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
from dataclasses import dataclass
from functools import lru_cache
from pathlib import Path

os.environ["PYTORCH_CUDA_ALLOC_CONF"] = "expandable_segments:True"
import torch
torch.empty(1, device='cuda', requires_grad=True).backward() # prevents a bug on some systems
from torch import Tensor, nn
import torch.nn.functional as F
import torch.distributed as dist
# use of FlexAttention contributed by @KoszarskyB
from torch.nn.attention.flex_attention import BlockMask, flex_attention
torch._inductor.config.coordinate_descent_tuning = True

# -----------------------------------------------------------------------------
# Custom operators

@torch.library.custom_op("nanogpt::mm", mutates_args=())
def mm_op(x: Tensor, w: Tensor, x_s: float, w_s: float, grad_s: float) -> tuple[Tensor, Tensor, Tensor]:
    @torch.compile
    def impl(x: Tensor, w: Tensor):
        assert x.is_contiguous() and w.is_contiguous()
        x_f8 = x.mul(x_s).to(torch.float8_e4m3fn)
        w_f8 = w.mul(w_s).to(torch.float8_e4m3fn)
        out = torch._scaled_mm(
            x_f8,
            w_f8.t(),
            out_dtype=torch.bfloat16,
            scale_a=x.new_tensor(1 / x_s, dtype=torch.float32),
            scale_b=x.new_tensor(1 / w_s, dtype=torch.float32),
            use_fast_accum=True,
        )
        return out, x_f8, w_f8

    return impl(x, w)

@mm_op.register_fake
def _(x: Tensor, w: Tensor, *_):
    assert x.ndim == w.ndim == 2
    assert x.shape[1] == w.shape[1]
    assert x.device == w.device
    assert x.is_contiguous() and w.is_contiguous()
    return x @ w.t(), x.to(torch.float8_e4m3fn), w.to(torch.float8_e4m3fn)

@torch.library.custom_op("nanogpt::mm_backward", mutates_args=())
def mm_backward_op(g: Tensor, x_f8: Tensor, w_f8: Tensor, x_s: float, w_s: float, grad_s: float) -> tuple[Tensor, Tensor]:
    @torch.compile
    def impl(grad: Tensor, x_f8: Tensor, w_f8: Tensor):
        assert grad.is_contiguous()
        x_inv_s = grad.new_tensor(1 / x_s, dtype=torch.float32)
        w_inv_s = grad.new_tensor(1 / w_s, dtype=torch.float32)
        grad_inv_s = grad.new_tensor(1 / grad_s, dtype=torch.float32)
        grad_f8 = grad.mul(grad_s).to(torch.float8_e5m2)
        grad_x = torch._scaled_mm(
            grad_f8,
            w_f8.t().contiguous().t(),
            out_dtype=torch.bfloat16,
            scale_a=grad_inv_s,
            scale_b=w_inv_s,
            use_fast_accum=False,
        )
        # faster than grad_f8_t @ x_f8, for (d_out, d_in) == (50304, 768)
        grad_w = torch._scaled_mm(
            x_f8.t().contiguous(),
            grad_f8.t().contiguous().t(),
            out_dtype=torch.float32,
            scale_a=x_inv_s,
            scale_b=grad_inv_s,
            use_fast_accum=False,
        ).t()
        return grad_x, grad_w

    return impl(g, x_f8, w_f8)

@mm_backward_op.register_fake
def _(g: Tensor, x_f8: Tensor, w_f8: Tensor, *_):
    return x_f8.to(torch.bfloat16), w_f8.to(torch.float32)

def backward(ctx, grad_out: Tensor, *_):
    x_f8, w_f8 = ctx.saved_tensors
    x_s, w_s, grad_s = ctx.scales
    grad_x, grad_w = torch.ops.nanogpt.mm_backward(
        grad_out, x_f8, w_f8, x_s, w_s, grad_s
    )
    return grad_x, grad_w, None, None, None

def setup_context(ctx: torch.autograd.function.FunctionCtx, inputs, output):
    *_, x_s, w_s, grad_s = inputs
    _, x_f8, w_f8 = output
    ctx.save_for_backward(x_f8, w_f8)
    ctx.scales = x_s, w_s, grad_s
    ctx.set_materialize_grads(False)

mm_op.register_autograd(backward, setup_context=setup_context)

def lm_head_fp8(x: Tensor, w: Tensor) -> Tensor:
    _x = x.flatten(0, -2)
    out: Tensor = torch.ops.nanogpt.mm(_x, w, x_s=2.0, w_s=32.0, grad_s=2.0**29)[0]
    return out.reshape(*x.shape[:-1], -1)

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

@torch.compile
def zeropower_via_newtonschulz5(G: Tensor, steps: int) -> Tensor:
    """
    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()
    if G.size(0) > G.size(1):
        X = X.T

    # Ensure spectral norm is at most 1
    X = X / (X.norm() + 1e-7)
    # Perform the NS iterations
    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, rank=0, world_size=1):
        self.rank = rank
        self.world_size = world_size
        defaults = dict(lr=lr, momentum=momentum, nesterov=nesterov, ns_steps=ns_steps)
        params: list[Tensor] = [*params]
        assert all(isinstance(p, Tensor) for p in params)
        sizes = {p.numel() for p in params}
        def create_update_buffer(size: int):
            b = torch.empty(self.world_size, size, dtype=torch.bfloat16, device="cuda")
            return dict(update_buffer=b, update_buffer_views=[b[i] for i in range(self.world_size)])
        param_groups = [
            dict(params=[p for p in params if p.numel() == size], **create_update_buffer(size)) for size in sizes]
        super().__init__(param_groups, defaults)

    @torch.no_grad()
    def step(self):
        for group in self.param_groups:
            lr = group['lr']
            momentum = group['momentum']
            nesterov = group['nesterov']
            ns_steps = group['ns_steps']
            update_buffer = group['update_buffer']
            update_buffer_views: list[Tensor] = group['update_buffer_views']
            # generate weight updates in distributed fashion
            params: list[Tensor] = group['params']
            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_buffer_views):
                    p_world.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]:
                if base_i + self.rank < len(params):
                    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: Tensor = 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()
                else:
                    g = update_buffer_views[self.rank]
                update_prev() # async all_gather instead of sync all_reduce by @YouJiacheng
                handle = dist.all_gather_into_tensor(update_buffer, 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: int, out_features: int):
        super().__init__(in_features, out_features, bias=False)

    def reset_parameters(self) -> None:
        std = 0.5 * (self.in_features ** -0.5) # 0.5 is a bit better than the default 1/sqrt(3)
        bound = (3 ** 0.5) * std
        with torch.no_grad():
            self.weight.uniform_(-bound, bound)

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

class Rotary(nn.Module):
    def __init__(self, dim: int, max_seq_len=65536):
        super().__init__()
        # half-truncate RoPE by @YouJiacheng (w/ base freq tuning)
        angular_freq = (1 / 1024) ** torch.linspace(0, 1, steps=dim//4, dtype=torch.float32)
        angular_freq = torch.cat([angular_freq, angular_freq.new_zeros(dim//4)])
        t = torch.arange(max_seq_len, dtype=torch.float32)
        theta = torch.einsum('i,j -> ij', t, angular_freq)
        self.cos = nn.Buffer(theta.cos(), persistent=False)
        self.sin = nn.Buffer(theta.sin(), persistent=False)

    def forward(self, x_BTHD: Tensor):
        assert self.cos.size(0) >= x_BTHD.size(-3)
        cos, sin = self.cos[None, :x_BTHD.size(-3), None, :], self.sin[None, :x_BTHD.size(-3), None, :]
        x1, x2 = x_BTHD.to(dtype=torch.float32).chunk(2, dim=-1)
        y1 = x1 * cos + x2 * sin
        y2 = x1 * (-sin) + x2 * cos
        return torch.cat((y1, y2), 3).type_as(x_BTHD)

class CausalSelfAttention(nn.Module):
    def __init__(self, dim: int, num_heads: int):
        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.detach().zero_() # zero init suggested by @Grad62304977

    def forward(self, x: Tensor, ve: Tensor | None, block_mask: BlockMask):
        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)
        if ve is not None:
            v = self.lambdas[0] * v + self.lambdas[1] * ve.view_as(v) # @KoszarskyB & @Grad62304977
        else: # skip mid-layers token value embeddings by @YouJiacheng
            v = self.lambdas[0] * v
        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)
        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.detach().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, model_dim: int, num_heads: int, layer_idx: int):
        super().__init__()
        # skip attention of blocks.7 (the 8th layer) by @YouJiacheng
        self.attn = CausalSelfAttention(model_dim, num_heads) if layer_idx != 7 else None
        self.mlp = MLP(model_dim)
        self.lambdas = nn.Parameter(torch.tensor([1., 0.]))

    def forward(self, x, ve, x0, block_mask):
        x = self.lambdas[0] * x + self.lambdas[1] * x0
        if self.attn is not None:
            x = x + self.attn(norm(x), ve, block_mask)
        x = x + self.mlp(norm(x))
        return x

class ValueEmbedding(nn.Module):
    def __init__(self, num_embeddings: int, embedding_dim: int):
        super().__init__()
        self.embed = nn.ModuleList([nn.Embedding(num_embeddings, embedding_dim) for _ in range(3)])

    def forward(self, input_seq) -> list[Tensor | None]:
        ve = [emb(input_seq) for emb in self.embed]
        # 012 ... 012 structure on token value embeddings by @YouJiacheng, improved on @leloykun's U-net structure
        ve = [ve[0], ve[1], ve[2], None, None, None, None, None, None, ve[0], ve[1], ve[2]]
        return ve

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

def next_multiple_of_n(v: float | int, *, n: int):
    return next(x for x in range(n, int(v) + 1 + n, n) if x >= v)

class GPT(nn.Module):
    def __init__(self, vocab_size: int, num_layers: int, num_heads: int, model_dim: int):
        super().__init__()
        self.embed = nn.Embedding(vocab_size, model_dim)
        # token value embeddings by @KoszarskyB - inspired by @Grad62304977's value residual learning
        self.value_embeds = ValueEmbedding(vocab_size, model_dim)
        self.blocks = nn.ModuleList([Block(model_dim, num_heads, layer_idx) for layer_idx in range(num_layers)])
        # U-net design by @brendanh0gan
        self.num_encoder_layers = num_layers // 2 # Half of the layers for encoder
        self.num_decoder_layers = 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))
        # 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.
        self.lm_head = CastedLinear(model_dim, next_multiple_of_n(vocab_size, n=128))
        self.lm_head.weight.detach().zero_() # @Grad62304977

    def forward(self, input_seq: Tensor, target_seq: Tensor, sliding_window_num_blocks: Tensor):
        BLOCK_SIZE = 128
        assert input_seq.ndim == 1
        assert len(input_seq) % BLOCK_SIZE == 0
        NUM_BLOCKS = len(input_seq) // BLOCK_SIZE
        docs = (input_seq == 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: 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()

        # manual block mask creation by @YouJiacheng
        def create_doc_swc_block_mask(sliding_window_num_blocks: Tensor):
            kv_idx = block_idx = torch.arange(NUM_BLOCKS, 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 # block-wise sliding window by @YouJiacheng
            # 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)

        x = x0 = norm(self.embed(input_seq)[None]) # use of norm here by @Grad62304977
        ve = self.value_embeds(input_seq)
        ve_enc, ve_dec = ve[:self.num_encoder_layers], ve[self.num_encoder_layers:]
        assert len(ve_enc) == self.num_encoder_layers and len(ve_dec) == self.num_decoder_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()
            x = self.blocks[self.num_encoder_layers + i](x, ve_dec[i], x0, block_mask)
        x = norm(x)
        logits = lm_head_fp8(x, self.lm_head.weight) if self.training else self.lm_head(x)
        # @Grad62304977 added tanh softcapping, @KoszarskyB reduced it from 30 to 15, @YouJiacheng shifted it by +15 (2*sigmoid(2*x)=tanh(x)+1)
        logits = 30 * torch.sigmoid(logits.float() / 7.5)
        loss = F.cross_entropy(logits.view(-1, logits.size(-1)), target_seq)
        return loss

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

def _load_data_shard(file: Path):
    header = torch.from_file(f"{file}", False, 256, dtype=torch.int32) # header is 256 int32
    assert header[0] == 20240520, 'magic number mismatch in the data .bin file'
    assert header[1] == 1, 'unsupported version'
    num_tokens = int(header[2]) # number of tokens (claimed)
    with file.open('rb', buffering=0) as f:
        tokens = torch.empty(num_tokens, dtype=torch.uint16, pin_memory=True) # avoid pin_memory copy by @YouJiacheng
        f.seek(256 * 4)
        nbytes = f.readinto(tokens.numpy()) # avoid bytes->array copy by @YouJiacheng
        assert nbytes == 2 * num_tokens, 'number of tokens read does not match header'
    return tokens

def distributed_data_generator(filename_pattern: str, batch_size: int, rank : int, world_size : int):
    files = sorted(Path.cwd().glob(filename_pattern))
    assert batch_size % world_size == 0
    local_batch_size = batch_size // world_size
    file_iter = iter(files) # use cycle(files) if you want to do multi-epoch training
    tokens, pos = _load_data_shard(next(file_iter)), 0
    while True:
        if pos + batch_size + 1 >= len(tokens):
            tokens, pos = _load_data_shard(next(file_iter)), 0
        buf = tokens[pos + rank * local_batch_size:][:local_batch_size + 1]
        inputs = buf[:-1].to(device="cuda", dtype=torch.int32, non_blocking=True) # no sync on host side;
        targets = buf[1:].to(device="cuda", dtype=torch.int64, non_blocking=True) # H2D in another stream isn't helpful.
        pos += batch_size
        yield inputs, targets

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

@dataclass
class Hyperparameters:
    # data
    train_files = 'data/fineweb10B/fineweb_train_*.bin' # input .bin to train on
    val_files = 'data/fineweb10B/fineweb_val_*.bin' # input .bin to eval validation loss on
    val_tokens = 10485760 # how many tokens of validation data? it's important to keep this fixed for consistent comparisons
    # optimization
    batch_size = 8*64*1024 # batch size in tokens
    num_iterations = 1395 # number of iterations to run
    cooldown_frac = 0.4 # fraction of training spent cooling down the learning rate
    # evaluation and logging
    val_loss_every = 125 # every how many steps to evaluate val loss? 0 for only at the end
    # implementation
    seq_len = 64*1024 # FlexAttention sequence length
    save_checkpoint = False
args = Hyperparameters()

# torchrun sets these env variables
rank = int(os.environ['RANK'])
world_size = int(os.environ['WORLD_SIZE'])
assert torch.cuda.is_available()
device = torch.device('cuda', int(os.environ['LOCAL_RANK']))
torch.cuda.set_device(device)
dist.init_process_group(backend='nccl', device_id=device)
dist.barrier()
master_process = (rank == 0) # this process will do logging, checkpointing etc.

