing the slope at
    zero even beyond the point where the iteration no longer converges all the way to one everywhere
    on the interval. This iteration therefore does not produce UV^T but rather something like US'V^T
    where S' is diagonal with S_{ii}' ~ Uniform(0.5, 1.5), which turns out not to hurt model
    performance at all relative to UV^T, where USV^T = G is the SVD.
    """
    assert len(G.shape) == 2
    a, b, c = (3.4445, -4.7750,  2.0315)
    X = G.bfloat16()
    X /= (X.norm() + eps) # ensure top singular value <= 1
    if G.size(0) > G.size(1):
        X = X.T
    for _ in range(steps):
        A = X @ X.T
        B = b * A + c * A @ A # adapted from suggestion by @jxbz, @leloykun, and @YouJiacheng
        X = a * X + B @ X
    if G.size(0) > G.size(1):
        X = X.T
    return X

class Muon(torch.optim.Optimizer):
    """
    Muon - MomentUm Orthogonalized by Newton-schulz

    Muon internally runs standard SGD-momentum, and then performs an orthogonalization post-
    processing step, in which each 2D parameter's update is replaced with the nearest orthogonal
    matrix. To efficiently orthogonalize each update, we use a Newton-Schulz iteration, which has
    the advantage that it can be stably run in bfloat16 on the GPU.

    Some warnings:
    - This optimizer assumes that all parameters passed in are 2D.
    - It should not be used for the embedding layer, the final fully connected layer, or any {0,1}-D
    parameters; those should all be optimized by a standard method (e.g., AdamW).
    - To use it with 4D convolutional filters, it works well to just flatten their last 3 dimensions.
    - We believe it is unlikely to work well for training with small batch size.
    - We believe it may not work well for finetuning pretrained models, but we haven't tested this.
    - We have not yet tried this optimizer for training scenarios larger than NanoGPT (124M).

    Arguments:
        lr: The learning rate used by the internal SGD.
        momentum: The momentum used by the internal SGD.
        nesterov: Whether to use Nesterov-style momentum in the internal SGD. (recommended)
        ns_steps: The number of Newton-Schulz iteration steps to use.
    """
    def __init__(self, params, lr=0.02, momentum=0.95, nesterov=True, ns_steps=5):
        self.world_size = int(os.environ['WORLD_SIZE'])
        self.rank = int(os.environ['RANK'])
        defaults = dict(lr=lr, momentum=momentum, nesterov=nesterov, ns_steps=ns_steps)
        params = list(params)
        assert all(isinstance(p, torch.Tensor) for p in params)
        sizes = {p.numel() for p in params}
        param_groups = [
            {
                'params': [p for p in params if p.numel() == size],
                'update_buffer': [
                    torch.empty(size, device='cuda', dtype=torch.bfloat16)
                    for _ in range(self.world_size)
                ],
            }
            for size in sizes
        ]
        super().__init__(param_groups, defaults)

    def step(self):

        for group in self.param_groups:

            lr = group['lr']
            momentum = group['momentum']
            nesterov = group['nesterov']
            ns_steps = group['ns_steps']
            update_buffers = group['update_buffer']
            # generate weight updates in distributed fashion
            params = group['params']
            assert len(params) % self.world_size == 0
            handle = None
            params_world = None
            def update_prev():
                if params_world is None:
                    return
                assert handle is not None
                handle.wait()
                for p_world, g_world in zip(params_world, update_buffers):
                    p_world.data.add_(
                        g_world.view_as(p_world),
                        alpha=-lr * max(1, p_world.size(0) / p_world.size(1)) ** 0.5,
                    )
            for base_i in range(len(params))[::self.world_size]:
                p = params[base_i + self.rank]
                g = p.grad
                assert g is not None
                state = self.state[p]
                if 'momentum_buffer' not in state:
                    state['momentum_buffer'] = torch.zeros_like(g)
                buf = state['momentum_buffer']
                buf.lerp_(g, 1 - momentum)
                g = g.lerp_(buf, momentum) if nesterov else buf
                g = zeropower_via_newtonschulz5(g, steps=ns_steps).flatten()
                update_prev()
                handle = dist.all_gather(update_buffers, g, async_op=True)
                params_world = params[base_i : base_i + self.world_size]
            update_prev()

# -----------------------------------------------------------------------------
# PyTorch nn.Module definitions for the GPT-2 model

def norm(x):
    return F.rms_norm(x, (x.size(-1),))

class CastedLinear(nn.Linear):

    def __init__(self, in_features, out_features):
        super().__init__(in_features, out_features, bias=False)

    def forward(self, x):
        return F.linear(x, self.weight.to(x.dtype))

class Rotary(torch.nn.Module):

    def __init__(self, dim, base=10000):
        super().__init__()
        self.register_buffer('inv_freq', (1 / base) ** (torch.arange(0, dim, 2) / dim))
        self.seq_len_cached = None
        self.cos_cached = None
        self.sin_cached = None

    def forward(self, x):
        seq_len = x.shape[1]
        if seq_len != self.seq_len_cached:
            t = torch.arange(seq_len, device=x.device)
            freqs = torch.outer(t, self.inv_freq)
            self.seq_len_cached = seq_len
            self.cos_cached = freqs.cos()
            self.sin_cached = freqs.sin()
        cos, sin = self.cos_cached[None, :, None, :], self.sin_cached[None, :, None, :]
        # apply_rotary_emb(x, cos, sin)
        x1, x2 = x.chunk(2, dim=3)
        y1 = x1 * cos + x2 * sin
        y2 = x1 * (-sin) + x2 * cos
        return torch.cat((y1, y2), 3).type_as(x)

class CausalSelfAttention(nn.Module):

    def __init__(self, dim, num_heads):
        super().__init__()
        assert dim % num_heads == 0
        self.num_heads = num_heads
        self.c_q = CastedLinear(dim, dim)
        self.c_k = CastedLinear(dim, dim)
        self.c_v = CastedLinear(dim, dim)
        self.lambdas = nn.Parameter(torch.tensor([0.5, 0.5]))
        self.rotary = Rotary(dim // num_heads) # dim // num_heads = head_dim
        self.c_proj = CastedLinear(dim, dim)
        self.c_proj.weight.data.zero_() # zero init suggested by @Grad62304977

    def forward(self, x, vi, block_mask):
        B, T = x.size(0), x.size(1) # batch size, sequence length
        assert B == 1, "Must use batch size = 1 for FlexAttention"
        q = self.c_q(x).view(B, T, self.num_heads, -1)
        k = self.c_k(x).view(B, T, self.num_heads, -1)
        v = self.c_v(x).view(B, T, self.num_heads, -1)
        v = self.lambdas[0] * v + self.lambdas[1] * vi.view_as(v) # @KoszarskyB & @Grad62304977
        q, k = norm(q), norm(k) # QK norm @Grad62304977
        q, k = self.rotary(q), self.rotary(k)
        y = flex_attention(q.transpose(1, 2), k.transpose(1, 2), v.transpose(1, 2), block_mask=block_mask, enable_gqa=True)
        y = y.transpose(1, 2).contiguous().view_as(x) # re-assemble all head outputs side by side
        y = self.c_proj(y)
        return y

class MLP(nn.Module):

    def __init__(self, dim):
        super().__init__()
        self.c_fc   = CastedLinear(dim, 4 * dim)
        self.c_proj = CastedLinear(4 * dim, dim)
        self.c_proj.weight.data.zero_() # zero init suggested by @Grad62304977

    def forward(self, x):
        x = self.c_fc(x)
        x = F.relu(x).square() # https://arxiv.org/abs/2109.08668v2; ~1-2% better than GELU; suggested by @SKYLINEZ007 and @Grad62304977
        x = self.c_proj(x)
        return x

class Block(nn.Module):

    def __init__(self, config):
        super().__init__()
        self.attn = CausalSelfAttention(config.model_dim, config.num_heads)
        self.mlp = MLP(config.model_dim)
        self.lambdas = nn.Parameter(torch.tensor([1., 0.]))

    def forward(self, x, vi, x0, block_mask):
        x = self.lambdas[0] * x + self.lambdas[1] * x0
        x = x + self.attn(norm(x), vi, block_mask)
        x = x + self.mlp(norm(x))
        return x

class ValueEmbedding(nn.Module):
    def __init__(self, config: "GPTConfig"):
        super().__init__()
        self.__setattr__
        self.embed = nn.ModuleList([
            nn.Embedding(config.vocab_size, config.model_dim)
            for _ in range(6)
        ])

    def forward(self, inputs) -> "list[torch.Tensor]":
        ve = [emb(inputs) for emb in self.embed]
        ve += reversed(ve)
        return ve


# -----------------------------------------------------------------------------
# The main GPT-2 model

@dataclass
class GPTConfig:
    vocab_size : int = 50304
    num_layers : int = 12
    num_heads : int = 6 # head dim 128 suggested by @Grad62304977
    model_dim : int = 768

class GPT(nn.Module):

    def __init__(self, config: GPTConfig):
        super().__init__()
        self.num_layers = config.num_layers

        # U-net design by @brendanh0gan
        self.num_encoder_layers = config.num_layers // 2 # Half of the layers for encoder
        self.num_decoder_layers = config.num_layers - self.num_encoder_layers # Remaining for decoder
        # Add learnable skip connection weights for decoder layers
        self.skip_weights = nn.Parameter(torch.ones(self.num_decoder_layers))

        self.embed = nn.Embedding(config.vocab_size, config.model_dim)
        self.blocks = nn.ModuleList([Block(config) for _ in range(config.num_layers)])
        # token value embeddings by @KoszarskyB - inspired by @Grad62304977's value residual learning
        # U-net structure on token value embeddings by @leloykun
        self.value_embeds = ValueEmbedding(config)
        self.lm_head = CastedLinear(config.model_dim, config.vocab_size)
        self.lm_head.weight.data.zero_() # @Grad62304977

    def forward(
        self,
        inputs: torch.Tensor,
        targets: torch.Tensor,
        sliding_window_num_blocks: torch.Tensor,
    ):
        BLOCK_SIZE = 128
        assert inputs.ndim == 1
        docs = (inputs == 50256).cumsum(0)
        docs_low = docs.view(-1, BLOCK_SIZE)[:, 0].contiguous()
        docs_high = docs.view(-1, BLOCK_SIZE)[:, -1].contiguous()

        def document_causal(b, h, q_idx, kv_idx):
            causal_mask = q_idx >= kv_idx
            document_mask = docs[q_idx] == docs[kv_idx]
            return causal_mask & document_mask

        def dense_to_ordered(dense_mask: torch.Tensor):
            num_blocks = dense_mask.sum(dim=-1, dtype=torch.int32)
            indices = dense_mask.argsort(dim=-1, descending=True, stable=True).to(torch.int32)
            return num_blocks[None, None].contiguous(), indices[None, None].contiguous()

        def create_doc_swc_block_mask(sliding_window_num_blocks: torch.Tensor):
            kv_idx = block_idx = torch.arange(512, dtype=torch.int32, device="cuda")
            q_idx = block_idx[:, None]
            causal_bm = q_idx >= kv_idx
            causal_full_bm = q_idx > kv_idx
            window_bm = q_idx - kv_idx < sliding_window_num_blocks
            window_full_bm = window_bm
            # document_bm = (docs_low[q_idx] <= docs_high[kv_idx]) & (docs_low[kv_idx] <= docs_high[q_idx])
            document_bm = (docs_low[:, None] <= docs_high) & (docs_low <= docs_high[:, None])
            document_full_bm = (docs_low[:, None] == docs_high) & (docs_low == docs_high[:, None])
            nonzero_bm = causal_bm & window_bm & document_bm
            full_bm  = causal_full_bm & window_full_bm & document_full_bm
            kv_num_blocks, kv_indices = dense_to_ordered(nonzero_bm ^ full_bm)
            full_kv_num_blocks, full_kv_indices = dense_to_ordered(full_bm)
            return BlockMask.from_kv_blocks(
                kv_num_blocks,
                kv_indices,
                full_kv_num_blocks,
                full_kv_indices,
                BLOCK_SIZE=BLOCK_SIZE,
                mask_mod=document_causal,
            )

        block_mask = create_doc_swc_block_mask(sliding_window_num_blocks)

        # forward the GPT model itself
        x = self.embed(inputs[None]) # token embeddings of shape (b, t, model_dim)
        x = norm(x) # @Grad62304977
        x0 = x
        ve = self.value_embeds(inputs)
        ve_enc, ve_dec = ve[:self.num_encoder_layers], ve[self.num_encoder_layers:]

        # Store outputs for U-Net skip connections
        skip_connections = []
        # Encoder pass - process only the first half of the blocks
        for i in range(self.num_encoder_layers):
            x = self.blocks[i](x, ve_enc[i], x0, block_mask)
            skip_connections.append(x)
        # Decoder pass - process the remaining blocks with weighted skip connections
        for i in range(self.num_decoder_layers):
            x = x + self.skip_weights[i] * skip_connections.pop()
            # U-net structure on token value embeddings by @leloykun
            x = self.blocks[self.num_encoder_layers + i](x, ve_dec[i], x0, block_mask)

        x = norm(x)
        logits = self.lm_head(x)
        logits = 30 * torch.tanh(logits / 30) # @Grad62304977
        logits = logits.float()
        loss = F.cross_entropy(logits.view(-1, logits.size(-1)), targets.view(-1))
        return loss

# -----------------------------------------------------------------------------
# Our own simple Distributed Data Loader

def _peek_data_shard(file: Path):
    # only reads the header, returns header data
    # header is 256 int32
    header = torch.from_file(f"{file}", False, 256, dtype=torch.int32)
    assert header[0] == 20240520, "magic number mismatch in the data .bin file"
    assert header[1] == 1, "unsupported version"
    return int(header[2]) # number of tokens (claimed)

def _load_data_shard(path: Path, num_tokens):
    with path.open("rb", buffering=0) as f:
        tokens = torch.empty(num_tokens, dtype=torch.uint16, pin_memory=True)
        f.seek(256 * 4)
        nbytes = f.readinto(tokens.numpy())
        assert nbytes == 2 * num_tokens, "number of tokens read does not match header?"
    return tokens

class DistributedDataLoader:
    def __init__(self, filename_pattern, seq_len, process_rank, num_processes):
        self.process_rank = process_rank
        self.num_processes = num_processes
        self.seq_len = seq_len

        # glob files that match the pattern
        self.files = sorted(Path.cwd().glob(filename_pattern))
        assert len(self.files) > 0, f"did not find any files that match the pattern {filename_pattern}"

        # load and validate all data shards, count number of tokens in total
        self.files_num_tokens = [_peek_data_shard(file) for file in self.files]
        assert min(self.files_num_tokens) >= num_processes * seq_len + 1
        self.total_num_tokens = sum(self.files_num_tokens)

        self.reset()

    def reset(self):
        self.current_shard = -1
        self.advance()

    def advance(self): # advance to next data shard
        self.current_shard = (self.current_shard + 1) % len(self.files)
        self.current_position = self.process_rank * self.seq_len
        self.tokens = _load_data_shard(self.files[self.current_shard], self.files_num_tokens[self.current_shard])

    def next_batch(self):
        batch_size = self.seq_len * self.num_processes
        buf = self.tokens[self.current_position:self.current_position+self.seq_len+1]
        # host side async is sufficient;
        # no performance improvement was observed when introducing a separate stream.
        inputs = buf[:-1].to(device="cuda", dtype=torch.int32, non_blocking=True) # inputs
        targets = buf[1:].to(device="cuda", dtype=torch.int64, non_blocking=True) # targets
        # advance current position and load next shard if necessary
        self.current_position += batch_size
        if self.current_position + batch_size + 1 >= len(self.tokens):
            self.advance()
        return inputs, targets

# -----------------------------------------------------------------------------
# int main

@dataclass
class Hyperparameters:
    # data hyperparams
    input_bin : str = 'data/fineweb10B/fineweb_train_*.bin' # input .bin to train on
    input_val_bin : str = 'data/fineweb10B/fineweb_val_*.bin' # input .bin to eval validation loss on
    # optimization hyperparams
    batch_size : int = 8 # batch size, in sequences, across all devices
    sequence_length : int = 64*1024 # sequence length, in tokens
    num_iterations : int = 1480 # number of iterations to run
    warmup_iters : int = 0
    cooldown_iters : int = 600 # number of iterations of linear warmup/cooldown for triangular or trapezoidal schedule
    weight_decay : float = 0
    # evaluation and logging hyperparams
    val_loss_every : int = 125 # every how many steps to evaluate val loss? 0 for only at the end
    val_tokens : int = 10485760 # how many tokens of validation data? it's important to keep this fixed for consistent comparisons
    save_every : int = 0 # every how many steps to save the checkpoint? 0 for only at the end
args = Hyperparameters()

# set up DDP (distributed data parallel). torchrun sets this env variable
ddp_rank = int(os.environ['RANK'])
ddp_local_rank = int(os.environ['LOCAL_RANK'])
ddp_world_size = int(os.environ['WORLD_SIZE'])
assert torch.cuda.is_available()
device = torch.device(f"cuda:{ddp_local_rank}")
torch.cuda.set_device(device)
print(f"using device: {device}")
dist.init_process_group(backend='nccl', device_id=device)
dist.barrier()
master_process = (ddp_rank == 0) # this process will do logging, checkpointing etc.

