Add files using upload-large-folder tool

checkpoints/quarantine/20260818T074500Z/official/langflow/owt/config.json

@@ -0,0 +1,29 @@
+{
+  "_name_or_path": "chumengl/langflow-owt",
+  "architectures": [
+    "LangFlow"
+  ],
+  "auto_map": {
+    "AutoConfig": "config.LangFlowConfig",
+    "AutoModelForMaskedLM": "model.LangFlow"
+  },
+  "model_type": "LangFlow",
+  "vocab_size": 50257,
+  "hidden_size": 768,
+  "cond_dim": 128,
+  "n_blocks": 12,
+  "n_heads": 12,
+  "dropout": 0.1,
+  "model_length": 1024,
+  "use_normalized_embedding": true,
+  "embedding_norm_method": "layernorm",
+  "self_conditioning": true,
+  "use_bias": true,
+  "gumbel_loc": 4.723,
+  "gumbel_scale": 0.852,
+  "gumbel_cutoff": 1e-5,
+  "gumbel_entropy": 7.02,
+  "return_dict": true,
+  "torch_dtype": "float32",
+  "transformers_version": "4.38.2"
+}

checkpoints/quarantine/20260818T074500Z/official/langflow/owt/model.py

@@ -0,0 +1,560 @@
+"""HuggingFace model implementation for LangFlow.
+
+LangFlow is a continuous diffusion language model that operates in embedding space.
+"""
+
+import math
+import typing
+
+import einops
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import transformers
+
+from .config import LangFlowConfig
+
+
+# Flags required to enable jit fusion kernels
+torch._C._jit_set_profiling_mode(False)
+torch._C._jit_set_profiling_executor(False)
+torch._C._jit_override_can_fuse_on_cpu(True)
+torch._C._jit_override_can_fuse_on_gpu(True)
+
+
+def bias_dropout_add_scale(
+    x: torch.Tensor,
+    bias: typing.Optional[torch.Tensor],
+    scale: torch.Tensor,
+    residual: typing.Optional[torch.Tensor],
+    prob: float,
+    training: bool) -> torch.Tensor:
+    if bias is not None:
+        out = scale * F.dropout(x + bias, p=prob, training=training)
+    else:
+        out = scale * F.dropout(x, p=prob, training=training)
+
+    if residual is not None:
+        out = residual + out
+    return out
+
+
+@torch.jit.script
+def bias_dropout_add_scale_fused_train(
+    x: torch.Tensor,
+    bias: typing.Optional[torch.Tensor],
+    scale: torch.Tensor,
+    residual: typing.Optional[torch.Tensor],
+    prob: float) -> torch.Tensor:
+    return bias_dropout_add_scale(x, bias, scale, residual, prob, True)
+
+
+@torch.jit.script
+def bias_dropout_add_scale_fused_inference(
+    x: torch.Tensor,
+    bias: typing.Optional[torch.Tensor],
+    scale: torch.Tensor,
+    residual: typing.Optional[torch.Tensor],
+    prob: float) -> torch.Tensor:
+    return bias_dropout_add_scale(x, bias, scale, residual, prob, False)
+
+
+@torch.jit.script
+def modulate_fused(x: torch.Tensor,
+                   shift: torch.Tensor,
+                   scale: torch.Tensor) -> torch.Tensor:
+    return x * (1 + scale) + shift
+
+
+class Rotary(nn.Module):
+    def __init__(self, dim, base=10_000):
+        super().__init__()
+        inv_freq = 1.0 / (base ** (torch.arange(0, dim, 2).float() / dim))
+        self.register_buffer('inv_freq', inv_freq)
+        self.seq_len_cached = None
+        self.cos_cached = None
+        self.sin_cached = None
+
+    def forward(self, x, seq_dim=1):
+        seq_len = x.shape[seq_dim]
+        if seq_len != self.seq_len_cached:
+            self.seq_len_cached = seq_len
+            t = torch.arange(x.shape[seq_dim], device=x.device).type_as(self.inv_freq)
+            freqs = torch.einsum("i,j->ij", t, self.inv_freq.clone())
+            emb = torch.cat((freqs, freqs), dim=-1).to(x.device)
+            self.cos_cached = emb.cos()[None, :, None, None, :].repeat(1, 1, 3, 1, 1)
+            self.sin_cached = emb.sin()[None, :, None, None, :].repeat(1, 1, 3, 1, 1)
+            self.cos_cached[:, :, 2, :, :].fill_(1.)
+            self.sin_cached[:, :, 2, :, :].fill_(0.)
+        return self.cos_cached, self.sin_cached
+
+
+def _apply_rotary_emb(x, cos, sin):
+    # x: [batch, seqlen, nheads, headdim]
+    # cos, sin: [seqlen, headdim//2]
+    ro_dim = cos.shape[-1] * 2
+    # Expand to [1, seqlen, 1, ro_dim] for broadcasting
+    cos = torch.cat([cos, cos], dim=-1)[None, :, None, :]
+    sin = torch.cat([sin, sin], dim=-1)[None, :, None, :]
+    x_rot = x[..., :ro_dim]
+    x1, x2 = x_rot.chunk(2, dim=-1)
+    x_rotated = torch.cat([-x2, x1], dim=-1)
+    return torch.cat([x_rot * cos + x_rotated * sin, x[..., ro_dim:]], dim=-1)
+
+
+def split_and_apply_rotary_pos_emb(qkv, rotary_cos_sin):
+    with torch.autocast(device_type='cuda', enabled=False):
+        cos, sin = rotary_cos_sin
+        cos = cos.to(qkv.dtype)
+        sin = sin.to(qkv.dtype)
+        cos = cos[0, :, 0, 0, :cos.shape[-1]//2]
+        sin = sin[0, :, 0, 0, :sin.shape[-1]//2]
+        q, k, v = qkv.chunk(3, dim=2)
+        q = _apply_rotary_emb(q.squeeze(dim=2), cos, sin)
+        k = _apply_rotary_emb(k.squeeze(dim=2), cos, sin)
+        v = v.squeeze(dim=2)
+    return q, k, v
+
+
+def regular_attention_multi_headed(q, k, v):
+    attention_output = F.scaled_dot_product_attention(
+        query=q.transpose(1, 2),
+        key=k.transpose(1, 2),
+        value=v.transpose(1, 2),
+        attn_mask=None,
+        dropout_p=0.0,
+        is_causal=False)
+    attention_output = attention_output.transpose(1, 2)
+    return einops.rearrange(attention_output, 'b s h d -> b s (h d)')
+
+
+class LayerNorm(nn.Module):
+    def __init__(self, dim):
+        super().__init__()
+        self.weight = nn.Parameter(torch.ones([dim]))
+        self.dim = dim
+
+    def forward(self, x):
+        with torch.autocast(device_type='cuda', enabled=False):
+            x = F.layer_norm(x.float(), [self.dim])
+        return x * self.weight[None, None, :]
+
+
+class TimestepEmbedder(nn.Module):
+    """Embeds scalar timesteps into vector representations."""
+    
+    def __init__(self, hidden_size, frequency_embedding_size=256):
+        super().__init__()
+        self.mlp = nn.Sequential(
+            nn.Linear(frequency_embedding_size, hidden_size, bias=True),
+            nn.SiLU(),
+            nn.Linear(hidden_size, hidden_size, bias=True))
+        self.frequency_embedding_size = frequency_embedding_size
+
+    @staticmethod
+    def timestep_embedding(t, dim, max_period=10000):
+        half = dim // 2
+        freqs = torch.exp(
+            -math.log(max_period)
+            * torch.arange(start=0, end=half, dtype=torch.float32, device=t.device)
+            / half)
+        args = t[:, None].float() * freqs[None]
+        embedding = torch.cat([torch.cos(args), torch.sin(args)], dim=-1)
+        if dim % 2:
+            embedding = torch.cat(
+                [embedding, torch.zeros_like(embedding[:, :1])], dim=-1)
+        return embedding
+
+    def forward(self, t):
+        t_freq = self.timestep_embedding(t, self.frequency_embedding_size)
+        t_emb = self.mlp(t_freq)
+        return t_emb
+
+
+class DDiTBlock(nn.Module):
+    def __init__(self, dim, n_heads, cond_dim, mlp_ratio=4, dropout=0.1):
+        super().__init__()
+        self.n_heads = n_heads
+
+        self.norm1 = LayerNorm(dim)
+        self.attn_qkv = nn.Linear(dim, 3 * dim, bias=False)
+        self.attn_out = nn.Linear(dim, dim, bias=False)
+
+        self.norm2 = LayerNorm(dim)
+        self.mlp = nn.Sequential(
+            nn.Linear(dim, mlp_ratio * dim, bias=True),
+            nn.GELU(approximate='tanh'),
+            nn.Linear(mlp_ratio * dim, dim, bias=True))
+        self.dropout = dropout
+
+        self.adaLN_modulation = nn.Linear(cond_dim, 6 * dim)
+        self.adaLN_modulation.weight.data.zero_()
+        self.adaLN_modulation.bias.data.zero_()
+
+    def _get_bias_dropout_scale(self):
+        if self.training:
+            return bias_dropout_add_scale_fused_train
+        else:
+            return bias_dropout_add_scale_fused_inference
+
+    def forward(self, x, rotary_cos_sin, c):
+        bias_dropout_scale_fn = self._get_bias_dropout_scale()
+
+        x_skip = x
+        x = self.norm1(x)
+
+        (shift_msa, scale_msa, gate_msa, shift_mlp,
+         scale_mlp, gate_mlp) = self.adaLN_modulation(c)[:, None].chunk(6, dim=2)
+        x = modulate_fused(x, shift_msa, scale_msa)
+
+        qkv = einops.rearrange(
+            self.attn_qkv(x),
+            'b s (three h d) -> b s three h d',
+            three=3,
+            h=self.n_heads)
+        q, k, v = split_and_apply_rotary_pos_emb(qkv, rotary_cos_sin)
+        x = regular_attention_multi_headed(q, k, v)
+
+        x = bias_dropout_scale_fn(self.attn_out(x), None, gate_msa, x_skip, self.dropout)
+        x = bias_dropout_scale_fn(
+            self.mlp(modulate_fused(self.norm2(x), shift_mlp, scale_mlp)),
+            None, gate_mlp, x, self.dropout)
+        return x
+
+
+def _normalize_embedding_layernorm(weight: torch.Tensor) -> torch.Tensor:
+    """Normalize embedding weights to unit norm per row, then scale by sqrt(dim)."""
+    normalized = F.normalize(weight.float(), dim=-1)
+    return (normalized * math.sqrt(weight.shape[-1])).to(weight.dtype)
+
+
+class EmbeddingLayer(nn.Module):
+    """Embedding layer with optional layernorm normalization."""
+    
+    def __init__(self, dim, vocab_dim, use_normalized_embedding=True):
+        super().__init__()
+        self.dim = dim
+        self.vocab_dim = vocab_dim
+        self.use_normalized_embedding = use_normalized_embedding
+        self.embedding = nn.Parameter(torch.empty((vocab_dim, dim)))
+        nn.init.kaiming_uniform_(self.embedding, a=math.sqrt(5))
+
+    def _get_embedding(self):
+        if self.use_normalized_embedding:
+            return _normalize_embedding_layernorm(self.embedding)
+        return self.embedding
+
+    def forward(self, x):
+        embedding = self._get_embedding()
+        if x.ndim == 2:
+            return embedding[x]
+        assert x.ndim == 3  # probabilities
+        return torch.einsum("blv,ve->ble", x.float(), embedding.float()).to(x.dtype)
+
+
+class DDiTFinalLayer(nn.Module):
+    def __init__(self, hidden_size, out_channels, cond_dim):
+        super().__init__()
+        self.norm_final = LayerNorm(hidden_size)
+        self.linear = nn.Linear(hidden_size, out_channels)
+        self.linear.weight.data.zero_()
+        self.linear.bias.data.zero_()
+        self.adaLN_modulation = nn.Linear(cond_dim, 2 * hidden_size, bias=True)
+        self.adaLN_modulation.weight.data.zero_()
+        self.adaLN_modulation.bias.data.zero_()
+
+    def forward(self, x, c):
+        x = self.norm_final(x)
+        shift, scale = self.adaLN_modulation(c)[:, None].chunk(2, dim=2)
+        x = modulate_fused(x, shift, scale)
+        x = self.linear(x)
+        return x
+
+
+class GumbelProposal(nn.Module):
+    """Learnable Gumbel distribution proposal for sampling gamma (log-SNR)."""
+    
+    def __init__(self, loc: float = 4.723, scale: float = 0.852, 
+                 cutoff: float = 1e-5, entropy: float = 7.02):
+        super().__init__()
+        self.loc = nn.Parameter(torch.tensor(loc))
+        self.scale = nn.Parameter(torch.tensor(scale))
+        self.cutoff = cutoff
+        self.entropy = nn.Parameter(torch.tensor(entropy))
+
+    def _get_distribution(self) -> torch.distributions.Gumbel:
+        return torch.distributions.Gumbel(self.loc, self.scale)
+
+    @property
+    def gamma_min(self) -> float:
+        return float(self.loc - math.log(-math.log(self.cutoff)) * self.scale)
+
+    @property
+    def gamma_max(self) -> float:
+        return float(self.loc - math.log(self.cutoff) * self.scale)
+
+    def forward(self, q: torch.Tensor) -> torch.Tensor:
+        """Convert uniform samples to gamma values via inverse CDF."""
+        gamma = self._get_distribution().icdf(q)
+        return gamma.clamp(min=self.gamma_min, max=self.gamma_max)
+
+    def log_pdf(self, gamma: torch.Tensor) -> torch.Tensor:
+        """Compute log probability density at gamma."""
+        return self._get_distribution().log_prob(gamma)
+
+
+class LangFlowBackbone(nn.Module):
+    """DiT backbone for LangFlow."""
+    
+    def __init__(self, config: LangFlowConfig):
+        super().__init__()
+        self.config = config
+        dim = config.hidden_size
+        cond_dim = config.cond_dim
+
+        self.vocab_embed = EmbeddingLayer(
+            dim, config.vocab_size,
+            use_normalized_embedding=config.use_normalized_embedding)
+        self.sigma_map = TimestepEmbedder(cond_dim)
+        self.rotary_emb = Rotary(dim // config.n_heads)
+
+        self.blocks = nn.ModuleList([
+            DDiTBlock(dim=dim, n_heads=config.n_heads, cond_dim=cond_dim, dropout=config.dropout)
+            for _ in range(config.n_blocks)
+        ])
+
+        self.output_layer = DDiTFinalLayer(
+            hidden_size=dim, out_channels=config.vocab_size, cond_dim=cond_dim)
+
+        # Self-conditioning projection
+        if config.self_conditioning:
+            self.self_cond_proj = nn.Linear(dim * 2, dim, bias=False)
+            nn.init.zeros_(self.self_cond_proj.weight)
+
+    def forward(self, x_embed, sigma, x_self_cond=None, output_hidden_states=False):
+        """Forward pass from embeddings.
+        
+        Args:
+            x_embed: [B, L, D] - Input embeddings (possibly noisy)
+            sigma: [B] - Gamma values (log-SNR)
+            x_self_cond: [B, L, D] - Self-conditioning embeddings (optional)
+            output_hidden_states: Whether to return all hidden states
+        
+        Returns:
+            logits: [B, L, vocab_size]
+            hidden_states: List of hidden states if output_hidden_states=True
+        """
+        all_hidden_states = []
+        x = x_embed
+        
+        if output_hidden_states:
+            all_hidden_states.append(x)
+
+        # Self-conditioning
+        if self.config.self_conditioning:
+            if x_self_cond is None:
+                x_self_cond = torch.zeros_like(x)
+            x = x + self.self_cond_proj(torch.cat([x, x_self_cond], dim=-1))
+
+        t_cond = F.silu(self.sigma_map(sigma))
+        rotary_cos_sin = self.rotary_emb(x)
+
+        with torch.autocast(device_type='cuda', dtype=torch.bfloat16):
+            for block in self.blocks:
+                x = block(x, rotary_cos_sin, c=t_cond)
+                if output_hidden_states:
+                    all_hidden_states.append(x)
+            x = self.output_layer(x, c=t_cond)
+
+        return x, all_hidden_states
+
+
+class LangFlow(transformers.PreTrainedModel):
+    """HuggingFace-compatible LangFlow model.
+    
+    LangFlow is a continuous diffusion language model that operates in embedding space.
+    It uses a DiT (Diffusion Transformer) backbone with:
+    - Self-conditioning: uses previous predictions as additional input
+    - Bias (preconditioning): skip connection for improved generation
+    - Normalized embeddings: layernorm on embedding vectors
+    - Learnable Gumbel proposal for gamma (log-SNR) sampling
+    """
+    config_class = LangFlowConfig
+    base_model_prefix = "langflow"
+
+    def __init__(self, config: LangFlowConfig):
+        super().__init__(config)
+        self.config = config
+        self.backbone = LangFlowBackbone(config)
+        self.proposal = GumbelProposal(
+            loc=config.gumbel_loc,
+            scale=config.gumbel_scale,
+            cutoff=config.gumbel_cutoff,
+            entropy=config.gumbel_entropy)
+
+    def _get_embedding_matrix(self) -> torch.Tensor:
+        """Get the embedding matrix for bias skip connection."""
+        return self.backbone.vocab_embed._get_embedding()
+
+    def _embed_tokens(self, x: torch.Tensor) -> torch.Tensor:
+        """Embed tokens or probabilities to continuous embeddings."""
+        return self.backbone.vocab_embed(x)
+
+    def _forward_diffusion(self, x_embed: torch.Tensor, 
+                           gamma: torch.Tensor) -> torch.Tensor:
+        """Add noise to embeddings (forward diffusion process)."""
+        gamma = gamma.float()
+        alpha = torch.sigmoid(-gamma).sqrt()[:, None, None]
+        sigma = torch.sigmoid(gamma).sqrt()[:, None, None]
+        noise = torch.randn_like(x_embed)
+        return (x_embed * alpha + noise * sigma).to(x_embed.dtype)
+
+    def _euler_edm_step(self, z: torch.Tensor, x_pred: torch.Tensor,
+                        t: torch.Tensor, s: torch.Tensor) -> torch.Tensor:
+        """Single Euler step for EDM sampling."""
+        t_ = t.double()
+        s_ = s.double()
+        cur = z.double() * ((F.softplus(t_) - F.softplus(s_)) / 2).exp()
+        end = torch.sigmoid(-s_).sqrt() * x_pred.double()
+        z = end.lerp(cur, ((s_ - t_) / 2).exp()).to(z.dtype)
+        return z
+
+    def forward(
+        self,
+        input_ids: typing.Optional[torch.LongTensor] = None,
+        noisy_embeds: typing.Optional[torch.FloatTensor] = None,
+        timesteps: typing.Optional[torch.FloatTensor] = None,
+        x_self_cond: typing.Optional[torch.FloatTensor] = None,
+        output_hidden_states: typing.Optional[bool] = None,
+        return_dict: typing.Optional[bool] = None,
+    ) -> typing.Union[torch.Tensor, typing.Tuple, transformers.modeling_outputs.MaskedLMOutput]:
+        """Forward pass for LangFlow.
+        
+        Args:
+            input_ids: [B, L] - Token IDs (will be embedded and noised if timesteps provided)
+            noisy_embeds: [B, L, D] - Pre-noised embeddings (alternative to input_ids)
+            timesteps: [B] - Gamma values (log-SNR) for conditioning
+            x_self_cond: [B, L, D] - Self-conditioning embeddings
+            output_hidden_states: Whether to return hidden states
+            return_dict: Whether to return MaskedLMOutput
+        
+        Returns:
+            logits or MaskedLMOutput
+        """
+        output_hidden_states = output_hidden_states if output_hidden_states is not None else False
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        # Get embeddings
+        if noisy_embeds is not None:
+            z = noisy_embeds
+        elif input_ids is not None:
+            x_embed = self._embed_tokens(input_ids)
+            if timesteps is not None:
+                z = self._forward_diffusion(x_embed, timesteps)
+            else:
+                z = x_embed
+        else:
+            raise ValueError("Either input_ids or noisy_embeds must be provided")
+
+        if timesteps is None:
+            # Use minimum gamma for clean input
+            timesteps = torch.full((z.shape[0],), self.proposal.gamma_min, device=z.device)
+
+        # Process sigma
+        sigma = timesteps
+        if sigma.ndim == 2:
+            sigma = sigma.mean(-1)
+
+        # Get model output
+        logits, all_hidden_states = self.backbone(
+            z, sigma, x_self_cond=x_self_cond, output_hidden_states=output_hidden_states)
+
+        # Add bias (preconditioning) skip connection
+        if self.config.use_bias:
+            c_skip = ((F.softplus(-sigma) - sigma) / 2).exp()
+            embedding = self._get_embedding_matrix()
+            skip_logits = torch.matmul(z.float(), embedding.t().float())
+            logits = logits + c_skip[:, None, None] * skip_logits.to(logits.dtype)
+
+        if return_dict:
+            return transformers.modeling_outputs.MaskedLMOutput(
+                logits=logits,
+                hidden_states=all_hidden_states if output_hidden_states else None,
+                loss=None)
+        elif output_hidden_states:
+            return logits, all_hidden_states
+        else:
+            return logits
+
+    @torch.no_grad()
+    def generate_samples(
+        self,
+        num_samples: int = 1,
+        seq_length: typing.Optional[int] = None,
+        num_steps: int = 128,
+        device: typing.Optional[torch.device] = None,
+    ) -> torch.LongTensor:
+        """Generate samples using Euler-EDM solver.
+        
+        Args:
+            num_samples: Number of samples to generate
+            seq_length: Sequence length (defaults to config.model_length)
+            num_steps: Number of denoising steps
+            device: Device to generate on
+        
+        Returns:
+            samples: [num_samples, seq_length] - Generated token IDs
+        """
+        if seq_length is None:
+            seq_length = self.config.model_length
+        if device is None:
+            device = next(self.parameters()).device
+
+        embed_dim = self.config.hidden_size
+        eps = 1e-5
+
+        # Initialize with Gaussian noise
+        z = torch.randn(num_samples, seq_length, embed_dim, device=device)
+
+        # Create gamma schedule from t=1-eps to t=eps
+        t = torch.linspace(1.0 - eps, eps, num_steps, device=device)
+        gamma = self.proposal(t)
+
+        # Self-conditioning state
+        x_self_cond = None
+
+        # Euler-EDM sampling loop
+        for i in range(len(gamma) - 1):
+            gamma_t = gamma[i]
+            gamma_s = gamma[i + 1]
+
+            # Get model prediction
+            gamma_expanded = gamma_t.unsqueeze(0).expand(num_samples)
+            logits = self.forward(
+                noisy_embeds=z,
+                timesteps=gamma_expanded,
+                x_self_cond=x_self_cond,
+                return_dict=False)
+
+            # Convert logits to embedding prediction
+            probs = F.softmax(logits.float(), dim=-1)
+            x_pred = self._embed_tokens(probs)
+
+            # Update self-conditioning
+            if self.config.self_conditioning:
+                x_self_cond = x_pred
+
+            # Euler step
+            z = self._euler_edm_step(z, x_pred, gamma_t, gamma_s)
+
+        # Final step: get logits and take argmax
+        gamma_final = gamma[-1]
+        gamma_expanded = gamma_final.unsqueeze(0).expand(num_samples)
+        logits = self.forward(
+            noisy_embeds=z,
+            timesteps=gamma_expanded,
+            x_self_cond=x_self_cond,
+            return_dict=False)
+        samples = logits.argmax(dim=-1)
+
+        return samples