# begin logging
logfile = None
if master_process:
    run_id = uuid.uuid4()
    os.makedirs('logs', exist_ok=True)
    logfile = f'logs/{run_id}.txt'
    print(logfile)
def print0(s, console=False):
    if master_process:
        with open(logfile, 'a') as f:
             if console:
                 print(s)
             print(s, file=f)

# begin by printing this file (the Python code)
print0(code)
print0('='*100)
# log information about the hardware/software environment this is running on
print0(f'Running Python {sys.version}')
print0(f'Running PyTorch {torch.version.__version__} compiled for CUDA {torch.version.cuda}')
def nvidia_smi():
    import subprocess  # avoid top level import
    return subprocess.run(['nvidia-smi'], stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True).stdout
print0(nvidia_smi())
print0('='*100)

# load data
train_loader = distributed_data_generator(args.train_files, args.batch_size, rank, world_size)

model = GPT(vocab_size=50257, num_layers=12, num_heads=6, model_dim=768).cuda()
for m in model.modules():
    if isinstance(m, nn.Embedding):
        m.bfloat16()
for param in model.parameters():
    dist.broadcast(param.detach(), 0)

# collect the parameters to optimize
hidden_matrix_params = [p for p in model.blocks.parameters() if p.ndim == 2]
embed_params = [model.embed.weight, *model.value_embeds.parameters()]
scalar_params = [p for p in model.parameters() if p.ndim < 2]
head_params = [model.lm_head.weight]

# init the optimizer(s)
adam_params = [dict(params=head_params, lr=0.008), dict(params=embed_params, lr=0.6), dict(params=scalar_params, lr=0.04)]
optimizer1 = torch.optim.Adam(adam_params, betas=(0.8, 0.95), fused=True)
optimizer2 = Muon(hidden_matrix_params, lr=0.05, momentum=0.95, rank=rank, world_size=world_size)
optimizers = [optimizer1, optimizer2]

# learning rate schedule: stable then decay
def get_lr(it: int):
    t = 1 - it / args.num_iterations # time remaining in training
    assert 1 >= t >= 0
    w = min(t / args.cooldown_frac, 1.0) # 1 -> 0
    return w * 1.0 + (1 - w) * 0.1
schedulers = [torch.optim.lr_scheduler.LambdaLR(opt, get_lr) for opt in optimizers]
@lru_cache(1)
def sw_num_blks(window_size: int):
    return torch.tensor(window_size // 128, dtype=torch.int32, pin_memory=True).cuda(non_blocking=True)

model: nn.Module = torch.compile(model)
training_time_ms = 0
# start the clock
torch.cuda.synchronize()
t0 = time.perf_counter()
# begin training
train_steps = args.num_iterations
for step in range(train_steps + 1):
    last_step = (step == train_steps)
    # 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 block-wise sliding window size over training 128 -> 1792:
    # increase by @fernbear.bsky.social; block-wise by @YouJiacheng
    window_size = next_multiple_of_n(1728 * step / train_steps, n=128)
    # --------------- VALIDATION SECTION -----------------
    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)
        model.eval()
        val_bs = world_size * args.seq_len
        assert args.val_tokens % val_bs == 0
        val_steps = args.val_tokens // val_bs
        val_loader = distributed_data_generator(args.val_files, val_bs, rank, world_size)
        val_loss = 0
        with torch.no_grad():
            for _ in range(val_steps):
                x, y = next(val_loader)
                val_loss += model(x, y, sw_num_blks(window_size))
        val_loss /= val_steps
        del val_loader
        dist.all_reduce(val_loss, op=dist.ReduceOp.AVG)
        print0(f'step:{step}/{train_steps} val_loss:{val_loss:.4f} train_time:{training_time_ms:.0f}ms step_avg:{training_time_ms/(timed_steps-1):.2f}ms', console=True)
        model.train()
        # start the clock again
        torch.cuda.synchronize()
        t0 = time.perf_counter()

    if last_step:
        if master_process and args.save_checkpoint:
            log = dict(step=step, code=code, model=model.state_dict(), optimizers=[opt.state_dict() for opt in optimizers])
            os.makedirs(f'logs/{run_id}', exist_ok=True)
            torch.save(log, f'logs/{run_id}/state_step{step:06d}.pt')
        # the last step only has the validation loop, so break to avoid training
        break

    # --------------- TRAINING SECTION BEGIN -----------------
    inputs, targets = next(train_loader)
    for input_seq, target_seq in zip(inputs.split(args.seq_len), targets.split(args.seq_len)):
        model(input_seq, target_seq, sw_num_blks(window_size)).backward()
    for param in model.parameters():
        dist.all_reduce(param.grad, op=dist.ReduceOp.AVG)
    # momentum warmup for Muon
    frac = min(step / 300, 1)
    for group in optimizer2.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)
    # logging
    approx_time = training_time_ms + 1000 * (time.perf_counter() - t0)
    print0(f'step:{step+1}/{train_steps} train_time:{approx_time:.0f}ms step_avg:{approx_time/timed_steps:.2f}ms', console=True)

print0(
    f"peak memory allocated: {torch.cuda.max_memory_allocated() // 1024 // 1024} MiB "
    f"reserved: {torch.cuda.max_memory_reserved() // 1024 // 1024} MiB"
)
dist.destroy_process_group()

====================================================================================================
Running Python 3.12.7 (main, Jan 16 2025, 08:58:39) [GCC 13.2.0]
Running PyTorch 2.7.0.dev20250110+cu126 compiled for CUDA 12.6
Thu Jan 16 10:26:28 2025       
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 550.127.05             Driver Version: 550.127.05     CUDA Version: 12.6     |
|-----------------------------------------+------------------------+----------------------+
| 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  |   00000000:61:00.0 Off |                    0 |
| N/A   31C    P0            133W /  700W |    7746MiB /  81559MiB |      0%      Default |
|                                         |                        |             Disabled |
+-----------------------------------------+------------------------+----------------------+
|   1  NVIDIA H100 80GB HBM3          On  |   00000000:62:00.0 Off |                    0 |
| N/A   32C    P0            125W /  700W |    3456MiB /  81559MiB |      0%      Default |
|                                         |                        |             Disabled |
+-----------------------------------------+------------------------+----------------------+
|   2  NVIDIA H100 80GB HBM3          On  |   00000000:63:00.0 Off |                    0 |
| N/A   34C    P0            127W /  700W |    3456MiB /  81559MiB |      0%      Default |
|                                         |                        |             Disabled |
+-----------------------------------------+------------------------+----------------------+
|   3  NVIDIA H100 80GB HBM3          On  |   00000000:64:00.0 Off |                    0 |
| N/A   30C    P0            116W /  700W |    3456MiB /  81559MiB |      0%      Default |
|                                         |                        |             Disabled |
+-----------------------------------------+------------------------+----------------------+
|   4  NVIDIA H100 80GB HBM3          On  |   00000000:6A:00.0 Off |                    0 |
| N/A   30C    P0            113W /  700W |    3456MiB /  81559MiB |      0%      Default |
|                                         |                        |             Disabled |
+-----------------------------------------+------------------------+----------------------+
|   5  NVIDIA H100 80GB HBM3          On  |   00000000:6B:00.0 Off |                    0 |
| N/A   35C    P0            124W /  700W |    3456MiB /  81559MiB |      0%      Default |
|                                         |                        |             Disabled |
+-----------------------------------------+------------------------+----------------------+
|   6  NVIDIA H100 80GB HBM3          On  |   00000000:6C:00.0 Off |                    0 |
| N/A   35C    P0            126W /  700W |    3456MiB /  81559MiB |      0%      Default |
|                                         |                        |             Disabled |
+-----------------------------------------+------------------------+----------------------+
|   7  NVIDIA H100 80GB HBM3          On  |   00000000:6D:00.0 Off |                    0 |
| N/A   30C    P0            126W /  700W |    3216MiB /  81559MiB |      0%      Default |
|                                         |                        |             Disabled |
+-----------------------------------------+------------------------+----------------------+
                                                                                         
+-----------------------------------------------------------------------------------------+
| Processes:                                                                              |
|  GPU   GI   CI        PID   Type   Process name                              GPU Memory |
|        ID   ID                                                               Usage      |
|=========================================================================================|
+-----------------------------------------------------------------------------------------+