# begin logging
logfile = None
if master_process:
    run_id = uuid.uuid4()
    logdir = Path("logs") / f"{run_id}"
    logdir.mkdir(exist_ok=True)
    logfile = Path("logs") / f"{run_id}.txt"
    print(logfile.stem)
    # create the log file
    with logfile.open("w") as f:
        # begin the log by printing this file (the Python code)
        print(code, file=f)
        print("=" * 100, file=f)
def print0(s, logonly=False):
    if master_process:
        with logfile.open("a") as f:
            if not logonly:
                print(s)
            print(s, file=f)
# log information about the hardware/software environment this is running on
# and print the full `nvidia-smi` to file
print0(f"Running python {sys.version}")
print0(f"Running pytorch {torch.version.__version__} compiled for CUDA {torch.version.cuda}\nnvidia-smi:")
import subprocess
result = subprocess.run(['nvidia-smi'], stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True)
print0(f'{result.stdout}', logonly=True)
print0('='*100, logonly=True)

# calculate the number of steps to take in the val loop.
assert args.val_tokens % (args.sequence_length * ddp_world_size) == 0
val_steps = args.val_tokens // (args.sequence_length * ddp_world_size)
# calculate the steps of gradient accumulation required to attain the desired global batch size.
assert args.batch_size % (ddp_world_size) == 0
train_accumulation_steps = args.batch_size // ddp_world_size

# load tokens
train_loader = DistributedDataLoader(args.input_bin, args.sequence_length, ddp_rank, ddp_world_size)
val_loader = DistributedDataLoader(args.input_val_bin, args.sequence_length, ddp_rank, ddp_world_size)
print0(f"Training DataLoader: total number of tokens: {train_loader.total_num_tokens} across {len(train_loader.files)} files")
print0(f"Validation DataLoader: total number of tokens: {val_loader.total_num_tokens} across {len(val_loader.files)} files")
print0('='*100, logonly=True)
inputs_train, targets_train = train_loader.next_batch()

# there are only 50257 unique GPT-2 tokens; we extend to nearest multiple of 128 for efficiency. suggested to me by @Grad62304977.
# this originates from Karpathy's experiments.
num_vocab = 50304
model = GPT(GPTConfig(vocab_size=num_vocab, num_layers=12, num_heads=6, model_dim=768))
model = model.cuda().bfloat16()
for m in model.modules():
    if isinstance(m, CastedLinear):
        m.float()
config.coordinate_descent_tuning = True # suggested by @Chillee
model = torch.compile(model)
# here we wrap model into DDP container
model = DDP(model, device_ids=[ddp_local_rank], broadcast_buffers=False, gradient_as_bucket_view=True)
raw_model = model.module # always contains the "raw" unwrapped model

# init the optimizer(s)
embed_params = [*raw_model.embed.parameters(), *raw_model.value_embeds.parameters()]
optimizer1 = torch.optim.Adam(embed_params, lr=0.6, betas=(0.8, 0.95), fused=True)
optimizer2 = torch.optim.Adam([raw_model.lm_head.weight], lr=0.008, betas=(0.8, 0.95), fused=True)
params = list(raw_model.blocks.parameters())
matrix_params = [p for p in params if p.ndim == 2]
scalar_params = [p for p in params if p.ndim < 2] + [raw_model.skip_weights]
optimizer3 = Muon(matrix_params, lr=0.05, momentum=0.95)
optimizer4 = torch.optim.Adam(scalar_params, lr=0.04, betas=(0.8, 0.95), fused=True)
optimizers = [optimizer1, optimizer2, optimizer3, optimizer4]
# learning rate decay scheduler (linear warmup and cooldown)
def get_lr(it):
    assert it <= args.num_iterations
    # 1) linear warmup for warmup_iters steps
    if it < args.warmup_iters:
        return (it+1) / args.warmup_iters
    # 2) constant lr for a while
    elif it < args.num_iterations - args.cooldown_iters:
        return 1.0
    # 3) linear cooldown
    else:
        decay_ratio = (args.num_iterations - it) / args.cooldown_iters
        return decay_ratio
schedulers = [torch.optim.lr_scheduler.LambdaLR(opt, get_lr) for opt in optimizers]

sliding_window_num_blocks = torch.tensor(1, dtype=torch.int32, device="cuda")
sw_num_blocks_prev = 1
# Start training loop
training_time_ms = 0
# start the clock
torch.cuda.synchronize()
t0 = time.perf_counter()
# begin training
for step in range(args.num_iterations + 1):
    last_step = (step == args.num_iterations)
    # This effectively ignores timing first 10 steps, which are slower for weird reasons.
    # Alternately, and slightly more correctly in terms of benchmarking, we could do 10
    # steps with dummy data first, and then re-initialize the model and reset the loader.
    if step == 10:
        training_time_ms = 0
        t0 = time.perf_counter()
    timed_steps = float('nan') if step <= 11 else (step - 10) + 1 # <= 11 to avoid bug in val

    # Linearly increase the sliding window size over training in chunks of 64 from 64 -> 1792. By @fernbear.bsky.social
    frac_done = step / args.num_iterations # training progress
    sw_num_blocks = int(((1 - frac_done) * 64 + frac_done * 1792 + 64) // 128)
    if sw_num_blocks != sw_num_blocks_prev:
        sliding_window_num_blocks.copy_(sw_num_blocks, non_blocking=True)
        sw_num_blocks_prev = sw_num_blocks

    # once in a while evaluate the validation dataset
    if (last_step or (args.val_loss_every > 0 and step % args.val_loss_every == 0)):
        # stop the clock
        torch.cuda.synchronize()
        training_time_ms += 1000 * (time.perf_counter() - t0)
        # run validation batches
        model.eval()
        val_loader.reset()
        val_loss = 0.0
        for _ in range(val_steps):
            with torch.no_grad():
                inputs_val, targets_val = val_loader.next_batch()
                val_loss += model(inputs_val, targets_val, sliding_window_num_blocks)
        dist.all_reduce(val_loss, op=dist.ReduceOp.AVG)
        val_loss /= val_steps
        # log val loss to console and to logfile
        print0(f'step:{step}/{args.num_iterations} val_loss:{val_loss:.4f} train_time:{training_time_ms:.0f}ms step_avg:{training_time_ms/(timed_steps-1):.2f}ms')
        # start the clock again
        torch.cuda.synchronize()
        t0 = time.perf_counter()

    if master_process and (last_step or (args.save_every > 0 and step % args.save_every == 0)):
        # stop the clock
        torch.cuda.synchronize()
        training_time_ms += 1000 * (time.perf_counter() - t0)
        # save the state of the training process
        log = dict(step=step, code=code, model=raw_model.state_dict(), optimizers=[opt.state_dict() for opt in optimizers])
        torch.save(log, 'logs/%s/state_step%06d.pt' % (run_id, step))
        # start the clock again
        torch.cuda.synchronize()
        t0 = time.perf_counter()

    # bit confusing: we want to make sure to eval on 0th iteration
    # but also after the very last iteration. so we loop for step <= num_iterations
    # instead of just < num_iterations (one extra due to <=), only to do
    # the validation/sampling one last time, and then we break right here as we're done.
    if last_step:
        break

    # --------------- TRAINING SECTION BEGIN -----------------
    model.train()
    for i in range(1, train_accumulation_steps + 1):
        with contextlib.ExitStack() as stack:
            if i < train_accumulation_steps: # there's no need to sync gradients every accumulation step
                stack.enter_context(model.no_sync())
            if step >= 5:
                stack.enter_context(torch.compiler.set_stance(skip_guard_eval_unsafe=True))
            model(inputs_train, targets_train, sliding_window_num_blocks).backward()
            inputs_train, targets_train = train_loader.next_batch()
    if train_accumulation_steps != 1:
        for p in model.parameters():
            p.grad /= train_accumulation_steps
    # momentum warmup for Muon
    frac = min(step/300, 1)
    for group in optimizer3.param_groups:
        group['momentum'] = (1 - frac) * 0.85 + frac * 0.95
    # step the optimizers and schedulers
    for opt, sched in zip(optimizers, schedulers):
        opt.step()
        sched.step()
    # null the gradients
    model.zero_grad(set_to_none=True)
    # --------------- TRAINING SECTION END -------------------
    # everything that follows now is just diagnostics, prints, logging, etc.
    approx_time = training_time_ms + 1000 * (time.perf_counter() - t0)
    print0(f"step:{step+1}/{args.num_iterations} train_time:{approx_time:.0f}ms step_avg:{approx_time/timed_steps:.2f}ms")

print0(f"peak memory consumption: {torch.cuda.max_memory_allocated() // 1024 // 1024} MiB")

# -------------------------------------------------------------------------
# clean up nice
dist.destroy_process_group()

====================================================================================================
Running python 3.10.12 (main, Sep 11 2024, 15:47:36) [GCC 11.4.0]
Running pytorch 2.6.0.dev20241203+cu124 compiled for CUDA 12.4
nvidia-smi:
Wed Dec 11 09:18:22 2024       
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 535.183.06             Driver Version: 535.183.06   CUDA Version: 12.2     |
|-----------------------------------------+----------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |         Memory-Usage | GPU-Util  Compute M. |
|                                         |                      |               MIG M. |
|=========================================+======================+======================|
|   0  NVIDIA H100 80GB HBM3          On  | 00000000:19:00.0 Off |                    0 |
| N/A   38C    P0             125W / 700W |   7084MiB / 81559MiB |      0%      Default |
|                                         |                      |             Disabled |
+-----------------------------------------+----------------------+----------------------+
|   1  NVIDIA H100 80GB HBM3          On  | 00000000:3B:00.0 Off |                    0 |
| N/A   30C    P0             116W / 700W |   3451MiB / 81559MiB |      0%      Default |
|                                         |                      |             Disabled |
+-----------------------------------------+----------------------+----------------------+
|   2  NVIDIA H100 80GB HBM3          On  | 00000000:4C:00.0 Off |                    0 |
| N/A   28C    P0             112W / 700W |   3451MiB / 81559MiB |      0%      Default |
|                                         |                      |             Disabled |
+-----------------------------------------+----------------------+----------------------+
|   3  NVIDIA H100 80GB HBM3          On  | 00000000:5D:00.0 Off |                    0 |
| N/A   36C    P0             114W / 700W |   3451MiB / 81559MiB |      0%      Default |
|                                         |                      |             Disabled |
+-----------------------------------------+----------------------+----------------------+
|   4  NVIDIA H100 80GB HBM3          On  | 00000000:9B:00.0 Off |                    0 |
| N/A   38C    P0             120W / 700W |   3451MiB / 81559MiB |      0%      Default |
|                                         |                      |             Disabled |
+-----------------------------------------+----------------------+----------------------+
|   5  NVIDIA H100 80GB HBM3          On  | 00000000:BB:00.0 Off |                    0 |
| N/A   30C    P0             118W / 700W |   3451MiB / 81559MiB |      0%      Default |
|                                         |                      |             Disabled |
+-----------------------------------------+----------------------+----------------------+
|   6  NVIDIA H100 80GB HBM3          On  | 00000000:CB:00.0 Off |                    0 |
| N/A   36C    P0             119W / 700W |   3451MiB / 81559MiB |      0%      Default |
|                                         |                      |             Disabled |
+-----------------------------------------+----------------------+----------------------+
|   7  NVIDIA H100 80GB HBM3          On  | 00000000:DB:00.0 Off |                    0 |
| N/A   30C    P0             118W / 700W |   3211MiB / 81559MiB |      0%      Default |
|                                         |                      |             Disabled |
+-----------------------------------------+----------------------+----------------------+
                                                                                         
+---------------------------------------------------------------------------------------+
| Processes:                                                                            |
|  GPU   GI   CI        PID   Type   Process name                            GPU Memory |
|        ID   ID                                                             Usage      |
|=======================================================================================|
+---------------------------------------------------------------------------------------+