results/conditional/matrix/benchmark_pending.tsv

@@ -0,0 +1,2 @@
+# schema=dlb-conditional-generation-matrix-v1
+task_id	category	model	dataset	steps	sample_count	seed	environment	adapter	source	provenance	protocol	conditioning_manifest	conditioning_manifest_sha256	sample_dir	metrics_path	timing_path

results/conditional/matrix/benchmark_pending_current.tsv

@@ -0,0 +1,14 @@
+# schema=dlb-conditional-generation-matrix-v1
+task_id	category	model	dataset	steps	sample_count	seed	environment	adapter	source	provenance	protocol	conditioning_manifest	conditioning_manifest_sha256	sample_dir	metrics_path	timing_path
+duo_di4c-lm1b-steps-1	few	duo_di4c	lm1b	1	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo_di4c/steps_1	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo_di4c/steps_1/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo_di4c/steps_1/timing.json
+duo_di4c-lm1b-steps-2	few	duo_di4c	lm1b	2	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo_di4c/steps_2	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo_di4c/steps_2/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo_di4c/steps_2/timing.json
+duo_di4c-lm1b-steps-4	few	duo_di4c	lm1b	4	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo_di4c/steps_4	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo_di4c/steps_4/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo_di4c/steps_4/timing.json
+duo_di4c-lm1b-steps-8	few	duo_di4c	lm1b	8	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo_di4c/steps_8	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo_di4c/steps_8/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo_di4c/steps_8/timing.json
+duo_di4c-lm1b-steps-16	few	duo_di4c	lm1b	16	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo_di4c/steps_16	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo_di4c/steps_16/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo_di4c/steps_16/timing.json
+duo_di4c-lm1b-steps-32	few	duo_di4c	lm1b	32	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo_di4c/steps_32	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo_di4c/steps_32/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo_di4c/steps_32/timing.json
+mdlm_di4c-lm1b-steps-1	few	mdlm_di4c	lm1b	1	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm_di4c/steps_1	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm_di4c/steps_1/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm_di4c/steps_1/timing.json
+mdlm_di4c-lm1b-steps-2	few	mdlm_di4c	lm1b	2	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm_di4c/steps_2	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm_di4c/steps_2/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm_di4c/steps_2/timing.json
+mdlm_di4c-lm1b-steps-4	few	mdlm_di4c	lm1b	4	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm_di4c/steps_4	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm_di4c/steps_4/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm_di4c/steps_4/timing.json
+mdlm_di4c-lm1b-steps-8	few	mdlm_di4c	lm1b	8	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm_di4c/steps_8	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm_di4c/steps_8/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm_di4c/steps_8/timing.json
+mdlm_di4c-lm1b-steps-16	few	mdlm_di4c	lm1b	16	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm_di4c/steps_16	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm_di4c/steps_16/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm_di4c/steps_16/timing.json
+mdlm_di4c-lm1b-steps-32	few	mdlm_di4c	lm1b	32	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm_di4c/steps_32	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm_di4c/steps_32/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm_di4c/steps_32/timing.json

results/conditional/matrix/duo_di4c_owt.tsv

@@ -0,0 +1,8 @@
+# schema=dlb-conditional-generation-matrix-v1
+task_id	category	model	dataset	steps	sample_count	seed	environment	adapter	source	provenance	protocol	conditioning_manifest	conditioning_manifest_sha256	sample_dir	metrics_path	timing_path
+duo_di4c-owt-steps-1	few	duo_di4c	owt	1	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_di4c/steps_1	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/duo_di4c/steps_1/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/duo_di4c/steps_1/timing.json
+duo_di4c-owt-steps-2	few	duo_di4c	owt	2	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_di4c/steps_2	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/duo_di4c/steps_2/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/duo_di4c/steps_2/timing.json
+duo_di4c-owt-steps-4	few	duo_di4c	owt	4	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_di4c/steps_4	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/duo_di4c/steps_4/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/duo_di4c/steps_4/timing.json
+duo_di4c-owt-steps-8	few	duo_di4c	owt	8	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_di4c/steps_8	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/duo_di4c/steps_8/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/duo_di4c/steps_8/timing.json
+duo_di4c-owt-steps-16	few	duo_di4c	owt	16	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_di4c/steps_16	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/duo_di4c/steps_16/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/duo_di4c/steps_16/timing.json
+duo_di4c-owt-steps-32	few	duo_di4c	owt	32	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_di4c/steps_32	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/duo_di4c/steps_32/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/duo_di4c/steps_32/timing.json

results/conditional/matrix/evaluate_pending.tsv

@@ -0,0 +1,23 @@
+# schema=dlb-conditional-generation-matrix-v1
+task_id	category	model	dataset	steps	sample_count	seed	environment	adapter	source	provenance	protocol	conditioning_manifest	conditioning_manifest_sha256	sample_dir	metrics_path	timing_path
+candi-lm1b-steps-1	many	candi	lm1b	1	2048	42	dlb-candi	candi	candi	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/candi/steps_1	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/candi/steps_1/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/candi/steps_1/timing.json
+candi-lm1b-steps-2	many	candi	lm1b	2	2048	42	dlb-candi	candi	candi	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/candi/steps_2	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/candi/steps_2/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/candi/steps_2/timing.json
+candi-lm1b-steps-4	many	candi	lm1b	4	2048	42	dlb-candi	candi	candi	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/candi/steps_4	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/candi/steps_4/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/candi/steps_4/timing.json
+candi-lm1b-steps-8	many	candi	lm1b	8	2048	42	dlb-candi	candi	candi	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/candi/steps_8	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/candi/steps_8/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/candi/steps_8/timing.json
+candi-lm1b-steps-16	many	candi	lm1b	16	2048	42	dlb-candi	candi	candi	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/candi/steps_16	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/candi/steps_16/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/candi/steps_16/timing.json
+candi-lm1b-steps-32	many	candi	lm1b	32	2048	42	dlb-candi	candi	candi	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/candi/steps_32	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/candi/steps_32/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/candi/steps_32/timing.json
+duo_di4c-lm1b-steps-1	few	duo_di4c	lm1b	1	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo_di4c/steps_1	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo_di4c/steps_1/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo_di4c/steps_1/timing.json
+duo_di4c-lm1b-steps-2	few	duo_di4c	lm1b	2	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo_di4c/steps_2	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo_di4c/steps_2/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo_di4c/steps_2/timing.json
+duo_di4c-lm1b-steps-4	few	duo_di4c	lm1b	4	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo_di4c/steps_4	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo_di4c/steps_4/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo_di4c/steps_4/timing.json
+duo_di4c-lm1b-steps-8	few	duo_di4c	lm1b	8	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo_di4c/steps_8	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo_di4c/steps_8/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo_di4c/steps_8/timing.json
+duo_di4c-lm1b-steps-16	few	duo_di4c	lm1b	16	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo_di4c/steps_16	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo_di4c/steps_16/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo_di4c/steps_16/timing.json
+duo_di4c-lm1b-steps-32	few	duo_di4c	lm1b	32	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo_di4c/steps_32	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo_di4c/steps_32/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo_di4c/steps_32/timing.json
+langflow-owt-steps-1	many	langflow	owt	1	2048	42	dlb-langflow	langflow	langflow	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/langflow/steps_1	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/langflow/steps_1/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/langflow/steps_1/timing.json
+mdlm_di4c-lm1b-steps-1	few	mdlm_di4c	lm1b	1	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm_di4c/steps_1	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm_di4c/steps_1/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm_di4c/steps_1/timing.json
+mdlm_di4c-lm1b-steps-2	few	mdlm_di4c	lm1b	2	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm_di4c/steps_2	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm_di4c/steps_2/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm_di4c/steps_2/timing.json
+mdlm_di4c-lm1b-steps-4	few	mdlm_di4c	lm1b	4	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm_di4c/steps_4	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm_di4c/steps_4/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm_di4c/steps_4/timing.json
+mdlm_di4c-lm1b-steps-8	few	mdlm_di4c	lm1b	8	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm_di4c/steps_8	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm_di4c/steps_8/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm_di4c/steps_8/timing.json
+mdlm_di4c-lm1b-steps-16	few	mdlm_di4c	lm1b	16	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm_di4c/steps_16	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm_di4c/steps_16/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm_di4c/steps_16/timing.json
+mdlm_di4c-lm1b-steps-32	few	mdlm_di4c	lm1b	32	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm_di4c/steps_32	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm_di4c/steps_32/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm_di4c/steps_32/timing.json
+rdlm-lm1b-steps-1000	many	rdlm	lm1b	1000	2048	42	dlb-rdlm	rdlm	rdlm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/rdlm/steps_1000	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/rdlm/steps_1000/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/rdlm/steps_1000/timing.json
+rdlm-lm1b-steps-1024	many	rdlm	lm1b	1024	2048	42	dlb-rdlm	rdlm	rdlm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/rdlm/steps_1024	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/rdlm/steps_1024/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/rdlm/steps_1024/timing.json