====================================================================================================
step:0/1395 val_loss:10.8258 train_time:0ms step_avg:nanms
step:1/1395 train_time:26881ms step_avg:nanms
step:2/1395 train_time:27284ms step_avg:nanms
step:3/1395 train_time:27403ms step_avg:nanms
step:4/1395 train_time:27526ms step_avg:nanms
step:5/1395 train_time:27650ms step_avg:nanms
step:6/1395 train_time:27773ms step_avg:nanms
step:7/1395 train_time:27896ms step_avg:nanms
step:8/1395 train_time:28018ms step_avg:nanms
step:9/1395 train_time:28142ms step_avg:nanms
step:10/1395 train_time:28267ms step_avg:nanms
step:11/1395 train_time:123ms step_avg:nanms
step:12/1395 train_time:247ms step_avg:nanms
step:13/1395 train_time:371ms step_avg:123.56ms
step:14/1395 train_time:495ms step_avg:123.70ms
step:15/1395 train_time:619ms step_avg:123.78ms
step:16/1395 train_time:742ms step_avg:123.65ms
step:17/1395 train_time:866ms step_avg:123.74ms
step:18/1395 train_time:991ms step_avg:123.93ms
step:19/1395 train_time:1117ms step_avg:124.10ms
step:20/1395 train_time:1241ms step_avg:124.07ms
step:21/1395 train_time:1365ms step_avg:124.09ms
step:22/1395 train_time:1490ms step_avg:124.14ms
step:23/1395 train_time:1614ms step_avg:124.17ms
step:24/1395 train_time:1737ms step_avg:124.09ms
step:25/1395 train_time:1860ms step_avg:124.03ms
step:26/1395 train_time:1985ms step_avg:124.09ms
step:27/1395 train_time:2109ms step_avg:124.05ms
step:28/1395 train_time:2233ms step_avg:124.04ms
step:29/1395 train_time:2357ms step_avg:124.03ms
step:30/1395 train_time:2480ms step_avg:124.02ms
step:31/1395 train_time:2603ms step_avg:123.97ms
step:32/1395 train_time:2727ms step_avg:123.95ms
step:33/1395 train_time:2850ms step_avg:123.93ms
step:34/1395 train_time:2975ms step_avg:123.98ms
step:35/1395 train_time:3100ms step_avg:123.98ms
step:36/1395 train_time:3224ms step_avg:123.99ms
step:37/1395 train_time:3349ms step_avg:124.04ms
step:38/1395 train_time:3473ms step_avg:124.03ms
step:39/1395 train_time:3598ms step_avg:124.07ms
step:40/1395 train_time:3723ms step_avg:124.09ms
step:41/1395 train_time:3847ms step_avg:124.08ms
step:42/1395 train_time:3971ms step_avg:124.08ms
step:43/1395 train_time:4095ms step_avg:124.08ms
step:44/1395 train_time:4220ms step_avg:124.11ms
step:45/1395 train_time:4343ms step_avg:124.07ms
step:46/1395 train_time:4469ms step_avg:124.15ms
step:47/1395 train_time:4593ms step_avg:124.15ms
step:48/1395 train_time:4717ms step_avg:124.13ms
step:49/1395 train_time:4842ms step_avg:124.14ms
step:50/1395 train_time:4965ms step_avg:124.12ms
step:51/1395 train_time:5090ms step_avg:124.14ms
step:52/1395 train_time:5215ms step_avg:124.17ms
step:53/1395 train_time:5339ms step_avg:124.16ms
step:54/1395 train_time:5463ms step_avg:124.15ms
step:55/1395 train_time:5587ms step_avg:124.16ms
step:56/1395 train_time:5711ms step_avg:124.16ms
step:57/1395 train_time:5836ms step_avg:124.17ms
step:58/1395 train_time:5960ms step_avg:124.17ms
step:59/1395 train_time:6084ms step_avg:124.17ms
step:60/1395 train_time:6210ms step_avg:124.19ms
step:61/1395 train_time:6333ms step_avg:124.18ms
step:62/1395 train_time:6457ms step_avg:124.18ms
step:63/1395 train_time:6581ms step_avg:124.17ms
step:64/1395 train_time:6705ms step_avg:124.17ms
step:65/1395 train_time:6831ms step_avg:124.20ms
step:66/1395 train_time:6955ms step_avg:124.20ms
step:67/1395 train_time:7080ms step_avg:124.20ms
step:68/1395 train_time:7203ms step_avg:124.20ms
step:69/1395 train_time:7327ms step_avg:124.18ms
step:70/1395 train_time:7452ms step_avg:124.20ms
step:71/1395 train_time:7576ms step_avg:124.20ms
step:72/1395 train_time:7700ms step_avg:124.20ms
step:73/1395 train_time:7824ms step_avg:124.19ms
step:74/1395 train_time:7949ms step_avg:124.20ms
step:75/1395 train_time:8073ms step_avg:124.20ms
step:76/1395 train_time:8198ms step_avg:124.21ms
step:77/1395 train_time:8322ms step_avg:124.21ms
step:78/1395 train_time:8445ms step_avg:124.19ms
step:79/1395 train_time:8569ms step_avg:124.18ms
step:80/1395 train_time:8693ms step_avg:124.18ms
step:81/1395 train_time:8818ms step_avg:124.20ms
step:82/1395 train_time:8941ms step_avg:124.18ms
step:83/1395 train_time:9065ms step_avg:124.17ms
step:84/1395 train_time:9190ms step_avg:124.19ms
step:85/1395 train_time:9314ms step_avg:124.19ms
step:86/1395 train_time:9439ms step_avg:124.19ms
step:87/1395 train_time:9562ms step_avg:124.19ms
step:88/1395 train_time:9688ms step_avg:124.20ms
step:89/1395 train_time:9812ms step_avg:124.21ms
step:90/1395 train_time:9936ms step_avg:124.20ms
step:91/1395 train_time:10059ms step_avg:124.18ms
step:92/1395 train_time:10182ms step_avg:124.18ms
step:93/1395 train_time:10306ms step_avg:124.17ms
step:94/1395 train_time:10431ms step_avg:124.18ms
step:95/1395 train_time:10556ms step_avg:124.19ms
step:96/1395 train_time:10681ms step_avg:124.20ms
step:97/1395 train_time:10806ms step_avg:124.21ms
step:98/1395 train_time:10930ms step_avg:124.21ms
step:99/1395 train_time:11054ms step_avg:124.20ms
step:100/1395 train_time:11178ms step_avg:124.20ms
step:101/1395 train_time:11301ms step_avg:124.19ms
step:102/1395 train_time:11425ms step_avg:124.19ms
step:103/1395 train_time:11551ms step_avg:124.20ms
step:104/1395 train_time:11676ms step_avg:124.22ms
step:105/1395 train_time:11802ms step_avg:124.23ms
step:106/1395 train_time:11928ms step_avg:124.25ms
step:107/1395 train_time:12055ms step_avg:124.28ms
step:108/1395 train_time:12182ms step_avg:124.30ms
step:109/1395 train_time:12308ms step_avg:124.32ms
step:110/1395 train_time:12434ms step_avg:124.34ms
step:111/1395 train_time:12561ms step_avg:124.37ms
step:112/1395 train_time:12690ms step_avg:124.41ms
step:113/1395 train_time:12817ms step_avg:124.44ms
step:114/1395 train_time:12944ms step_avg:124.46ms
step:115/1395 train_time:13071ms step_avg:124.49ms
step:116/1395 train_time:13199ms step_avg:124.51ms
step:117/1395 train_time:13325ms step_avg:124.53ms
step:118/1395 train_time:13451ms step_avg:124.54ms
step:119/1395 train_time:13578ms step_avg:124.57ms
step:120/1395 train_time:13705ms step_avg:124.59ms
step:121/1395 train_time:13833ms step_avg:124.62ms
step:122/1395 train_time:13960ms step_avg:124.64ms
step:123/1395 train_time:14087ms step_avg:124.66ms
step:124/1395 train_time:14214ms step_avg:124.68ms
step:125/1395 train_time:14341ms step_avg:124.70ms
step:125/1395 val_loss:4.3745 train_time:14441ms step_avg:125.57ms
step:126/1395 train_time:14475ms step_avg:124.78ms
step:127/1395 train_time:14609ms step_avg:124.86ms
step:128/1395 train_time:14736ms step_avg:124.88ms
step:129/1395 train_time:14862ms step_avg:124.89ms
step:130/1395 train_time:14988ms step_avg:124.90ms
step:131/1395 train_time:15115ms step_avg:124.91ms
step:132/1395 train_time:15241ms step_avg:124.93ms
step:133/1395 train_time:15368ms step_avg:124.94ms
step:134/1395 train_time:15494ms step_avg:124.95ms
step:135/1395 train_time:15621ms step_avg:124.96ms
step:136/1395 train_time:15748ms step_avg:124.99ms
step:137/1395 train_time:15877ms step_avg:125.02ms
step:138/1395 train_time:16004ms step_avg:125.03ms
step:139/1395 train_time:16130ms step_avg:125.04ms
step:140/1395 train_time:16255ms step_avg:125.04ms
step:141/1395 train_time:16382ms step_avg:125.05ms
step:142/1395 train_time:16508ms step_avg:125.06ms
step:143/1395 train_time:16634ms step_avg:125.07ms
step:144/1395 train_time:16762ms step_avg:125.09ms
step:145/1395 train_time:16890ms step_avg:125.11ms
step:146/1395 train_time:17016ms step_avg:125.12ms
step:147/1395 train_time:17143ms step_avg:125.13ms
step:148/1395 train_time:17270ms step_avg:125.14ms
step:149/1395 train_time:17397ms step_avg:125.16ms
step:150/1395 train_time:17524ms step_avg:125.17ms
step:151/1395 train_time:17650ms step_avg:125.18ms
step:152/1395 train_time:17776ms step_avg:125.19ms
step:153/1395 train_time:17903ms step_avg:125.19ms
step:154/1395 train_time:18029ms step_avg:125.20ms
step:155/1395 train_time:18155ms step_avg:125.21ms
step:156/1395 train_time:18282ms step_avg:125.22ms
step:157/1395 train_time:18409ms step_avg:125.23ms
step:158/1395 train_time:18535ms step_avg:125.24ms
step:159/1395 train_time:18662ms step_avg:125.25ms
step:160/1395 train_time:18789ms step_avg:125.26ms
step:161/1395 train_time:18916ms step_avg:125.27ms
step:162/1395 train_time:19042ms step_avg:125.27ms
step:163/1395 train_time:19169ms step_avg:125.29ms
step:164/1395 train_time:19295ms step_avg:125.29ms
step:165/1395 train_time:19422ms step_avg:125.30ms
step:166/1395 train_time:19549ms step_avg:125.31ms
step:167/1395 train_time:19675ms step_avg:125.32ms
step:168/1395 train_time:19802ms step_avg:125.33ms
step:169/1395 train_time:19928ms step_avg:125.34ms
step:170/1395 train_time:20055ms step_avg:125.34ms
step:171/1395 train_time:20182ms step_avg:125.35ms
step:172/1395 train_time:20309ms step_avg:125.36ms
step:173/1395 train_time:20435ms step_avg:125.37ms
step:174/1395 train_time:20561ms step_avg:125.37ms
step:175/1395 train_time:20688ms step_avg:125.38ms
step:176/1395 train_time:20814ms step_avg:125.39ms
step:177/1395 train_time:20941ms step_avg:125.40ms
step:178/1395 train_time:21068ms step_avg:125.40ms
step:179/1395 train_time:21194ms step_avg:125.41ms
step:180/1395 train_time:21321ms step_avg:125.42ms
step:181/1395 train_time:21448ms step_avg:125.42ms
step:182/1395 train_time:21574ms step_avg:125.43ms
step:183/1395 train_time:21701ms step_avg:125.44ms
step:184/1395 train_time:21827ms step_avg:125.45ms
step:185/1395 train_time:21954ms step_avg:125.45ms
step:186/1395 train_time:22080ms step_avg:125.46ms
step:187/1395 train_time:22208ms step_avg:125.47ms
step:188/1395 train_time:22334ms step_avg:125.47ms
step:189/1395 train_time:22462ms step_avg:125.49ms
step:190/1395 train_time:22589ms step_avg:125.50ms
step:191/1395 train_time:22716ms step_avg:125.50ms
step:192/1395 train_time:22843ms step_avg:125.51ms
step:193/1395 train_time:22971ms step_avg:125.52ms
step:194/1395 train_time:23097ms step_avg:125.53ms
step:195/1395 train_time:23223ms step_avg:125.53ms
step:196/1395 train_time:23349ms step_avg:125.53ms
step:197/1395 train_time:23475ms step_avg:125.53ms
step:198/1395 train_time:23602ms step_avg:125.54ms
step:199/1395 train_time:23728ms step_avg:125.55ms
step:200/1395 train_time:23855ms step_avg:125.55ms
step:201/1395 train_time:23983ms step_avg:125.57ms
step:202/1395 train_time:24111ms step_avg:125.58ms
step:203/1395 train_time:24238ms step_avg:125.59ms
step:204/1395 train_time:24365ms step_avg:125.59ms
step:205/1395 train_time:24492ms step_avg:125.60ms
step:206/1395 train_time:24617ms step_avg:125.60ms
step:207/1395 train_time:24744ms step_avg:125.60ms
step:208/1395 train_time:24871ms step_avg:125.61ms
step:209/1395 train_time:25000ms step_avg:125.63ms
step:210/1395 train_time:25129ms step_avg:125.64ms
step:211/1395 train_time:25257ms step_avg:125.66ms
step:212/1395 train_time:25387ms step_avg:125.68ms
step:213/1395 train_time:25516ms step_avg:125.70ms
step:214/1395 train_time:25645ms step_avg:125.71ms
step:215/1395 train_time:25775ms step_avg:125.73ms
step:216/1395 train_time:25904ms step_avg:125.75ms
step:217/1395 train_time:26032ms step_avg:125.76ms
step:218/1395 train_time:26159ms step_avg:125.76ms
step:219/1395 train_time:26289ms step_avg:125.79ms
step:220/1395 train_time:26418ms step_avg:125.80ms
step:221/1395 train_time:26548ms step_avg:125.82ms
step:222/1395 train_time:26678ms step_avg:125.84ms
step:223/1395 train_time:26807ms step_avg:125.86ms
step:224/1395 train_time:26937ms step_avg:125.87ms
step:225/1395 train_time:27066ms step_avg:125.89ms
step:226/1395 train_time:27196ms step_avg:125.91ms
step:227/1395 train_time:27325ms step_avg:125.92ms
step:228/1395 train_time:27455ms step_avg:125.94ms
step:229/1395 train_time:27584ms step_avg:125.95ms
step:230/1395 train_time:27714ms step_avg:125.97ms
step:231/1395 train_time:27843ms step_avg:125.99ms
step:232/1395 train_time:27972ms step_avg:126.00ms
step:233/1395 train_time:28101ms step_avg:126.02ms
step:234/1395 train_time:28230ms step_avg:126.03ms
step:235/1395 train_time:28359ms step_avg:126.04ms
step:236/1395 train_time:28489ms step_avg:126.06ms
step:237/1395 train_time:28619ms step_avg:126.08ms
step:238/1395 train_time:28748ms step_avg:126.09ms
step:239/1395 train_time:28878ms step_avg:126.10ms