====================================================================================================
Training DataLoader: total number of tokens: 1000000000 across 10 files
Validation DataLoader: total number of tokens: 100000000 across 1 files
====================================================================================================
step:0/1480 val_loss:10.8258 train_time:0ms step_avg:nanms
step:1/1480 train_time:28760ms step_avg:nanms
step:2/1480 train_time:28949ms step_avg:nanms
step:3/1480 train_time:29075ms step_avg:nanms
step:4/1480 train_time:29214ms step_avg:nanms
step:5/1480 train_time:29357ms step_avg:nanms
step:6/1480 train_time:29496ms step_avg:nanms
step:7/1480 train_time:29637ms step_avg:nanms
step:8/1480 train_time:29778ms step_avg:nanms
step:9/1480 train_time:29919ms step_avg:nanms
step:10/1480 train_time:30062ms step_avg:nanms
step:11/1480 train_time:144ms step_avg:nanms
step:12/1480 train_time:285ms step_avg:nanms
step:13/1480 train_time:426ms step_avg:142.14ms
step:14/1480 train_time:568ms step_avg:141.98ms
step:15/1480 train_time:709ms step_avg:141.86ms
step:16/1480 train_time:852ms step_avg:142.04ms
step:17/1480 train_time:994ms step_avg:142.02ms
step:18/1480 train_time:1138ms step_avg:142.21ms
step:19/1480 train_time:1281ms step_avg:142.38ms
step:20/1480 train_time:1425ms step_avg:142.45ms
step:21/1480 train_time:1567ms step_avg:142.47ms
step:22/1480 train_time:1710ms step_avg:142.47ms
step:23/1480 train_time:1851ms step_avg:142.42ms
step:24/1480 train_time:1994ms step_avg:142.44ms
step:25/1480 train_time:2136ms step_avg:142.43ms
step:26/1480 train_time:2280ms step_avg:142.48ms
step:27/1480 train_time:2422ms step_avg:142.50ms
step:28/1480 train_time:2567ms step_avg:142.61ms
step:29/1480 train_time:2709ms step_avg:142.60ms
step:30/1480 train_time:3233ms step_avg:161.63ms
step:31/1480 train_time:3336ms step_avg:158.84ms
step:32/1480 train_time:3480ms step_avg:158.16ms
step:33/1480 train_time:3621ms step_avg:157.45ms
step:34/1480 train_time:3765ms step_avg:156.88ms
step:35/1480 train_time:3906ms step_avg:156.26ms
step:36/1480 train_time:4050ms step_avg:155.77ms
step:37/1480 train_time:4192ms step_avg:155.24ms
step:38/1480 train_time:4334ms step_avg:154.79ms
step:39/1480 train_time:4478ms step_avg:154.40ms
step:40/1480 train_time:4621ms step_avg:154.03ms
step:41/1480 train_time:4764ms step_avg:153.69ms
step:42/1480 train_time:4907ms step_avg:153.33ms
step:43/1480 train_time:5049ms step_avg:153.00ms
step:44/1480 train_time:5191ms step_avg:152.67ms
step:45/1480 train_time:5333ms step_avg:152.37ms
step:46/1480 train_time:5476ms step_avg:152.12ms
step:47/1480 train_time:5619ms step_avg:151.87ms
step:48/1480 train_time:5762ms step_avg:151.63ms
step:49/1480 train_time:5906ms step_avg:151.43ms
step:50/1480 train_time:6047ms step_avg:151.19ms
step:51/1480 train_time:6189ms step_avg:150.96ms
step:52/1480 train_time:6331ms step_avg:150.73ms
step:53/1480 train_time:6473ms step_avg:150.53ms
step:54/1480 train_time:6616ms step_avg:150.36ms
step:55/1480 train_time:6759ms step_avg:150.21ms
step:56/1480 train_time:6902ms step_avg:150.05ms
step:57/1480 train_time:7044ms step_avg:149.87ms
step:58/1480 train_time:7187ms step_avg:149.73ms
step:59/1480 train_time:7330ms step_avg:149.59ms
step:60/1480 train_time:7472ms step_avg:149.45ms
step:61/1480 train_time:7615ms step_avg:149.32ms
step:62/1480 train_time:7760ms step_avg:149.23ms
step:63/1480 train_time:7903ms step_avg:149.12ms
step:64/1480 train_time:8047ms step_avg:149.02ms
step:65/1480 train_time:8189ms step_avg:148.88ms
step:66/1480 train_time:8331ms step_avg:148.77ms
step:67/1480 train_time:8474ms step_avg:148.66ms
step:68/1480 train_time:8617ms step_avg:148.58ms
step:69/1480 train_time:8762ms step_avg:148.50ms
step:70/1480 train_time:8905ms step_avg:148.41ms
step:71/1480 train_time:9047ms step_avg:148.31ms
step:72/1480 train_time:9190ms step_avg:148.22ms
step:73/1480 train_time:9332ms step_avg:148.12ms
step:74/1480 train_time:9475ms step_avg:148.04ms
step:75/1480 train_time:9617ms step_avg:147.95ms
step:76/1480 train_time:9762ms step_avg:147.90ms
step:77/1480 train_time:9905ms step_avg:147.84ms
step:78/1480 train_time:10048ms step_avg:147.77ms
step:79/1480 train_time:10191ms step_avg:147.70ms
step:80/1480 train_time:10739ms step_avg:153.42ms
step:81/1480 train_time:10843ms step_avg:152.72ms
step:82/1480 train_time:10986ms step_avg:152.58ms
step:83/1480 train_time:11127ms step_avg:152.43ms
step:84/1480 train_time:11269ms step_avg:152.29ms
step:85/1480 train_time:11410ms step_avg:152.14ms
step:86/1480 train_time:11552ms step_avg:152.00ms
step:87/1480 train_time:11693ms step_avg:151.86ms
step:88/1480 train_time:11841ms step_avg:151.80ms
step:89/1480 train_time:11988ms step_avg:151.74ms
step:90/1480 train_time:12130ms step_avg:151.63ms
step:91/1480 train_time:12272ms step_avg:151.51ms
step:92/1480 train_time:12414ms step_avg:151.39ms
step:93/1480 train_time:12557ms step_avg:151.29ms
step:94/1480 train_time:12700ms step_avg:151.19ms
step:95/1480 train_time:12843ms step_avg:151.10ms
step:96/1480 train_time:13355ms step_avg:155.29ms
step:97/1480 train_time:13459ms step_avg:154.70ms
step:98/1480 train_time:13602ms step_avg:154.57ms
step:99/1480 train_time:13745ms step_avg:154.44ms
step:100/1480 train_time:13888ms step_avg:154.31ms
step:101/1480 train_time:14031ms step_avg:154.19ms
step:102/1480 train_time:14171ms step_avg:154.04ms
step:103/1480 train_time:14312ms step_avg:153.90ms
step:104/1480 train_time:14456ms step_avg:153.79ms
step:105/1480 train_time:14601ms step_avg:153.70ms
step:106/1480 train_time:14744ms step_avg:153.58ms
step:107/1480 train_time:14886ms step_avg:153.47ms
step:108/1480 train_time:15030ms step_avg:153.36ms
step:109/1480 train_time:15171ms step_avg:153.25ms
step:110/1480 train_time:15313ms step_avg:153.13ms
step:111/1480 train_time:15459ms step_avg:153.06ms
step:112/1480 train_time:15605ms step_avg:152.99ms
step:113/1480 train_time:15752ms step_avg:152.93ms
step:114/1480 train_time:15897ms step_avg:152.86ms
step:115/1480 train_time:16044ms step_avg:152.80ms
step:116/1480 train_time:16189ms step_avg:152.73ms
step:117/1480 train_time:16334ms step_avg:152.65ms
step:118/1480 train_time:16481ms step_avg:152.60ms
step:119/1480 train_time:16626ms step_avg:152.53ms
step:120/1480 train_time:16771ms step_avg:152.47ms
step:121/1480 train_time:16917ms step_avg:152.41ms
step:122/1480 train_time:17063ms step_avg:152.35ms
step:123/1480 train_time:17209ms step_avg:152.29ms
step:124/1480 train_time:17355ms step_avg:152.24ms
step:125/1480 train_time:17501ms step_avg:152.18ms
step:125/1480 val_loss:4.4117 train_time:17566ms step_avg:152.74ms
step:126/1480 train_time:17657ms step_avg:152.21ms
step:127/1480 train_time:17804ms step_avg:152.18ms
step:128/1480 train_time:17951ms step_avg:152.13ms
step:129/1480 train_time:18096ms step_avg:152.06ms
step:130/1480 train_time:18241ms step_avg:152.01ms
step:131/1480 train_time:18387ms step_avg:151.96ms
step:132/1480 train_time:18532ms step_avg:151.90ms
step:133/1480 train_time:18677ms step_avg:151.84ms
step:134/1480 train_time:18826ms step_avg:151.82ms
step:135/1480 train_time:18971ms step_avg:151.77ms
step:136/1480 train_time:19117ms step_avg:151.72ms
step:137/1480 train_time:19263ms step_avg:151.68ms
step:138/1480 train_time:19410ms step_avg:151.64ms
step:139/1480 train_time:19554ms step_avg:151.58ms
step:140/1480 train_time:19700ms step_avg:151.54ms
step:141/1480 train_time:19850ms step_avg:151.53ms
step:142/1480 train_time:19996ms step_avg:151.48ms
step:143/1480 train_time:20141ms step_avg:151.43ms
step:144/1480 train_time:20287ms step_avg:151.39ms
step:145/1480 train_time:20432ms step_avg:151.35ms
step:146/1480 train_time:20577ms step_avg:151.30ms
step:147/1480 train_time:20723ms step_avg:151.26ms
step:148/1480 train_time:20869ms step_avg:151.22ms
step:149/1480 train_time:21015ms step_avg:151.19ms
step:150/1480 train_time:21161ms step_avg:151.15ms
step:151/1480 train_time:21307ms step_avg:151.12ms
step:152/1480 train_time:21452ms step_avg:151.07ms
step:153/1480 train_time:21598ms step_avg:151.03ms
step:154/1480 train_time:21744ms step_avg:151.00ms
step:155/1480 train_time:21891ms step_avg:150.97ms
step:156/1480 train_time:22035ms step_avg:150.93ms
step:157/1480 train_time:22181ms step_avg:150.89ms
step:158/1480 train_time:22328ms step_avg:150.86ms
step:159/1480 train_time:22472ms step_avg:150.82ms
step:160/1480 train_time:22618ms step_avg:150.78ms
step:161/1480 train_time:22764ms step_avg:150.75ms
step:162/1480 train_time:22910ms step_avg:150.72ms
step:163/1480 train_time:23056ms step_avg:150.69ms
step:164/1480 train_time:23202ms step_avg:150.67ms
step:165/1480 train_time:23349ms step_avg:150.64ms
step:166/1480 train_time:23494ms step_avg:150.60ms
step:167/1480 train_time:23640ms step_avg:150.57ms
step:168/1480 train_time:23786ms step_avg:150.55ms
step:169/1480 train_time:23932ms step_avg:150.52ms
step:170/1480 train_time:24077ms step_avg:150.48ms
step:171/1480 train_time:24224ms step_avg:150.46ms
step:172/1480 train_time:24370ms step_avg:150.43ms
step:173/1480 train_time:24515ms step_avg:150.40ms
step:174/1480 train_time:24662ms step_avg:150.38ms
step:175/1480 train_time:24808ms step_avg:150.35ms
step:176/1480 train_time:24953ms step_avg:150.32ms
step:177/1480 train_time:25100ms step_avg:150.30ms
step:178/1480 train_time:25247ms step_avg:150.28ms
step:179/1480 train_time:25393ms step_avg:150.26ms
step:180/1480 train_time:25538ms step_avg:150.22ms
step:181/1480 train_time:25685ms step_avg:150.20ms
step:182/1480 train_time:25831ms step_avg:150.18ms
step:183/1480 train_time:25977ms step_avg:150.16ms
step:184/1480 train_time:26124ms step_avg:150.14ms
step:185/1480 train_time:26270ms step_avg:150.11ms
step:186/1480 train_time:26416ms step_avg:150.09ms
step:187/1480 train_time:26563ms step_avg:150.07ms
step:188/1480 train_time:26709ms step_avg:150.05ms
step:189/1480 train_time:26881ms step_avg:150.17ms
step:190/1480 train_time:27000ms step_avg:150.00ms
step:191/1480 train_time:27147ms step_avg:149.98ms
step:192/1480 train_time:27292ms step_avg:149.95ms
step:193/1480 train_time:27437ms step_avg:149.93ms
step:194/1480 train_time:27583ms step_avg:149.91ms
step:195/1480 train_time:27730ms step_avg:149.89ms
step:196/1480 train_time:27874ms step_avg:149.86ms
step:197/1480 train_time:28020ms step_avg:149.84ms
step:198/1480 train_time:28166ms step_avg:149.82ms
step:199/1480 train_time:28312ms step_avg:149.80ms
step:200/1480 train_time:28457ms step_avg:149.78ms
step:201/1480 train_time:28604ms step_avg:149.76ms
step:202/1480 train_time:28750ms step_avg:149.74ms
step:203/1480 train_time:28895ms step_avg:149.72ms
step:204/1480 train_time:29041ms step_avg:149.69ms
step:205/1480 train_time:29187ms step_avg:149.68ms
step:206/1480 train_time:29332ms step_avg:149.65ms
step:207/1480 train_time:29477ms step_avg:149.63ms
step:208/1480 train_time:29623ms step_avg:149.61ms
step:209/1480 train_time:29769ms step_avg:149.59ms
step:210/1480 train_time:29916ms step_avg:149.58ms
step:211/1480 train_time:30062ms step_avg:149.56ms
step:212/1480 train_time:30208ms step_avg:149.55ms
step:213/1480 train_time:30353ms step_avg:149.52ms
step:214/1480 train_time:30499ms step_avg:149.50ms
step:215/1480 train_time:30646ms step_avg:149.49ms
step:216/1480 train_time:30792ms step_avg:149.47ms
step:217/1480 train_time:30937ms step_avg:149.46ms
step:218/1480 train_time:31084ms step_avg:149.44ms
step:219/1480 train_time:31231ms step_avg:149.43ms
step:220/1480 train_time:31376ms step_avg:149.41ms
step:221/1480 train_time:31928ms step_avg:151.32ms
step:222/1480 train_time:32039ms step_avg:151.13ms
step:223/1480 train_time:32186ms step_avg:151.11ms
step:224/1480 train_time:32334ms step_avg:151.09ms
step:225/1480 train_time:32482ms step_avg:151.08ms
step:226/1480 train_time:32630ms step_avg:151.07ms
step:227/1480 train_time:32778ms step_avg:151.05ms
step:228/1480 train_time:32928ms step_avg:151.04ms
step:229/1480 train_time:33076ms step_avg:151.03ms
step:230/1480 train_time:33225ms step_avg:151.02ms
step:231/1480 train_time:33372ms step_avg:151.01ms
step:232/1480 train_time:33522ms step_avg:151.00ms
step:233/1480 train_time:33670ms step_avg:150.99ms
step:234/1480 train_time:33818ms step_avg:150.97ms
step:235/1480 train_time:33967ms step_avg:150.96ms
step:236/1480 train_time:34115ms step_avg:150.95ms
step:237/1480 train_time:34264ms step_avg:150.94ms
step:238/1480 train_time:34412ms step_avg:150.93ms
step:239/1480 train_time:34560ms step_avg:150.92ms
step:240/1480 train_time:34709ms step_avg:150.91ms
step:241/1480 train_time:34856ms step_avg:150.89ms
step:242/1480 train_time:35006ms step_avg:150.89ms
step:243/1480 train_time:35153ms step_avg:150.87ms
step:244/1480 train_time:35302ms step_avg:150.86ms
step:245/1480 train_time:35451ms step_avg:150.86ms
step:246/1480 train_time:35600ms step_avg:150.85ms
step:247/1480 train_time:35748ms step_avg:150.84ms
step:248/1480 train_time:35896ms step_avg:150.82ms
step:249/1480 train_time:36046ms step_avg:150.82ms