results/conditional/matrix/generation.tsv

@@ -0,0 +1,134 @@
+# schema=dlb-conditional-generation-matrix-v1
+task_id	category	model	dataset	steps	sample_count	seed	environment	adapter	source	provenance	protocol	conditioning_manifest	conditioning_manifest_sha256	sample_dir	metrics_path	timing_path
+candi-lm1b-steps-1	many	candi	lm1b	1	2048	42	dlb-candi	candi	candi	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/candi/steps_1	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/candi/steps_1/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/candi/steps_1/timing.json
+candi-lm1b-steps-2	many	candi	lm1b	2	2048	42	dlb-candi	candi	candi	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/candi/steps_2	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/candi/steps_2/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/candi/steps_2/timing.json
+candi-lm1b-steps-4	many	candi	lm1b	4	2048	42	dlb-candi	candi	candi	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/candi/steps_4	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/candi/steps_4/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/candi/steps_4/timing.json
+candi-lm1b-steps-8	many	candi	lm1b	8	2048	42	dlb-candi	candi	candi	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/candi/steps_8	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/candi/steps_8/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/candi/steps_8/timing.json
+candi-lm1b-steps-16	many	candi	lm1b	16	2048	42	dlb-candi	candi	candi	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/candi/steps_16	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/candi/steps_16/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/candi/steps_16/timing.json
+candi-lm1b-steps-32	many	candi	lm1b	32	2048	42	dlb-candi	candi	candi	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/candi/steps_32	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/candi/steps_32/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/candi/steps_32/timing.json
+candi-lm1b-steps-1024	many	candi	lm1b	1024	2048	42	dlb-candi	candi	candi	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/candi/steps_1024	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/candi/steps_1024/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/candi/steps_1024/timing.json
+candi-owt-steps-1	many	candi	owt	1	2048	42	dlb-candi	candi	candi	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/candi/steps_1	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/candi/steps_1/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/candi/steps_1/timing.json
+candi-owt-steps-2	many	candi	owt	2	2048	42	dlb-candi	candi	candi	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/candi/steps_2	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/candi/steps_2/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/candi/steps_2/timing.json
+candi-owt-steps-4	many	candi	owt	4	2048	42	dlb-candi	candi	candi	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/candi/steps_4	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/candi/steps_4/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/candi/steps_4/timing.json
+candi-owt-steps-8	many	candi	owt	8	2048	42	dlb-candi	candi	candi	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/candi/steps_8	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/candi/steps_8/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/candi/steps_8/timing.json
+candi-owt-steps-16	many	candi	owt	16	2048	42	dlb-candi	candi	candi	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/candi/steps_16	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/candi/steps_16/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/candi/steps_16/timing.json
+candi-owt-steps-32	many	candi	owt	32	2048	42	dlb-candi	candi	candi	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/candi/steps_32	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/candi/steps_32/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/candi/steps_32/timing.json
+candi-owt-steps-1024	many	candi	owt	1024	2048	42	dlb-candi	candi	candi	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/candi/steps_1024	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/candi/steps_1024/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/candi/steps_1024/timing.json
+duo-lm1b-steps-1	many	duo	lm1b	1	2048	42	dlb-duo	duo	duo	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo/steps_1	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo/steps_1/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo/steps_1/timing.json
+duo-lm1b-steps-2	many	duo	lm1b	2	2048	42	dlb-duo	duo	duo	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo/steps_2	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo/steps_2/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo/steps_2/timing.json
+duo-lm1b-steps-4	many	duo	lm1b	4	2048	42	dlb-duo	duo	duo	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo/steps_4	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo/steps_4/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo/steps_4/timing.json
+duo-lm1b-steps-8	many	duo	lm1b	8	2048	42	dlb-duo	duo	duo	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo/steps_8	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo/steps_8/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo/steps_8/timing.json
+duo-lm1b-steps-16	many	duo	lm1b	16	2048	42	dlb-duo	duo	duo	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo/steps_16	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo/steps_16/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo/steps_16/timing.json
+duo-lm1b-steps-32	many	duo	lm1b	32	2048	42	dlb-duo	duo	duo	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo/steps_32	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo/steps_32/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo/steps_32/timing.json
+duo-lm1b-steps-1024	many	duo	lm1b	1024	2048	42	dlb-duo	duo	duo	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo/steps_1024	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo/steps_1024/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo/steps_1024/timing.json
+duo-owt-steps-1	many	duo	owt	1	2048	42	dlb-duo	duo	duo	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo/steps_1	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/duo/steps_1/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/duo/steps_1/timing.json
+duo-owt-steps-2	many	duo	owt	2	2048	42	dlb-duo	duo	duo	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo/steps_2	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/duo/steps_2/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/duo/steps_2/timing.json
+duo-owt-steps-4	many	duo	owt	4	2048	42	dlb-duo	duo	duo	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo/steps_4	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/duo/steps_4/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/duo/steps_4/timing.json
+duo-owt-steps-8	many	duo	owt	8	2048	42	dlb-duo	duo	duo	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo/steps_8	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/duo/steps_8/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/duo/steps_8/timing.json
+duo-owt-steps-16	many	duo	owt	16	2048	42	dlb-duo	duo	duo	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo/steps_16	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/duo/steps_16/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/duo/steps_16/timing.json
+duo-owt-steps-32	many	duo	owt	32	2048	42	dlb-duo	duo	duo	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo/steps_32	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/duo/steps_32/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/duo/steps_32/timing.json
+duo-owt-steps-1024	many	duo	owt	1024	2048	42	dlb-duo	duo	duo	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo/steps_1024	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/duo/steps_1024/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/duo/steps_1024/timing.json
+duo_dcd-lm1b-steps-1	few	duo_dcd	lm1b	1	2048	42	dlb-duo	duo	duo	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo_dcd/steps_1	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo_dcd/steps_1/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo_dcd/steps_1/timing.json
+duo_dcd-lm1b-steps-2	few	duo_dcd	lm1b	2	2048	42	dlb-duo	duo	duo	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo_dcd/steps_2	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo_dcd/steps_2/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo_dcd/steps_2/timing.json
+duo_dcd-lm1b-steps-4	few	duo_dcd	lm1b	4	2048	42	dlb-duo	duo	duo	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo_dcd/steps_4	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo_dcd/steps_4/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo_dcd/steps_4/timing.json
+duo_dcd-lm1b-steps-8	few	duo_dcd	lm1b	8	2048	42	dlb-duo	duo	duo	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo_dcd/steps_8	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo_dcd/steps_8/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo_dcd/steps_8/timing.json
+duo_dcd-lm1b-steps-16	few	duo_dcd	lm1b	16	2048	42	dlb-duo	duo	duo	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo_dcd/steps_16	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo_dcd/steps_16/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo_dcd/steps_16/timing.json
+duo_dcd-lm1b-steps-32	few	duo_dcd	lm1b	32	2048	42	dlb-duo	duo	duo	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo_dcd/steps_32	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo_dcd/steps_32/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo_dcd/steps_32/timing.json
+duo_dcd-owt-steps-1	few	duo_dcd	owt	1	2048	42	dlb-duo	duo	duo	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_dcd/steps_1	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/duo_dcd/steps_1/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/duo_dcd/steps_1/timing.json
+duo_dcd-owt-steps-2	few	duo_dcd	owt	2	2048	42	dlb-duo	duo	duo	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_dcd/steps_2	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/duo_dcd/steps_2/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/duo_dcd/steps_2/timing.json
+duo_dcd-owt-steps-4	few	duo_dcd	owt	4	2048	42	dlb-duo	duo	duo	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_dcd/steps_4	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/duo_dcd/steps_4/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/duo_dcd/steps_4/timing.json
+duo_dcd-owt-steps-8	few	duo_dcd	owt	8	2048	42	dlb-duo	duo	duo	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_dcd/steps_8	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/duo_dcd/steps_8/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/duo_dcd/steps_8/timing.json
+duo_dcd-owt-steps-16	few	duo_dcd	owt	16	2048	42	dlb-duo	duo	duo	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_dcd/steps_16	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/duo_dcd/steps_16/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/duo_dcd/steps_16/timing.json
+duo_dcd-owt-steps-32	few	duo_dcd	owt	32	2048	42	dlb-duo	duo	duo	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_dcd/steps_32	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/duo_dcd/steps_32/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/duo_dcd/steps_32/timing.json
+duo_di4c-lm1b-steps-1	few	duo_di4c	lm1b	1	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo_di4c/steps_1	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo_di4c/steps_1/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo_di4c/steps_1/timing.json
+duo_di4c-lm1b-steps-2	few	duo_di4c	lm1b	2	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo_di4c/steps_2	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo_di4c/steps_2/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo_di4c/steps_2/timing.json
+duo_di4c-lm1b-steps-4	few	duo_di4c	lm1b	4	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo_di4c/steps_4	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo_di4c/steps_4/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo_di4c/steps_4/timing.json
+duo_di4c-lm1b-steps-8	few	duo_di4c	lm1b	8	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo_di4c/steps_8	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo_di4c/steps_8/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo_di4c/steps_8/timing.json
+duo_di4c-lm1b-steps-16	few	duo_di4c	lm1b	16	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo_di4c/steps_16	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo_di4c/steps_16/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo_di4c/steps_16/timing.json
+duo_di4c-lm1b-steps-32	few	duo_di4c	lm1b	32	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/duo_di4c/steps_32	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/duo_di4c/steps_32/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/duo_di4c/steps_32/timing.json
+duo_di4c-owt-steps-1	few	duo_di4c	owt	1	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_di4c/steps_1	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/duo_di4c/steps_1/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/duo_di4c/steps_1/timing.json
+duo_di4c-owt-steps-2	few	duo_di4c	owt	2	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_di4c/steps_2	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/duo_di4c/steps_2/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/duo_di4c/steps_2/timing.json
+duo_di4c-owt-steps-4	few	duo_di4c	owt	4	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_di4c/steps_4	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/duo_di4c/steps_4/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/duo_di4c/steps_4/timing.json
+duo_di4c-owt-steps-8	few	duo_di4c	owt	8	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_di4c/steps_8	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/duo_di4c/steps_8/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/duo_di4c/steps_8/timing.json
+duo_di4c-owt-steps-16	few	duo_di4c	owt	16	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_di4c/steps_16	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/duo_di4c/steps_16/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/duo_di4c/steps_16/timing.json
+duo_di4c-owt-steps-32	few	duo_di4c	owt	32	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_di4c/steps_32	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/duo_di4c/steps_32/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/duo_di4c/steps_32/timing.json
+flm-lm1b-steps-1	many	flm	lm1b	1	2048	42	dlb-flm	flm	flm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/flm/steps_1	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/flm/steps_1/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/flm/steps_1/timing.json
+flm-lm1b-steps-2	many	flm	lm1b	2	2048	42	dlb-flm	flm	flm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/flm/steps_2	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/flm/steps_2/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/flm/steps_2/timing.json
+flm-lm1b-steps-4	many	flm	lm1b	4	2048	42	dlb-flm	flm	flm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/flm/steps_4	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/flm/steps_4/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/flm/steps_4/timing.json
+flm-lm1b-steps-8	many	flm	lm1b	8	2048	42	dlb-flm	flm	flm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/flm/steps_8	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/flm/steps_8/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/flm/steps_8/timing.json
+flm-lm1b-steps-16	many	flm	lm1b	16	2048	42	dlb-flm	flm	flm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/flm/steps_16	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/flm/steps_16/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/flm/steps_16/timing.json
+flm-lm1b-steps-32	many	flm	lm1b	32	2048	42	dlb-flm	flm	flm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/flm/steps_32	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/flm/steps_32/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/flm/steps_32/timing.json
+flm-lm1b-steps-1024	many	flm	lm1b	1024	2048	42	dlb-flm	flm	flm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/flm/steps_1024	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/flm/steps_1024/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/flm/steps_1024/timing.json
+flm-owt-steps-1	many	flm	owt	1	2048	42	dlb-flm	flm	flm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/flm/steps_1	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/flm/steps_1/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/flm/steps_1/timing.json
+flm-owt-steps-2	many	flm	owt	2	2048	42	dlb-flm	flm	flm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/flm/steps_2	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/flm/steps_2/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/flm/steps_2/timing.json
+flm-owt-steps-4	many	flm	owt	4	2048	42	dlb-flm	flm	flm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/flm/steps_4	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/flm/steps_4/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/flm/steps_4/timing.json
+flm-owt-steps-8	many	flm	owt	8	2048	42	dlb-flm	flm	flm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/flm/steps_8	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/flm/steps_8/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/flm/steps_8/timing.json
+flm-owt-steps-16	many	flm	owt	16	2048	42	dlb-flm	flm	flm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/flm/steps_16	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/flm/steps_16/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/flm/steps_16/timing.json
+flm-owt-steps-32	many	flm	owt	32	2048	42	dlb-flm	flm	flm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/flm/steps_32	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/flm/steps_32/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/flm/steps_32/timing.json
+flm-owt-steps-1024	many	flm	owt	1024	2048	42	dlb-flm	flm	flm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/flm/steps_1024	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/flm/steps_1024/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/flm/steps_1024/timing.json
+fmlm-lm1b-steps-1	few	fmlm	lm1b	1	2048	42	dlb-flm	flm	flm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/fmlm/steps_1	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/fmlm/steps_1/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/fmlm/steps_1/timing.json
+fmlm-lm1b-steps-2	few	fmlm	lm1b	2	2048	42	dlb-flm	flm	flm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/fmlm/steps_2	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/fmlm/steps_2/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/fmlm/steps_2/timing.json