step:240/1395 train_time:29007ms step_avg:126.12ms
step:241/1395 train_time:29136ms step_avg:126.13ms
step:242/1395 train_time:29264ms step_avg:126.14ms
step:243/1395 train_time:29394ms step_avg:126.15ms
step:244/1395 train_time:29523ms step_avg:126.17ms
step:245/1395 train_time:29652ms step_avg:126.18ms
step:246/1395 train_time:29781ms step_avg:126.19ms
step:247/1395 train_time:29911ms step_avg:126.21ms
step:248/1395 train_time:30039ms step_avg:126.21ms
step:249/1395 train_time:30168ms step_avg:126.23ms
step:250/1395 train_time:30299ms step_avg:126.25ms
step:250/1395 val_loss:3.9501 train_time:30403ms step_avg:126.68ms
step:251/1395 train_time:30437ms step_avg:126.29ms
step:252/1395 train_time:30573ms step_avg:126.33ms
step:253/1395 train_time:30701ms step_avg:126.34ms
step:254/1395 train_time:30830ms step_avg:126.35ms
step:255/1395 train_time:30958ms step_avg:126.36ms
step:256/1395 train_time:31086ms step_avg:126.37ms
step:257/1395 train_time:31215ms step_avg:126.38ms
step:258/1395 train_time:31343ms step_avg:126.38ms
step:259/1395 train_time:31473ms step_avg:126.40ms
step:260/1395 train_time:31603ms step_avg:126.41ms
step:261/1395 train_time:31733ms step_avg:126.43ms
step:262/1395 train_time:31863ms step_avg:126.44ms
step:263/1395 train_time:31991ms step_avg:126.45ms
step:264/1395 train_time:32120ms step_avg:126.46ms
step:265/1395 train_time:32249ms step_avg:126.47ms
step:266/1395 train_time:32378ms step_avg:126.47ms
step:267/1395 train_time:32507ms step_avg:126.49ms
step:268/1395 train_time:32637ms step_avg:126.50ms
step:269/1395 train_time:32766ms step_avg:126.51ms
step:270/1395 train_time:32895ms step_avg:126.52ms
step:271/1395 train_time:33024ms step_avg:126.53ms
step:272/1395 train_time:33154ms step_avg:126.54ms
step:273/1395 train_time:33283ms step_avg:126.55ms
step:274/1395 train_time:33412ms step_avg:126.56ms
step:275/1395 train_time:33541ms step_avg:126.57ms
step:276/1395 train_time:33671ms step_avg:126.58ms
step:277/1395 train_time:33799ms step_avg:126.59ms
step:278/1395 train_time:33929ms step_avg:126.60ms
step:279/1395 train_time:34058ms step_avg:126.61ms
step:280/1395 train_time:34186ms step_avg:126.62ms
step:281/1395 train_time:34316ms step_avg:126.63ms
step:282/1395 train_time:34446ms step_avg:126.64ms
step:283/1395 train_time:34575ms step_avg:126.65ms
step:284/1395 train_time:34706ms step_avg:126.66ms
step:285/1395 train_time:34836ms step_avg:126.68ms
step:286/1395 train_time:34965ms step_avg:126.68ms
step:287/1395 train_time:35093ms step_avg:126.69ms
step:288/1395 train_time:35222ms step_avg:126.70ms
step:289/1395 train_time:35352ms step_avg:126.71ms
step:290/1395 train_time:35480ms step_avg:126.72ms
step:291/1395 train_time:35611ms step_avg:126.73ms
step:292/1395 train_time:35741ms step_avg:126.74ms
step:293/1395 train_time:35872ms step_avg:126.75ms
step:294/1395 train_time:36001ms step_avg:126.77ms
step:295/1395 train_time:36131ms step_avg:126.78ms
step:296/1395 train_time:36261ms step_avg:126.79ms
step:297/1395 train_time:36390ms step_avg:126.79ms
step:298/1395 train_time:36518ms step_avg:126.80ms
step:299/1395 train_time:36648ms step_avg:126.81ms
step:300/1395 train_time:36778ms step_avg:126.82ms
step:301/1395 train_time:36908ms step_avg:126.83ms
step:302/1395 train_time:37039ms step_avg:126.85ms
step:303/1395 train_time:37169ms step_avg:126.85ms
step:304/1395 train_time:37298ms step_avg:126.86ms
step:305/1395 train_time:37427ms step_avg:126.87ms
step:306/1395 train_time:37558ms step_avg:126.88ms
step:307/1395 train_time:37686ms step_avg:126.89ms
step:308/1395 train_time:37817ms step_avg:126.90ms
step:309/1395 train_time:37947ms step_avg:126.91ms
step:310/1395 train_time:38077ms step_avg:126.92ms
step:311/1395 train_time:38206ms step_avg:126.93ms
step:312/1395 train_time:38336ms step_avg:126.94ms
step:313/1395 train_time:38468ms step_avg:126.96ms
step:314/1395 train_time:38600ms step_avg:126.97ms
step:315/1395 train_time:38730ms step_avg:126.98ms
step:316/1395 train_time:38862ms step_avg:127.00ms
step:317/1395 train_time:38993ms step_avg:127.01ms
step:318/1395 train_time:39123ms step_avg:127.02ms
step:319/1395 train_time:39257ms step_avg:127.04ms
step:320/1395 train_time:39388ms step_avg:127.06ms
step:321/1395 train_time:39519ms step_avg:127.07ms
step:322/1395 train_time:39650ms step_avg:127.08ms
step:323/1395 train_time:39780ms step_avg:127.09ms
step:324/1395 train_time:39912ms step_avg:127.11ms
step:325/1395 train_time:40043ms step_avg:127.12ms
step:326/1395 train_time:40175ms step_avg:127.14ms
step:327/1395 train_time:40306ms step_avg:127.15ms
step:328/1395 train_time:40438ms step_avg:127.16ms
step:329/1395 train_time:40571ms step_avg:127.18ms
step:330/1395 train_time:40701ms step_avg:127.19ms
step:331/1395 train_time:40833ms step_avg:127.21ms
step:332/1395 train_time:40966ms step_avg:127.22ms
step:333/1395 train_time:41097ms step_avg:127.23ms
step:334/1395 train_time:41228ms step_avg:127.25ms
step:335/1395 train_time:41359ms step_avg:127.26ms
step:336/1395 train_time:41490ms step_avg:127.27ms
step:337/1395 train_time:41620ms step_avg:127.28ms
step:338/1395 train_time:41754ms step_avg:127.30ms
step:339/1395 train_time:41884ms step_avg:127.31ms
step:340/1395 train_time:42016ms step_avg:127.32ms
step:341/1395 train_time:42148ms step_avg:127.34ms
step:342/1395 train_time:42279ms step_avg:127.35ms
step:343/1395 train_time:42410ms step_avg:127.36ms
step:344/1395 train_time:42542ms step_avg:127.37ms
step:345/1395 train_time:42673ms step_avg:127.38ms
step:346/1395 train_time:42804ms step_avg:127.39ms
step:347/1395 train_time:42937ms step_avg:127.41ms
step:348/1395 train_time:43069ms step_avg:127.42ms
step:349/1395 train_time:43200ms step_avg:127.43ms
step:350/1395 train_time:43331ms step_avg:127.44ms
step:351/1395 train_time:43462ms step_avg:127.45ms
step:352/1395 train_time:43593ms step_avg:127.46ms
step:353/1395 train_time:43724ms step_avg:127.48ms
step:354/1395 train_time:43857ms step_avg:127.49ms
step:355/1395 train_time:43990ms step_avg:127.51ms
step:356/1395 train_time:44122ms step_avg:127.52ms
step:357/1395 train_time:44254ms step_avg:127.53ms
step:358/1395 train_time:44385ms step_avg:127.54ms
step:359/1395 train_time:44517ms step_avg:127.56ms
step:360/1395 train_time:44650ms step_avg:127.57ms
step:361/1395 train_time:44781ms step_avg:127.58ms
step:362/1395 train_time:44914ms step_avg:127.60ms
step:363/1395 train_time:45045ms step_avg:127.61ms
step:364/1395 train_time:45178ms step_avg:127.62ms
step:365/1395 train_time:45310ms step_avg:127.63ms
step:366/1395 train_time:45441ms step_avg:127.64ms
step:367/1395 train_time:45573ms step_avg:127.66ms
step:368/1395 train_time:45703ms step_avg:127.66ms
step:369/1395 train_time:45834ms step_avg:127.67ms
step:370/1395 train_time:45966ms step_avg:127.68ms
step:371/1395 train_time:46098ms step_avg:127.70ms
step:372/1395 train_time:46229ms step_avg:127.71ms
step:373/1395 train_time:46360ms step_avg:127.71ms
step:374/1395 train_time:46491ms step_avg:127.72ms
step:375/1395 train_time:46622ms step_avg:127.73ms
step:375/1395 val_loss:3.7704 train_time:46727ms step_avg:128.02ms
step:376/1395 train_time:46761ms step_avg:127.76ms
step:377/1395 train_time:46896ms step_avg:127.78ms
step:378/1395 train_time:47027ms step_avg:127.79ms
step:379/1395 train_time:47158ms step_avg:127.80ms
step:380/1395 train_time:47289ms step_avg:127.81ms
step:381/1395 train_time:47421ms step_avg:127.82ms
step:382/1395 train_time:47552ms step_avg:127.83ms
step:383/1395 train_time:47684ms step_avg:127.84ms
step:384/1395 train_time:47818ms step_avg:127.86ms
step:385/1395 train_time:47950ms step_avg:127.87ms
step:386/1395 train_time:48082ms step_avg:127.88ms
step:387/1395 train_time:48214ms step_avg:127.89ms
step:388/1395 train_time:48344ms step_avg:127.89ms
step:389/1395 train_time:48475ms step_avg:127.90ms
step:390/1395 train_time:48606ms step_avg:127.91ms
step:391/1395 train_time:48736ms step_avg:127.92ms
step:392/1395 train_time:48868ms step_avg:127.93ms
step:393/1395 train_time:49001ms step_avg:127.94ms
step:394/1395 train_time:49133ms step_avg:127.95ms
step:395/1395 train_time:49265ms step_avg:127.96ms
step:396/1395 train_time:49397ms step_avg:127.97ms
step:397/1395 train_time:49528ms step_avg:127.98ms
step:398/1395 train_time:49661ms step_avg:127.99ms
step:399/1395 train_time:49793ms step_avg:128.00ms
step:400/1395 train_time:49925ms step_avg:128.01ms
step:401/1395 train_time:50056ms step_avg:128.02ms
step:402/1395 train_time:50188ms step_avg:128.03ms
step:403/1395 train_time:50320ms step_avg:128.04ms
step:404/1395 train_time:50451ms step_avg:128.05ms
step:405/1395 train_time:50582ms step_avg:128.06ms
step:406/1395 train_time:50714ms step_avg:128.07ms
step:407/1395 train_time:50846ms step_avg:128.08ms
step:408/1395 train_time:50979ms step_avg:128.09ms
step:409/1395 train_time:51108ms step_avg:128.09ms
step:410/1395 train_time:51242ms step_avg:128.10ms
step:411/1395 train_time:51375ms step_avg:128.12ms
step:412/1395 train_time:51506ms step_avg:128.13ms
step:413/1395 train_time:51638ms step_avg:128.13ms
step:414/1395 train_time:51769ms step_avg:128.14ms
step:415/1395 train_time:51903ms step_avg:128.16ms
step:416/1395 train_time:52036ms step_avg:128.17ms
step:417/1395 train_time:52168ms step_avg:128.18ms
step:418/1395 train_time:52302ms step_avg:128.19ms
step:419/1395 train_time:52435ms step_avg:128.20ms
step:420/1395 train_time:52567ms step_avg:128.21ms
step:421/1395 train_time:52702ms step_avg:128.23ms
step:422/1395 train_time:52834ms step_avg:128.24ms
step:423/1395 train_time:52967ms step_avg:128.25ms
step:424/1395 train_time:53101ms step_avg:128.26ms
step:425/1395 train_time:53233ms step_avg:128.27ms
step:426/1395 train_time:53367ms step_avg:128.29ms
step:427/1395 train_time:53500ms step_avg:128.30ms
step:428/1395 train_time:53632ms step_avg:128.31ms
step:429/1395 train_time:53766ms step_avg:128.32ms
step:430/1395 train_time:53900ms step_avg:128.33ms
step:431/1395 train_time:54033ms step_avg:128.34ms
step:432/1395 train_time:54167ms step_avg:128.36ms
step:433/1395 train_time:54301ms step_avg:128.37ms
step:434/1395 train_time:54433ms step_avg:128.38ms
step:435/1395 train_time:54567ms step_avg:128.39ms
step:436/1395 train_time:54702ms step_avg:128.41ms
step:437/1395 train_time:54835ms step_avg:128.42ms
step:438/1395 train_time:54967ms step_avg:128.43ms
step:439/1395 train_time:55101ms step_avg:128.44ms
step:440/1395 train_time:55233ms step_avg:128.45ms
step:441/1395 train_time:55366ms step_avg:128.46ms
step:442/1395 train_time:55499ms step_avg:128.47ms
step:443/1395 train_time:55631ms step_avg:128.48ms
step:444/1395 train_time:55765ms step_avg:128.49ms
step:445/1395 train_time:55898ms step_avg:128.50ms
step:446/1395 train_time:56031ms step_avg:128.51ms
step:447/1395 train_time:56166ms step_avg:128.53ms
step:448/1395 train_time:56300ms step_avg:128.54ms
step:449/1395 train_time:56433ms step_avg:128.55ms
step:450/1395 train_time:56567ms step_avg:128.56ms
step:451/1395 train_time:56701ms step_avg:128.57ms
step:452/1395 train_time:56833ms step_avg:128.58ms
step:453/1395 train_time:56967ms step_avg:128.59ms
step:454/1395 train_time:57100ms step_avg:128.60ms
step:455/1395 train_time:57233ms step_avg:128.61ms
step:456/1395 train_time:57367ms step_avg:128.63ms
step:457/1395 train_time:57500ms step_avg:128.64ms
step:458/1395 train_time:57633ms step_avg:128.65ms
step:459/1395 train_time:57767ms step_avg:128.66ms
step:460/1395 train_time:57901ms step_avg:128.67ms
step:461/1395 train_time:58034ms step_avg:128.68ms
step:462/1395 train_time:58167ms step_avg:128.69ms
step:463/1395 train_time:58301ms step_avg:128.70ms
step:464/1395 train_time:58434ms step_avg:128.71ms
step:465/1395 train_time:58567ms step_avg:128.72ms
step:466/1395 train_time:58701ms step_avg:128.73ms
step:467/1395 train_time:58834ms step_avg:128.74ms
step:468/1395 train_time:58967ms step_avg:128.75ms
step:469/1395 train_time:59102ms step_avg:128.76ms
step:470/1395 train_time:59236ms step_avg:128.77ms
step:471/1395 train_time:59368ms step_avg:128.78ms
step:472/1395 train_time:59503ms step_avg:128.79ms
step:473/1395 train_time:59636ms step_avg:128.80ms
step:474/1395 train_time:59770ms step_avg:128.81ms