step:250/1480 train_time:36194ms step_avg:150.81ms
step:250/1480 val_loss:3.9867 train_time:36261ms step_avg:151.09ms
step:251/1480 train_time:36352ms step_avg:150.84ms
step:252/1480 train_time:36502ms step_avg:150.83ms
step:253/1480 train_time:36651ms step_avg:150.83ms
step:254/1480 train_time:36797ms step_avg:150.81ms
step:255/1480 train_time:36945ms step_avg:150.80ms
step:256/1480 train_time:37094ms step_avg:150.79ms
step:257/1480 train_time:37242ms step_avg:150.78ms
step:258/1480 train_time:37391ms step_avg:150.77ms
step:259/1480 train_time:37540ms step_avg:150.76ms
step:260/1480 train_time:37689ms step_avg:150.76ms
step:261/1480 train_time:37836ms step_avg:150.74ms
step:262/1480 train_time:37985ms step_avg:150.73ms
step:263/1480 train_time:38134ms step_avg:150.73ms
step:264/1480 train_time:38282ms step_avg:150.71ms
step:265/1480 train_time:38431ms step_avg:150.71ms
step:266/1480 train_time:38579ms step_avg:150.70ms
step:267/1480 train_time:38728ms step_avg:150.69ms
step:268/1480 train_time:38876ms step_avg:150.68ms
step:269/1480 train_time:39025ms step_avg:150.67ms
step:270/1480 train_time:39174ms step_avg:150.67ms
step:271/1480 train_time:39322ms step_avg:150.66ms
step:272/1480 train_time:39472ms step_avg:150.66ms
step:273/1480 train_time:39621ms step_avg:150.65ms
step:274/1480 train_time:39770ms step_avg:150.64ms
step:275/1480 train_time:39918ms step_avg:150.63ms
step:276/1480 train_time:40066ms step_avg:150.62ms
step:277/1480 train_time:40215ms step_avg:150.62ms
step:278/1480 train_time:40364ms step_avg:150.61ms
step:279/1480 train_time:40514ms step_avg:150.61ms
step:280/1480 train_time:40661ms step_avg:150.60ms
step:281/1480 train_time:40810ms step_avg:150.59ms
step:282/1480 train_time:40958ms step_avg:150.58ms
step:283/1480 train_time:41107ms step_avg:150.58ms
step:284/1480 train_time:41256ms step_avg:150.57ms
step:285/1480 train_time:41405ms step_avg:150.56ms
step:286/1480 train_time:41554ms step_avg:150.56ms
step:287/1480 train_time:41703ms step_avg:150.55ms
step:288/1480 train_time:41853ms step_avg:150.55ms
step:289/1480 train_time:42000ms step_avg:150.54ms
step:290/1480 train_time:42149ms step_avg:150.53ms
step:291/1480 train_time:42297ms step_avg:150.52ms
step:292/1480 train_time:42446ms step_avg:150.52ms
step:293/1480 train_time:42595ms step_avg:150.51ms
step:294/1480 train_time:42743ms step_avg:150.50ms
step:295/1480 train_time:42892ms step_avg:150.50ms
step:296/1480 train_time:43040ms step_avg:150.49ms
step:297/1480 train_time:43188ms step_avg:150.48ms
step:298/1480 train_time:43337ms step_avg:150.48ms
step:299/1480 train_time:43487ms step_avg:150.47ms
step:300/1480 train_time:43636ms step_avg:150.47ms
step:301/1480 train_time:43786ms step_avg:150.47ms
step:302/1480 train_time:43933ms step_avg:150.46ms
step:303/1480 train_time:44082ms step_avg:150.45ms
step:304/1480 train_time:44230ms step_avg:150.44ms
step:305/1480 train_time:44378ms step_avg:150.43ms
step:306/1480 train_time:44527ms step_avg:150.43ms
step:307/1480 train_time:44676ms step_avg:150.42ms
step:308/1480 train_time:44824ms step_avg:150.42ms
step:309/1480 train_time:44974ms step_avg:150.41ms
step:310/1480 train_time:45123ms step_avg:150.41ms
step:311/1480 train_time:45272ms step_avg:150.41ms
step:312/1480 train_time:45420ms step_avg:150.40ms
step:313/1480 train_time:45570ms step_avg:150.40ms
step:314/1480 train_time:45718ms step_avg:150.39ms
step:315/1480 train_time:45868ms step_avg:150.39ms
step:316/1480 train_time:46016ms step_avg:150.38ms
step:317/1480 train_time:46165ms step_avg:150.37ms
step:318/1480 train_time:46314ms step_avg:150.37ms
step:319/1480 train_time:46462ms step_avg:150.36ms
step:320/1480 train_time:46610ms step_avg:150.36ms
step:321/1480 train_time:46757ms step_avg:150.34ms
step:322/1480 train_time:46906ms step_avg:150.34ms
step:323/1480 train_time:47055ms step_avg:150.34ms
step:324/1480 train_time:47205ms step_avg:150.34ms
step:325/1480 train_time:47354ms step_avg:150.33ms
step:326/1480 train_time:47503ms step_avg:150.33ms
step:327/1480 train_time:47653ms step_avg:150.32ms
step:328/1480 train_time:47800ms step_avg:150.31ms
step:329/1480 train_time:47949ms step_avg:150.31ms
step:330/1480 train_time:48098ms step_avg:150.31ms
step:331/1480 train_time:48251ms step_avg:150.31ms
step:332/1480 train_time:48401ms step_avg:150.31ms
step:333/1480 train_time:48552ms step_avg:150.32ms
step:334/1480 train_time:48703ms step_avg:150.32ms
step:335/1480 train_time:48855ms step_avg:150.32ms
step:336/1480 train_time:49005ms step_avg:150.32ms
step:337/1480 train_time:49157ms step_avg:150.33ms
step:338/1480 train_time:49308ms step_avg:150.33ms
step:339/1480 train_time:49458ms step_avg:150.33ms
step:340/1480 train_time:49610ms step_avg:150.33ms
step:341/1480 train_time:49759ms step_avg:150.33ms
step:342/1480 train_time:49910ms step_avg:150.33ms
step:343/1480 train_time:50059ms step_avg:150.33ms
step:344/1480 train_time:50211ms step_avg:150.33ms
step:345/1480 train_time:50361ms step_avg:150.33ms
step:346/1480 train_time:50513ms step_avg:150.34ms
step:347/1480 train_time:50664ms step_avg:150.34ms
step:348/1480 train_time:50816ms step_avg:150.34ms
step:349/1480 train_time:50965ms step_avg:150.34ms
step:350/1480 train_time:51116ms step_avg:150.34ms
step:351/1480 train_time:51268ms step_avg:150.34ms
step:352/1480 train_time:51418ms step_avg:150.35ms
step:353/1480 train_time:51570ms step_avg:150.35ms
step:354/1480 train_time:51720ms step_avg:150.35ms
step:355/1480 train_time:51873ms step_avg:150.36ms
step:356/1480 train_time:52023ms step_avg:150.35ms
step:357/1480 train_time:52174ms step_avg:150.36ms
step:358/1480 train_time:52324ms step_avg:150.36ms
step:359/1480 train_time:52475ms step_avg:150.36ms
step:360/1480 train_time:52627ms step_avg:150.36ms
step:361/1480 train_time:52779ms step_avg:150.37ms
step:362/1480 train_time:52931ms step_avg:150.37ms
step:363/1480 train_time:53081ms step_avg:150.37ms
step:364/1480 train_time:53232ms step_avg:150.37ms
step:365/1480 train_time:53382ms step_avg:150.37ms
step:366/1480 train_time:53533ms step_avg:150.37ms
step:367/1480 train_time:53684ms step_avg:150.38ms
step:368/1480 train_time:53835ms step_avg:150.38ms
step:369/1480 train_time:53986ms step_avg:150.38ms
step:370/1480 train_time:54137ms step_avg:150.38ms
step:371/1480 train_time:54288ms step_avg:150.38ms
step:372/1480 train_time:54438ms step_avg:150.38ms
step:373/1480 train_time:54589ms step_avg:150.38ms
step:374/1480 train_time:54739ms step_avg:150.38ms
step:375/1480 train_time:54891ms step_avg:150.39ms
step:375/1480 val_loss:3.8036 train_time:54959ms step_avg:150.57ms
step:376/1480 train_time:55050ms step_avg:150.41ms
step:377/1480 train_time:55199ms step_avg:150.41ms
step:378/1480 train_time:55350ms step_avg:150.41ms
step:379/1480 train_time:55522ms step_avg:150.47ms
step:380/1480 train_time:55650ms step_avg:150.41ms
step:381/1480 train_time:55801ms step_avg:150.41ms
step:382/1480 train_time:55952ms step_avg:150.41ms
step:383/1480 train_time:56104ms step_avg:150.41ms
step:384/1480 train_time:56257ms step_avg:150.42ms
step:385/1480 train_time:56408ms step_avg:150.42ms
step:386/1480 train_time:56559ms step_avg:150.42ms
step:387/1480 train_time:56709ms step_avg:150.42ms
step:388/1480 train_time:56860ms step_avg:150.42ms
step:389/1480 train_time:57010ms step_avg:150.42ms
step:390/1480 train_time:57162ms step_avg:150.43ms
step:391/1480 train_time:57314ms step_avg:150.43ms
step:392/1480 train_time:57464ms step_avg:150.43ms
step:393/1480 train_time:57615ms step_avg:150.43ms
step:394/1480 train_time:57765ms step_avg:150.43ms
step:395/1480 train_time:57917ms step_avg:150.43ms
step:396/1480 train_time:58067ms step_avg:150.43ms
step:397/1480 train_time:58219ms step_avg:150.44ms
step:398/1480 train_time:58369ms step_avg:150.44ms
step:399/1480 train_time:58521ms step_avg:150.44ms
step:400/1480 train_time:58671ms step_avg:150.44ms
step:401/1480 train_time:58822ms step_avg:150.44ms
step:402/1480 train_time:58972ms step_avg:150.44ms
step:403/1480 train_time:59123ms step_avg:150.44ms
step:404/1480 train_time:59275ms step_avg:150.44ms
step:405/1480 train_time:59426ms step_avg:150.44ms
step:406/1480 train_time:59577ms step_avg:150.45ms
step:407/1480 train_time:59727ms step_avg:150.44ms
step:408/1480 train_time:59878ms step_avg:150.45ms
step:409/1480 train_time:60029ms step_avg:150.45ms
step:410/1480 train_time:60180ms step_avg:150.45ms
step:411/1480 train_time:60331ms step_avg:150.45ms
step:412/1480 train_time:60482ms step_avg:150.45ms
step:413/1480 train_time:60633ms step_avg:150.45ms
step:414/1480 train_time:60784ms step_avg:150.46ms
step:415/1480 train_time:60936ms step_avg:150.46ms
step:416/1480 train_time:61086ms step_avg:150.46ms
step:417/1480 train_time:61238ms step_avg:150.46ms
step:418/1480 train_time:61389ms step_avg:150.46ms
step:419/1480 train_time:61540ms step_avg:150.46ms
step:420/1480 train_time:61690ms step_avg:150.46ms
step:421/1480 train_time:61842ms step_avg:150.47ms
step:422/1480 train_time:61992ms step_avg:150.47ms
step:423/1480 train_time:62143ms step_avg:150.47ms
step:424/1480 train_time:62293ms step_avg:150.47ms
step:425/1480 train_time:62444ms step_avg:150.47ms
step:426/1480 train_time:62595ms step_avg:150.47ms
step:427/1480 train_time:62746ms step_avg:150.47ms
step:428/1480 train_time:62897ms step_avg:150.47ms
step:429/1480 train_time:63048ms step_avg:150.47ms
step:430/1480 train_time:63199ms step_avg:150.47ms
step:431/1480 train_time:63351ms step_avg:150.48ms
step:432/1480 train_time:63502ms step_avg:150.48ms
step:433/1480 train_time:63653ms step_avg:150.48ms
step:434/1480 train_time:63804ms step_avg:150.48ms
step:435/1480 train_time:63956ms step_avg:150.48ms
step:436/1480 train_time:64106ms step_avg:150.48ms
step:437/1480 train_time:64257ms step_avg:150.49ms
step:438/1480 train_time:64408ms step_avg:150.49ms
step:439/1480 train_time:64560ms step_avg:150.49ms
step:440/1480 train_time:64711ms step_avg:150.49ms
step:441/1480 train_time:64864ms step_avg:150.50ms
step:442/1480 train_time:65016ms step_avg:150.50ms
step:443/1480 train_time:65169ms step_avg:150.51ms
step:444/1480 train_time:65323ms step_avg:150.51ms
step:445/1480 train_time:65475ms step_avg:150.52ms
step:446/1480 train_time:65627ms step_avg:150.52ms
step:447/1480 train_time:65781ms step_avg:150.53ms
step:448/1480 train_time:65934ms step_avg:150.53ms
step:449/1480 train_time:66087ms step_avg:150.54ms
step:450/1480 train_time:66241ms step_avg:150.55ms
step:451/1480 train_time:66393ms step_avg:150.55ms
step:452/1480 train_time:66546ms step_avg:150.56ms
step:453/1480 train_time:66698ms step_avg:150.56ms
step:454/1480 train_time:66851ms step_avg:150.57ms
step:455/1480 train_time:67004ms step_avg:150.57ms
step:456/1480 train_time:67157ms step_avg:150.58ms
step:457/1480 train_time:67310ms step_avg:150.58ms
step:458/1480 train_time:67463ms step_avg:150.59ms
step:459/1480 train_time:67617ms step_avg:150.59ms
step:460/1480 train_time:67769ms step_avg:150.60ms
step:461/1480 train_time:67922ms step_avg:150.60ms
step:462/1480 train_time:68075ms step_avg:150.61ms
step:463/1480 train_time:68228ms step_avg:150.61ms
step:464/1480 train_time:68382ms step_avg:150.62ms
step:465/1480 train_time:68535ms step_avg:150.63ms
step:466/1480 train_time:68687ms step_avg:150.63ms
step:467/1480 train_time:68841ms step_avg:150.64ms
step:468/1480 train_time:68994ms step_avg:150.64ms
step:469/1480 train_time:69147ms step_avg:150.65ms
step:470/1480 train_time:69299ms step_avg:150.65ms
step:471/1480 train_time:69454ms step_avg:150.66ms
step:472/1480 train_time:69606ms step_avg:150.66ms
step:473/1480 train_time:69760ms step_avg:150.67ms
step:474/1480 train_time:69912ms step_avg:150.67ms
step:475/1480 train_time:70065ms step_avg:150.68ms
step:476/1480 train_time:70219ms step_avg:150.68ms
step:477/1480 train_time:70371ms step_avg:150.69ms
step:478/1480 train_time:70523ms step_avg:150.69ms
step:479/1480 train_time:70676ms step_avg:150.70ms
step:480/1480 train_time:70829ms step_avg:150.70ms
step:481/1480 train_time:70982ms step_avg:150.71ms
step:482/1480 train_time:71137ms step_avg:150.71ms
step:483/1480 train_time:71289ms step_avg:150.72ms
step:484/1480 train_time:71442ms step_avg:150.72ms
step:485/1480 train_time:71594ms step_avg:150.72ms
step:486/1480 train_time:71746ms step_avg:150.73ms
step:487/1480 train_time:71901ms step_avg:150.73ms
step:488/1480 train_time:72055ms step_avg:150.74ms
step:489/1480 train_time:72208ms step_avg:150.75ms
step:490/1480 train_time:72361ms step_avg:150.75ms
step:491/1480 train_time:72514ms step_avg:150.76ms
step:492/1480 train_time:72665ms step_avg:150.76ms
step:493/1480 train_time:72819ms step_avg:150.76ms
step:494/1480 train_time:72970ms step_avg:150.76ms
step:495/1480 train_time:73123ms step_avg:150.77ms
step:496/1480 train_time:73277ms step_avg:150.78ms
step:497/1480 train_time:73429ms step_avg:150.78ms
step:498/1480 train_time:73582ms step_avg:150.78ms
step:499/1480 train_time:73736ms step_avg:150.79ms
step:500/1480 train_time:73889ms step_avg:150.79ms
step:500/1480 val_loss:3.6837 train_time:73958ms step_avg:150.93ms