+fmlm-lm1b-steps-4	few	fmlm	lm1b	4	2048	42	dlb-flm	flm	flm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/fmlm/steps_4	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/fmlm/steps_4/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/fmlm/steps_4/timing.json
+fmlm-lm1b-steps-8	few	fmlm	lm1b	8	2048	42	dlb-flm	flm	flm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/fmlm/steps_8	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/fmlm/steps_8/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/fmlm/steps_8/timing.json
+fmlm-lm1b-steps-16	few	fmlm	lm1b	16	2048	42	dlb-flm	flm	flm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/fmlm/steps_16	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/fmlm/steps_16/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/fmlm/steps_16/timing.json
+fmlm-lm1b-steps-32	few	fmlm	lm1b	32	2048	42	dlb-flm	flm	flm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/fmlm/steps_32	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/fmlm/steps_32/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/fmlm/steps_32/timing.json
+fmlm-owt-steps-1	few	fmlm	owt	1	2048	42	dlb-flm	flm	flm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/fmlm/steps_1	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/fmlm/steps_1/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/fmlm/steps_1/timing.json
+fmlm-owt-steps-2	few	fmlm	owt	2	2048	42	dlb-flm	flm	flm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/fmlm/steps_2	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/fmlm/steps_2/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/fmlm/steps_2/timing.json
+fmlm-owt-steps-4	few	fmlm	owt	4	2048	42	dlb-flm	flm	flm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/fmlm/steps_4	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/fmlm/steps_4/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/fmlm/steps_4/timing.json
+fmlm-owt-steps-8	few	fmlm	owt	8	2048	42	dlb-flm	flm	flm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/fmlm/steps_8	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/fmlm/steps_8/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/fmlm/steps_8/timing.json
+fmlm-owt-steps-16	few	fmlm	owt	16	2048	42	dlb-flm	flm	flm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/fmlm/steps_16	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/fmlm/steps_16/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/fmlm/steps_16/timing.json
+fmlm-owt-steps-32	few	fmlm	owt	32	2048	42	dlb-flm	flm	flm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/fmlm/steps_32	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/fmlm/steps_32/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/fmlm/steps_32/timing.json
+langflow-lm1b-steps-1	many	langflow	lm1b	1	2048	42	dlb-langflow	langflow	langflow	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/langflow/steps_1	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/langflow/steps_1/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/langflow/steps_1/timing.json
+langflow-lm1b-steps-2	many	langflow	lm1b	2	2048	42	dlb-langflow	langflow	langflow	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/langflow/steps_2	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/langflow/steps_2/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/langflow/steps_2/timing.json
+langflow-lm1b-steps-4	many	langflow	lm1b	4	2048	42	dlb-langflow	langflow	langflow	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/langflow/steps_4	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/langflow/steps_4/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/langflow/steps_4/timing.json
+langflow-lm1b-steps-8	many	langflow	lm1b	8	2048	42	dlb-langflow	langflow	langflow	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/langflow/steps_8	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/langflow/steps_8/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/langflow/steps_8/timing.json
+langflow-lm1b-steps-16	many	langflow	lm1b	16	2048	42	dlb-langflow	langflow	langflow	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/langflow/steps_16	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/langflow/steps_16/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/langflow/steps_16/timing.json
+langflow-lm1b-steps-32	many	langflow	lm1b	32	2048	42	dlb-langflow	langflow	langflow	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/langflow/steps_32	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/langflow/steps_32/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/langflow/steps_32/timing.json
+langflow-lm1b-steps-1024	many	langflow	lm1b	1024	2048	42	dlb-langflow	langflow	langflow	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/langflow/steps_1024	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/langflow/steps_1024/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/langflow/steps_1024/timing.json
+langflow-owt-steps-1	many	langflow	owt	1	2048	42	dlb-langflow	langflow	langflow	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/langflow/steps_1	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/langflow/steps_1/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/langflow/steps_1/timing.json
+langflow-owt-steps-2	many	langflow	owt	2	2048	42	dlb-langflow	langflow	langflow	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/langflow/steps_2	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/langflow/steps_2/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/langflow/steps_2/timing.json
+langflow-owt-steps-4	many	langflow	owt	4	2048	42	dlb-langflow	langflow	langflow	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/langflow/steps_4	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/langflow/steps_4/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/langflow/steps_4/timing.json
+langflow-owt-steps-8	many	langflow	owt	8	2048	42	dlb-langflow	langflow	langflow	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/langflow/steps_8	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/langflow/steps_8/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/langflow/steps_8/timing.json
+langflow-owt-steps-16	many	langflow	owt	16	2048	42	dlb-langflow	langflow	langflow	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/langflow/steps_16	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/langflow/steps_16/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/langflow/steps_16/timing.json
+langflow-owt-steps-32	many	langflow	owt	32	2048	42	dlb-langflow	langflow	langflow	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/langflow/steps_32	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/langflow/steps_32/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/langflow/steps_32/timing.json
+langflow-owt-steps-1024	many	langflow	owt	1024	2048	42	dlb-langflow	langflow	langflow	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/langflow/steps_1024	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/langflow/steps_1024/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/langflow/steps_1024/timing.json
+mdlm-lm1b-steps-1	many	mdlm	lm1b	1	2048	42	dlb-mdlm	mdlm	mdlm	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm/steps_1	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm/steps_1/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm/steps_1/timing.json
+mdlm-lm1b-steps-2	many	mdlm	lm1b	2	2048	42	dlb-mdlm	mdlm	mdlm	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm/steps_2	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm/steps_2/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm/steps_2/timing.json
+mdlm-lm1b-steps-4	many	mdlm	lm1b	4	2048	42	dlb-mdlm	mdlm	mdlm	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm/steps_4	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm/steps_4/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm/steps_4/timing.json
+mdlm-lm1b-steps-8	many	mdlm	lm1b	8	2048	42	dlb-mdlm	mdlm	mdlm	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm/steps_8	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm/steps_8/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm/steps_8/timing.json
+mdlm-lm1b-steps-16	many	mdlm	lm1b	16	2048	42	dlb-mdlm	mdlm	mdlm	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm/steps_16	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm/steps_16/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm/steps_16/timing.json
+mdlm-lm1b-steps-32	many	mdlm	lm1b	32	2048	42	dlb-mdlm	mdlm	mdlm	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm/steps_32	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm/steps_32/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm/steps_32/timing.json
+mdlm-lm1b-steps-1024	many	mdlm	lm1b	1024	2048	42	dlb-mdlm	mdlm	mdlm	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm/steps_1024	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm/steps_1024/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm/steps_1024/timing.json
+mdlm-owt-steps-1	many	mdlm	owt	1	2048	42	dlb-mdlm	mdlm	mdlm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/mdlm/steps_1	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/mdlm/steps_1/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/mdlm/steps_1/timing.json
+mdlm-owt-steps-2	many	mdlm	owt	2	2048	42	dlb-mdlm	mdlm	mdlm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/mdlm/steps_2	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/mdlm/steps_2/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/mdlm/steps_2/timing.json
+mdlm-owt-steps-4	many	mdlm	owt	4	2048	42	dlb-mdlm	mdlm	mdlm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/mdlm/steps_4	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/mdlm/steps_4/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/mdlm/steps_4/timing.json
+mdlm-owt-steps-8	many	mdlm	owt	8	2048	42	dlb-mdlm	mdlm	mdlm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/mdlm/steps_8	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/mdlm/steps_8/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/mdlm/steps_8/timing.json
+mdlm-owt-steps-16	many	mdlm	owt	16	2048	42	dlb-mdlm	mdlm	mdlm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/mdlm/steps_16	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/mdlm/steps_16/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/mdlm/steps_16/timing.json
+mdlm-owt-steps-32	many	mdlm	owt	32	2048	42	dlb-mdlm	mdlm	mdlm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/mdlm/steps_32	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/mdlm/steps_32/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/mdlm/steps_32/timing.json
+mdlm-owt-steps-1024	many	mdlm	owt	1024	2048	42	dlb-mdlm	mdlm	mdlm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/mdlm/steps_1024	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/mdlm/steps_1024/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/mdlm/steps_1024/timing.json
+mdlm_di4c-lm1b-steps-1	few	mdlm_di4c	lm1b	1	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm_di4c/steps_1	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm_di4c/steps_1/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm_di4c/steps_1/timing.json
+mdlm_di4c-lm1b-steps-2	few	mdlm_di4c	lm1b	2	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm_di4c/steps_2	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm_di4c/steps_2/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm_di4c/steps_2/timing.json
+mdlm_di4c-lm1b-steps-4	few	mdlm_di4c	lm1b	4	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm_di4c/steps_4	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm_di4c/steps_4/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm_di4c/steps_4/timing.json
+mdlm_di4c-lm1b-steps-8	few	mdlm_di4c	lm1b	8	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm_di4c/steps_8	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm_di4c/steps_8/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm_di4c/steps_8/timing.json
+mdlm_di4c-lm1b-steps-16	few	mdlm_di4c	lm1b	16	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm_di4c/steps_16	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm_di4c/steps_16/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm_di4c/steps_16/timing.json
+mdlm_di4c-lm1b-steps-32	few	mdlm_di4c	lm1b	32	2048	42	dlb-di4c	di4c	di4c	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm_di4c/steps_32	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm_di4c/steps_32/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm_di4c/steps_32/timing.json
+mdlm_di4c-owt-steps-1	few	mdlm_di4c	owt	1	2048	42	dlb-di4c	di4c	di4c	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/mdlm_di4c/steps_1	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/mdlm_di4c/steps_1/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/mdlm_di4c/steps_1/timing.json
+mdlm_di4c-owt-steps-2	few	mdlm_di4c	owt	2	2048	42	dlb-di4c	di4c	di4c	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/mdlm_di4c/steps_2	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/mdlm_di4c/steps_2/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/mdlm_di4c/steps_2/timing.json
+mdlm_di4c-owt-steps-4	few	mdlm_di4c	owt	4	2048	42	dlb-di4c	di4c	di4c	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/mdlm_di4c/steps_4	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/mdlm_di4c/steps_4/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/mdlm_di4c/steps_4/timing.json
+mdlm_di4c-owt-steps-8	few	mdlm_di4c	owt	8	2048	42	dlb-di4c	di4c	di4c	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/mdlm_di4c/steps_8	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/mdlm_di4c/steps_8/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/mdlm_di4c/steps_8/timing.json
+mdlm_di4c-owt-steps-16	few	mdlm_di4c	owt	16	2048	42	dlb-di4c	di4c	di4c	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/mdlm_di4c/steps_16	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/mdlm_di4c/steps_16/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/mdlm_di4c/steps_16/timing.json
+mdlm_di4c-owt-steps-32	few	mdlm_di4c	owt	32	2048	42	dlb-di4c	di4c	di4c	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/mdlm_di4c/steps_32	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/mdlm_di4c/steps_32/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/mdlm_di4c/steps_32/timing.json
+mdlm_sdtt-lm1b-steps-1	few	mdlm_sdtt	lm1b	1	2048	42	dlb-sdtt	sdtt	sdtt	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm_sdtt/steps_1	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm_sdtt/steps_1/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm_sdtt/steps_1/timing.json
+mdlm_sdtt-lm1b-steps-2	few	mdlm_sdtt	lm1b	2	2048	42	dlb-sdtt	sdtt	sdtt	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm_sdtt/steps_2	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm_sdtt/steps_2/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm_sdtt/steps_2/timing.json
+mdlm_sdtt-lm1b-steps-4	few	mdlm_sdtt	lm1b	4	2048	42	dlb-sdtt	sdtt	sdtt	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm_sdtt/steps_4	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm_sdtt/steps_4/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm_sdtt/steps_4/timing.json
+mdlm_sdtt-lm1b-steps-8	few	mdlm_sdtt	lm1b	8	2048	42	dlb-sdtt	sdtt	sdtt	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm_sdtt/steps_8	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm_sdtt/steps_8/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm_sdtt/steps_8/timing.json
+mdlm_sdtt-lm1b-steps-16	few	mdlm_sdtt	lm1b	16	2048	42	dlb-sdtt	sdtt	sdtt	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm_sdtt/steps_16	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm_sdtt/steps_16/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm_sdtt/steps_16/timing.json
+mdlm_sdtt-lm1b-steps-32	few	mdlm_sdtt	lm1b	32	2048	42	dlb-sdtt	sdtt	sdtt	reference_reproduction	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/mdlm_sdtt/steps_32	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/mdlm_sdtt/steps_32/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/mdlm_sdtt/steps_32/timing.json
+mdlm_sdtt-owt-steps-1	few	mdlm_sdtt	owt	1	2048	42	dlb-sdtt	sdtt	sdtt	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/mdlm_sdtt/steps_1	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/mdlm_sdtt/steps_1/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/mdlm_sdtt/steps_1/timing.json
+mdlm_sdtt-owt-steps-2	few	mdlm_sdtt	owt	2	2048	42	dlb-sdtt	sdtt	sdtt	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/mdlm_sdtt/steps_2	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/mdlm_sdtt/steps_2/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/mdlm_sdtt/steps_2/timing.json
+mdlm_sdtt-owt-steps-4	few	mdlm_sdtt	owt	4	2048	42	dlb-sdtt	sdtt	sdtt	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/mdlm_sdtt/steps_4	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/mdlm_sdtt/steps_4/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/mdlm_sdtt/steps_4/timing.json
+mdlm_sdtt-owt-steps-8	few	mdlm_sdtt	owt	8	2048	42	dlb-sdtt	sdtt	sdtt	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/mdlm_sdtt/steps_8	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/mdlm_sdtt/steps_8/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/mdlm_sdtt/steps_8/timing.json
+mdlm_sdtt-owt-steps-16	few	mdlm_sdtt	owt	16	2048	42	dlb-sdtt	sdtt	sdtt	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/mdlm_sdtt/steps_16	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/mdlm_sdtt/steps_16/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/mdlm_sdtt/steps_16/timing.json
+mdlm_sdtt-owt-steps-32	few	mdlm_sdtt	owt	32	2048	42	dlb-sdtt	sdtt	sdtt	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json	7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32	/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/mdlm_sdtt/steps_32	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/owt/mdlm_sdtt/steps_32/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/owt/mdlm_sdtt/steps_32/timing.json
+rdlm-lm1b-steps-1000	many	rdlm	lm1b	1000	2048	42	dlb-rdlm	rdlm	rdlm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/rdlm/steps_1000	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/rdlm/steps_1000/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/rdlm/steps_1000/timing.json
+rdlm-lm1b-steps-1024	many	rdlm	lm1b	1024	2048	42	dlb-rdlm	rdlm	rdlm	official	c64_zs_v1	/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json	0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039	/home/ruizzhao/diffusion_baseline/results/conditional/samples/lm1b/rdlm/steps_1024	/home/ruizzhao/diffusion_baseline/results/conditional/metrics/lm1b/rdlm/steps_1024/metrics.json	/home/ruizzhao/diffusion_baseline/results/conditional/timing/lm1b/rdlm/steps_1024/timing.json