step:475/1395 train_time:59905ms step_avg:128.83ms
step:476/1395 train_time:60039ms step_avg:128.84ms
step:477/1395 train_time:60172ms step_avg:128.85ms
step:478/1395 train_time:60306ms step_avg:128.86ms
step:479/1395 train_time:60437ms step_avg:128.86ms
step:480/1395 train_time:60569ms step_avg:128.87ms
step:481/1395 train_time:60703ms step_avg:128.88ms
step:482/1395 train_time:60836ms step_avg:128.89ms
step:483/1395 train_time:60970ms step_avg:128.90ms
step:484/1395 train_time:61105ms step_avg:128.91ms
step:485/1395 train_time:61240ms step_avg:128.93ms
step:486/1395 train_time:61374ms step_avg:128.94ms
step:487/1395 train_time:61508ms step_avg:128.95ms
step:488/1395 train_time:61640ms step_avg:128.95ms
step:489/1395 train_time:61772ms step_avg:128.96ms
step:490/1395 train_time:61906ms step_avg:128.97ms
step:491/1395 train_time:62040ms step_avg:128.98ms
step:492/1395 train_time:62172ms step_avg:128.99ms
step:493/1395 train_time:62306ms step_avg:129.00ms
step:494/1395 train_time:62439ms step_avg:129.01ms
step:495/1395 train_time:62571ms step_avg:129.01ms
step:496/1395 train_time:62706ms step_avg:129.02ms
step:497/1395 train_time:62837ms step_avg:129.03ms
step:498/1395 train_time:62970ms step_avg:129.04ms
step:499/1395 train_time:63103ms step_avg:129.05ms
step:500/1395 train_time:63235ms step_avg:129.05ms
step:500/1395 val_loss:3.6542 train_time:63342ms step_avg:129.27ms
step:501/1395 train_time:63376ms step_avg:129.07ms
step:502/1395 train_time:63513ms step_avg:129.09ms
step:503/1395 train_time:63647ms step_avg:129.10ms
step:504/1395 train_time:63779ms step_avg:129.11ms
step:505/1395 train_time:63912ms step_avg:129.11ms
step:506/1395 train_time:64044ms step_avg:129.12ms
step:507/1395 train_time:64176ms step_avg:129.13ms
step:508/1395 train_time:64310ms step_avg:129.14ms
step:509/1395 train_time:64444ms step_avg:129.15ms
step:510/1395 train_time:64579ms step_avg:129.16ms
step:511/1395 train_time:64711ms step_avg:129.16ms
step:512/1395 train_time:64845ms step_avg:129.17ms
step:513/1395 train_time:64978ms step_avg:129.18ms
step:514/1395 train_time:65110ms step_avg:129.19ms
step:515/1395 train_time:65242ms step_avg:129.19ms
step:516/1395 train_time:65377ms step_avg:129.20ms
step:517/1395 train_time:65512ms step_avg:129.22ms
step:518/1395 train_time:65646ms step_avg:129.22ms
step:519/1395 train_time:65780ms step_avg:129.23ms
step:520/1395 train_time:65916ms step_avg:129.25ms
step:521/1395 train_time:66051ms step_avg:129.26ms
step:522/1395 train_time:66185ms step_avg:129.27ms
step:523/1395 train_time:66320ms step_avg:129.28ms
step:524/1395 train_time:66455ms step_avg:129.29ms
step:525/1395 train_time:66590ms step_avg:129.30ms
step:526/1395 train_time:66724ms step_avg:129.31ms
step:527/1395 train_time:66859ms step_avg:129.32ms
step:528/1395 train_time:66993ms step_avg:129.33ms
step:529/1395 train_time:67128ms step_avg:129.34ms
step:530/1395 train_time:67262ms step_avg:129.35ms
step:531/1395 train_time:67398ms step_avg:129.36ms
step:532/1395 train_time:67535ms step_avg:129.38ms
step:533/1395 train_time:67670ms step_avg:129.39ms
step:534/1395 train_time:67806ms step_avg:129.40ms
step:535/1395 train_time:67941ms step_avg:129.41ms
step:536/1395 train_time:68078ms step_avg:129.43ms
step:537/1395 train_time:68214ms step_avg:129.44ms
step:538/1395 train_time:68348ms step_avg:129.45ms
step:539/1395 train_time:68484ms step_avg:129.46ms
step:540/1395 train_time:68619ms step_avg:129.47ms
step:541/1395 train_time:68755ms step_avg:129.48ms
step:542/1395 train_time:68892ms step_avg:129.50ms
step:543/1395 train_time:69026ms step_avg:129.50ms
step:544/1395 train_time:69160ms step_avg:129.51ms
step:545/1395 train_time:69294ms step_avg:129.52ms
step:546/1395 train_time:69428ms step_avg:129.53ms
step:547/1395 train_time:69562ms step_avg:129.54ms
step:548/1395 train_time:69698ms step_avg:129.55ms
step:549/1395 train_time:69834ms step_avg:129.56ms
step:550/1395 train_time:69967ms step_avg:129.57ms
step:551/1395 train_time:70102ms step_avg:129.58ms
step:552/1395 train_time:70238ms step_avg:129.59ms
step:553/1395 train_time:70372ms step_avg:129.60ms
step:554/1395 train_time:70507ms step_avg:129.61ms
step:555/1395 train_time:70642ms step_avg:129.62ms
step:556/1395 train_time:70778ms step_avg:129.63ms
step:557/1395 train_time:70914ms step_avg:129.64ms
step:558/1395 train_time:71048ms step_avg:129.65ms
step:559/1395 train_time:71182ms step_avg:129.66ms
step:560/1395 train_time:71318ms step_avg:129.67ms
step:561/1395 train_time:71454ms step_avg:129.68ms
step:562/1395 train_time:71589ms step_avg:129.69ms
step:563/1395 train_time:71722ms step_avg:129.70ms
step:564/1395 train_time:71858ms step_avg:129.71ms
step:565/1395 train_time:71992ms step_avg:129.72ms
step:566/1395 train_time:72128ms step_avg:129.73ms
step:567/1395 train_time:72262ms step_avg:129.73ms
step:568/1395 train_time:72398ms step_avg:129.75ms
step:569/1395 train_time:72536ms step_avg:129.76ms
step:570/1395 train_time:72670ms step_avg:129.77ms
step:571/1395 train_time:72804ms step_avg:129.78ms
step:572/1395 train_time:72941ms step_avg:129.79ms
step:573/1395 train_time:73076ms step_avg:129.80ms
step:574/1395 train_time:73212ms step_avg:129.81ms
step:575/1395 train_time:73346ms step_avg:129.82ms
step:576/1395 train_time:73481ms step_avg:129.83ms
step:577/1395 train_time:73617ms step_avg:129.84ms
step:578/1395 train_time:73751ms step_avg:129.84ms
step:579/1395 train_time:73886ms step_avg:129.85ms
step:580/1395 train_time:74020ms step_avg:129.86ms
step:581/1395 train_time:74155ms step_avg:129.87ms
step:582/1395 train_time:74289ms step_avg:129.88ms
step:583/1395 train_time:74424ms step_avg:129.88ms
step:584/1395 train_time:74560ms step_avg:129.90ms
step:585/1395 train_time:74694ms step_avg:129.90ms
step:586/1395 train_time:74832ms step_avg:129.92ms
step:587/1395 train_time:74964ms step_avg:129.92ms
step:588/1395 train_time:75100ms step_avg:129.93ms
step:589/1395 train_time:75235ms step_avg:129.94ms
step:590/1395 train_time:75372ms step_avg:129.95ms
step:591/1395 train_time:75506ms step_avg:129.96ms
step:592/1395 train_time:75642ms step_avg:129.97ms
step:593/1395 train_time:75777ms step_avg:129.98ms
step:594/1395 train_time:75913ms step_avg:129.99ms
step:595/1395 train_time:76051ms step_avg:130.00ms
step:596/1395 train_time:76184ms step_avg:130.01ms
step:597/1395 train_time:76319ms step_avg:130.01ms
step:598/1395 train_time:76455ms step_avg:130.03ms
step:599/1395 train_time:76589ms step_avg:130.03ms
step:600/1395 train_time:76724ms step_avg:130.04ms
step:601/1395 train_time:76860ms step_avg:130.05ms
step:602/1395 train_time:76997ms step_avg:130.06ms
step:603/1395 train_time:77135ms step_avg:130.08ms
step:604/1395 train_time:77270ms step_avg:130.08ms
step:605/1395 train_time:77404ms step_avg:130.09ms
step:606/1395 train_time:77540ms step_avg:130.10ms
step:607/1395 train_time:77675ms step_avg:130.11ms
step:608/1395 train_time:77809ms step_avg:130.12ms
step:609/1395 train_time:77943ms step_avg:130.12ms
step:610/1395 train_time:78079ms step_avg:130.13ms
step:611/1395 train_time:78214ms step_avg:130.14ms
step:612/1395 train_time:78351ms step_avg:130.15ms
step:613/1395 train_time:78485ms step_avg:130.16ms
step:614/1395 train_time:78620ms step_avg:130.17ms
step:615/1395 train_time:78754ms step_avg:130.17ms
step:616/1395 train_time:78888ms step_avg:130.18ms
step:617/1395 train_time:79023ms step_avg:130.19ms
step:618/1395 train_time:79159ms step_avg:130.20ms
step:619/1395 train_time:79295ms step_avg:130.20ms
step:620/1395 train_time:79430ms step_avg:130.21ms
step:621/1395 train_time:79565ms step_avg:130.22ms
step:622/1395 train_time:79702ms step_avg:130.23ms
step:623/1395 train_time:79840ms step_avg:130.25ms
step:624/1395 train_time:79977ms step_avg:130.26ms
step:625/1395 train_time:80113ms step_avg:130.27ms
step:625/1395 val_loss:3.5718 train_time:80223ms step_avg:130.44ms
step:626/1395 train_time:80257ms step_avg:130.29ms
step:627/1395 train_time:80395ms step_avg:130.30ms
step:628/1395 train_time:80532ms step_avg:130.31ms
step:629/1395 train_time:80670ms step_avg:130.32ms
step:630/1395 train_time:80805ms step_avg:130.33ms
step:631/1395 train_time:80940ms step_avg:130.34ms
step:632/1395 train_time:81076ms step_avg:130.35ms
step:633/1395 train_time:81213ms step_avg:130.36ms
step:634/1395 train_time:81354ms step_avg:130.38ms
step:635/1395 train_time:81491ms step_avg:130.39ms
step:636/1395 train_time:81628ms step_avg:130.40ms
step:637/1395 train_time:81762ms step_avg:130.40ms
step:638/1395 train_time:81897ms step_avg:130.41ms
step:639/1395 train_time:82034ms step_avg:130.42ms
step:640/1395 train_time:82171ms step_avg:130.43ms
step:641/1395 train_time:82308ms step_avg:130.44ms
step:642/1395 train_time:82443ms step_avg:130.45ms
step:643/1395 train_time:82579ms step_avg:130.46ms
step:644/1395 train_time:82714ms step_avg:130.46ms
step:645/1395 train_time:82849ms step_avg:130.47ms
step:646/1395 train_time:82984ms step_avg:130.48ms
step:647/1395 train_time:83119ms step_avg:130.48ms
step:648/1395 train_time:83257ms step_avg:130.50ms
step:649/1395 train_time:83392ms step_avg:130.50ms
step:650/1395 train_time:83530ms step_avg:130.52ms
step:651/1395 train_time:83668ms step_avg:130.53ms
step:652/1395 train_time:83804ms step_avg:130.54ms
step:653/1395 train_time:83939ms step_avg:130.54ms
step:654/1395 train_time:84077ms step_avg:130.55ms
step:655/1395 train_time:84213ms step_avg:130.56ms
step:656/1395 train_time:84350ms step_avg:130.57ms
step:657/1395 train_time:84487ms step_avg:130.58ms
step:658/1395 train_time:84623ms step_avg:130.59ms
step:659/1395 train_time:84760ms step_avg:130.60ms
step:660/1395 train_time:84895ms step_avg:130.61ms
step:661/1395 train_time:85034ms step_avg:130.62ms
step:662/1395 train_time:85171ms step_avg:130.63ms
step:663/1395 train_time:85308ms step_avg:130.64ms
step:664/1395 train_time:85444ms step_avg:130.65ms
step:665/1395 train_time:85579ms step_avg:130.65ms
step:666/1395 train_time:85715ms step_avg:130.66ms
step:667/1395 train_time:85853ms step_avg:130.67ms
step:668/1395 train_time:85989ms step_avg:130.68ms
step:669/1395 train_time:86127ms step_avg:130.69ms
step:670/1395 train_time:86261ms step_avg:130.70ms
step:671/1395 train_time:86397ms step_avg:130.71ms
step:672/1395 train_time:86535ms step_avg:130.72ms
step:673/1395 train_time:86671ms step_avg:130.73ms
step:674/1395 train_time:86806ms step_avg:130.73ms
step:675/1395 train_time:86942ms step_avg:130.74ms
step:676/1395 train_time:87078ms step_avg:130.75ms
step:677/1395 train_time:87216ms step_avg:130.76ms
step:678/1395 train_time:87353ms step_avg:130.77ms
step:679/1395 train_time:87490ms step_avg:130.78ms
step:680/1395 train_time:87627ms step_avg:130.79ms
step:681/1395 train_time:87762ms step_avg:130.79ms
step:682/1395 train_time:87898ms step_avg:130.80ms
step:683/1395 train_time:88037ms step_avg:130.81ms
step:684/1395 train_time:88175ms step_avg:130.82ms
step:685/1395 train_time:88313ms step_avg:130.83ms
step:686/1395 train_time:88450ms step_avg:130.84ms
step:687/1395 train_time:88587ms step_avg:130.85ms
step:688/1395 train_time:88722ms step_avg:130.86ms
step:689/1395 train_time:88860ms step_avg:130.87ms
step:690/1395 train_time:88996ms step_avg:130.88ms
step:691/1395 train_time:89132ms step_avg:130.88ms
step:692/1395 train_time:89269ms step_avg:130.89ms
step:693/1395 train_time:89405ms step_avg:130.90ms
step:694/1395 train_time:89542ms step_avg:130.91ms
step:695/1395 train_time:89677ms step_avg:130.91ms
step:696/1395 train_time:89813ms step_avg:130.92ms
step:697/1395 train_time:89949ms step_avg:130.93ms
step:698/1395 train_time:90084ms step_avg:130.94ms
step:699/1395 train_time:90223ms step_avg:130.95ms
step:700/1395 train_time:90358ms step_avg:130.95ms
step:701/1395 train_time:90497ms step_avg:130.96ms
step:702/1395 train_time:90635ms step_avg:130.98ms
step:703/1395 train_time:90771ms step_avg:130.98ms
step:704/1395 train_time:90907ms step_avg:130.99ms
step:705/1395 train_time:91042ms step_avg:131.00ms
step:706/1395 train_time:91177ms step_avg:131.00ms
step:707/1395 train_time:91313ms step_avg:131.01ms
step:708/1395 train_time:91452ms step_avg:131.02ms
step:709/1395 train_time:91591ms step_avg:131.03ms
step:710/1395 train_time:91727ms step_avg:131.04ms