step:501/1480 train_time:74048ms step_avg:150.81ms
step:502/1480 train_time:74202ms step_avg:150.82ms
step:503/1480 train_time:74356ms step_avg:150.82ms
step:504/1480 train_time:74507ms step_avg:150.82ms
step:505/1480 train_time:74659ms step_avg:150.83ms
step:506/1480 train_time:74812ms step_avg:150.83ms
step:507/1480 train_time:74963ms step_avg:150.83ms
step:508/1480 train_time:75118ms step_avg:150.84ms
step:509/1480 train_time:75271ms step_avg:150.84ms
step:510/1480 train_time:75424ms step_avg:150.85ms
step:511/1480 train_time:75577ms step_avg:150.85ms
step:512/1480 train_time:75732ms step_avg:150.86ms
step:513/1480 train_time:75884ms step_avg:150.86ms
step:514/1480 train_time:76037ms step_avg:150.87ms
step:515/1480 train_time:76190ms step_avg:150.87ms
step:516/1480 train_time:76343ms step_avg:150.88ms
step:517/1480 train_time:76497ms step_avg:150.88ms
step:518/1480 train_time:76651ms step_avg:150.89ms
step:519/1480 train_time:76805ms step_avg:150.89ms
step:520/1480 train_time:76958ms step_avg:150.90ms
step:521/1480 train_time:77111ms step_avg:150.90ms
step:522/1480 train_time:77263ms step_avg:150.90ms
step:523/1480 train_time:77416ms step_avg:150.91ms
step:524/1480 train_time:77569ms step_avg:150.91ms
step:525/1480 train_time:77723ms step_avg:150.92ms
step:526/1480 train_time:77877ms step_avg:150.92ms
step:527/1480 train_time:78031ms step_avg:150.93ms
step:528/1480 train_time:78184ms step_avg:150.93ms
step:529/1480 train_time:78337ms step_avg:150.94ms
step:530/1480 train_time:78491ms step_avg:150.94ms
step:531/1480 train_time:78643ms step_avg:150.95ms
step:532/1480 train_time:78796ms step_avg:150.95ms
step:533/1480 train_time:78949ms step_avg:150.96ms
step:534/1480 train_time:79103ms step_avg:150.96ms
step:535/1480 train_time:79257ms step_avg:150.96ms
step:536/1480 train_time:79409ms step_avg:150.97ms
step:537/1480 train_time:79563ms step_avg:150.97ms
step:538/1480 train_time:79716ms step_avg:150.98ms
step:539/1480 train_time:79869ms step_avg:150.98ms
step:540/1480 train_time:80024ms step_avg:150.99ms
step:541/1480 train_time:80177ms step_avg:150.99ms
step:542/1480 train_time:80330ms step_avg:151.00ms
step:543/1480 train_time:80482ms step_avg:151.00ms
step:544/1480 train_time:80635ms step_avg:151.00ms
step:545/1480 train_time:80787ms step_avg:151.00ms
step:546/1480 train_time:80939ms step_avg:151.01ms
step:547/1480 train_time:81093ms step_avg:151.01ms
step:548/1480 train_time:81248ms step_avg:151.02ms
step:549/1480 train_time:81400ms step_avg:151.02ms
step:550/1480 train_time:81556ms step_avg:151.03ms
step:551/1480 train_time:81710ms step_avg:151.03ms
step:552/1480 train_time:81865ms step_avg:151.04ms
step:553/1480 train_time:82020ms step_avg:151.05ms
step:554/1480 train_time:82174ms step_avg:151.06ms
step:555/1480 train_time:82329ms step_avg:151.06ms
step:556/1480 train_time:82483ms step_avg:151.07ms
step:557/1480 train_time:82638ms step_avg:151.08ms
step:558/1480 train_time:82792ms step_avg:151.08ms
step:559/1480 train_time:82946ms step_avg:151.09ms
step:560/1480 train_time:83101ms step_avg:151.09ms
step:561/1480 train_time:83256ms step_avg:151.10ms
step:562/1480 train_time:83410ms step_avg:151.11ms
step:563/1480 train_time:83564ms step_avg:151.11ms
step:564/1480 train_time:83719ms step_avg:151.12ms
step:565/1480 train_time:83873ms step_avg:151.12ms
step:566/1480 train_time:84028ms step_avg:151.13ms
step:567/1480 train_time:84183ms step_avg:151.14ms
step:568/1480 train_time:84338ms step_avg:151.14ms
step:569/1480 train_time:84515ms step_avg:151.19ms
step:570/1480 train_time:84646ms step_avg:151.15ms
step:571/1480 train_time:84800ms step_avg:151.16ms
step:572/1480 train_time:84955ms step_avg:151.17ms
step:573/1480 train_time:85109ms step_avg:151.17ms
step:574/1480 train_time:85267ms step_avg:151.18ms
step:575/1480 train_time:85422ms step_avg:151.19ms
step:576/1480 train_time:85577ms step_avg:151.20ms
step:577/1480 train_time:85731ms step_avg:151.20ms
step:578/1480 train_time:85885ms step_avg:151.21ms
step:579/1480 train_time:86040ms step_avg:151.21ms
step:580/1480 train_time:86194ms step_avg:151.22ms
step:581/1480 train_time:86351ms step_avg:151.23ms
step:582/1480 train_time:86506ms step_avg:151.23ms
step:583/1480 train_time:86661ms step_avg:151.24ms
step:584/1480 train_time:86816ms step_avg:151.25ms
step:585/1480 train_time:86970ms step_avg:151.25ms
step:586/1480 train_time:87124ms step_avg:151.26ms
step:587/1480 train_time:87280ms step_avg:151.26ms
step:588/1480 train_time:87435ms step_avg:151.27ms
step:589/1480 train_time:87589ms step_avg:151.28ms
step:590/1480 train_time:87744ms step_avg:151.28ms
step:591/1480 train_time:87899ms step_avg:151.29ms
step:592/1480 train_time:88055ms step_avg:151.30ms
step:593/1480 train_time:88211ms step_avg:151.31ms
step:594/1480 train_time:88367ms step_avg:151.31ms
step:595/1480 train_time:88523ms step_avg:151.32ms
step:596/1480 train_time:88679ms step_avg:151.33ms
step:597/1480 train_time:88835ms step_avg:151.34ms
step:598/1480 train_time:88990ms step_avg:151.34ms
step:599/1480 train_time:89144ms step_avg:151.35ms
step:600/1480 train_time:89298ms step_avg:151.35ms
step:601/1480 train_time:89455ms step_avg:151.36ms
step:602/1480 train_time:89610ms step_avg:151.37ms
step:603/1480 train_time:89764ms step_avg:151.37ms
step:604/1480 train_time:89919ms step_avg:151.38ms
step:605/1480 train_time:90074ms step_avg:151.39ms
step:606/1480 train_time:90230ms step_avg:151.39ms
step:607/1480 train_time:90386ms step_avg:151.40ms
step:608/1480 train_time:90540ms step_avg:151.40ms
step:609/1480 train_time:90694ms step_avg:151.41ms
step:610/1480 train_time:90849ms step_avg:151.42ms
step:611/1480 train_time:91004ms step_avg:151.42ms
step:612/1480 train_time:91158ms step_avg:151.43ms
step:613/1480 train_time:91313ms step_avg:151.43ms
step:614/1480 train_time:91468ms step_avg:151.44ms
step:615/1480 train_time:91623ms step_avg:151.44ms
step:616/1480 train_time:91777ms step_avg:151.45ms
step:617/1480 train_time:91933ms step_avg:151.46ms
step:618/1480 train_time:92088ms step_avg:151.46ms
step:619/1480 train_time:92243ms step_avg:151.47ms
step:620/1480 train_time:92398ms step_avg:151.47ms
step:621/1480 train_time:92554ms step_avg:151.48ms
step:622/1480 train_time:92709ms step_avg:151.49ms
step:623/1480 train_time:92864ms step_avg:151.49ms
step:624/1480 train_time:93019ms step_avg:151.50ms
step:625/1480 train_time:93174ms step_avg:151.50ms
step:625/1480 val_loss:3.6023 train_time:93245ms step_avg:151.62ms
step:626/1480 train_time:93335ms step_avg:151.52ms
step:627/1480 train_time:93491ms step_avg:151.52ms
step:628/1480 train_time:93646ms step_avg:151.53ms
step:629/1480 train_time:93800ms step_avg:151.53ms
step:630/1480 train_time:93954ms step_avg:151.54ms
step:631/1480 train_time:94108ms step_avg:151.54ms
step:632/1480 train_time:94262ms step_avg:151.55ms
step:633/1480 train_time:94418ms step_avg:151.55ms
step:634/1480 train_time:94573ms step_avg:151.56ms
step:635/1480 train_time:94727ms step_avg:151.56ms
step:636/1480 train_time:94882ms step_avg:151.57ms
step:637/1480 train_time:95038ms step_avg:151.58ms
step:638/1480 train_time:95193ms step_avg:151.58ms
step:639/1480 train_time:95347ms step_avg:151.58ms
step:640/1480 train_time:95501ms step_avg:151.59ms
step:641/1480 train_time:95656ms step_avg:151.59ms
step:642/1480 train_time:95811ms step_avg:151.60ms
step:643/1480 train_time:95967ms step_avg:151.61ms
step:644/1480 train_time:96122ms step_avg:151.61ms
step:645/1480 train_time:96276ms step_avg:151.62ms
step:646/1480 train_time:96430ms step_avg:151.62ms
step:647/1480 train_time:96585ms step_avg:151.62ms
step:648/1480 train_time:96740ms step_avg:151.63ms
step:649/1480 train_time:96895ms step_avg:151.64ms
step:650/1480 train_time:97052ms step_avg:151.64ms
step:651/1480 train_time:97207ms step_avg:151.65ms
step:652/1480 train_time:97362ms step_avg:151.65ms
step:653/1480 train_time:97518ms step_avg:151.66ms
step:654/1480 train_time:97674ms step_avg:151.67ms
step:655/1480 train_time:97828ms step_avg:151.67ms
step:656/1480 train_time:97983ms step_avg:151.68ms
step:657/1480 train_time:98137ms step_avg:151.68ms
step:658/1480 train_time:98292ms step_avg:151.69ms
step:659/1480 train_time:98447ms step_avg:151.69ms
step:660/1480 train_time:98604ms step_avg:151.70ms
step:661/1480 train_time:98760ms step_avg:151.71ms
step:662/1480 train_time:98917ms step_avg:151.71ms
step:663/1480 train_time:99073ms step_avg:151.72ms
step:664/1480 train_time:99229ms step_avg:151.73ms
step:665/1480 train_time:99385ms step_avg:151.73ms
step:666/1480 train_time:99541ms step_avg:151.74ms
step:667/1480 train_time:99698ms step_avg:151.75ms
step:668/1480 train_time:99855ms step_avg:151.76ms
step:669/1480 train_time:100014ms step_avg:151.77ms
step:670/1480 train_time:100171ms step_avg:151.77ms
step:671/1480 train_time:100326ms step_avg:151.78ms
step:672/1480 train_time:100482ms step_avg:151.78ms
step:673/1480 train_time:100639ms step_avg:151.79ms
step:674/1480 train_time:100796ms step_avg:151.80ms
step:675/1480 train_time:100953ms step_avg:151.81ms
step:676/1480 train_time:101111ms step_avg:151.82ms
step:677/1480 train_time:101268ms step_avg:151.83ms
step:678/1480 train_time:101423ms step_avg:151.83ms
step:679/1480 train_time:101579ms step_avg:151.84ms
step:680/1480 train_time:101738ms step_avg:151.85ms
step:681/1480 train_time:101892ms step_avg:151.85ms
step:682/1480 train_time:102049ms step_avg:151.86ms
step:683/1480 train_time:102205ms step_avg:151.86ms
step:684/1480 train_time:102363ms step_avg:151.87ms
step:685/1480 train_time:102520ms step_avg:151.88ms
step:686/1480 train_time:102677ms step_avg:151.89ms
step:687/1480 train_time:102834ms step_avg:151.90ms
step:688/1480 train_time:102992ms step_avg:151.91ms
step:689/1480 train_time:103150ms step_avg:151.91ms
step:690/1480 train_time:103307ms step_avg:151.92ms
step:691/1480 train_time:103464ms step_avg:151.93ms
step:692/1480 train_time:103621ms step_avg:151.94ms
step:693/1480 train_time:103777ms step_avg:151.94ms
step:694/1480 train_time:103934ms step_avg:151.95ms
step:695/1480 train_time:104090ms step_avg:151.96ms
step:696/1480 train_time:104245ms step_avg:151.96ms
step:697/1480 train_time:104401ms step_avg:151.97ms
step:698/1480 train_time:104556ms step_avg:151.97ms
step:699/1480 train_time:104714ms step_avg:151.98ms
step:700/1480 train_time:104871ms step_avg:151.99ms
step:701/1480 train_time:105027ms step_avg:151.99ms
step:702/1480 train_time:105182ms step_avg:152.00ms
step:703/1480 train_time:105339ms step_avg:152.00ms
step:704/1480 train_time:105495ms step_avg:152.01ms
step:705/1480 train_time:105650ms step_avg:152.01ms
step:706/1480 train_time:105808ms step_avg:152.02ms
step:707/1480 train_time:105963ms step_avg:152.03ms
step:708/1480 train_time:106119ms step_avg:152.03ms
step:709/1480 train_time:106275ms step_avg:152.04ms
step:710/1480 train_time:106431ms step_avg:152.04ms
step:711/1480 train_time:106586ms step_avg:152.05ms
step:712/1480 train_time:106742ms step_avg:152.05ms
step:713/1480 train_time:106900ms step_avg:152.06ms
step:714/1480 train_time:107057ms step_avg:152.07ms
step:715/1480 train_time:107212ms step_avg:152.07ms
step:716/1480 train_time:107367ms step_avg:152.08ms
step:717/1480 train_time:107523ms step_avg:152.08ms
step:718/1480 train_time:107678ms step_avg:152.09ms
step:719/1480 train_time:107835ms step_avg:152.09ms
step:720/1480 train_time:107994ms step_avg:152.10ms
step:721/1480 train_time:108152ms step_avg:152.11ms
step:722/1480 train_time:108308ms step_avg:152.12ms
step:723/1480 train_time:108465ms step_avg:152.12ms
step:724/1480 train_time:108621ms step_avg:152.13ms
step:725/1480 train_time:108777ms step_avg:152.14ms
step:726/1480 train_time:108934ms step_avg:152.14ms
step:727/1480 train_time:109092ms step_avg:152.15ms
step:728/1480 train_time:109248ms step_avg:152.16ms
step:729/1480 train_time:109404ms step_avg:152.16ms
step:730/1480 train_time:109562ms step_avg:152.17ms
step:731/1480 train_time:109719ms step_avg:152.18ms
step:732/1480 train_time:109874ms step_avg:152.18ms
step:733/1480 train_time:110031ms step_avg:152.19ms
step:734/1480 train_time:110187ms step_avg:152.19ms
step:735/1480 train_time:110343ms step_avg:152.20ms
step:736/1480 train_time:110500ms step_avg:152.20ms
step:737/1480 train_time:110655ms step_avg:152.21ms
step:738/1480 train_time:110811ms step_avg:152.21ms
step:739/1480 train_time:110967ms step_avg:152.22ms
step:740/1480 train_time:111124ms step_avg:152.22ms
step:741/1480 train_time:111281ms step_avg:152.23ms
step:742/1480 train_time:111438ms step_avg:152.24ms
step:743/1480 train_time:111593ms step_avg:152.24ms
step:744/1480 train_time:111749ms step_avg:152.25ms
step:745/1480 train_time:111907ms step_avg:152.25ms
step:746/1480 train_time:112063ms step_avg:152.26ms
step:747/1480 train_time:112220ms step_avg:152.27ms
step:748/1480 train_time:112378ms step_avg:152.27ms
step:749/1480 train_time:112536ms step_avg:152.28ms
step:750/1480 train_time:112692ms step_avg:152.29ms
step:750/1480 val_loss:3.5468 train_time:112765ms step_avg:152.38ms