results/conditional/matrix/unsupported.tsv

@@ -0,0 +1,3 @@
+# schema=dlb-conditional-generation-matrix-v1-unsupported
+status	model	dataset	category	reason
+unsupported	rdlm	owt	few	The official RDLM release and paper do not provide an OWT model.

results/conditional/samples/owt/duo_dcd/steps_4/run_metadata.json

@@ -0,0 +1,98 @@
+{
+  "checkpoint_lock_id": "duo_dcd_owt_hf:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152:all",
+  "checkpoint_selection": {
+    "path": null,
+    "resource": "duo_dcd_owt_hf",
+    "teacher_family": null
+  },
+  "checkpoint_sha256": "74770bf9f7305ba80f4cd37d5b0ba78e09a887a7480894a4c6567278e9f6148f",
+  "checkpoint_teacher_family": "uniform_duo",
+  "command": [
+    "/opt/anaconda3/envs/dlb-duo/bin/python",
+    "-u",
+    "-m",
+    "dlb.adapters.capture",
+    "--upstream-entrypoint=/home/ruizzhao/diffusion_baseline/upstreams/duo/main.py",
+    "--capture-path=/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_dcd/steps_4/upstream_token_ids.json",
+    "--generation-mode=conditional_prefix",
+    "--conditioning-manifest=/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json",
+    "--conditioning-manifest-sha256=7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32",
+    "--conditioning-config-sha256=89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114",
+    "--prefix-length=64",
+    "--evaluation-continuation-length=64",
+    "--prompt-count=1024",
+    "--diversity-prompt-count=256",
+    "--completions-per-diversity-prompt=5",
+    "--completion-schedule=c0:p0-1023;c1-4:p0-255",
+    "--",
+    "mode=sample_eval",
+    "seed=42",
+    "data=openwebtext-split",
+    "model=small",
+    "model.length=1024",
+    "loader.batch_size=2",
+    "loader.eval_batch_size=8",
+    "sampling.num_sample_batches=256",
+    "sampling.steps=4",
+    "eval.checkpoint_path=/home/ruizzhao/diffusion_baseline/checkpoints/official/duo_dcd/owt",
+    "trainer.devices=1",
+    "hydra.run.dir=/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_dcd/steps_4/upstream_hydra",
+    "checkpointing.save_dir=/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_dcd/steps_4/upstream_hydra",
+    "+wandb.offline=true",
+    "eval.generated_samples_path=/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_dcd/steps_4/upstream_samples.json",
+    "algo=duo_base",
+    "algo.backbone=hf_dit",
+    "sampling.predictor=ancestral",
+    "sampling.noise_removal=ancestral",
+    "eval.disable_ema=False"
+  ],
+  "command_sha256": "7c8879b6749c894f1baf622ce566f975b6a6ce162184f0169c19a52bea48a3e9",
+  "config_sha256": "6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f",
+  "environment": {
+    "cuda_version": null,
+    "cuda_visible_devices": "4",
+    "hostname": "DS210029",
+    "platform": "Linux-5.15.0-186-generic-x86_64-with-glibc2.35",
+    "python": "3.12.13",
+    "torch_version": "2.5.1"
+  },
+  "finished_at": "2026-08-27T03:11:59.611297Z",
+  "identity": {
+    "adapter_identity": "dlb.adapters.duo:v1|dlb.adapters.duo:DuoAdapter|8878182eb10d7d1e3f18219cb62c2153084842fdb0ab532b1a3c7df0e10da3c6",
+    "checkpoint_lock_id": "duo_dcd_owt_hf:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152:all",
+    "checkpoint_selection": {
+      "path": null,
+      "resource": "duo_dcd_owt_hf",
+      "teacher_family": null
+    },
+    "checkpoint_sha256": "74770bf9f7305ba80f4cd37d5b0ba78e09a887a7480894a4c6567278e9f6148f",
+    "checkpoint_teacher_family": "uniform_duo",
+    "command_sha256": "7c8879b6749c894f1baf622ce566f975b6a6ce162184f0169c19a52bea48a3e9",
+    "completion_schedule": "c0:p0-1023;c1-4:p0-255",
+    "completions_per_diversity_prompt": 5,
+    "conditioning_config_sha256": "89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114",
+    "conditioning_manifest": "/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json",
+    "conditioning_manifest_sha256": "7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32",
+    "config_sha256": "6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f",
+    "dataset_id": "owt",
+    "device": null,
+    "diversity_prompt_count": 256,
+    "environment": "dlb-duo",
+    "evaluation_continuation_length": 64,
+    "generation_mode": "conditional_prefix",
+    "model_id": "duo_dcd",
+    "prefix_length": 64,
+    "prompt_count": 1024,
+    "results_root": "/home/ruizzhao/diffusion_baseline/results/conditional",
+    "run_id": "duo_dcd-owt-steps-4",
+    "sample_count": 2048,
+    "seed": 42,
+    "source_sha256": "7c9b498f5b717de064d6fad7e2509c866e6cb620",
+    "step_count": 4
+  },
+  "source_sha256": "7c9b498f5b717de064d6fad7e2509c866e6cb620",
+  "started_at": "2026-08-27T03:04:17.151285Z",
+  "status": "succeeded",
+  "stderr_log": "/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_dcd/steps_4/stderr.log",
+  "stdout_log": "/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_dcd/steps_4/stdout.log"
+}

results/conditional/samples/owt/duo_dcd/steps_4/upstream_samples.json

@@ -0,0 +1,7 @@
+{
+    "generative_ppl": 176.0227508544922,
+    "entropy": 5.410013675689697,
+    "generated_seqs": [
+        "<|endoftext|>\ufffdWhen it brought to the coast they were better of us; faster be they, to bring the new ones to old ones and you didn\u2019t have to go with a top-tier dealer that\u2019s just really, going,.\u201dIn II\u2019kinen the children lived in the York but never came to life so the Z\u00e9---, so now it is \ufffd local! for Iniche St. I that\u2019s the Roots mell theoleras from life with perfect space to go through the bathroom.this is me) and urban plague Labor, on the other kept him company. He cut corners. All you had to do was assemble the walls, his path wasn\u2019t easy.\u201dI don't have to say to Amoto Lady me am more, everyone else chose The and is, and my Major went to old girl\u2019s stories. hear lots of us go round with see she the front page story that all hima have and the foot doesn\u2019t exist. I what changed her character, went back to grand, looky, with pure is, the one that she still races and is the m. of ofIt to it going to be something I\u2019m very proud of.\u201d it is that herself huh? It is.\" K.R. McCarthy with permission.\n\nInternational Pk\n\nLUCKY GOOD, FRIEND, TREMANT, QUICK PARTIES,\n\nOld we, seriously, realize they were mean; the year ended in June, they're are going the same. That's 61 times what just about when you were to a more 10 school year before, Only 40% of this. The so-said, stay strict for the year. The fact that now covering about 45 to 3 is an essential, hours a year amount, as the said productivity chief, and who didn\u2019s be employed. It, just in the way the\n\nJustice Committee was ill plated during a hike up the Somali parliament spared have by its existence by good work.\n\nIn the article I covered the other technocrats down before I I observed one son of f***s\n one other accused Barack Obama of\n\n\nmaking monetary does\n\nHe have been around this the task, a few years without being public speaking.\n\nThe polls\n\n\n the National Security Council leading him to finally think this guy is actually serious policy.\n\nThis in the middle of the tantrums is going to go really.\n\nBack to say Obama was in his D.C. address at the\n\nThis article is 2 years old\n\nGeorge Bush FUSH's a like..., and Chris Christie his need is, according to a Rose Garden House Press briefing that shared with users it is the\nthat burger that Gibbs, a prominent Obama opposition panelister\n\nLock GPS, place a program on place clamp quirk regulations in business.\n\nthe person behind the president, to get a look.\n\nI to have the truth because you\u2019ll admit it was NOT turned into a fan \u2014publicized, disinterested, ButHet a consensus most Americans where opposed that.\ufffd said sought to applause in his closing speech by citing the ongoing loss of jobs \u2013 and that sentiment was drastically threatened in butter chatter in Lord Obama\u2019s mood during the most contentious campaign.\n\nNow I\u2019 I think he might point out his expertise in the age of New York. And he is a motivated brouder to take Bobby\u2019s request seriously. For his part to he simply forgot explaining to to crowd that he didn\u2019t want gun control.\n\nMAJ TELL WHATEVER?\n\nInching that President Obama means having a top fundroomose in-shells for global warming research, also indicated that he\u2019s avoided extreme disruption.\n\n\"When you have this genuine average in yourself, gut-measuring and when it comes to your political achievements; line between freedom the day, much commitment and to to action every day is a huge and then you just say \u2018this is something we\u2019ve endured and fought and recomm forward to on every day,\u201d Obama wrote on Facebook page.\n\nOn Sunday night, he also pointed out to the long history he\u2019d extensive pushing villagers.\n\nThe school district passed passed a rule created protections, food and water.\n\nBut the\n\nyou, in a little bit my head,\" he said in a werebeds extract. \u201cAnd People break down and says this. \u2018How much would you do for a cost?\u2019 You do.. brilliant people who are what you are. It\u2019s something what \u2019s going to be. Remember dancing\n\n\nFollow Business Insider, on Twitter of views expressed here the Wensch iss cry not on necessarily those of editorial, but do follow them on feeling it is the impression of one that<|endoftext|>"
+    ]
+}

results/conditional/samples/owt/duo_dcd/steps_4/upstream_token_ids.json

@@ -0,0 +1,4 @@
+{
+  "samples": [],
+  "schema": "dlb-upstream-token-capture-v1"
+}

results/conditional/samples/owt/duo_dcd/steps_8/run_metadata.json

@@ -0,0 +1,98 @@
+{
+  "checkpoint_lock_id": "duo_dcd_owt_hf:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152:all",
+  "checkpoint_selection": {
+    "path": null,
+    "resource": "duo_dcd_owt_hf",
+    "teacher_family": null
+  },
+  "checkpoint_sha256": "74770bf9f7305ba80f4cd37d5b0ba78e09a887a7480894a4c6567278e9f6148f",
+  "checkpoint_teacher_family": "uniform_duo",
+  "command": [
+    "/opt/anaconda3/envs/dlb-duo/bin/python",
+    "-u",
+    "-m",
+    "dlb.adapters.capture",
+    "--upstream-entrypoint=/home/ruizzhao/diffusion_baseline/upstreams/duo/main.py",
+    "--capture-path=/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_dcd/steps_8/upstream_token_ids.json",
+    "--generation-mode=conditional_prefix",
+    "--conditioning-manifest=/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json",
+    "--conditioning-manifest-sha256=7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32",
+    "--conditioning-config-sha256=89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114",
+    "--prefix-length=64",
+    "--evaluation-continuation-length=64",
+    "--prompt-count=1024",
+    "--diversity-prompt-count=256",
+    "--completions-per-diversity-prompt=5",
+    "--completion-schedule=c0:p0-1023;c1-4:p0-255",
+    "--",
+    "mode=sample_eval",
+    "seed=42",
+    "data=openwebtext-split",
+    "model=small",
+    "model.length=1024",
+    "loader.batch_size=2",
+    "loader.eval_batch_size=8",
+    "sampling.num_sample_batches=256",
+    "sampling.steps=8",
+    "eval.checkpoint_path=/home/ruizzhao/diffusion_baseline/checkpoints/official/duo_dcd/owt",
+    "trainer.devices=1",
+    "hydra.run.dir=/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_dcd/steps_8/upstream_hydra",
+    "checkpointing.save_dir=/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_dcd/steps_8/upstream_hydra",
+    "+wandb.offline=true",
+    "eval.generated_samples_path=/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_dcd/steps_8/upstream_samples.json",
+    "algo=duo_base",
+    "algo.backbone=hf_dit",
+    "sampling.predictor=ancestral",
+    "sampling.noise_removal=ancestral",
+    "eval.disable_ema=False"
+  ],
+  "command_sha256": "83c00e17c5c9344e20182ba4fea59e8ceb3aea0a2747efa8fed3fb76cb535223",
+  "config_sha256": "6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f",
+  "environment": {
+    "cuda_version": null,
+    "cuda_visible_devices": "4",
+    "hostname": "DS210029",
+    "platform": "Linux-5.15.0-186-generic-x86_64-with-glibc2.35",
+    "python": "3.12.13",
+    "torch_version": "2.5.1"
+  },
+  "finished_at": "2026-08-27T03:20:38.580080Z",
+  "identity": {
+    "adapter_identity": "dlb.adapters.duo:v1|dlb.adapters.duo:DuoAdapter|8878182eb10d7d1e3f18219cb62c2153084842fdb0ab532b1a3c7df0e10da3c6",
+    "checkpoint_lock_id": "duo_dcd_owt_hf:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152:all",
+    "checkpoint_selection": {
+      "path": null,
+      "resource": "duo_dcd_owt_hf",
+      "teacher_family": null
+    },
+    "checkpoint_sha256": "74770bf9f7305ba80f4cd37d5b0ba78e09a887a7480894a4c6567278e9f6148f",
+    "checkpoint_teacher_family": "uniform_duo",
+    "command_sha256": "83c00e17c5c9344e20182ba4fea59e8ceb3aea0a2747efa8fed3fb76cb535223",
+    "completion_schedule": "c0:p0-1023;c1-4:p0-255",
+    "completions_per_diversity_prompt": 5,
+    "conditioning_config_sha256": "89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114",
+    "conditioning_manifest": "/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json",
+    "conditioning_manifest_sha256": "7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32",
+    "config_sha256": "6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f",
+    "dataset_id": "owt",
+    "device": null,
+    "diversity_prompt_count": 256,
+    "environment": "dlb-duo",
+    "evaluation_continuation_length": 64,
+    "generation_mode": "conditional_prefix",
+    "model_id": "duo_dcd",
+    "prefix_length": 64,
+    "prompt_count": 1024,
+    "results_root": "/home/ruizzhao/diffusion_baseline/results/conditional",
+    "run_id": "duo_dcd-owt-steps-8",
+    "sample_count": 2048,
+    "seed": 42,
+    "source_sha256": "7c9b498f5b717de064d6fad7e2509c866e6cb620",
+    "step_count": 8
+  },
+  "source_sha256": "7c9b498f5b717de064d6fad7e2509c866e6cb620",
+  "started_at": "2026-08-27T03:12:01.731024Z",
+  "status": "succeeded",
+  "stderr_log": "/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_dcd/steps_8/stderr.log",
+  "stdout_log": "/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_dcd/steps_8/stdout.log"
+}

results/conditional/samples/owt/duo_dcd/steps_8/upstream_hydra/.hydra/config.yaml

@@ -0,0 +1,151 @@
+mode: sample_eval
+seed: 42
+loader:
+  global_batch_size: 512
+  eval_global_batch_size: ${.global_batch_size}
+  batch_size: 2
+  eval_batch_size: 1
+  num_workers: ${eval:"len(__import__('os').sched_getaffinity(0))"}
+  pin_memory: true
+sampling:
+  predictor: ancestral
+  steps: 8
+  noise_removal: ancestral
+  use_float64: true
+  p_nucleus: 1.0
+  num_sample_batches: 1
+  num_sample_log: 2
+  semi_ar: false
+  stride_length: 1
+  num_strides: 1
+  guid_weight: null
+  psi:
+    time_profile: linear
+    high_mode: pure-posterior
+    middle_mode: pure-posterior
+    low_mode: pure-posterior
+    high_frac: 0.2
+    middle_frac: 0.6
+training:
+  ema: 0.9999
+  antithetic_sampling: true
+  importance_sampling: false
+  sampling_eps: 0.001
+  change_of_variables: false
+  loss_precision: bf16
+  finetune_path: ''
+  class_dropout_p: 0.1
+eval:
+  checkpoint_path: /home/ruizzhao/diffusion_baseline/checkpoints/official/duo_dcd/owt
+  disable_ema: false
+  compute_generative_perplexity: false
+  perplexity_batch_size: 8
+  compute_perplexity_on_sanity: false
+  gen_ppl_eval_model_name_or_path: gpt2-large
+  generate_samples: true
+  generated_samples_path: /home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_dcd/steps_8/upstream_samples.json
+optim:
+  weight_decay: 0
+  lr: 0.0003
+  beta1: 0.9
+  beta2: 0.999
+  eps: 1.0e-08
+trainer:
+  _target_: lightning.Trainer
+  accelerator: cuda
+  num_nodes: 1
+  devices: 1
+  accumulate_grad_batches: ${div_up:${loader.global_batch_size}, ${eval:${trainer.devices}
+    * ${loader.batch_size} * ${trainer.num_nodes}}}
+  gradient_clip_val: 1.0
+  precision: bf16
+  num_sanity_val_steps: 2
+  max_steps: 1000000
+  log_every_n_steps: 100
+  limit_train_batches: 1.0
+  limit_val_batches: 1.0
+  val_check_interval: 5000
+  check_val_every_n_epoch: null
+wandb:
+  project: duo
+  group: null
+  job_type: null
+  name: null
+  id: ${.name}_${seed}
+  tags:
+  - ${noise.type}
+  - ${data.train}
+  - ${data.valid}
+  - ${algo.name}
+  offline: true
+checkpointing:
+  save_dir: /home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_dcd/steps_8/upstream_hydra
+  resume_from_ckpt: true
+  resume_ckpt_path: ${.save_dir}/checkpoints/last.ckpt
+callbacks:
+  checkpoint_every_n_steps:
+    _target_: lightning.pytorch.callbacks.ModelCheckpoint
+    save_top_k: -1
+    save_last: true
+    dirpath: ${checkpointing.save_dir}/checkpoints
+    verbose: true
+    auto_insert_metric_name: false
+    every_n_train_steps: 500
+  checkpoint_monitor:
+    _target_: lightning.pytorch.callbacks.ModelCheckpoint
+    monitor: val/nll
+    mode: min
+    save_top_k: 1
+    save_last: false
+    dirpath: ${checkpointing.save_dir}/checkpoints
+    filename: best
+    auto_insert_metric_name: false
+    verbose: true
+  learning_rate_monitor:
+    _target_: lightning.pytorch.callbacks.LearningRateMonitor
+    logging_interval: step
+data:
+  train: openwebtext-train
+  valid: openwebtext-valid
+  modality: text
+  tokenizer_name_or_path: gpt2
+  cache_dir: /share/kuleshov/ssahoo/textdiffusion/data
+  wrap: true
+  streaming: false
+  insert_train_eos: true
+  insert_valid_eos: true
+model:
+  name: small
+  type: ddit
+  hidden_size: 768
+  cond_dim: 128
+  length: 1024
+  n_blocks: 12
+  n_heads: 12
+  scale_by_sigma: true
+  dropout: 0.1
+  tie_word_embeddings: false
+  vocab_lookup: true
+strategy:
+  _target_: lightning.pytorch.strategies.DDPStrategy
+  find_unused_parameters: false
+noise:
+  type: log-linear
+  eps: 0.001
+lr_scheduler:
+  _target_: transformers.get_constant_schedule_with_warmup
+  num_warmup_steps: 2500
+prior:
+  type: none
+  latent_width: 0
+  latent_height: 0
+algo:
+  name: duo_base
+  backbone: hf_dit
+  parameterization: mean
+  time_conditioning: true
+  T: 0
+  subs_masking: false
+  causal_attention: false
+  ignore_bos: false
+  loss_type: elbo