step:711/1395 train_time:91862ms step_avg:131.04ms
step:712/1395 train_time:92000ms step_avg:131.05ms
step:713/1395 train_time:92137ms step_avg:131.06ms
step:714/1395 train_time:92273ms step_avg:131.07ms
step:715/1395 train_time:92408ms step_avg:131.08ms
step:716/1395 train_time:92545ms step_avg:131.08ms
step:717/1395 train_time:92681ms step_avg:131.09ms
step:718/1395 train_time:92817ms step_avg:131.10ms
step:719/1395 train_time:92955ms step_avg:131.11ms
step:720/1395 train_time:93092ms step_avg:131.12ms
step:721/1395 train_time:93229ms step_avg:131.12ms
step:722/1395 train_time:93364ms step_avg:131.13ms
step:723/1395 train_time:93499ms step_avg:131.13ms
step:724/1395 train_time:93637ms step_avg:131.14ms
step:725/1395 train_time:93774ms step_avg:131.15ms
step:726/1395 train_time:93912ms step_avg:131.16ms
step:727/1395 train_time:94053ms step_avg:131.18ms
step:728/1395 train_time:94189ms step_avg:131.18ms
step:729/1395 train_time:94326ms step_avg:131.19ms
step:730/1395 train_time:94464ms step_avg:131.20ms
step:731/1395 train_time:94601ms step_avg:131.21ms
step:732/1395 train_time:94737ms step_avg:131.22ms
step:733/1395 train_time:94877ms step_avg:131.23ms
step:734/1395 train_time:95014ms step_avg:131.24ms
step:735/1395 train_time:95151ms step_avg:131.24ms
step:736/1395 train_time:95288ms step_avg:131.25ms
step:737/1395 train_time:95429ms step_avg:131.26ms
step:738/1395 train_time:95569ms step_avg:131.28ms
step:739/1395 train_time:95706ms step_avg:131.28ms
step:740/1395 train_time:95846ms step_avg:131.30ms
step:741/1395 train_time:95985ms step_avg:131.31ms
step:742/1395 train_time:96125ms step_avg:131.32ms
step:743/1395 train_time:96263ms step_avg:131.33ms
step:744/1395 train_time:96399ms step_avg:131.33ms
step:745/1395 train_time:96538ms step_avg:131.34ms
step:746/1395 train_time:96674ms step_avg:131.35ms
step:747/1395 train_time:96813ms step_avg:131.36ms
step:748/1395 train_time:96952ms step_avg:131.37ms
step:749/1395 train_time:97092ms step_avg:131.38ms
step:750/1395 train_time:97229ms step_avg:131.39ms
step:750/1395 val_loss:3.5197 train_time:97342ms step_avg:131.54ms
step:751/1395 train_time:97376ms step_avg:131.41ms
step:752/1395 train_time:97515ms step_avg:131.42ms
step:753/1395 train_time:97652ms step_avg:131.43ms
step:754/1395 train_time:97789ms step_avg:131.44ms
step:755/1395 train_time:97928ms step_avg:131.45ms
step:756/1395 train_time:98066ms step_avg:131.46ms
step:757/1395 train_time:98205ms step_avg:131.47ms
step:758/1395 train_time:98341ms step_avg:131.47ms
step:759/1395 train_time:98479ms step_avg:131.48ms
step:760/1395 train_time:98617ms step_avg:131.49ms
step:761/1395 train_time:98756ms step_avg:131.50ms
step:762/1395 train_time:98894ms step_avg:131.51ms
step:763/1395 train_time:99031ms step_avg:131.51ms
step:764/1395 train_time:99169ms step_avg:131.52ms
step:765/1395 train_time:99307ms step_avg:131.53ms
step:766/1395 train_time:99445ms step_avg:131.54ms
step:767/1395 train_time:99582ms step_avg:131.55ms
step:768/1395 train_time:99722ms step_avg:131.56ms
step:769/1395 train_time:99862ms step_avg:131.57ms
step:770/1395 train_time:99999ms step_avg:131.58ms
step:771/1395 train_time:100139ms step_avg:131.59ms
step:772/1395 train_time:100278ms step_avg:131.60ms
step:773/1395 train_time:100417ms step_avg:131.61ms
step:774/1395 train_time:100557ms step_avg:131.62ms
step:775/1395 train_time:100695ms step_avg:131.63ms
step:776/1395 train_time:100834ms step_avg:131.64ms
step:777/1395 train_time:100970ms step_avg:131.64ms
step:778/1395 train_time:101109ms step_avg:131.65ms
step:779/1395 train_time:101246ms step_avg:131.66ms
step:780/1395 train_time:101383ms step_avg:131.67ms
step:781/1395 train_time:101521ms step_avg:131.67ms
step:782/1395 train_time:101660ms step_avg:131.68ms
step:783/1395 train_time:101796ms step_avg:131.69ms
step:784/1395 train_time:101933ms step_avg:131.70ms
step:785/1395 train_time:102071ms step_avg:131.70ms
step:786/1395 train_time:102210ms step_avg:131.71ms
step:787/1395 train_time:102348ms step_avg:131.72ms
step:788/1395 train_time:102484ms step_avg:131.73ms
step:789/1395 train_time:102621ms step_avg:131.73ms
step:790/1395 train_time:102760ms step_avg:131.74ms
step:791/1395 train_time:102897ms step_avg:131.75ms
step:792/1395 train_time:103037ms step_avg:131.76ms
step:793/1395 train_time:103175ms step_avg:131.77ms
step:794/1395 train_time:103315ms step_avg:131.78ms
step:795/1395 train_time:103454ms step_avg:131.79ms
step:796/1395 train_time:103594ms step_avg:131.80ms
step:797/1395 train_time:103732ms step_avg:131.81ms
step:798/1395 train_time:103869ms step_avg:131.81ms
step:799/1395 train_time:104007ms step_avg:131.82ms
step:800/1395 train_time:104143ms step_avg:131.83ms
step:801/1395 train_time:104281ms step_avg:131.83ms
step:802/1395 train_time:104420ms step_avg:131.84ms
step:803/1395 train_time:104556ms step_avg:131.85ms
step:804/1395 train_time:104695ms step_avg:131.86ms
step:805/1395 train_time:104833ms step_avg:131.87ms
step:806/1395 train_time:104973ms step_avg:131.88ms
step:807/1395 train_time:105111ms step_avg:131.88ms
step:808/1395 train_time:105250ms step_avg:131.89ms
step:809/1395 train_time:105386ms step_avg:131.90ms
step:810/1395 train_time:105523ms step_avg:131.90ms
step:811/1395 train_time:105660ms step_avg:131.91ms
step:812/1395 train_time:105800ms step_avg:131.92ms
step:813/1395 train_time:105938ms step_avg:131.93ms
step:814/1395 train_time:106074ms step_avg:131.93ms
step:815/1395 train_time:106211ms step_avg:131.94ms
step:816/1395 train_time:106348ms step_avg:131.95ms
step:817/1395 train_time:106485ms step_avg:131.95ms
step:818/1395 train_time:106621ms step_avg:131.96ms
step:819/1395 train_time:106761ms step_avg:131.97ms
step:820/1395 train_time:106901ms step_avg:131.98ms
step:821/1395 train_time:107038ms step_avg:131.98ms
step:822/1395 train_time:107176ms step_avg:131.99ms
step:823/1395 train_time:107312ms step_avg:132.00ms
step:824/1395 train_time:107448ms step_avg:132.00ms
step:825/1395 train_time:107584ms step_avg:132.00ms
step:826/1395 train_time:107723ms step_avg:132.01ms
step:827/1395 train_time:107861ms step_avg:132.02ms
step:828/1395 train_time:108000ms step_avg:132.03ms
step:829/1395 train_time:108139ms step_avg:132.04ms
step:830/1395 train_time:108278ms step_avg:132.05ms
step:831/1395 train_time:108416ms step_avg:132.05ms
step:832/1395 train_time:108558ms step_avg:132.07ms
step:833/1395 train_time:108697ms step_avg:132.07ms
step:834/1395 train_time:108837ms step_avg:132.08ms
step:835/1395 train_time:108978ms step_avg:132.09ms
step:836/1395 train_time:109118ms step_avg:132.10ms
step:837/1395 train_time:109257ms step_avg:132.11ms
step:838/1395 train_time:109394ms step_avg:132.12ms
step:839/1395 train_time:109531ms step_avg:132.12ms
step:840/1395 train_time:109667ms step_avg:132.13ms
step:841/1395 train_time:109805ms step_avg:132.14ms
step:842/1395 train_time:109945ms step_avg:132.15ms
step:843/1395 train_time:110082ms step_avg:132.15ms
step:844/1395 train_time:110221ms step_avg:132.16ms
step:845/1395 train_time:110358ms step_avg:132.17ms
step:846/1395 train_time:110499ms step_avg:132.18ms
step:847/1395 train_time:110637ms step_avg:132.18ms
step:848/1395 train_time:110773ms step_avg:132.19ms
step:849/1395 train_time:110914ms step_avg:132.20ms
step:850/1395 train_time:111054ms step_avg:132.21ms
step:851/1395 train_time:111194ms step_avg:132.22ms
step:852/1395 train_time:111333ms step_avg:132.22ms
step:853/1395 train_time:111469ms step_avg:132.23ms
step:854/1395 train_time:111607ms step_avg:132.24ms
step:855/1395 train_time:111744ms step_avg:132.24ms
step:856/1395 train_time:111882ms step_avg:132.25ms
step:857/1395 train_time:112020ms step_avg:132.26ms
step:858/1395 train_time:112162ms step_avg:132.27ms
step:859/1395 train_time:112301ms step_avg:132.27ms
step:860/1395 train_time:112437ms step_avg:132.28ms
step:861/1395 train_time:112578ms step_avg:132.29ms
step:862/1395 train_time:112717ms step_avg:132.30ms
step:863/1395 train_time:112859ms step_avg:132.31ms
step:864/1395 train_time:112998ms step_avg:132.32ms
step:865/1395 train_time:113135ms step_avg:132.32ms
step:866/1395 train_time:113281ms step_avg:132.34ms
step:867/1395 train_time:113421ms step_avg:132.35ms
step:868/1395 train_time:113559ms step_avg:132.35ms
step:869/1395 train_time:113696ms step_avg:132.36ms
step:870/1395 train_time:113834ms step_avg:132.37ms
step:871/1395 train_time:113975ms step_avg:132.38ms
step:872/1395 train_time:114113ms step_avg:132.38ms
step:873/1395 train_time:114256ms step_avg:132.39ms
step:874/1395 train_time:114396ms step_avg:132.40ms
step:875/1395 train_time:114537ms step_avg:132.41ms
step:875/1395 val_loss:3.4735 train_time:114648ms step_avg:132.54ms
step:876/1395 train_time:114682ms step_avg:132.43ms
step:877/1395 train_time:114827ms step_avg:132.44ms
step:878/1395 train_time:114966ms step_avg:132.45ms
step:879/1395 train_time:115103ms step_avg:132.45ms
step:880/1395 train_time:115240ms step_avg:132.46ms
step:881/1395 train_time:115379ms step_avg:132.47ms
step:882/1395 train_time:115516ms step_avg:132.47ms
step:883/1395 train_time:115654ms step_avg:132.48ms
step:884/1395 train_time:115794ms step_avg:132.49ms
step:885/1395 train_time:115934ms step_avg:132.50ms
step:886/1395 train_time:116073ms step_avg:132.50ms
step:887/1395 train_time:116210ms step_avg:132.51ms
step:888/1395 train_time:116351ms step_avg:132.52ms
step:889/1395 train_time:116492ms step_avg:132.53ms
step:890/1395 train_time:116628ms step_avg:132.53ms
step:891/1395 train_time:116767ms step_avg:132.54ms
step:892/1395 train_time:116905ms step_avg:132.55ms
step:893/1395 train_time:117043ms step_avg:132.55ms
step:894/1395 train_time:117182ms step_avg:132.56ms
step:895/1395 train_time:117320ms step_avg:132.56ms
step:896/1395 train_time:117459ms step_avg:132.57ms
step:897/1395 train_time:117599ms step_avg:132.58ms
step:898/1395 train_time:117740ms step_avg:132.59ms
step:899/1395 train_time:117880ms step_avg:132.60ms
step:900/1395 train_time:118018ms step_avg:132.60ms
step:901/1395 train_time:118157ms step_avg:132.61ms
step:902/1395 train_time:118294ms step_avg:132.62ms
step:903/1395 train_time:118437ms step_avg:132.63ms
step:904/1395 train_time:118577ms step_avg:132.64ms
step:905/1395 train_time:118715ms step_avg:132.64ms
step:906/1395 train_time:118854ms step_avg:132.65ms
step:907/1395 train_time:118994ms step_avg:132.66ms
step:908/1395 train_time:119133ms step_avg:132.66ms
step:909/1395 train_time:119271ms step_avg:132.67ms
step:910/1395 train_time:119412ms step_avg:132.68ms
step:911/1395 train_time:119550ms step_avg:132.69ms
step:912/1395 train_time:119687ms step_avg:132.69ms
step:913/1395 train_time:119825ms step_avg:132.70ms
step:914/1395 train_time:119964ms step_avg:132.70ms
step:915/1395 train_time:120106ms step_avg:132.71ms
step:916/1395 train_time:120244ms step_avg:132.72ms
step:917/1395 train_time:120384ms step_avg:132.73ms
step:918/1395 train_time:120525ms step_avg:132.74ms
step:919/1395 train_time:120666ms step_avg:132.75ms
step:920/1395 train_time:120804ms step_avg:132.75ms
step:921/1395 train_time:120945ms step_avg:132.76ms
step:922/1395 train_time:121084ms step_avg:132.77ms
step:923/1395 train_time:121221ms step_avg:132.77ms
step:924/1395 train_time:121362ms step_avg:132.78ms
step:925/1395 train_time:121500ms step_avg:132.79ms
step:926/1395 train_time:121639ms step_avg:132.79ms
step:927/1395 train_time:121779ms step_avg:132.80ms
step:928/1395 train_time:121919ms step_avg:132.81ms
step:929/1395 train_time:122057ms step_avg:132.82ms
step:930/1395 train_time:122197ms step_avg:132.82ms
step:931/1395 train_time:122337ms step_avg:132.83ms
step:932/1395 train_time:122477ms step_avg:132.84ms
step:933/1395 train_time:122619ms step_avg:132.85ms
step:934/1395 train_time:122760ms step_avg:132.86ms
step:935/1395 train_time:122903ms step_avg:132.87ms
step:936/1395 train_time:123041ms step_avg:132.87ms
step:937/1395 train_time:123188ms step_avg:132.89ms
step:938/1395 train_time:123328ms step_avg:132.90ms
step:939/1395 train_time:123468ms step_avg:132.90ms
step:940/1395 train_time:123610ms step_avg:132.91ms
step:941/1395 train_time:123748ms step_avg:132.92ms
step:942/1395 train_time:123888ms step_avg:132.93ms
step:943/1395 train_time:124031ms step_avg:132.94ms