step:751/1480 train_time:112861ms step_avg:152.31ms
step:752/1480 train_time:113012ms step_avg:152.31ms
step:753/1480 train_time:113168ms step_avg:152.31ms
step:754/1480 train_time:113324ms step_avg:152.32ms
step:755/1480 train_time:113479ms step_avg:152.32ms
step:756/1480 train_time:113635ms step_avg:152.33ms
step:757/1480 train_time:113794ms step_avg:152.33ms
step:758/1480 train_time:113950ms step_avg:152.34ms
step:759/1480 train_time:114126ms step_avg:152.37ms
step:760/1480 train_time:114267ms step_avg:152.36ms
step:761/1480 train_time:114422ms step_avg:152.36ms
step:762/1480 train_time:114578ms step_avg:152.36ms
step:763/1480 train_time:114734ms step_avg:152.37ms
step:764/1480 train_time:114891ms step_avg:152.38ms
step:765/1480 train_time:115048ms step_avg:152.38ms
step:766/1480 train_time:115207ms step_avg:152.39ms
step:767/1480 train_time:115364ms step_avg:152.40ms
step:768/1480 train_time:115520ms step_avg:152.40ms
step:769/1480 train_time:115677ms step_avg:152.41ms
step:770/1480 train_time:115835ms step_avg:152.41ms
step:771/1480 train_time:115994ms step_avg:152.42ms
step:772/1480 train_time:116150ms step_avg:152.43ms
step:773/1480 train_time:116309ms step_avg:152.44ms
step:774/1480 train_time:116467ms step_avg:152.44ms
step:775/1480 train_time:116624ms step_avg:152.45ms
step:776/1480 train_time:116783ms step_avg:152.46ms
step:777/1480 train_time:116943ms step_avg:152.47ms
step:778/1480 train_time:117101ms step_avg:152.48ms
step:779/1480 train_time:117258ms step_avg:152.48ms
step:780/1480 train_time:117416ms step_avg:152.49ms
step:781/1480 train_time:117573ms step_avg:152.49ms
step:782/1480 train_time:117731ms step_avg:152.50ms
step:783/1480 train_time:117889ms step_avg:152.51ms
step:784/1480 train_time:118047ms step_avg:152.52ms
step:785/1480 train_time:118207ms step_avg:152.52ms
step:786/1480 train_time:118365ms step_avg:152.53ms
step:787/1480 train_time:118523ms step_avg:152.54ms
step:788/1480 train_time:118682ms step_avg:152.55ms
step:789/1480 train_time:118838ms step_avg:152.55ms
step:790/1480 train_time:118996ms step_avg:152.56ms
step:791/1480 train_time:119156ms step_avg:152.57ms
step:792/1480 train_time:119315ms step_avg:152.58ms
step:793/1480 train_time:119472ms step_avg:152.58ms
step:794/1480 train_time:119631ms step_avg:152.59ms
step:795/1480 train_time:119791ms step_avg:152.60ms
step:796/1480 train_time:119952ms step_avg:152.61ms
step:797/1480 train_time:120113ms step_avg:152.62ms
step:798/1480 train_time:120270ms step_avg:152.63ms
step:799/1480 train_time:120433ms step_avg:152.64ms
step:800/1480 train_time:120591ms step_avg:152.65ms
step:801/1480 train_time:120749ms step_avg:152.65ms
step:802/1480 train_time:120909ms step_avg:152.66ms
step:803/1480 train_time:121068ms step_avg:152.67ms
step:804/1480 train_time:121225ms step_avg:152.68ms
step:805/1480 train_time:121384ms step_avg:152.68ms
step:806/1480 train_time:121541ms step_avg:152.69ms
step:807/1480 train_time:121698ms step_avg:152.70ms
step:808/1480 train_time:121855ms step_avg:152.70ms
step:809/1480 train_time:122012ms step_avg:152.71ms
step:810/1480 train_time:122169ms step_avg:152.71ms
step:811/1480 train_time:122325ms step_avg:152.72ms
step:812/1480 train_time:122482ms step_avg:152.72ms
step:813/1480 train_time:122639ms step_avg:152.73ms
step:814/1480 train_time:122797ms step_avg:152.73ms
step:815/1480 train_time:122954ms step_avg:152.74ms
step:816/1480 train_time:123114ms step_avg:152.75ms
step:817/1480 train_time:123272ms step_avg:152.75ms
step:818/1480 train_time:123429ms step_avg:152.76ms
step:819/1480 train_time:123588ms step_avg:152.77ms
step:820/1480 train_time:123747ms step_avg:152.77ms
step:821/1480 train_time:123904ms step_avg:152.78ms
step:822/1480 train_time:124061ms step_avg:152.78ms
step:823/1480 train_time:124219ms step_avg:152.79ms
step:824/1480 train_time:124376ms step_avg:152.80ms
step:825/1480 train_time:124535ms step_avg:152.80ms
step:826/1480 train_time:124694ms step_avg:152.81ms
step:827/1480 train_time:124852ms step_avg:152.82ms
step:828/1480 train_time:125013ms step_avg:152.83ms
step:829/1480 train_time:125172ms step_avg:152.83ms
step:830/1480 train_time:125331ms step_avg:152.84ms
step:831/1480 train_time:125490ms step_avg:152.85ms
step:832/1480 train_time:125648ms step_avg:152.86ms
step:833/1480 train_time:125804ms step_avg:152.86ms
step:834/1480 train_time:125965ms step_avg:152.87ms
step:835/1480 train_time:126122ms step_avg:152.87ms
step:836/1480 train_time:126281ms step_avg:152.88ms
step:837/1480 train_time:126438ms step_avg:152.89ms
step:838/1480 train_time:126596ms step_avg:152.89ms
step:839/1480 train_time:126752ms step_avg:152.90ms
step:840/1480 train_time:126912ms step_avg:152.91ms
step:841/1480 train_time:127069ms step_avg:152.91ms
step:842/1480 train_time:127227ms step_avg:152.92ms
step:843/1480 train_time:127386ms step_avg:152.92ms
step:844/1480 train_time:127542ms step_avg:152.93ms
step:845/1480 train_time:127699ms step_avg:152.93ms
step:846/1480 train_time:127857ms step_avg:152.94ms
step:847/1480 train_time:128016ms step_avg:152.95ms
step:848/1480 train_time:128173ms step_avg:152.95ms
step:849/1480 train_time:128331ms step_avg:152.96ms
step:850/1480 train_time:128490ms step_avg:152.96ms
step:851/1480 train_time:128650ms step_avg:152.97ms
step:852/1480 train_time:128808ms step_avg:152.98ms
step:853/1480 train_time:128967ms step_avg:152.99ms
step:854/1480 train_time:129124ms step_avg:152.99ms
step:855/1480 train_time:129281ms step_avg:152.99ms
step:856/1480 train_time:129437ms step_avg:153.00ms
step:857/1480 train_time:129595ms step_avg:153.01ms
step:858/1480 train_time:129755ms step_avg:153.01ms
step:859/1480 train_time:129913ms step_avg:153.02ms
step:860/1480 train_time:130070ms step_avg:153.02ms
step:861/1480 train_time:130230ms step_avg:153.03ms
step:862/1480 train_time:130393ms step_avg:153.04ms
step:863/1480 train_time:130552ms step_avg:153.05ms
step:864/1480 train_time:130711ms step_avg:153.06ms
step:865/1480 train_time:130869ms step_avg:153.06ms
step:866/1480 train_time:131027ms step_avg:153.07ms
step:867/1480 train_time:131187ms step_avg:153.08ms
step:868/1480 train_time:131345ms step_avg:153.08ms
step:869/1480 train_time:131502ms step_avg:153.09ms
step:870/1480 train_time:131660ms step_avg:153.09ms
step:871/1480 train_time:131817ms step_avg:153.10ms
step:872/1480 train_time:131975ms step_avg:153.10ms
step:873/1480 train_time:132132ms step_avg:153.11ms
step:874/1480 train_time:132293ms step_avg:153.12ms
step:875/1480 train_time:132452ms step_avg:153.12ms
step:875/1480 val_loss:3.5011 train_time:132524ms step_avg:153.21ms
step:876/1480 train_time:132615ms step_avg:153.14ms
step:877/1480 train_time:132772ms step_avg:153.14ms
step:878/1480 train_time:132930ms step_avg:153.14ms
step:879/1480 train_time:133088ms step_avg:153.15ms
step:880/1480 train_time:133246ms step_avg:153.16ms
step:881/1480 train_time:133404ms step_avg:153.16ms
step:882/1480 train_time:133562ms step_avg:153.17ms
step:883/1480 train_time:133720ms step_avg:153.17ms
step:884/1480 train_time:133881ms step_avg:153.18ms
step:885/1480 train_time:134040ms step_avg:153.19ms
step:886/1480 train_time:134202ms step_avg:153.20ms
step:887/1480 train_time:134361ms step_avg:153.20ms
step:888/1480 train_time:134523ms step_avg:153.22ms
step:889/1480 train_time:134683ms step_avg:153.22ms
step:890/1480 train_time:134841ms step_avg:153.23ms
step:891/1480 train_time:135000ms step_avg:153.23ms
step:892/1480 train_time:135159ms step_avg:153.24ms
step:893/1480 train_time:135317ms step_avg:153.25ms
step:894/1480 train_time:135477ms step_avg:153.25ms
step:895/1480 train_time:135640ms step_avg:153.27ms
step:896/1480 train_time:135799ms step_avg:153.27ms
step:897/1480 train_time:135961ms step_avg:153.28ms
step:898/1480 train_time:136120ms step_avg:153.29ms
step:899/1480 train_time:136279ms step_avg:153.29ms
step:900/1480 train_time:136437ms step_avg:153.30ms
step:901/1480 train_time:136599ms step_avg:153.31ms
step:902/1480 train_time:136757ms step_avg:153.32ms
step:903/1480 train_time:136918ms step_avg:153.32ms
step:904/1480 train_time:137076ms step_avg:153.33ms
step:905/1480 train_time:137235ms step_avg:153.34ms
step:906/1480 train_time:137393ms step_avg:153.34ms
step:907/1480 train_time:137556ms step_avg:153.35ms
step:908/1480 train_time:137714ms step_avg:153.36ms
step:909/1480 train_time:137875ms step_avg:153.36ms
step:910/1480 train_time:138040ms step_avg:153.38ms
step:911/1480 train_time:138198ms step_avg:153.38ms
step:912/1480 train_time:138358ms step_avg:153.39ms
step:913/1480 train_time:138519ms step_avg:153.40ms
step:914/1480 train_time:138679ms step_avg:153.41ms
step:915/1480 train_time:138843ms step_avg:153.42ms
step:916/1480 train_time:139003ms step_avg:153.43ms
step:917/1480 train_time:139163ms step_avg:153.43ms
step:918/1480 train_time:139323ms step_avg:153.44ms
step:919/1480 train_time:139484ms step_avg:153.45ms
step:920/1480 train_time:139643ms step_avg:153.45ms
step:921/1480 train_time:139803ms step_avg:153.46ms
step:922/1480 train_time:139965ms step_avg:153.47ms
step:923/1480 train_time:140123ms step_avg:153.48ms
step:924/1480 train_time:140282ms step_avg:153.48ms
step:925/1480 train_time:140441ms step_avg:153.49ms
step:926/1480 train_time:140600ms step_avg:153.49ms
step:927/1480 train_time:140757ms step_avg:153.50ms
step:928/1480 train_time:140917ms step_avg:153.50ms
step:929/1480 train_time:141077ms step_avg:153.51ms
step:930/1480 train_time:141238ms step_avg:153.52ms
step:931/1480 train_time:141397ms step_avg:153.53ms
step:932/1480 train_time:141556ms step_avg:153.53ms
step:933/1480 train_time:141716ms step_avg:153.54ms
step:934/1480 train_time:141876ms step_avg:153.55ms
step:935/1480 train_time:142037ms step_avg:153.55ms
step:936/1480 train_time:142195ms step_avg:153.56ms
step:937/1480 train_time:142357ms step_avg:153.57ms
step:938/1480 train_time:142518ms step_avg:153.58ms
step:939/1480 train_time:142679ms step_avg:153.58ms
step:940/1480 train_time:142840ms step_avg:153.59ms
step:941/1480 train_time:142999ms step_avg:153.60ms
step:942/1480 train_time:143158ms step_avg:153.60ms
step:943/1480 train_time:143317ms step_avg:153.61ms
step:944/1480 train_time:143482ms step_avg:153.62ms
step:945/1480 train_time:143640ms step_avg:153.63ms
step:946/1480 train_time:143801ms step_avg:153.63ms
step:947/1480 train_time:143962ms step_avg:153.64ms
step:948/1480 train_time:144120ms step_avg:153.65ms
step:949/1480 train_time:144300ms step_avg:153.67ms
step:950/1480 train_time:144440ms step_avg:153.66ms
step:951/1480 train_time:144602ms step_avg:153.67ms
step:952/1480 train_time:144761ms step_avg:153.67ms
step:953/1480 train_time:144921ms step_avg:153.68ms
step:954/1480 train_time:145083ms step_avg:153.69ms
step:955/1480 train_time:145241ms step_avg:153.69ms
step:956/1480 train_time:145399ms step_avg:153.70ms
step:957/1480 train_time:145561ms step_avg:153.71ms
step:958/1480 train_time:145722ms step_avg:153.72ms
step:959/1480 train_time:145882ms step_avg:153.72ms
step:960/1480 train_time:146041ms step_avg:153.73ms
step:961/1480 train_time:146201ms step_avg:153.73ms
step:962/1480 train_time:146361ms step_avg:153.74ms
step:963/1480 train_time:146520ms step_avg:153.75ms
step:964/1480 train_time:146682ms step_avg:153.75ms
step:965/1480 train_time:146841ms step_avg:153.76ms
step:966/1480 train_time:146999ms step_avg:153.77ms
step:967/1480 train_time:147158ms step_avg:153.77ms
step:968/1480 train_time:147317ms step_avg:153.78ms
step:969/1480 train_time:147477ms step_avg:153.78ms
step:970/1480 train_time:147635ms step_avg:153.79ms
step:971/1480 train_time:147796ms step_avg:153.79ms
step:972/1480 train_time:147954ms step_avg:153.80ms
step:973/1480 train_time:148114ms step_avg:153.80ms
step:974/1480 train_time:148275ms step_avg:153.81ms
step:975/1480 train_time:148437ms step_avg:153.82ms
step:976/1480 train_time:148596ms step_avg:153.83ms
step:977/1480 train_time:148756ms step_avg:153.83ms
step:978/1480 train_time:148916ms step_avg:153.84ms
step:979/1480 train_time:149078ms step_avg:153.85ms
step:980/1480 train_time:149238ms step_avg:153.85ms
step:981/1480 train_time:149401ms step_avg:153.86ms
step:982/1480 train_time:149560ms step_avg:153.87ms
step:983/1480 train_time:149719ms step_avg:153.87ms
step:984/1480 train_time:149879ms step_avg:153.88ms
step:985/1480 train_time:150039ms step_avg:153.89ms
step:986/1480 train_time:150198ms step_avg:153.89ms
step:987/1480 train_time:150356ms step_avg:153.90ms
step:988/1480 train_time:150515ms step_avg:153.90ms
step:989/1480 train_time:150676ms step_avg:153.91ms
step:990/1480 train_time:150838ms step_avg:153.92ms
step:991/1480 train_time:150999ms step_avg:153.92ms
step:992/1480 train_time:151165ms step_avg:153.94ms
step:993/1480 train_time:151331ms step_avg:153.95ms
step:994/1480 train_time:151491ms step_avg:153.95ms
step:995/1480 train_time:151650ms step_avg:153.96ms
step:996/1480 train_time:151808ms step_avg:153.96ms
step:997/1480 train_time:151966ms step_avg:153.97ms
step:998/1480 train_time:152124ms step_avg:153.97ms