results/conditional/samples/owt/duo_dcd/steps_8/upstream_hydra/config_tree.txt

@@ -0,0 +1,169 @@
+CONFIG
+├── mode
+│   └── sample_eval                                                             
+├── seed
+│   └── 42                                                                      
+├── loader
+│   └── global_batch_size: 512                                                  
+│       eval_global_batch_size: 512                                             
+│       batch_size: 2                                                           
+│       eval_batch_size: 1                                                      
+│       num_workers: 512                                                        
+│       pin_memory: true                                                        
+│                                                                               
+├── sampling
+│   └── predictor: ancestral                                                    
+│       steps: 8                                                                
+│       noise_removal: ancestral                                                
+│       use_float64: true                                                       
+│       p_nucleus: 1.0                                                          
+│       num_sample_batches: 1                                                   
+│       num_sample_log: 2                                                       
+│       semi_ar: false                                                          
+│       stride_length: 1                                                        
+│       num_strides: 1                                                          
+│       guid_weight: null                                                       
+│       psi:                                                                    
+│         time_profile: linear                                                  
+│         high_mode: pure-posterior                                             
+│         middle_mode: pure-posterior                                           
+│         low_mode: pure-posterior                                              
+│         high_frac: 0.2                                                        
+│         middle_frac: 0.6                                                      
+│                                                                               
+├── training
+│   └── ema: 0.9999                                                             
+│       antithetic_sampling: true                                               
+│       importance_sampling: false                                              
+│       sampling_eps: 0.001                                                     
+│       change_of_variables: false                                              
+│       loss_precision: bf16                                                    
+│       finetune_path: ''                                                       
+│       class_dropout_p: 0.1                                                    
+│                                                                               
+├── eval
+│   └── checkpoint_path: /home/ruizzhao/diffusion_baseline/checkpoints/official/
+│       disable_ema: false                                                      
+│       compute_generative_perplexity: false                                    
+│       perplexity_batch_size: 8                                                
+│       compute_perplexity_on_sanity: false                                     
+│       gen_ppl_eval_model_name_or_path: gpt2-large                             
+│       generate_samples: true                                                  
+│       generated_samples_path: /home/ruizzhao/diffusion_baseline/results/condit
+│                                                                               
+├── optim
+│   └── weight_decay: 0                                                         
+│       lr: 0.0003                                                              
+│       beta1: 0.9                                                              
+│       beta2: 0.999                                                            
+│       eps: 1.0e-08                                                            
+│                                                                               
+├── trainer
+│   └── _target_: lightning.Trainer                                             
+│       accelerator: cuda                                                       
+│       num_nodes: 1                                                            
+│       devices: 1                                                              
+│       accumulate_grad_batches: 256                                            
+│       gradient_clip_val: 1.0                                                  
+│       precision: bf16                                                         
+│       num_sanity_val_steps: 2                                                 
+│       max_steps: 1000000                                                      
+│       log_every_n_steps: 100                                                  
+│       limit_train_batches: 1.0                                                
+│       limit_val_batches: 1.0                                                  
+│       val_check_interval: 5000                                                
+│       check_val_every_n_epoch: null                                           
+│                                                                               
+├── wandb
+│   └── project: duo                                                            
+│       group: null                                                             
+│       job_type: null                                                          
+│       name: null                                                              
+│       id: None_42                                                             
+│       tags:                                                                   
+│       - log-linear                                                            
+│       - openwebtext-train                                                     
+│       - openwebtext-valid                                                     
+│       - duo_base                                                              
+│       offline: true                                                           
+│                                                                               
+├── checkpointing
+│   └── save_dir: /home/ruizzhao/diffusion_baseline/results/conditional/samples/
+│       resume_from_ckpt: true                                                  
+│       resume_ckpt_path: /home/ruizzhao/diffusion_baseline/results/conditional/
+│                                                                               
+├── callbacks
+│   └── checkpoint_every_n_steps:                                               
+│         _target_: lightning.pytorch.callbacks.ModelCheckpoint                 
+│         save_top_k: -1                                                        
+│         save_last: true                                                       
+│         dirpath: /home/ruizzhao/diffusion_baseline/results/conditional/samples
+│         verbose: true                                                         
+│         auto_insert_metric_name: false                                        
+│         every_n_train_steps: 500                                              
+│       checkpoint_monitor:                                                     
+│         _target_: lightning.pytorch.callbacks.ModelCheckpoint                 
+│         monitor: val/nll                                                      
+│         mode: min                                                             
+│         save_top_k: 1                                                         
+│         save_last: false                                                      
+│         dirpath: /home/ruizzhao/diffusion_baseline/results/conditional/samples
+│         filename: best                                                        
+│         auto_insert_metric_name: false                                        
+│         verbose: true                                                         
+│       learning_rate_monitor:                                                  
+│         _target_: lightning.pytorch.callbacks.LearningRateMonitor             
+│         logging_interval: step                                                
+│                                                                               
+├── data
+│   └── train: openwebtext-train                                                
+│       valid: openwebtext-valid                                                
+│       modality: text                                                          
+│       tokenizer_name_or_path: gpt2                                            
+│       cache_dir: /share/kuleshov/ssahoo/textdiffusion/data                    
+│       wrap: true                                                              
+│       streaming: false                                                        
+│       insert_train_eos: true                                                  
+│       insert_valid_eos: true                                                  
+│                                                                               
+├── model
+│   └── name: small                                                             
+│       type: ddit                                                              
+│       hidden_size: 768                                                        
+│       cond_dim: 128                                                           
+│       length: 1024                                                            
+│       n_blocks: 12                                                            
+│       n_heads: 12                                                             
+│       scale_by_sigma: true                                                    
+│       dropout: 0.1                                                            
+│       tie_word_embeddings: false                                              
+│       vocab_lookup: true                                                      
+│                                                                               
+├── strategy
+│   └── _target_: lightning.pytorch.strategies.DDPStrategy                      
+│       find_unused_parameters: false                                           
+│                                                                               
+├── noise
+│   └── type: log-linear                                                        
+│       eps: 0.001                                                              
+│                                                                               
+├── lr_scheduler
+│   └── _target_: transformers.get_constant_schedule_with_warmup                
+│       num_warmup_steps: 2500                                                  
+│                                                                               
+├── prior
+│   └── type: none                                                              
+│       latent_width: 0                                                         
+│       latent_height: 0                                                        
+│                                                                               
+└── algo
+    └── name: duo_base                                                          
+        backbone: hf_dit                                                        
+        parameterization: mean                                                  
+        time_conditioning: true                                                 
+        T: 0                                                                    
+        subs_masking: false                                                     
+        causal_attention: false                                                 
+        ignore_bos: false                                                       
+        loss_type: elbo                                                         
+                                                                                

results/conditional/samples/owt/duo_di4c/steps_8/conversion_metadata.json

@@ -0,0 +1,33 @@
+{
+  "checkpoint_lock_id": "recipe:duo_di4c_owt:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152",
+  "checkpoint_selection": {
+    "output": "checkpoints/reference_reproduction/duo_di4c/owt",
+    "recipe": "duo_di4c_owt",
+    "sampling_checkpoint": "student_checkpoints/50000.ckpt",
+    "sampling_config": "config.yaml",
+    "source": "di4c",
+    "source_commit": "ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3",
+    "teacher_adapter": "uniform_to_absorbing",
+    "teacher_family": "uniform_duo"
+  },
+  "checkpoint_sha256": "41d65d63b4a886200cb53efbb83318d300430c628f1569b4c7aab66aae3cdef9",
+  "conditioning_implementation": "native_projection",
+  "exclude_from_latency": true,
+  "format": "dlb-conditional-token-capture-v1",
+  "generated_samples": 2048,
+  "generation_seconds_sentinel": 0.0,
+  "generation_seconds_source": "unavailable_excluded_sentinel",
+  "paper_conditional_training_reproduced": false,
+  "requested_samples": 2048,
+  "teacher_family": "uniform_duo",
+  "token_ids_source": "upstream_conditional_capture",
+  "token_ids_transformation": {
+    "max_length": 1024,
+    "operation": "validated_exact_length"
+  },
+  "tokenizer": "gpt2",
+  "tokenizer_revision": "607a30d783dfa663caf39e06633721c8d4cfcd7e",
+  "trim_policy": "none_exact_conditional_schedule",
+  "trimmed_samples": 0,
+  "upstream_output": "/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_di4c/steps_8/upstream_token_ids.json"
+}

results/conditional/samples/owt/duo_di4c/steps_8/run_metadata.json

@@ -0,0 +1,133 @@
+{
+  "checkpoint_lock_id": "recipe:duo_di4c_owt:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152",
+  "checkpoint_selection": {
+    "output": "checkpoints/reference_reproduction/duo_di4c/owt",
+    "recipe": "duo_di4c_owt",
+    "sampling_checkpoint": "student_checkpoints/50000.ckpt",
+    "sampling_config": "config.yaml",
+    "source": "di4c",
+    "source_commit": "ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3",
+    "teacher_adapter": "uniform_to_absorbing",
+    "teacher_family": "uniform_duo"
+  },
+  "checkpoint_sha256": "41d65d63b4a886200cb53efbb83318d300430c628f1569b4c7aab66aae3cdef9",
+  "checkpoint_teacher_family": "uniform_duo",
+  "command": [
+    "/opt/anaconda3/envs/dlb-di4c/bin/python",
+    "-B",
+    "-u",
+    "/home/ruizzhao/diffusion_baseline/adapters/sample_di4c.py",
+    "--upstream-root",
+    "/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt",
+    "--checkpoint",
+    "/home/ruizzhao/diffusion_baseline/checkpoints/reference_reproduction/duo_di4c/owt/student_checkpoints/50000.ckpt",
+    "--config",
+    "/home/ruizzhao/diffusion_baseline/checkpoints/reference_reproduction/duo_di4c/owt/config.yaml",
+    "--checkpoint-sha256",
+    "9ce6481909aba45321170d89ac8cda97ad99611d52f37149b569e331cbbb0e6f",
+    "--config-sha256",
+    "b46b04e22137f51adffccb0d7bd3f093109c5df02e3ce2d93bb36565df8fcee4",
+    "--data-config",
+    "/home/ruizzhao/diffusion_baseline/artifacts/data.yaml",
+    "--downloads-manifest",
+    "/home/ruizzhao/diffusion_baseline/data/manifests/downloads.json",
+    "--tokenizer-snapshot",
+    "/home/ruizzhao/diffusion_baseline/data/raw/huggingface/hub/models--gpt2/snapshots/607a30d783dfa663caf39e06633721c8d4cfcd7e",
+    "--output",
+    "/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_di4c/steps_8/upstream_token_ids.json",
+    "--sample-count",
+    "2048",
+    "--batch-size",
+    "16",
+    "--num-steps",
+    "8",
+    "--seq-len",
+    "1024",
+    "--seed",
+    "42",
+    "--sampler",
+    "ancestral",
+    "--teacher-family",
+    "uniform_duo",
+    "--dataset",
+    "owt",
+    "--generation-mode",
+    "conditional_prefix",
+    "--conditioning-manifest",
+    "/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json",
+    "--conditioning-manifest-sha256",
+    "7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32",
+    "--conditioning-config-sha256",
+    "89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114",
+    "--prefix-length",
+    "64",
+    "--evaluation-continuation-length",
+    "64",
+    "--prompt-count",
+    "1024",
+    "--diversity-prompt-count",
+    "256",
+    "--completions-per-diversity-prompt",
+    "5",
+    "--completion-schedule",
+    "c0:p0-1023;c1-4:p0-255",
+    "--offline",
+    "true",
+    "--allow-missing-embedded-config",
+    "true"
+  ],
+  "command_sha256": "189cb944c78db758e8ef63ab3e500cb4732b29e063c0bfd1341390891681102c",
+  "config_sha256": "6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f",
+  "environment": {
+    "cuda_version": null,
+    "cuda_visible_devices": "0",
+    "hostname": "DS210029",
+    "platform": "Linux-5.15.0-186-generic-x86_64-with-glibc2.35",
+    "python": "3.10.20",
+    "torch_version": "2.5.1"
+  },
+  "finished_at": "2026-09-02T09:18:15.928820Z",
+  "identity": {
+    "adapter_identity": "dlb.adapters.di4c:v3|dlb.adapters.di4c:Di4CAdapter|e1636b46628a6015944553de2d5d8a1b829bb6bf3f9c6b789d829cb6db222ec9",
+    "checkpoint_lock_id": "recipe:duo_di4c_owt:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152",
+    "checkpoint_selection": {
+      "output": "checkpoints/reference_reproduction/duo_di4c/owt",
+      "recipe": "duo_di4c_owt",
+      "sampling_checkpoint": "student_checkpoints/50000.ckpt",
+      "sampling_config": "config.yaml",
+      "source": "di4c",
+      "source_commit": "ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3",
+      "teacher_adapter": "uniform_to_absorbing",
+      "teacher_family": "uniform_duo"
+    },
+    "checkpoint_sha256": "41d65d63b4a886200cb53efbb83318d300430c628f1569b4c7aab66aae3cdef9",
+    "checkpoint_teacher_family": "uniform_duo",
+    "command_sha256": "189cb944c78db758e8ef63ab3e500cb4732b29e063c0bfd1341390891681102c",
+    "completion_schedule": "c0:p0-1023;c1-4:p0-255",
+    "completions_per_diversity_prompt": 5,
+    "conditioning_config_sha256": "89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114",
+    "conditioning_manifest": "/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json",
+    "conditioning_manifest_sha256": "7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32",
+    "config_sha256": "6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f",
+    "dataset_id": "owt",
+    "device": null,
+    "diversity_prompt_count": 256,
+    "environment": "dlb-di4c",
+    "evaluation_continuation_length": 64,
+    "generation_mode": "conditional_prefix",
+    "model_id": "duo_di4c",
+    "prefix_length": 64,
+    "prompt_count": 1024,
+    "results_root": "/home/ruizzhao/diffusion_baseline/results/conditional",
+    "run_id": "duo_di4c-owt-steps-8",
+    "sample_count": 2048,
+    "seed": 42,
+    "source_sha256": "ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3",
+    "step_count": 8
+  },
+  "source_sha256": "ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3",
+  "started_at": "2026-09-02T09:15:34.579550Z",
+  "status": "succeeded",
+  "stderr_log": "/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_di4c/steps_8/stderr.log",
+  "stdout_log": "/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/duo_di4c/steps_8/stdout.log"
+}

results/conditional/samples/owt/fmlm/steps_16/stderr.log

@@ -0,0 +1,4 @@
+Seed set to 42
+/opt/anaconda3/envs/dlb-flm/lib/python3.11/site-packages/lightning/pytorch/utilities/migration/utils.py:56: The loaded checkpoint was produced with Lightning v2.6.1, which is newer than your current Lightning version: v2.6.0
+/home/ruizzhao/diffusion_baseline/upstreams/flm/models/dit.py:476: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead.
+  with torch.cuda.amp.autocast(enabled=False):