step:944/1395 train_time:124180ms step_avg:132.95ms
step:945/1395 train_time:124321ms step_avg:132.96ms
step:946/1395 train_time:124465ms step_avg:132.98ms
step:947/1395 train_time:124607ms step_avg:132.99ms
step:948/1395 train_time:124748ms step_avg:132.99ms
step:949/1395 train_time:124887ms step_avg:133.00ms
step:950/1395 train_time:125025ms step_avg:133.01ms
step:951/1395 train_time:125167ms step_avg:133.01ms
step:952/1395 train_time:125306ms step_avg:133.02ms
step:953/1395 train_time:125447ms step_avg:133.03ms
step:954/1395 train_time:125584ms step_avg:133.03ms
step:955/1395 train_time:125726ms step_avg:133.04ms
step:956/1395 train_time:125867ms step_avg:133.05ms
step:957/1395 train_time:126009ms step_avg:133.06ms
step:958/1395 train_time:126149ms step_avg:133.07ms
step:959/1395 train_time:126291ms step_avg:133.08ms
step:960/1395 train_time:126433ms step_avg:133.09ms
step:961/1395 train_time:126571ms step_avg:133.09ms
step:962/1395 train_time:126710ms step_avg:133.10ms
step:963/1395 train_time:126855ms step_avg:133.11ms
step:964/1395 train_time:126994ms step_avg:133.12ms
step:965/1395 train_time:127133ms step_avg:133.12ms
step:966/1395 train_time:127271ms step_avg:133.13ms
step:967/1395 train_time:127411ms step_avg:133.14ms
step:968/1395 train_time:127550ms step_avg:133.14ms
step:969/1395 train_time:127690ms step_avg:133.15ms
step:970/1395 train_time:127830ms step_avg:133.16ms
step:971/1395 train_time:127969ms step_avg:133.16ms
step:972/1395 train_time:128107ms step_avg:133.17ms
step:973/1395 train_time:128247ms step_avg:133.17ms
step:974/1395 train_time:128388ms step_avg:133.18ms
step:975/1395 train_time:128525ms step_avg:133.19ms
step:976/1395 train_time:128665ms step_avg:133.19ms
step:977/1395 train_time:128803ms step_avg:133.20ms
step:978/1395 train_time:128943ms step_avg:133.21ms
step:979/1395 train_time:129083ms step_avg:133.21ms
step:980/1395 train_time:129223ms step_avg:133.22ms
step:981/1395 train_time:129362ms step_avg:133.23ms
step:982/1395 train_time:129501ms step_avg:133.23ms
step:983/1395 train_time:129639ms step_avg:133.24ms
step:984/1395 train_time:129780ms step_avg:133.24ms
step:985/1395 train_time:129921ms step_avg:133.25ms
step:986/1395 train_time:130067ms step_avg:133.27ms
step:987/1395 train_time:130207ms step_avg:133.27ms
step:988/1395 train_time:130347ms step_avg:133.28ms
step:989/1395 train_time:130485ms step_avg:133.28ms
step:990/1395 train_time:130626ms step_avg:133.29ms
step:991/1395 train_time:130765ms step_avg:133.30ms
step:992/1395 train_time:130906ms step_avg:133.31ms
step:993/1395 train_time:131051ms step_avg:133.32ms
step:994/1395 train_time:131189ms step_avg:133.32ms
step:995/1395 train_time:131326ms step_avg:133.33ms
step:996/1395 train_time:131464ms step_avg:133.33ms
step:997/1395 train_time:131602ms step_avg:133.34ms
step:998/1395 train_time:131740ms step_avg:133.34ms
step:999/1395 train_time:131880ms step_avg:133.35ms
step:1000/1395 train_time:132020ms step_avg:133.35ms
step:1000/1395 val_loss:3.4101 train_time:132138ms step_avg:133.47ms
step:1001/1395 train_time:132171ms step_avg:133.37ms
step:1002/1395 train_time:132309ms step_avg:133.38ms
step:1003/1395 train_time:132451ms step_avg:133.38ms
step:1004/1395 train_time:132591ms step_avg:133.39ms
step:1005/1395 train_time:132732ms step_avg:133.40ms
step:1006/1395 train_time:132872ms step_avg:133.41ms
step:1007/1395 train_time:133012ms step_avg:133.41ms
step:1008/1395 train_time:133156ms step_avg:133.42ms
step:1009/1395 train_time:133299ms step_avg:133.43ms
step:1010/1395 train_time:133437ms step_avg:133.44ms
step:1011/1395 train_time:133577ms step_avg:133.44ms
step:1012/1395 train_time:133715ms step_avg:133.45ms
step:1013/1395 train_time:133858ms step_avg:133.46ms
step:1014/1395 train_time:133997ms step_avg:133.46ms
step:1015/1395 train_time:134135ms step_avg:133.47ms
step:1016/1395 train_time:134276ms step_avg:133.48ms
step:1017/1395 train_time:134418ms step_avg:133.48ms
step:1018/1395 train_time:134557ms step_avg:133.49ms
step:1019/1395 train_time:134698ms step_avg:133.50ms
step:1020/1395 train_time:134841ms step_avg:133.51ms
step:1021/1395 train_time:134979ms step_avg:133.51ms
step:1022/1395 train_time:135118ms step_avg:133.52ms
step:1023/1395 train_time:135259ms step_avg:133.52ms
step:1024/1395 train_time:135397ms step_avg:133.53ms
step:1025/1395 train_time:135538ms step_avg:133.53ms
step:1026/1395 train_time:135676ms step_avg:133.54ms
step:1027/1395 train_time:135817ms step_avg:133.55ms
step:1028/1395 train_time:135960ms step_avg:133.56ms
step:1029/1395 train_time:136103ms step_avg:133.57ms
step:1030/1395 train_time:136243ms step_avg:133.57ms
step:1031/1395 train_time:136381ms step_avg:133.58ms
step:1032/1395 train_time:136520ms step_avg:133.58ms
step:1033/1395 train_time:136657ms step_avg:133.58ms
step:1034/1395 train_time:136798ms step_avg:133.59ms
step:1035/1395 train_time:136941ms step_avg:133.60ms
step:1036/1395 train_time:137080ms step_avg:133.61ms
step:1037/1395 train_time:137224ms step_avg:133.62ms
step:1038/1395 train_time:137369ms step_avg:133.63ms
step:1039/1395 train_time:137510ms step_avg:133.63ms
step:1040/1395 train_time:137650ms step_avg:133.64ms
step:1041/1395 train_time:137791ms step_avg:133.65ms
step:1042/1395 train_time:137933ms step_avg:133.66ms
step:1043/1395 train_time:138074ms step_avg:133.66ms
step:1044/1395 train_time:138217ms step_avg:133.67ms
step:1045/1395 train_time:138362ms step_avg:133.68ms
step:1046/1395 train_time:138503ms step_avg:133.69ms
step:1047/1395 train_time:138643ms step_avg:133.70ms
step:1048/1395 train_time:138782ms step_avg:133.70ms
step:1049/1395 train_time:138920ms step_avg:133.71ms
step:1050/1395 train_time:139059ms step_avg:133.71ms
step:1051/1395 train_time:139201ms step_avg:133.72ms
step:1052/1395 train_time:139342ms step_avg:133.73ms
step:1053/1395 train_time:139480ms step_avg:133.73ms
step:1054/1395 train_time:139623ms step_avg:133.74ms
step:1055/1395 train_time:139763ms step_avg:133.74ms
step:1056/1395 train_time:139900ms step_avg:133.75ms
step:1057/1395 train_time:140039ms step_avg:133.75ms
step:1058/1395 train_time:140181ms step_avg:133.76ms
step:1059/1395 train_time:140323ms step_avg:133.77ms
step:1060/1395 train_time:140465ms step_avg:133.78ms
step:1061/1395 train_time:140605ms step_avg:133.78ms
step:1062/1395 train_time:140749ms step_avg:133.79ms
step:1063/1395 train_time:140889ms step_avg:133.80ms
step:1064/1395 train_time:141032ms step_avg:133.81ms
step:1065/1395 train_time:141173ms step_avg:133.81ms
step:1066/1395 train_time:141315ms step_avg:133.82ms
step:1067/1395 train_time:141458ms step_avg:133.83ms
step:1068/1395 train_time:141598ms step_avg:133.84ms
step:1069/1395 train_time:141740ms step_avg:133.84ms
step:1070/1395 train_time:141879ms step_avg:133.85ms
step:1071/1395 train_time:142024ms step_avg:133.86ms
step:1072/1395 train_time:142164ms step_avg:133.86ms
step:1073/1395 train_time:142302ms step_avg:133.87ms
step:1074/1395 train_time:142442ms step_avg:133.87ms
step:1075/1395 train_time:142585ms step_avg:133.88ms
step:1076/1395 train_time:142727ms step_avg:133.89ms
step:1077/1395 train_time:142869ms step_avg:133.90ms
step:1078/1395 train_time:143012ms step_avg:133.91ms
step:1079/1395 train_time:143158ms step_avg:133.92ms
step:1080/1395 train_time:143299ms step_avg:133.92ms
step:1081/1395 train_time:143439ms step_avg:133.93ms
step:1082/1395 train_time:143578ms step_avg:133.93ms
step:1083/1395 train_time:143718ms step_avg:133.94ms
step:1084/1395 train_time:143864ms step_avg:133.95ms
step:1085/1395 train_time:144004ms step_avg:133.96ms
step:1086/1395 train_time:144146ms step_avg:133.96ms
step:1087/1395 train_time:144284ms step_avg:133.97ms
step:1088/1395 train_time:144427ms step_avg:133.98ms
step:1089/1395 train_time:144570ms step_avg:133.99ms
step:1090/1395 train_time:144714ms step_avg:133.99ms
step:1091/1395 train_time:144855ms step_avg:134.00ms
step:1092/1395 train_time:144997ms step_avg:134.01ms
step:1093/1395 train_time:145136ms step_avg:134.01ms
step:1094/1395 train_time:145277ms step_avg:134.02ms
step:1095/1395 train_time:145417ms step_avg:134.02ms
step:1096/1395 train_time:145561ms step_avg:134.03ms
step:1097/1395 train_time:145703ms step_avg:134.04ms
step:1098/1395 train_time:145845ms step_avg:134.05ms
step:1099/1395 train_time:145986ms step_avg:134.05ms
step:1100/1395 train_time:146126ms step_avg:134.06ms
step:1101/1395 train_time:146265ms step_avg:134.07ms
step:1102/1395 train_time:146406ms step_avg:134.07ms
step:1103/1395 train_time:146549ms step_avg:134.08ms
step:1104/1395 train_time:146690ms step_avg:134.09ms
step:1105/1395 train_time:146836ms step_avg:134.10ms
step:1106/1395 train_time:146978ms step_avg:134.10ms
step:1107/1395 train_time:147117ms step_avg:134.11ms
step:1108/1395 train_time:147262ms step_avg:134.12ms
step:1109/1395 train_time:147402ms step_avg:134.12ms
step:1110/1395 train_time:147542ms step_avg:134.13ms
step:1111/1395 train_time:147682ms step_avg:134.13ms
step:1112/1395 train_time:147820ms step_avg:134.14ms
step:1113/1395 train_time:147958ms step_avg:134.14ms
step:1114/1395 train_time:148104ms step_avg:134.15ms
step:1115/1395 train_time:148249ms step_avg:134.16ms
step:1116/1395 train_time:148389ms step_avg:134.17ms
step:1117/1395 train_time:148529ms step_avg:134.17ms
step:1118/1395 train_time:148676ms step_avg:134.18ms
step:1119/1395 train_time:148814ms step_avg:134.19ms
step:1120/1395 train_time:148957ms step_avg:134.20ms
step:1121/1395 train_time:149099ms step_avg:134.20ms
step:1122/1395 train_time:149239ms step_avg:134.21ms
step:1123/1395 train_time:149379ms step_avg:134.21ms
step:1124/1395 train_time:149519ms step_avg:134.22ms
step:1125/1395 train_time:149658ms step_avg:134.22ms
step:1125/1395 val_loss:3.3609 train_time:149772ms step_avg:134.32ms
step:1126/1395 train_time:149805ms step_avg:134.23ms
step:1127/1395 train_time:149950ms step_avg:134.24ms
step:1128/1395 train_time:150093ms step_avg:134.25ms
step:1129/1395 train_time:150240ms step_avg:134.26ms
step:1130/1395 train_time:150382ms step_avg:134.27ms
step:1131/1395 train_time:150529ms step_avg:134.28ms
step:1132/1395 train_time:150668ms step_avg:134.29ms
step:1133/1395 train_time:150808ms step_avg:134.29ms
step:1134/1395 train_time:150949ms step_avg:134.30ms
step:1135/1395 train_time:151088ms step_avg:134.30ms
step:1136/1395 train_time:151234ms step_avg:134.31ms
step:1137/1395 train_time:151374ms step_avg:134.32ms
step:1138/1395 train_time:151518ms step_avg:134.32ms
step:1139/1395 train_time:151659ms step_avg:134.33ms
step:1140/1395 train_time:151800ms step_avg:134.34ms
step:1141/1395 train_time:151940ms step_avg:134.34ms
step:1142/1395 train_time:152081ms step_avg:134.35ms
step:1143/1395 train_time:152224ms step_avg:134.36ms
step:1144/1395 train_time:152365ms step_avg:134.36ms
step:1145/1395 train_time:152504ms step_avg:134.36ms
step:1146/1395 train_time:152645ms step_avg:134.37ms
step:1147/1395 train_time:152790ms step_avg:134.38ms
step:1148/1395 train_time:152933ms step_avg:134.39ms
step:1149/1395 train_time:153075ms step_avg:134.39ms
step:1150/1395 train_time:153216ms step_avg:134.40ms
step:1151/1395 train_time:153362ms step_avg:134.41ms
step:1152/1395 train_time:153503ms step_avg:134.42ms
step:1153/1395 train_time:153649ms step_avg:134.43ms
step:1154/1395 train_time:153788ms step_avg:134.43ms
step:1155/1395 train_time:153929ms step_avg:134.44ms
step:1156/1395 train_time:154079ms step_avg:134.45ms
step:1157/1395 train_time:154221ms step_avg:134.46ms
step:1158/1395 train_time:154361ms step_avg:134.46ms
step:1159/1395 train_time:154503ms step_avg:134.47ms
step:1160/1395 train_time:154644ms step_avg:134.47ms
step:1161/1395 train_time:154787ms step_avg:134.48ms
step:1162/1395 train_time:154929ms step_avg:134.49ms
step:1163/1395 train_time:155070ms step_avg:134.49ms
step:1164/1395 train_time:155212ms step_avg:134.50ms
step:1165/1395 train_time:155352ms step_avg:134.50ms
step:1166/1395 train_time:155493ms step_avg:134.51ms
step:1167/1395 train_time:155635ms step_avg:134.52ms
step:1168/1395 train_time:155776ms step_avg:134.52ms
step:1169/1395 train_time:155918ms step_avg:134.53ms
step:1170/1395 train_time:156059ms step_avg:134.53ms