step:999/1480 train_time:152284ms step_avg:153.98ms
step:1000/1480 train_time:152445ms step_avg:153.98ms
step:1000/1480 val_loss:3.4382 train_time:152517ms step_avg:154.06ms
step:1001/1480 train_time:152608ms step_avg:153.99ms
step:1002/1480 train_time:152768ms step_avg:154.00ms
step:1003/1480 train_time:152932ms step_avg:154.01ms
step:1004/1480 train_time:153093ms step_avg:154.02ms
step:1005/1480 train_time:153253ms step_avg:154.02ms
step:1006/1480 train_time:153413ms step_avg:154.03ms
step:1007/1480 train_time:153574ms step_avg:154.04ms
step:1008/1480 train_time:153735ms step_avg:154.04ms
step:1009/1480 train_time:153901ms step_avg:154.05ms
step:1010/1480 train_time:154060ms step_avg:154.06ms
step:1011/1480 train_time:154219ms step_avg:154.06ms
step:1012/1480 train_time:154376ms step_avg:154.07ms
step:1013/1480 train_time:154536ms step_avg:154.07ms
step:1014/1480 train_time:154695ms step_avg:154.08ms
step:1015/1480 train_time:154855ms step_avg:154.08ms
step:1016/1480 train_time:155014ms step_avg:154.09ms
step:1017/1480 train_time:155176ms step_avg:154.10ms
step:1018/1480 train_time:155336ms step_avg:154.10ms
step:1019/1480 train_time:155498ms step_avg:154.11ms
step:1020/1480 train_time:155657ms step_avg:154.12ms
step:1021/1480 train_time:155816ms step_avg:154.12ms
step:1022/1480 train_time:155975ms step_avg:154.13ms
step:1023/1480 train_time:156137ms step_avg:154.13ms
step:1024/1480 train_time:156298ms step_avg:154.14ms
step:1025/1480 train_time:156462ms step_avg:154.15ms
step:1026/1480 train_time:156623ms step_avg:154.16ms
step:1027/1480 train_time:156783ms step_avg:154.16ms
step:1028/1480 train_time:156947ms step_avg:154.17ms
step:1029/1480 train_time:157110ms step_avg:154.18ms
step:1030/1480 train_time:157270ms step_avg:154.19ms
step:1031/1480 train_time:157430ms step_avg:154.19ms
step:1032/1480 train_time:157596ms step_avg:154.20ms
step:1033/1480 train_time:157754ms step_avg:154.21ms
step:1034/1480 train_time:157914ms step_avg:154.21ms
step:1035/1480 train_time:158075ms step_avg:154.22ms
step:1036/1480 train_time:158234ms step_avg:154.22ms
step:1037/1480 train_time:158395ms step_avg:154.23ms
step:1038/1480 train_time:158554ms step_avg:154.24ms
step:1039/1480 train_time:158716ms step_avg:154.24ms
step:1040/1480 train_time:158875ms step_avg:154.25ms
step:1041/1480 train_time:159035ms step_avg:154.25ms
step:1042/1480 train_time:159193ms step_avg:154.26ms
step:1043/1480 train_time:159353ms step_avg:154.26ms
step:1044/1480 train_time:159513ms step_avg:154.27ms
step:1045/1480 train_time:159673ms step_avg:154.27ms
step:1046/1480 train_time:159832ms step_avg:154.28ms
step:1047/1480 train_time:159992ms step_avg:154.28ms
step:1048/1480 train_time:160152ms step_avg:154.29ms
step:1049/1480 train_time:160313ms step_avg:154.30ms
step:1050/1480 train_time:160475ms step_avg:154.30ms
step:1051/1480 train_time:160638ms step_avg:154.31ms
step:1052/1480 train_time:160797ms step_avg:154.32ms
step:1053/1480 train_time:160956ms step_avg:154.32ms
step:1054/1480 train_time:161117ms step_avg:154.33ms
step:1055/1480 train_time:161277ms step_avg:154.33ms
step:1056/1480 train_time:161435ms step_avg:154.34ms
step:1057/1480 train_time:161595ms step_avg:154.34ms
step:1058/1480 train_time:161757ms step_avg:154.35ms
step:1059/1480 train_time:161919ms step_avg:154.36ms
step:1060/1480 train_time:162082ms step_avg:154.36ms
step:1061/1480 train_time:162239ms step_avg:154.37ms
step:1062/1480 train_time:162399ms step_avg:154.37ms
step:1063/1480 train_time:162560ms step_avg:154.38ms
step:1064/1480 train_time:162718ms step_avg:154.38ms
step:1065/1480 train_time:162879ms step_avg:154.39ms
step:1066/1480 train_time:163040ms step_avg:154.39ms
step:1067/1480 train_time:163206ms step_avg:154.40ms
step:1068/1480 train_time:163367ms step_avg:154.41ms
step:1069/1480 train_time:163532ms step_avg:154.42ms
step:1070/1480 train_time:163691ms step_avg:154.43ms
step:1071/1480 train_time:163855ms step_avg:154.43ms
step:1072/1480 train_time:164014ms step_avg:154.44ms
step:1073/1480 train_time:164172ms step_avg:154.44ms
step:1074/1480 train_time:164332ms step_avg:154.45ms
step:1075/1480 train_time:164493ms step_avg:154.45ms
step:1076/1480 train_time:164652ms step_avg:154.46ms
step:1077/1480 train_time:164811ms step_avg:154.46ms
step:1078/1480 train_time:164976ms step_avg:154.47ms
step:1079/1480 train_time:165140ms step_avg:154.48ms
step:1080/1480 train_time:165301ms step_avg:154.49ms
step:1081/1480 train_time:165462ms step_avg:154.49ms
step:1082/1480 train_time:165623ms step_avg:154.50ms
step:1083/1480 train_time:165785ms step_avg:154.51ms
step:1084/1480 train_time:165946ms step_avg:154.51ms
step:1085/1480 train_time:166107ms step_avg:154.52ms
step:1086/1480 train_time:166268ms step_avg:154.52ms
step:1087/1480 train_time:166430ms step_avg:154.53ms
step:1088/1480 train_time:166590ms step_avg:154.54ms
step:1089/1480 train_time:166755ms step_avg:154.55ms
step:1090/1480 train_time:166919ms step_avg:154.55ms
step:1091/1480 train_time:167078ms step_avg:154.56ms
step:1092/1480 train_time:167238ms step_avg:154.56ms
step:1093/1480 train_time:167400ms step_avg:154.57ms
step:1094/1480 train_time:167560ms step_avg:154.58ms
step:1095/1480 train_time:167718ms step_avg:154.58ms
step:1096/1480 train_time:167880ms step_avg:154.59ms
step:1097/1480 train_time:168043ms step_avg:154.59ms
step:1098/1480 train_time:168206ms step_avg:154.60ms
step:1099/1480 train_time:168368ms step_avg:154.61ms
step:1100/1480 train_time:168533ms step_avg:154.62ms
step:1101/1480 train_time:168696ms step_avg:154.62ms
step:1102/1480 train_time:168858ms step_avg:154.63ms
step:1103/1480 train_time:169024ms step_avg:154.64ms
step:1104/1480 train_time:169187ms step_avg:154.65ms
step:1105/1480 train_time:169350ms step_avg:154.66ms
step:1106/1480 train_time:169511ms step_avg:154.66ms
step:1107/1480 train_time:169671ms step_avg:154.67ms
step:1108/1480 train_time:169832ms step_avg:154.67ms
step:1109/1480 train_time:169993ms step_avg:154.68ms
step:1110/1480 train_time:170152ms step_avg:154.68ms
step:1111/1480 train_time:170313ms step_avg:154.69ms
step:1112/1480 train_time:170475ms step_avg:154.70ms
step:1113/1480 train_time:170642ms step_avg:154.71ms
step:1114/1480 train_time:170805ms step_avg:154.72ms
step:1115/1480 train_time:170968ms step_avg:154.72ms
step:1116/1480 train_time:171128ms step_avg:154.73ms
step:1117/1480 train_time:171293ms step_avg:154.74ms
step:1118/1480 train_time:171456ms step_avg:154.74ms
step:1119/1480 train_time:171616ms step_avg:154.75ms
step:1120/1480 train_time:171778ms step_avg:154.75ms
step:1121/1480 train_time:171941ms step_avg:154.76ms
step:1122/1480 train_time:172101ms step_avg:154.77ms
step:1123/1480 train_time:172262ms step_avg:154.77ms
step:1124/1480 train_time:172425ms step_avg:154.78ms
step:1125/1480 train_time:172588ms step_avg:154.79ms
step:1125/1480 val_loss:3.3817 train_time:172663ms step_avg:154.85ms
step:1126/1480 train_time:172755ms step_avg:154.80ms
step:1127/1480 train_time:172913ms step_avg:154.80ms
step:1128/1480 train_time:173075ms step_avg:154.81ms
step:1129/1480 train_time:173238ms step_avg:154.82ms
step:1130/1480 train_time:173398ms step_avg:154.82ms
step:1131/1480 train_time:173566ms step_avg:154.83ms
step:1132/1480 train_time:173726ms step_avg:154.84ms
step:1133/1480 train_time:173889ms step_avg:154.84ms
step:1134/1480 train_time:174054ms step_avg:154.85ms
step:1135/1480 train_time:174215ms step_avg:154.86ms
step:1136/1480 train_time:174378ms step_avg:154.86ms
step:1137/1480 train_time:174538ms step_avg:154.87ms
step:1138/1480 train_time:174702ms step_avg:154.88ms
step:1139/1480 train_time:174879ms step_avg:154.90ms
step:1140/1480 train_time:175023ms step_avg:154.89ms
step:1141/1480 train_time:175187ms step_avg:154.90ms
step:1142/1480 train_time:175347ms step_avg:154.90ms
step:1143/1480 train_time:175512ms step_avg:154.91ms
step:1144/1480 train_time:175675ms step_avg:154.92ms
step:1145/1480 train_time:175835ms step_avg:154.92ms
step:1146/1480 train_time:175998ms step_avg:154.93ms
step:1147/1480 train_time:176160ms step_avg:154.93ms
step:1148/1480 train_time:176320ms step_avg:154.94ms
step:1149/1480 train_time:176484ms step_avg:154.95ms
step:1150/1480 train_time:176643ms step_avg:154.95ms
step:1151/1480 train_time:176810ms step_avg:154.96ms
step:1152/1480 train_time:176975ms step_avg:154.97ms
step:1153/1480 train_time:177140ms step_avg:154.98ms
step:1154/1480 train_time:177299ms step_avg:154.98ms
step:1155/1480 train_time:177462ms step_avg:154.99ms
step:1156/1480 train_time:177629ms step_avg:155.00ms
step:1157/1480 train_time:177793ms step_avg:155.01ms
step:1158/1480 train_time:177955ms step_avg:155.01ms
step:1159/1480 train_time:178116ms step_avg:155.02ms
step:1160/1480 train_time:178276ms step_avg:155.02ms
step:1161/1480 train_time:178438ms step_avg:155.03ms
step:1162/1480 train_time:178600ms step_avg:155.03ms
step:1163/1480 train_time:178761ms step_avg:155.04ms
step:1164/1480 train_time:178923ms step_avg:155.05ms
step:1165/1480 train_time:179082ms step_avg:155.05ms
step:1166/1480 train_time:179244ms step_avg:155.06ms
step:1167/1480 train_time:179403ms step_avg:155.06ms
step:1168/1480 train_time:179567ms step_avg:155.07ms
step:1169/1480 train_time:179731ms step_avg:155.07ms
step:1170/1480 train_time:179893ms step_avg:155.08ms
step:1171/1480 train_time:180054ms step_avg:155.09ms
step:1172/1480 train_time:180215ms step_avg:155.09ms
step:1173/1480 train_time:180378ms step_avg:155.10ms
step:1174/1480 train_time:180546ms step_avg:155.11ms
step:1175/1480 train_time:180709ms step_avg:155.12ms
step:1176/1480 train_time:180872ms step_avg:155.12ms
step:1177/1480 train_time:181039ms step_avg:155.13ms
step:1178/1480 train_time:181198ms step_avg:155.14ms
step:1179/1480 train_time:181358ms step_avg:155.14ms
step:1180/1480 train_time:181529ms step_avg:155.15ms
step:1181/1480 train_time:181693ms step_avg:155.16ms
step:1182/1480 train_time:181853ms step_avg:155.17ms
step:1183/1480 train_time:182014ms step_avg:155.17ms
step:1184/1480 train_time:182177ms step_avg:155.18ms
step:1185/1480 train_time:182342ms step_avg:155.18ms
step:1186/1480 train_time:182504ms step_avg:155.19ms
step:1187/1480 train_time:182678ms step_avg:155.21ms
step:1188/1480 train_time:182838ms step_avg:155.21ms
step:1189/1480 train_time:183000ms step_avg:155.22ms
step:1190/1480 train_time:183160ms step_avg:155.22ms
step:1191/1480 train_time:183323ms step_avg:155.23ms
step:1192/1480 train_time:183484ms step_avg:155.23ms
step:1193/1480 train_time:183645ms step_avg:155.24ms
step:1194/1480 train_time:183807ms step_avg:155.24ms
step:1195/1480 train_time:183970ms step_avg:155.25ms
step:1196/1480 train_time:184141ms step_avg:155.26ms
step:1197/1480 train_time:184302ms step_avg:155.27ms
step:1198/1480 train_time:184472ms step_avg:155.28ms
step:1199/1480 train_time:184635ms step_avg:155.29ms
step:1200/1480 train_time:184795ms step_avg:155.29ms
step:1201/1480 train_time:184957ms step_avg:155.30ms
step:1202/1480 train_time:185126ms step_avg:155.31ms
step:1203/1480 train_time:185292ms step_avg:155.32ms
step:1204/1480 train_time:185457ms step_avg:155.32ms
step:1205/1480 train_time:185617ms step_avg:155.33ms
step:1206/1480 train_time:185779ms step_avg:155.33ms
step:1207/1480 train_time:185939ms step_avg:155.34ms
step:1208/1480 train_time:186099ms step_avg:155.34ms
step:1209/1480 train_time:186262ms step_avg:155.35ms
step:1210/1480 train_time:186429ms step_avg:155.36ms
step:1211/1480 train_time:186592ms step_avg:155.36ms
step:1212/1480 train_time:186756ms step_avg:155.37ms
step:1213/1480 train_time:186920ms step_avg:155.38ms
step:1214/1480 train_time:187084ms step_avg:155.39ms
step:1215/1480 train_time:187249ms step_avg:155.39ms
step:1216/1480 train_time:187410ms step_avg:155.40ms
step:1217/1480 train_time:187575ms step_avg:155.41ms
step:1218/1480 train_time:187738ms step_avg:155.41ms
step:1219/1480 train_time:187907ms step_avg:155.42ms
step:1220/1480 train_time:188071ms step_avg:155.43ms
step:1221/1480 train_time:188233ms step_avg:155.44ms
step:1222/1480 train_time:188394ms step_avg:155.44ms
step:1223/1480 train_time:188558ms step_avg:155.45ms
step:1224/1480 train_time:188723ms step_avg:155.46ms
step:1225/1480 train_time:188887ms step_avg:155.46ms
step:1226/1480 train_time:189052ms step_avg:155.47ms
step:1227/1480 train_time:189217ms step_avg:155.48ms
step:1228/1480 train_time:189378ms step_avg:155.48ms
step:1229/1480 train_time:189541ms step_avg:155.49ms
step:1230/1480 train_time:189711ms step_avg:155.50ms
step:1231/1480 train_time:189878ms step_avg:155.51ms
step:1232/1480 train_time:190043ms step_avg:155.52ms
step:1233/1480 train_time:190203ms step_avg:155.52ms
step:1234/1480 train_time:190365ms step_avg:155.53ms
step:1235/1480 train_time:190531ms step_avg:155.54ms
step:1236/1480 train_time:190692ms step_avg:155.54ms
step:1237/1480 train_time:190855ms step_avg:155.55ms
step:1238/1480 train_time:191026ms step_avg:155.56ms
step:1239/1480 train_time:191189ms step_avg:155.56ms
step:1240/1480 train_time:191353ms step_avg:155.57ms