results/conditional/samples/owt/fmlm/steps_16/upstream_samples.json

@@ -0,0 +1,7 @@
+{
+    "generative_ppl": 51.365478515625,
+    "entropy": 5.192865371704102,
+    "generated_seqs": [
+        "<|endoftext|> at least two women. No, yes, yes, a meeting with the United Nations, Russia, Clinton, and the president for that. I assure you this is an interesting campaign:\n\nYou simply said none of words.\n\nNo; not as much as in Mar-a-hush and in meetings.\n\nNo, not at a reporter you\u2019ve asked for months.\n\nNo, when someone else asked, laughed or laughed at the report toward you.\n\nKissay\u2026\n\nBut more\u2026\n\nYou have kept saying this: There is a control, by the president, an often personal, and ongoing operation, you know, and that is reaping you and me \u2013 all with you, but not at the top of any of the other agencies.\n\nThere\u2019s a lot of good, and sourced reporting that on that subject, what made me feel it hasn\u2019t been made public.\n\nYou people have repeatedly said that the president was not in a private or Mar-a-Lago to report matters to and the press would investigate it.\n\nNow, see if you\u2019ve got the sense that\u2019s what I\u2019ve read in your experience that I was going to be reading. I trust one another with her words, but then I have come at things that didn\u2019t you have actually said even what\u2019s to them.\n\nNow, I will admit, that kind of stuff has always had some merit. But the Trump administration, cannot understand that the country no one else is waking up from itself, just a mess. He has taken a hard time focusing on the issue.\n\nNow, if the issue is political, I can understand.\n\nSo, while you said that hasn\u2019t happened by now, the Russians are now making a terrible, bad report.\n\nHere is the thing which doesn\u2019t quite that either: the fact that there are a lot of facts tucked away in the way of the 90\u2019s \u2013 90% of that \u2013 which was obvious to me as well. According to your campaign\u2019s rhetoric there are Hillary\u2019s playing questions of all control, and with whom, none. As always being a Democrat, it\u2019s on the Clintons and the leaders of all parties\u2019 campaigns that are willing and willing to always lead the money to their advantage. So let\u2019s get out back on early Friday.\n\nRelated:<|endoftext|>Unfortunately for decades, the wars among countries have become so fraught with mass us. The fact that many of us in the left will take more and more part in fighting the \"revolution\" has been deemed even attention to, although not self-serving. In two years the elections have passed and the presidency of the new United States began in earnest. I wanted to share my not-so with protest on behalf of an anti-globalitarian campaign that involves onts up the chain of democracy rather than down on the part of the ruling elite, who eventually will come home, and lead a book of peace.\n\nThe citizens of most of their time are being paid by the press and the press to tell the truth.\n\nThe fact that journalists and journalists will now have to report on the actions of the government they claim have been the way out of the 2014 U.S. elections and also the large swhes of Americans in the post-election War continues to take home to give in touch with the most dangerous by-products of democracy.\n\nShortly after the president went into office a quarter of a century that the U.S. President learned a step or two from describing his government\u2019s total abuse of power as a thinly affiment to social justice and a sign of trying to undermine the president\u2019s tyrannocratic system of law by promising to pass a new code of conduct, the quickly expanded to include the judiciary.\n\nAccordingly, he proposed that Congress spent a few weeks between the Constitution and the Bill of Rights as \u201ca couple of innocent children \u2013 short on moralness\u201d that would release pregnant women from bondage to \u201ca society of such force as the lusts of the people.\u201d As Orwell wrote:\n\nThis proposal by the extreme right means that Congress and the Supreme Party itself are still to blame from the crisis. This is the day when, when faced with the poor and Congress, the government does nothing in a way that givesates to the citizens of control. Based on first such occasion, the government would be left with the condition that a constitution and a code of social justice out of necessity be established. The government, the press and Congress would now take the move forward.\n\nRightsous protests that are being gets to the point as the members of Congress do upon public pressure to overthrow the government. The police are trying to prove the point that since their inception have shown the press what they know and are afraid. Before the authorities even have done<|endoftext|>"
+    ]
+}

results/conditional/samples/owt/fmlm/steps_2/run_metadata.json

@@ -0,0 +1,110 @@
+{
+  "checkpoint_lock_id": "recipe:fmlm_owt_official_ckpt:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152",
+  "checkpoint_selection": {
+    "output": "checkpoints/official/fmlm_ckpt/owt",
+    "recipe": "fmlm_owt_official_ckpt",
+    "sampling_checkpoint": "owt_fmlm.ckpt",
+    "sampling_config": null,
+    "source": "flm",
+    "source_commit": "a1918d5164e5038e37d0b7a4fb2010ce75b863b3",
+    "teacher_adapter": null,
+    "teacher_family": "continuous_flm"
+  },
+  "checkpoint_sha256": "3ab473f823f6f30d7617708759e5f8edf38159fb7751d4afedc96f88c2d312b4",
+  "checkpoint_teacher_family": "continuous_flm",
+  "command": [
+    "/opt/anaconda3/envs/dlb-flm/bin/python",
+    "-u",
+    "-m",
+    "dlb.adapters.capture",
+    "--upstream-entrypoint=/home/ruizzhao/diffusion_baseline/upstreams/flm/main.py",
+    "--capture-path=/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/fmlm/steps_2/upstream_token_ids.json",
+    "--generation-mode=conditional_prefix",
+    "--conditioning-manifest=/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json",
+    "--conditioning-manifest-sha256=7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32",
+    "--conditioning-config-sha256=89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114",
+    "--prefix-length=64",
+    "--evaluation-continuation-length=64",
+    "--prompt-count=1024",
+    "--diversity-prompt-count=256",
+    "--completions-per-diversity-prompt=5",
+    "--completion-schedule=c0:p0-1023;c1-4:p0-255",
+    "--",
+    "mode=sample_eval",
+    "seed=42",
+    "data=openwebtext-split",
+    "model=small",
+    "model.length=1024",
+    "loader.batch_size=2",
+    "loader.eval_batch_size=16",
+    "sampling.num_sample_batches=128",
+    "sampling.steps=2",
+    "eval.checkpoint_path=/home/ruizzhao/diffusion_baseline/checkpoints/official/fmlm_ckpt/owt/owt_fmlm.ckpt",
+    "trainer.devices=1",
+    "hydra.run.dir=/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/fmlm/steps_2/upstream_hydra",
+    "checkpointing.save_dir=/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/fmlm/steps_2/upstream_hydra",
+    "+wandb.offline=true",
+    "eval.generated_samples_path=/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/fmlm/steps_2/upstream_samples.json",
+    "algo=fmlm",
+    "algo.backbone=dit",
+    "eval.disable_ema=False",
+    "sampling.solver=euler",
+    "algo.double_temb=True",
+    "algo.learnable_loss_weighting=False",
+    "sampling.gamma=1.0"
+  ],
+  "command_sha256": "ad7748be2099ffdd10a46054d77dcccabf3f3875673387293f9f0f1ed0b85151",
+  "config_sha256": "6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f",
+  "environment": {
+    "cuda_version": null,
+    "cuda_visible_devices": "7",
+    "hostname": "DS210029",
+    "platform": "Linux-5.15.0-186-generic-x86_64-with-glibc2.35",
+    "python": "3.11.15",
+    "torch_version": "2.5.1"
+  },
+  "finished_at": "2026-08-27T06:16:11.155324Z",
+  "identity": {
+    "adapter_identity": "dlb.adapters.flm:v2|dlb.adapters.flm:FLMAdapter|c3a6a4573200fecf38a47fc6e7bc78254ca362fff3b96079dc2efa9f8e27aff0",
+    "checkpoint_lock_id": "recipe:fmlm_owt_official_ckpt:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152",
+    "checkpoint_selection": {
+      "output": "checkpoints/official/fmlm_ckpt/owt",
+      "recipe": "fmlm_owt_official_ckpt",
+      "sampling_checkpoint": "owt_fmlm.ckpt",
+      "sampling_config": null,
+      "source": "flm",
+      "source_commit": "a1918d5164e5038e37d0b7a4fb2010ce75b863b3",
+      "teacher_adapter": null,
+      "teacher_family": "continuous_flm"
+    },
+    "checkpoint_sha256": "3ab473f823f6f30d7617708759e5f8edf38159fb7751d4afedc96f88c2d312b4",
+    "checkpoint_teacher_family": "continuous_flm",
+    "command_sha256": "ad7748be2099ffdd10a46054d77dcccabf3f3875673387293f9f0f1ed0b85151",
+    "completion_schedule": "c0:p0-1023;c1-4:p0-255",
+    "completions_per_diversity_prompt": 5,
+    "conditioning_config_sha256": "89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114",
+    "conditioning_manifest": "/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json",
+    "conditioning_manifest_sha256": "7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32",
+    "config_sha256": "6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f",
+    "dataset_id": "owt",
+    "device": null,
+    "diversity_prompt_count": 256,
+    "environment": "dlb-flm",
+    "evaluation_continuation_length": 64,
+    "generation_mode": "conditional_prefix",
+    "model_id": "fmlm",
+    "prefix_length": 64,
+    "prompt_count": 1024,
+    "results_root": "/home/ruizzhao/diffusion_baseline/results/conditional",
+    "run_id": "fmlm-owt-steps-2",
+    "sample_count": 2048,
+    "seed": 42,
+    "source_sha256": "a1918d5164e5038e37d0b7a4fb2010ce75b863b3",
+    "step_count": 2
+  },
+  "source_sha256": "a1918d5164e5038e37d0b7a4fb2010ce75b863b3",
+  "started_at": "2026-08-27T06:11:06.393521Z",
+  "status": "succeeded",
+  "stderr_log": "/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/fmlm/steps_2/stderr.log",
+  "stdout_log": "/home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/fmlm/steps_2/stdout.log"
+}

results/conditional/samples/owt/fmlm/steps_32/upstream_samples.json

@@ -0,0 +1,7 @@
+{
+    "generative_ppl": 39.642539978027344,
+    "entropy": 5.057096004486084,
+    "generated_seqs": [
+        "<|endoftext|> in that, you can\u2019t tell the difference between young men beating you up on as old as sex, love, and a child. No one is there to protect you with this.\n\nPart of that story was happening outside the church, something that we did witness together.\n\nFor only a few months now, that led to the attention of the hundreds of thousands of young girls trying to express its message of love, especially for children, and for women. It\u2019s also a good example of the value of the internet, a huge problem that is growing. And a small minority of kids is a safe, sometimesless, and dangerous way away because it figures out how to deal with person, child, or child.\n\nFair enough.\n\nSo you believe the ending of that story is the only step we can begin to take?\n\nThe other way to answer that is by creating a welfare system. We cannot welfare as long as we have to know what a person\u2019s life needs. Socialicians are not a solution. Any So that responds to this, everyone supported by society and everyone that gets supports, will social [ize] more as a result. This may not actually be, but it\u2019s a form of it that can be used. For some, there is nothing but moral poverty.\n\nAs a result, though, we have to still be getting more political attention, and those who oppose us ignore the fact that the price of care has increased during the campaign. So, legalization would help to increase the ability of one\u2019s life to hold on to.\n\nWhat is the definition of a child right?\n\nWell, being involved with a crime or never allowed to. They can be, or get along with it, for their actions, one has to work in a process that isn\u2019t as long as the circumstances are right, and the record and the focus are on the issue. If you keep the kids, if you let them go, they will come back to the table.\n\nBut what changes will be for other churches, and how they operate, to go from isolated church to country and church that not only have the same message or do the same, but you can never spend as much money and even to educate and support.\n\nHow do you get behind it?\n\nRegardless of the story, it may not be free in our system, that\u2019s just the music. The music is the same, sometimes it loud enough to make things worse, so the government can keep it on it for years. But for years of listening reform, we\u2019re just trying a little more pop music.\n\n\u2014\n\nL. Mair is one of those artists.<|endoftext|>This is a little story we\u2019ve got into in World of a Thousand Economy, with a focus on the internet and its world order.\n\nIt has been difficult. You can see the public reaction most of the time. The political differences have been there, but it was in the U.S. and Canada for not making a treaty (or a deal), but because of that disconnect as well.\n\nWe all meet a friend, player, or coach in the family, and we have to hear from them because they got a game through different experiences with players, and so have an issue. This is not going to be a rant, and to be found by them of course, but our own experiences of it are, unfortunately.\n\nBeingership is not easy to have, and will always be difficult to speak for everybody. It is, in fact \u2014 it is stuff that people fight all the time, and for one another, all over the time around me.\n\nBut meeting the other of them is different and cuter. Games are important for business. To assume that people actually care for them is more than just on a U.S. government website, so that there is a safe place and running into problems with another for free.\n\nBut that\u2019s not true. There is no way of being financially responsible for your business, but being there, it also must be really good to be \u201cfair.\u201d\n\nNow, if all of that stuff had to come from someone who\u2019s been trying to adapt to the game, and is questioning at times a company\u2019s business model, the ability of players to care would be just not as good as they could.\n\nAnd that\u2019s how we created the whole process.\n\nSome of you, and others, won\u2019t get involved at all when the other guys are left outside. So, ignore all that and leave it to here, the same as thoughist X and the other that most people will find.\n\nWhat will we learn? We can be there for all three nations on Earth, and the same goes for the world\u2019s governments and losers \u2014 the two of our United States and Canada<|endoftext|>"
+    ]
+}