step:1171/1395 train_time:156200ms step_avg:134.54ms
step:1172/1395 train_time:156344ms step_avg:134.55ms
step:1173/1395 train_time:156485ms step_avg:134.55ms
step:1174/1395 train_time:156638ms step_avg:134.57ms
step:1175/1395 train_time:156781ms step_avg:134.58ms
step:1176/1395 train_time:156923ms step_avg:134.58ms
step:1177/1395 train_time:157070ms step_avg:134.59ms
step:1178/1395 train_time:157211ms step_avg:134.60ms
step:1179/1395 train_time:157350ms step_avg:134.60ms
step:1180/1395 train_time:157497ms step_avg:134.61ms
step:1181/1395 train_time:157641ms step_avg:134.62ms
step:1182/1395 train_time:157781ms step_avg:134.63ms
step:1183/1395 train_time:157923ms step_avg:134.63ms
step:1184/1395 train_time:158065ms step_avg:134.64ms
step:1185/1395 train_time:158207ms step_avg:134.64ms
step:1186/1395 train_time:158350ms step_avg:134.65ms
step:1187/1395 train_time:158502ms step_avg:134.67ms
step:1188/1395 train_time:158643ms step_avg:134.67ms
step:1189/1395 train_time:158791ms step_avg:134.68ms
step:1190/1395 train_time:158932ms step_avg:134.69ms
step:1191/1395 train_time:159074ms step_avg:134.69ms
step:1192/1395 train_time:159213ms step_avg:134.70ms
step:1193/1395 train_time:159352ms step_avg:134.70ms
step:1194/1395 train_time:159493ms step_avg:134.71ms
step:1195/1395 train_time:159636ms step_avg:134.71ms
step:1196/1395 train_time:159779ms step_avg:134.72ms
step:1197/1395 train_time:159921ms step_avg:134.73ms
step:1198/1395 train_time:160070ms step_avg:134.74ms
step:1199/1395 train_time:160211ms step_avg:134.74ms
step:1200/1395 train_time:160351ms step_avg:134.75ms
step:1201/1395 train_time:160490ms step_avg:134.75ms
step:1202/1395 train_time:160643ms step_avg:134.77ms
step:1203/1395 train_time:160792ms step_avg:134.78ms
step:1204/1395 train_time:160938ms step_avg:134.79ms
step:1205/1395 train_time:161080ms step_avg:134.80ms
step:1206/1395 train_time:161224ms step_avg:134.80ms
step:1207/1395 train_time:161368ms step_avg:134.81ms
step:1208/1395 train_time:161513ms step_avg:134.82ms
step:1209/1395 train_time:161656ms step_avg:134.83ms
step:1210/1395 train_time:161801ms step_avg:134.83ms
step:1211/1395 train_time:161947ms step_avg:134.84ms
step:1212/1395 train_time:162089ms step_avg:134.85ms
step:1213/1395 train_time:162231ms step_avg:134.86ms
step:1214/1395 train_time:162375ms step_avg:134.86ms
step:1215/1395 train_time:162520ms step_avg:134.87ms
step:1216/1395 train_time:162659ms step_avg:134.88ms
step:1217/1395 train_time:162803ms step_avg:134.88ms
step:1218/1395 train_time:162944ms step_avg:134.89ms
step:1219/1395 train_time:163085ms step_avg:134.89ms
step:1220/1395 train_time:163228ms step_avg:134.90ms
step:1221/1395 train_time:163367ms step_avg:134.90ms
step:1222/1395 train_time:163508ms step_avg:134.91ms
step:1223/1395 train_time:163648ms step_avg:134.91ms
step:1224/1395 train_time:163791ms step_avg:134.92ms
step:1225/1395 train_time:163935ms step_avg:134.93ms
step:1226/1395 train_time:164075ms step_avg:134.93ms
step:1227/1395 train_time:164215ms step_avg:134.93ms
step:1228/1395 train_time:164357ms step_avg:134.94ms
step:1229/1395 train_time:164499ms step_avg:134.95ms
step:1230/1395 train_time:164647ms step_avg:134.96ms
step:1231/1395 train_time:164791ms step_avg:134.96ms
step:1232/1395 train_time:164933ms step_avg:134.97ms
step:1233/1395 train_time:165075ms step_avg:134.98ms
step:1234/1395 train_time:165217ms step_avg:134.98ms
step:1235/1395 train_time:165360ms step_avg:134.99ms
step:1236/1395 train_time:165502ms step_avg:134.99ms
step:1237/1395 train_time:165645ms step_avg:135.00ms
step:1238/1395 train_time:165796ms step_avg:135.01ms
step:1239/1395 train_time:165939ms step_avg:135.02ms
step:1240/1395 train_time:166083ms step_avg:135.03ms
step:1241/1395 train_time:166228ms step_avg:135.04ms
step:1242/1395 train_time:166369ms step_avg:135.04ms
step:1243/1395 train_time:166517ms step_avg:135.05ms
step:1244/1395 train_time:166658ms step_avg:135.05ms
step:1245/1395 train_time:166799ms step_avg:135.06ms
step:1246/1395 train_time:166942ms step_avg:135.07ms
step:1247/1395 train_time:167086ms step_avg:135.07ms
step:1248/1395 train_time:167226ms step_avg:135.08ms
step:1249/1395 train_time:167366ms step_avg:135.08ms
step:1250/1395 train_time:167508ms step_avg:135.09ms
step:1250/1395 val_loss:3.3151 train_time:167627ms step_avg:135.18ms
step:1251/1395 train_time:167660ms step_avg:135.10ms
step:1252/1395 train_time:167805ms step_avg:135.11ms
step:1253/1395 train_time:167944ms step_avg:135.11ms
step:1254/1395 train_time:168084ms step_avg:135.12ms
step:1255/1395 train_time:168241ms step_avg:135.13ms
step:1256/1395 train_time:168382ms step_avg:135.14ms
step:1257/1395 train_time:168524ms step_avg:135.14ms
step:1258/1395 train_time:168670ms step_avg:135.15ms
step:1259/1395 train_time:168818ms step_avg:135.16ms
step:1260/1395 train_time:168957ms step_avg:135.17ms
step:1261/1395 train_time:169102ms step_avg:135.17ms
step:1262/1395 train_time:169245ms step_avg:135.18ms
step:1263/1395 train_time:169392ms step_avg:135.19ms
step:1264/1395 train_time:169536ms step_avg:135.20ms
step:1265/1395 train_time:169676ms step_avg:135.20ms
step:1266/1395 train_time:169821ms step_avg:135.21ms
step:1267/1395 train_time:169965ms step_avg:135.22ms
step:1268/1395 train_time:170110ms step_avg:135.22ms
step:1269/1395 train_time:170257ms step_avg:135.23ms
step:1270/1395 train_time:170401ms step_avg:135.24ms
step:1271/1395 train_time:170541ms step_avg:135.24ms
step:1272/1395 train_time:170682ms step_avg:135.25ms
step:1273/1395 train_time:170824ms step_avg:135.25ms
step:1274/1395 train_time:170967ms step_avg:135.26ms
step:1275/1395 train_time:171112ms step_avg:135.27ms
step:1276/1395 train_time:171253ms step_avg:135.27ms
step:1277/1395 train_time:171395ms step_avg:135.28ms
step:1278/1395 train_time:171537ms step_avg:135.28ms
step:1279/1395 train_time:171679ms step_avg:135.29ms
step:1280/1395 train_time:171831ms step_avg:135.30ms
step:1281/1395 train_time:171973ms step_avg:135.31ms
step:1282/1395 train_time:172115ms step_avg:135.31ms
step:1283/1395 train_time:172255ms step_avg:135.31ms
step:1284/1395 train_time:172397ms step_avg:135.32ms
step:1285/1395 train_time:172541ms step_avg:135.33ms
step:1286/1395 train_time:172683ms step_avg:135.33ms
step:1287/1395 train_time:172827ms step_avg:135.34ms
step:1288/1395 train_time:172967ms step_avg:135.34ms
step:1289/1395 train_time:173119ms step_avg:135.35ms
step:1290/1395 train_time:173267ms step_avg:135.36ms
step:1291/1395 train_time:173412ms step_avg:135.37ms
step:1292/1395 train_time:173557ms step_avg:135.38ms
step:1293/1395 train_time:173706ms step_avg:135.39ms
step:1294/1395 train_time:173846ms step_avg:135.39ms
step:1295/1395 train_time:173988ms step_avg:135.40ms
step:1296/1395 train_time:174137ms step_avg:135.41ms
step:1297/1395 train_time:174283ms step_avg:135.42ms
step:1298/1395 train_time:174425ms step_avg:135.42ms
step:1299/1395 train_time:174566ms step_avg:135.43ms
step:1300/1395 train_time:174705ms step_avg:135.43ms
step:1301/1395 train_time:174846ms step_avg:135.43ms
step:1302/1395 train_time:174987ms step_avg:135.44ms
step:1303/1395 train_time:175132ms step_avg:135.45ms
step:1304/1395 train_time:175277ms step_avg:135.45ms
step:1305/1395 train_time:175419ms step_avg:135.46ms
step:1306/1395 train_time:175561ms step_avg:135.46ms
step:1307/1395 train_time:175703ms step_avg:135.47ms
step:1308/1395 train_time:175847ms step_avg:135.48ms
step:1309/1395 train_time:175992ms step_avg:135.48ms
step:1310/1395 train_time:176135ms step_avg:135.49ms
step:1311/1395 train_time:176276ms step_avg:135.49ms
step:1312/1395 train_time:176418ms step_avg:135.50ms
step:1313/1395 train_time:176561ms step_avg:135.50ms
step:1314/1395 train_time:176704ms step_avg:135.51ms
step:1315/1395 train_time:176848ms step_avg:135.52ms
step:1316/1395 train_time:176989ms step_avg:135.52ms
step:1317/1395 train_time:177132ms step_avg:135.53ms
step:1318/1395 train_time:177280ms step_avg:135.53ms
step:1319/1395 train_time:177424ms step_avg:135.54ms
step:1320/1395 train_time:177566ms step_avg:135.55ms
step:1321/1395 train_time:177706ms step_avg:135.55ms
step:1322/1395 train_time:177856ms step_avg:135.56ms
step:1323/1395 train_time:178000ms step_avg:135.57ms
step:1324/1395 train_time:178141ms step_avg:135.57ms
step:1325/1395 train_time:178284ms step_avg:135.58ms
step:1326/1395 train_time:178430ms step_avg:135.59ms
step:1327/1395 train_time:178572ms step_avg:135.59ms
step:1328/1395 train_time:178711ms step_avg:135.59ms
step:1329/1395 train_time:178865ms step_avg:135.61ms
step:1330/1395 train_time:179012ms step_avg:135.61ms
step:1331/1395 train_time:179158ms step_avg:135.62ms
step:1332/1395 train_time:179308ms step_avg:135.63ms
step:1333/1395 train_time:179452ms step_avg:135.64ms
step:1334/1395 train_time:179597ms step_avg:135.65ms
step:1335/1395 train_time:179736ms step_avg:135.65ms
step:1336/1395 train_time:179890ms step_avg:135.66ms
step:1337/1395 train_time:180035ms step_avg:135.67ms
step:1338/1395 train_time:180178ms step_avg:135.68ms
step:1339/1395 train_time:180320ms step_avg:135.68ms
step:1340/1395 train_time:180465ms step_avg:135.69ms
step:1341/1395 train_time:180605ms step_avg:135.69ms
step:1342/1395 train_time:180746ms step_avg:135.70ms
step:1343/1395 train_time:180889ms step_avg:135.70ms
step:1344/1395 train_time:181033ms step_avg:135.71ms
step:1345/1395 train_time:181176ms step_avg:135.71ms
step:1346/1395 train_time:181319ms step_avg:135.72ms
step:1347/1395 train_time:181465ms step_avg:135.73ms
step:1348/1395 train_time:181607ms step_avg:135.73ms
step:1349/1395 train_time:181748ms step_avg:135.73ms
step:1350/1395 train_time:181889ms step_avg:135.74ms
step:1351/1395 train_time:182032ms step_avg:135.74ms
step:1352/1395 train_time:182183ms step_avg:135.76ms
step:1353/1395 train_time:182329ms step_avg:135.76ms
step:1354/1395 train_time:182472ms step_avg:135.77ms
step:1355/1395 train_time:182614ms step_avg:135.77ms
step:1356/1395 train_time:182756ms step_avg:135.78ms
step:1357/1395 train_time:182900ms step_avg:135.78ms
step:1358/1395 train_time:183044ms step_avg:135.79ms
step:1359/1395 train_time:183185ms step_avg:135.79ms
step:1360/1395 train_time:183329ms step_avg:135.80ms
step:1361/1395 train_time:183475ms step_avg:135.81ms
step:1362/1395 train_time:183622ms step_avg:135.81ms
step:1363/1395 train_time:183775ms step_avg:135.83ms
step:1364/1395 train_time:183920ms step_avg:135.83ms
step:1365/1395 train_time:184059ms step_avg:135.84ms
step:1366/1395 train_time:184205ms step_avg:135.84ms
step:1367/1395 train_time:184347ms step_avg:135.85ms
step:1368/1395 train_time:184491ms step_avg:135.85ms
step:1369/1395 train_time:184642ms step_avg:135.87ms
step:1370/1395 train_time:184791ms step_avg:135.88ms
step:1371/1395 train_time:184936ms step_avg:135.88ms
step:1372/1395 train_time:185083ms step_avg:135.89ms
step:1373/1395 train_time:185225ms step_avg:135.89ms
step:1374/1395 train_time:185373ms step_avg:135.90ms
step:1375/1395 train_time:185518ms step_avg:135.91ms
step:1375/1395 val_loss:3.2814 train_time:185631ms step_avg:135.99ms
step:1376/1395 train_time:185665ms step_avg:135.92ms
step:1377/1395 train_time:185816ms step_avg:135.93ms
step:1378/1395 train_time:185958ms step_avg:135.93ms
step:1379/1395 train_time:186101ms step_avg:135.94ms
step:1380/1395 train_time:186245ms step_avg:135.95ms
step:1381/1395 train_time:186394ms step_avg:135.95ms
step:1382/1395 train_time:186537ms step_avg:135.96ms
step:1383/1395 train_time:186678ms step_avg:135.96ms
step:1384/1395 train_time:186828ms step_avg:135.97ms
step:1385/1395 train_time:186968ms step_avg:135.98ms
step:1386/1395 train_time:187113ms step_avg:135.98ms
step:1387/1395 train_time:187256ms step_avg:135.99ms
step:1388/1395 train_time:187397ms step_avg:135.99ms
step:1389/1395 train_time:187540ms step_avg:136.00ms
step:1390/1395 train_time:187683ms step_avg:136.00ms
step:1391/1395 train_time:187825ms step_avg:136.01ms
step:1392/1395 train_time:187969ms step_avg:136.01ms
step:1393/1395 train_time:188114ms step_avg:136.02ms
step:1394/1395 train_time:188257ms step_avg:136.02ms
step:1395/1395 train_time:188397ms step_avg:136.03ms
step:1395/1395 val_loss:3.2770 train_time:188512ms step_avg:136.11ms
peak memory allocated: 37620 MiB reserved: 39014 MiB