step:1241/1480 train_time:191517ms step_avg:155.58ms
step:1242/1480 train_time:191679ms step_avg:155.58ms
step:1243/1480 train_time:191842ms step_avg:155.59ms
step:1244/1480 train_time:192002ms step_avg:155.59ms
step:1245/1480 train_time:192164ms step_avg:155.60ms
step:1246/1480 train_time:192327ms step_avg:155.60ms
step:1247/1480 train_time:192490ms step_avg:155.61ms
step:1248/1480 train_time:192651ms step_avg:155.62ms
step:1249/1480 train_time:192813ms step_avg:155.62ms
step:1250/1480 train_time:192976ms step_avg:155.63ms
step:1250/1480 val_loss:3.3327 train_time:193051ms step_avg:155.69ms
step:1251/1480 train_time:193146ms step_avg:155.64ms
step:1252/1480 train_time:193309ms step_avg:155.64ms
step:1253/1480 train_time:193469ms step_avg:155.65ms
step:1254/1480 train_time:193631ms step_avg:155.65ms
step:1255/1480 train_time:193801ms step_avg:155.66ms
step:1256/1480 train_time:193966ms step_avg:155.67ms
step:1257/1480 train_time:194127ms step_avg:155.68ms
step:1258/1480 train_time:194292ms step_avg:155.68ms
step:1259/1480 train_time:194457ms step_avg:155.69ms
step:1260/1480 train_time:194618ms step_avg:155.69ms
step:1261/1480 train_time:194780ms step_avg:155.70ms
step:1262/1480 train_time:194945ms step_avg:155.71ms
step:1263/1480 train_time:195111ms step_avg:155.72ms
step:1264/1480 train_time:195272ms step_avg:155.72ms
step:1265/1480 train_time:195431ms step_avg:155.72ms
step:1266/1480 train_time:195595ms step_avg:155.73ms
step:1267/1480 train_time:195755ms step_avg:155.73ms
step:1268/1480 train_time:195918ms step_avg:155.74ms
step:1269/1480 train_time:196084ms step_avg:155.75ms
step:1270/1480 train_time:196247ms step_avg:155.75ms
step:1271/1480 train_time:196410ms step_avg:155.76ms
step:1272/1480 train_time:196571ms step_avg:155.76ms
step:1273/1480 train_time:196734ms step_avg:155.77ms
step:1274/1480 train_time:196899ms step_avg:155.77ms
step:1275/1480 train_time:197059ms step_avg:155.78ms
step:1276/1480 train_time:197219ms step_avg:155.78ms
step:1277/1480 train_time:197382ms step_avg:155.79ms
step:1278/1480 train_time:197542ms step_avg:155.79ms
step:1279/1480 train_time:197702ms step_avg:155.79ms
step:1280/1480 train_time:197870ms step_avg:155.80ms
step:1281/1480 train_time:198031ms step_avg:155.81ms
step:1282/1480 train_time:198192ms step_avg:155.81ms
step:1283/1480 train_time:198354ms step_avg:155.82ms
step:1284/1480 train_time:198519ms step_avg:155.82ms
step:1285/1480 train_time:198680ms step_avg:155.83ms
step:1286/1480 train_time:198842ms step_avg:155.83ms
step:1287/1480 train_time:199004ms step_avg:155.84ms
step:1288/1480 train_time:199166ms step_avg:155.84ms
step:1289/1480 train_time:199336ms step_avg:155.85ms
step:1290/1480 train_time:199504ms step_avg:155.86ms
step:1291/1480 train_time:199670ms step_avg:155.87ms
step:1292/1480 train_time:199834ms step_avg:155.88ms
step:1293/1480 train_time:199999ms step_avg:155.88ms
step:1294/1480 train_time:200162ms step_avg:155.89ms
step:1295/1480 train_time:200325ms step_avg:155.90ms
step:1296/1480 train_time:200489ms step_avg:155.90ms
step:1297/1480 train_time:200653ms step_avg:155.91ms
step:1298/1480 train_time:200817ms step_avg:155.91ms
step:1299/1480 train_time:200979ms step_avg:155.92ms
step:1300/1480 train_time:201139ms step_avg:155.92ms
step:1301/1480 train_time:201301ms step_avg:155.93ms
step:1302/1480 train_time:201467ms step_avg:155.93ms
step:1303/1480 train_time:201634ms step_avg:155.94ms
step:1304/1480 train_time:201799ms step_avg:155.95ms
step:1305/1480 train_time:201960ms step_avg:155.95ms
step:1306/1480 train_time:202125ms step_avg:155.96ms
step:1307/1480 train_time:202287ms step_avg:155.96ms
step:1308/1480 train_time:202450ms step_avg:155.97ms
step:1309/1480 train_time:202615ms step_avg:155.98ms
step:1310/1480 train_time:202778ms step_avg:155.98ms
step:1311/1480 train_time:202939ms step_avg:155.99ms
step:1312/1480 train_time:203104ms step_avg:155.99ms
step:1313/1480 train_time:203266ms step_avg:156.00ms
step:1314/1480 train_time:203430ms step_avg:156.00ms
step:1315/1480 train_time:203595ms step_avg:156.01ms
step:1316/1480 train_time:203755ms step_avg:156.01ms
step:1317/1480 train_time:203917ms step_avg:156.02ms
step:1318/1480 train_time:204083ms step_avg:156.03ms
step:1319/1480 train_time:204249ms step_avg:156.03ms
step:1320/1480 train_time:204418ms step_avg:156.04ms
step:1321/1480 train_time:204581ms step_avg:156.05ms
step:1322/1480 train_time:204752ms step_avg:156.06ms
step:1323/1480 train_time:204916ms step_avg:156.07ms
step:1324/1480 train_time:205080ms step_avg:156.07ms
step:1325/1480 train_time:205250ms step_avg:156.08ms
step:1326/1480 train_time:205415ms step_avg:156.09ms
step:1327/1480 train_time:205578ms step_avg:156.10ms
step:1328/1480 train_time:205740ms step_avg:156.10ms
step:1329/1480 train_time:205931ms step_avg:156.13ms
step:1330/1480 train_time:206088ms step_avg:156.13ms
step:1331/1480 train_time:206252ms step_avg:156.13ms
step:1332/1480 train_time:206416ms step_avg:156.14ms
step:1333/1480 train_time:206581ms step_avg:156.15ms
step:1334/1480 train_time:206744ms step_avg:156.15ms
step:1335/1480 train_time:206904ms step_avg:156.15ms
step:1336/1480 train_time:207074ms step_avg:156.16ms
step:1337/1480 train_time:207240ms step_avg:156.17ms
step:1338/1480 train_time:207404ms step_avg:156.18ms
step:1339/1480 train_time:207569ms step_avg:156.18ms
step:1340/1480 train_time:207733ms step_avg:156.19ms
step:1341/1480 train_time:207895ms step_avg:156.19ms
step:1342/1480 train_time:208061ms step_avg:156.20ms
step:1343/1480 train_time:208221ms step_avg:156.20ms
step:1344/1480 train_time:208384ms step_avg:156.21ms
step:1345/1480 train_time:208552ms step_avg:156.22ms
step:1346/1480 train_time:208715ms step_avg:156.22ms
step:1347/1480 train_time:208878ms step_avg:156.23ms
step:1348/1480 train_time:209038ms step_avg:156.23ms
step:1349/1480 train_time:209200ms step_avg:156.24ms
step:1350/1480 train_time:209366ms step_avg:156.24ms
step:1351/1480 train_time:209527ms step_avg:156.25ms
step:1352/1480 train_time:209691ms step_avg:156.25ms
step:1353/1480 train_time:209856ms step_avg:156.26ms
step:1354/1480 train_time:210019ms step_avg:156.26ms
step:1355/1480 train_time:210181ms step_avg:156.27ms
step:1356/1480 train_time:210346ms step_avg:156.27ms
step:1357/1480 train_time:210509ms step_avg:156.28ms
step:1358/1480 train_time:210672ms step_avg:156.29ms
step:1359/1480 train_time:210837ms step_avg:156.29ms
step:1360/1480 train_time:211004ms step_avg:156.30ms
step:1361/1480 train_time:211172ms step_avg:156.31ms
step:1362/1480 train_time:211337ms step_avg:156.31ms
step:1363/1480 train_time:211505ms step_avg:156.32ms
step:1364/1480 train_time:211667ms step_avg:156.33ms
step:1365/1480 train_time:211826ms step_avg:156.33ms
step:1366/1480 train_time:211991ms step_avg:156.34ms
step:1367/1480 train_time:212153ms step_avg:156.34ms
step:1368/1480 train_time:212319ms step_avg:156.35ms
step:1369/1480 train_time:212490ms step_avg:156.36ms
step:1370/1480 train_time:212657ms step_avg:156.37ms
step:1371/1480 train_time:212820ms step_avg:156.37ms
step:1372/1480 train_time:212989ms step_avg:156.38ms
step:1373/1480 train_time:213151ms step_avg:156.38ms
step:1374/1480 train_time:213318ms step_avg:156.39ms
step:1375/1480 train_time:213480ms step_avg:156.40ms
step:1375/1480 val_loss:3.2942 train_time:213554ms step_avg:156.45ms
step:1376/1480 train_time:213647ms step_avg:156.40ms
step:1377/1480 train_time:213810ms step_avg:156.41ms
step:1378/1480 train_time:213972ms step_avg:156.41ms
step:1379/1480 train_time:214138ms step_avg:156.42ms
step:1380/1480 train_time:214303ms step_avg:156.43ms
step:1381/1480 train_time:214471ms step_avg:156.43ms
step:1382/1480 train_time:214633ms step_avg:156.44ms
step:1383/1480 train_time:214796ms step_avg:156.44ms
step:1384/1480 train_time:214963ms step_avg:156.45ms
step:1385/1480 train_time:215124ms step_avg:156.45ms
step:1386/1480 train_time:215288ms step_avg:156.46ms
step:1387/1480 train_time:215454ms step_avg:156.47ms
step:1388/1480 train_time:215614ms step_avg:156.47ms
step:1389/1480 train_time:215781ms step_avg:156.48ms
step:1390/1480 train_time:215943ms step_avg:156.48ms
step:1391/1480 train_time:216105ms step_avg:156.48ms
step:1392/1480 train_time:216268ms step_avg:156.49ms
step:1393/1480 train_time:216431ms step_avg:156.49ms
step:1394/1480 train_time:216592ms step_avg:156.50ms
step:1395/1480 train_time:216754ms step_avg:156.50ms
step:1396/1480 train_time:216915ms step_avg:156.50ms
step:1397/1480 train_time:217075ms step_avg:156.51ms
step:1398/1480 train_time:217236ms step_avg:156.51ms
step:1399/1480 train_time:217396ms step_avg:156.51ms
step:1400/1480 train_time:217563ms step_avg:156.52ms
step:1401/1480 train_time:217724ms step_avg:156.52ms
step:1402/1480 train_time:217886ms step_avg:156.53ms
step:1403/1480 train_time:218053ms step_avg:156.53ms
step:1404/1480 train_time:218215ms step_avg:156.54ms
step:1405/1480 train_time:218382ms step_avg:156.55ms
step:1406/1480 train_time:218548ms step_avg:156.55ms
step:1407/1480 train_time:218709ms step_avg:156.56ms
step:1408/1480 train_time:218870ms step_avg:156.56ms
step:1409/1480 train_time:219042ms step_avg:156.57ms
step:1410/1480 train_time:219206ms step_avg:156.58ms
step:1411/1480 train_time:219369ms step_avg:156.58ms
step:1412/1480 train_time:219531ms step_avg:156.58ms
step:1413/1480 train_time:219693ms step_avg:156.59ms
step:1414/1480 train_time:219857ms step_avg:156.59ms
step:1415/1480 train_time:220022ms step_avg:156.60ms
step:1416/1480 train_time:220195ms step_avg:156.61ms
step:1417/1480 train_time:220360ms step_avg:156.62ms
step:1418/1480 train_time:220525ms step_avg:156.62ms
step:1419/1480 train_time:220689ms step_avg:156.63ms
step:1420/1480 train_time:220854ms step_avg:156.63ms
step:1421/1480 train_time:221020ms step_avg:156.64ms
step:1422/1480 train_time:221186ms step_avg:156.65ms
step:1423/1480 train_time:221348ms step_avg:156.65ms
step:1424/1480 train_time:221514ms step_avg:156.66ms
step:1425/1480 train_time:221684ms step_avg:156.67ms
step:1426/1480 train_time:221849ms step_avg:156.67ms
step:1427/1480 train_time:222014ms step_avg:156.68ms
step:1428/1480 train_time:222175ms step_avg:156.68ms
step:1429/1480 train_time:222336ms step_avg:156.69ms
step:1430/1480 train_time:222501ms step_avg:156.69ms
step:1431/1480 train_time:222668ms step_avg:156.70ms
step:1432/1480 train_time:222835ms step_avg:156.71ms
step:1433/1480 train_time:223005ms step_avg:156.71ms
step:1434/1480 train_time:223174ms step_avg:156.72ms
step:1435/1480 train_time:223339ms step_avg:156.73ms
step:1436/1480 train_time:223505ms step_avg:156.74ms
step:1437/1480 train_time:223667ms step_avg:156.74ms
step:1438/1480 train_time:223829ms step_avg:156.74ms
step:1439/1480 train_time:223995ms step_avg:156.75ms
step:1440/1480 train_time:224158ms step_avg:156.75ms
step:1441/1480 train_time:224324ms step_avg:156.76ms
step:1442/1480 train_time:224489ms step_avg:156.77ms
step:1443/1480 train_time:224662ms step_avg:156.78ms
step:1444/1480 train_time:224827ms step_avg:156.78ms
step:1445/1480 train_time:224989ms step_avg:156.79ms
step:1446/1480 train_time:225155ms step_avg:156.79ms
step:1447/1480 train_time:225324ms step_avg:156.80ms
step:1448/1480 train_time:225486ms step_avg:156.81ms
step:1449/1480 train_time:225650ms step_avg:156.81ms
step:1450/1480 train_time:225814ms step_avg:156.82ms
step:1451/1480 train_time:225978ms step_avg:156.82ms
step:1452/1480 train_time:226143ms step_avg:156.83ms
step:1453/1480 train_time:226306ms step_avg:156.83ms
step:1454/1480 train_time:226469ms step_avg:156.83ms
step:1455/1480 train_time:226637ms step_avg:156.84ms
step:1456/1480 train_time:226801ms step_avg:156.85ms
step:1457/1480 train_time:226964ms step_avg:156.85ms
step:1458/1480 train_time:227129ms step_avg:156.86ms
step:1459/1480 train_time:227294ms step_avg:156.86ms
step:1460/1480 train_time:227456ms step_avg:156.87ms
step:1461/1480 train_time:227620ms step_avg:156.87ms
step:1462/1480 train_time:227787ms step_avg:156.88ms
step:1463/1480 train_time:227953ms step_avg:156.88ms
step:1464/1480 train_time:228120ms step_avg:156.89ms
step:1465/1480 train_time:228284ms step_avg:156.90ms
step:1466/1480 train_time:228447ms step_avg:156.90ms
step:1467/1480 train_time:228611ms step_avg:156.91ms
step:1468/1480 train_time:228773ms step_avg:156.91ms
step:1469/1480 train_time:228936ms step_avg:156.91ms
step:1470/1480 train_time:229105ms step_avg:156.92ms
step:1471/1480 train_time:229274ms step_avg:156.93ms
step:1472/1480 train_time:229445ms step_avg:156.94ms
step:1473/1480 train_time:229609ms step_avg:156.94ms
step:1474/1480 train_time:229776ms step_avg:156.95ms
step:1475/1480 train_time:229945ms step_avg:156.96ms
step:1476/1480 train_time:230109ms step_avg:156.96ms
step:1477/1480 train_time:230276ms step_avg:156.97ms
step:1478/1480 train_time:230448ms step_avg:156.98ms
step:1479/1480 train_time:230613ms step_avg:156.99ms
step:1480/1480 train_time:230774ms step_avg:156.99ms
step:1480/1480 val_loss:3.2753 train_time:230850ms step_avg:157.04ms
peak memory consumption: 34239 MiB