results/conditional/samples/owt/fmlm/steps_4/upstream_hydra/.hydra/config.yaml

@@ -0,0 +1,174 @@
+mode: sample_eval
+is_di4c: false
+is_di4c_deterministic: true
+seed: 42
+loader:
+  global_batch_size: 512
+  eval_global_batch_size: ${.global_batch_size}
+  batch_size: 2
+  eval_batch_size: 1
+  num_workers: 4
+  pin_memory: true
+sampling:
+  predictor: ancestral
+  steps: 4
+  noise_removal: ancestral
+  use_float64: true
+  p_nucleus: 1.0
+  num_sample_batches: 1
+  num_sample_log: 2
+  semi_ar: false
+  stride_length: 1
+  num_strides: 1
+  num_reflow_samples: 50000
+  temperature: 1.0
+  solver: euler
+  gamma: 1.0
+training:
+  loss_type: flow
+  pred_type: x0
+  ema: 0.9999
+  antithetic_sampling: true
+  importance_sampling: false
+  sampling_eps: 0.001
+  change_of_variables: false
+  loss_precision: bf16
+  finetune_path: ''
+  not_sampling_t: false
+  load_ema: false
+  t_curriculum_steps: 100000
+  t_curriculum_min: 0.0
+  t_curriculum_max: 1.0
+eval:
+  checkpoint_path: /home/ruizzhao/diffusion_baseline/checkpoints/official/fmlm_ckpt/owt/owt_fmlm.ckpt
+  disable_ema: false
+  ema_decay: null
+  compute_generative_perplexity: true
+  perplexity_batch_size: 8
+  compute_perplexity_on_sanity: false
+  gen_ppl_eval_model_name_or_path: gpt2-large
+  generate_samples: true
+  generated_samples_path: /home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/fmlm/steps_4/upstream_samples.json
+optim:
+  weight_decay: 0
+  lr: 0.0003
+  beta1: 0.9
+  beta2: 0.999
+  eps: 1.0e-08
+  ln_tune: none
+trainer:
+  _target_: lightning.Trainer
+  accelerator: cuda
+  num_nodes: 1
+  devices: 1
+  accumulate_grad_batches: ${div_up:${loader.global_batch_size}, ${eval:${trainer.devices}
+    * ${loader.batch_size} * ${trainer.num_nodes}}}
+  gradient_clip_val: 1.0
+  precision: bf16
+  num_sanity_val_steps: 2
+  max_steps: 1000000
+  log_every_n_steps: 100
+  limit_train_batches: 1.0
+  limit_val_batches: 1.0
+  val_check_interval: 5000
+wandb:
+  project: flm
+  notes: FLM
+  group: null
+  job_type: null
+  name: null
+  id: ${.name}_${seed}
+  tags:
+  - ${data.train}
+  - ${algo.name}
+  offline: true
+checkpointing:
+  save_dir: /home/ruizzhao/diffusion_baseline/results/conditional/samples/owt/fmlm/steps_4/upstream_hydra
+  resume_from_ckpt: false
+  resume_ckpt_path: ${.save_dir}/checkpoints/last.ckpt
+callbacks:
+  checkpoint_every_n_steps:
+    _target_: lightning.pytorch.callbacks.ModelCheckpoint
+    save_top_k: -1
+    save_last: false
+    dirpath: ${checkpointing.save_dir}/checkpoints
+    verbose: true
+    auto_insert_metric_name: false
+    every_n_train_steps: 25000
+  checkpoint_monitor:
+    _target_: lightning.pytorch.callbacks.ModelCheckpoint
+    monitor: val/nll
+    mode: min
+    save_top_k: 1
+    save_last: true
+    dirpath: ${checkpointing.save_dir}/checkpoints
+    filename: best_nll
+    auto_insert_metric_name: false
+    verbose: true
+  learning_rate_monitor:
+    _target_: lightning.pytorch.callbacks.LearningRateMonitor
+    logging_interval: step
+data:
+  train: openwebtext-train
+  valid: openwebtext-valid
+  tokenizer_name_or_path: gpt2
+  cache_dir: ${hydra:runtime.cwd}/datasets/openwebtext
+  wrap: true
+  streaming: false
+  insert_train_eos: true
+  insert_valid_eos: true
+model:
+  name: small
+  type: ddit
+  hidden_size: 768
+  cond_dim: 128
+  length: 1024
+  n_blocks: 12
+  n_heads: 12
+  scale_by_sigma: true
+  dropout: 0.1
+  tie_word_embeddings: false
+  vocab_lookup: true
+strategy:
+  _target_: lightning.pytorch.strategies.DDPStrategy
+  find_unused_parameters: false
+noise:
+  type: log-linear
+  parameterization: log-linear
+  eps: 0
+  denoiser_latent_conditioning: -1
+  freeze_encoder: false
+  freeze_decoder: false
+lr_scheduler:
+  _target_: transformers.get_constant_schedule_with_warmup
+  num_warmup_steps: 2500
+prior:
+  type: none
+  latent_width: 0
+  latent_height: 0
+algo:
+  name: fmlm
+  backbone: dit
+  parameterization: mean
+  time_conditioning: true
+  T: 0
+  subs_masking: false
+  causal_attention: false
+  t_max: 1.0
+  t_min: 0.0
+  double_temb: true
+  add_boundary: fixed
+  diagonal_fraction: 0.5
+  offdiagonal_sampling: uniform_diff
+  set_midpoint: midpoint
+  boundary_prob: 32
+  distillation_method: PSD
+  entmax_temp_lsd: 2.0
+  backprop_entmax_temp_lsd: false
+  learnable_loss_weighting: false
+  use_mse_loss_psd: false
+  rescale_offdiag_loss_psd: false
+  use_ema_for_psd_target: false
+  use_teacher_for_D_t_lsd: false
+  initialize_student_from_teacher: true
+  teacher_path: ''

results/conditional/samples/owt/fmlm/steps_4/upstream_hydra/config_tree.txt

@@ -0,0 +1,194 @@
+CONFIG
+├── mode
+│   └── sample_eval                                                             
+├── is_di4c
+│   └── False                                                                   
+├── is_di4c_deterministic
+│   └── True                                                                    
+├── seed
+│   └── 42                                                                      
+├── loader
+│   └── global_batch_size: 512                                                  
+│       eval_global_batch_size: 512                                             
+│       batch_size: 2                                                           
+│       eval_batch_size: 1                                                      
+│       num_workers: 4                                                          
+│       pin_memory: true                                                        
+│                                                                               
+├── sampling
+│   └── predictor: ancestral                                                    
+│       steps: 4                                                                
+│       noise_removal: ancestral                                                
+│       use_float64: true                                                       
+│       p_nucleus: 1.0                                                          
+│       num_sample_batches: 1                                                   
+│       num_sample_log: 2                                                       
+│       semi_ar: false                                                          
+│       stride_length: 1                                                        
+│       num_strides: 1                                                          
+│       num_reflow_samples: 50000                                               
+│       temperature: 1.0                                                        
+│       solver: euler                                                           
+│       gamma: 1.0                                                              
+│                                                                               
+├── training
+│   └── loss_type: flow                                                         
+│       pred_type: x0                                                           
+│       ema: 0.9999                                                             
+│       antithetic_sampling: true                                               
+│       importance_sampling: false                                              
+│       sampling_eps: 0.001                                                     
+│       change_of_variables: false                                              
+│       loss_precision: bf16                                                    
+│       finetune_path: ''                                                       
+│       not_sampling_t: false                                                   
+│       load_ema: false                                                         
+│       t_curriculum_steps: 100000                                              
+│       t_curriculum_min: 0.0                                                   
+│       t_curriculum_max: 1.0                                                   
+│                                                                               
+├── eval
+│   └── checkpoint_path: /home/ruizzhao/diffusion_baseline/checkpoints/official/
+│       disable_ema: false                                                      
+│       ema_decay: null                                                         
+│       compute_generative_perplexity: true                                     
+│       perplexity_batch_size: 8                                                
+│       compute_perplexity_on_sanity: false                                     
+│       gen_ppl_eval_model_name_or_path: gpt2-large                             
+│       generate_samples: true                                                  
+│       generated_samples_path: /home/ruizzhao/diffusion_baseline/results/condit
+│                                                                               
+├── optim
+│   └── weight_decay: 0                                                         
+│       lr: 0.0003                                                              
+│       beta1: 0.9                                                              
+│       beta2: 0.999                                                            
+│       eps: 1.0e-08                                                            
+│       ln_tune: none                                                           
+│                                                                               
+├── trainer
+│   └── _target_: lightning.Trainer                                             
+│       accelerator: cuda                                                       
+│       num_nodes: 1                                                            
+│       devices: 1                                                              
+│       accumulate_grad_batches: 256                                            
+│       gradient_clip_val: 1.0                                                  
+│       precision: bf16                                                         
+│       num_sanity_val_steps: 2                                                 
+│       max_steps: 1000000                                                      
+│       log_every_n_steps: 100                                                  
+│       limit_train_batches: 1.0                                                
+│       limit_val_batches: 1.0                                                  
+│       val_check_interval: 5000                                                
+│                                                                               
+├── wandb
+│   └── project: flm                                                            
+│       notes: FLM                                                              
+│       group: null                                                             
+│       job_type: null                                                          
+│       name: null                                                              
+│       id: None_42                                                             
+│       tags:                                                                   
+│       - openwebtext-train                                                     
+│       - fmlm                                                                  
+│       offline: true                                                           
+│                                                                               
+├── checkpointing
+│   └── save_dir: /home/ruizzhao/diffusion_baseline/results/conditional/samples/
+│       resume_from_ckpt: false                                                 
+│       resume_ckpt_path: /home/ruizzhao/diffusion_baseline/results/conditional/
+│                                                                               
+├── callbacks
+│   └── checkpoint_every_n_steps:                                               
+│         _target_: lightning.pytorch.callbacks.ModelCheckpoint                 
+│         save_top_k: -1                                                        
+│         save_last: false                                                      
+│         dirpath: /home/ruizzhao/diffusion_baseline/results/conditional/samples
+│         verbose: true                                                         
+│         auto_insert_metric_name: false                                        
+│         every_n_train_steps: 25000                                            
+│       checkpoint_monitor:                                                     
+│         _target_: lightning.pytorch.callbacks.ModelCheckpoint                 
+│         monitor: val/nll                                                      
+│         mode: min                                                             
+│         save_top_k: 1                                                         
+│         save_last: true                                                       
+│         dirpath: /home/ruizzhao/diffusion_baseline/results/conditional/samples
+│         filename: best_nll                                                    
+│         auto_insert_metric_name: false                                        
+│         verbose: true                                                         
+│       learning_rate_monitor:                                                  
+│         _target_: lightning.pytorch.callbacks.LearningRateMonitor             
+│         logging_interval: step                                                
+│                                                                               
+├── data
+│   └── train: openwebtext-train                                                
+│       valid: openwebtext-valid                                                
+│       tokenizer_name_or_path: gpt2                                            
+│       cache_dir: /home/ruizzhao/diffusion_baseline/datasets/openwebtext       
+│       wrap: true                                                              
+│       streaming: false                                                        
+│       insert_train_eos: true                                                  
+│       insert_valid_eos: true                                                  
+│                                                                               
+├── model
+│   └── name: small                                                             
+│       type: ddit                                                              
+│       hidden_size: 768                                                        
+│       cond_dim: 128                                                           
+│       length: 1024                                                            
+│       n_blocks: 12                                                            
+│       n_heads: 12                                                             
+│       scale_by_sigma: true                                                    
+│       dropout: 0.1                                                            
+│       tie_word_embeddings: false                                              
+│       vocab_lookup: true                                                      
+│                                                                               
+├── strategy
+│   └── _target_: lightning.pytorch.strategies.DDPStrategy                      
+│       find_unused_parameters: false                                           
+│                                                                               
+├── noise
+│   └── type: log-linear                                                        
+│       parameterization: log-linear                                            
+│       eps: 0                                                                  
+│       denoiser_latent_conditioning: -1                                        
+│       freeze_encoder: false                                                   
+│       freeze_decoder: false                                                   
+│                                                                               
+├── lr_scheduler
+│   └── _target_: transformers.get_constant_schedule_with_warmup                
+│       num_warmup_steps: 2500                                                  
+│                                                                               
+├── prior
+│   └── type: none                                                              
+│       latent_width: 0                                                         
+│       latent_height: 0                                                        
+│                                                                               
+└── algo
+    └── name: fmlm                                                              
+        backbone: dit                                                           
+        parameterization: mean                                                  
+        time_conditioning: true                                                 
+        T: 0                                                                    
+        subs_masking: false                                                     
+        causal_attention: false                                                 
+        t_max: 1.0                                                              
+        t_min: 0.0                                                              
+        double_temb: true                                                       
+        add_boundary: fixed                                                     
+        diagonal_fraction: 0.5                                                  
+        offdiagonal_sampling: uniform_diff                                      
+        set_midpoint: midpoint                                                  
+        boundary_prob: 32                                                       
+        distillation_method: PSD                                                
+        entmax_temp_lsd: 2.0                                                    
+        backprop_entmax_temp_lsd: false                                         
+        learnable_loss_weighting: false                                         
+        use_mse_loss_psd: false                                                 
+        rescale_offdiag_loss_psd: false                                         
+        use_ema_for_psd_target: false                                           
+        use_teacher_for_D_t_lsd: false                                          
+        initialize_student_from_teacher: true                                   
+        teacher_path: ''                                                        
+                                                                                

results/conditional/summary_partial51/conditional_partial51_results_long.csv

@@ -0,0 +1,409 @@
+task_id,model,dataset,steps,metric,value
+duo-lm1b-steps-1,duo,lm1b,1,conditional_generative_perplexity,1103.1731695265846
+duo-lm1b-steps-1,duo,lm1b,1,mauve_suffix,0.004769036437965002
+duo-lm1b-steps-1,duo,lm1b,1,sample_entropy,3.832398496448956
+duo-lm1b-steps-1,duo,lm1b,1,reference_entropy,3.8171525992110147
+duo-lm1b-steps-1,duo,lm1b,1,sample_entropy_delta,0.015245897237941453
+duo-lm1b-steps-1,duo,lm1b,1,self_bleu,0.10883688196510578
+duo-lm1b-steps-1,duo,lm1b,1,prefix_exact_match,1.0
+duo-lm1b-steps-1,duo,lm1b,1,seconds_per_sample,0.01323219874757342
+duo-lm1b-steps-2,duo,lm1b,2,conditional_generative_perplexity,403.0607128601672
+duo-lm1b-steps-2,duo,lm1b,2,mauve_suffix,0.007948864164280891
+duo-lm1b-steps-2,duo,lm1b,2,sample_entropy,3.763820682140112
+duo-lm1b-steps-2,duo,lm1b,2,reference_entropy,3.8171525992110147
+duo-lm1b-steps-2,duo,lm1b,2,sample_entropy_delta,-0.05333191707090279
+duo-lm1b-steps-2,duo,lm1b,2,self_bleu,0.14471178741354485
+duo-lm1b-steps-2,duo,lm1b,2,prefix_exact_match,1.0
+duo-lm1b-steps-2,duo,lm1b,2,seconds_per_sample,0.020430842341738753
+duo-lm1b-steps-4,duo,lm1b,4,conditional_generative_perplexity,163.8759815021547
+duo-lm1b-steps-4,duo,lm1b,4,mauve_suffix,0.06012401068746828
+duo-lm1b-steps-4,duo,lm1b,4,sample_entropy,3.762677442509309
+duo-lm1b-steps-4,duo,lm1b,4,reference_entropy,3.8171525992110147
+duo-lm1b-steps-4,duo,lm1b,4,sample_entropy_delta,-0.054475156701705885
+duo-lm1b-steps-4,duo,lm1b,4,self_bleu,0.1634691173142037
+duo-lm1b-steps-4,duo,lm1b,4,prefix_exact_match,1.0
+duo-lm1b-steps-4,duo,lm1b,4,seconds_per_sample,0.032152149695320986
+duo-lm1b-steps-8,duo,lm1b,8,conditional_generative_perplexity,99.97514021433763
+duo-lm1b-steps-8,duo,lm1b,8,mauve_suffix,0.2538963124843746
+duo-lm1b-steps-8,duo,lm1b,8,sample_entropy,3.7804684293475046
+duo-lm1b-steps-8,duo,lm1b,8,reference_entropy,3.8171525992110147
+duo-lm1b-steps-8,duo,lm1b,8,sample_entropy_delta,-0.036684169863510085
+duo-lm1b-steps-8,duo,lm1b,8,self_bleu,0.1678868639465062
+duo-lm1b-steps-8,duo,lm1b,8,prefix_exact_match,1.0
+duo-lm1b-steps-8,duo,lm1b,8,seconds_per_sample,0.06817099513136782
+duo-lm1b-steps-16,duo,lm1b,16,conditional_generative_perplexity,80.94347004980234
+duo-lm1b-steps-16,duo,lm1b,16,mauve_suffix,0.28999537223397975
+duo-lm1b-steps-16,duo,lm1b,16,sample_entropy,3.7954451758645353
+duo-lm1b-steps-16,duo,lm1b,16,reference_entropy,3.8171525992110147
+duo-lm1b-steps-16,duo,lm1b,16,sample_entropy_delta,-0.02170742334647935
+duo-lm1b-steps-16,duo,lm1b,16,self_bleu,0.16888220466713877
+duo-lm1b-steps-16,duo,lm1b,16,prefix_exact_match,1.0
+duo-lm1b-steps-16,duo,lm1b,16,seconds_per_sample,0.12841798868612386
+duo-lm1b-steps-32,duo,lm1b,32,conditional_generative_perplexity,76.69792739955646
+duo-lm1b-steps-32,duo,lm1b,32,mauve_suffix,0.37896998414643557
+duo-lm1b-steps-32,duo,lm1b,32,sample_entropy,3.800785349969878
+duo-lm1b-steps-32,duo,lm1b,32,reference_entropy,3.8171525992110147
+duo-lm1b-steps-32,duo,lm1b,32,sample_entropy_delta,-0.01636724924113686
+duo-lm1b-steps-32,duo,lm1b,32,self_bleu,0.16634681861084277
+duo-lm1b-steps-32,duo,lm1b,32,prefix_exact_match,1.0
+duo-lm1b-steps-32,duo,lm1b,32,seconds_per_sample,0.22576639815815724
+duo-lm1b-steps-1024,duo,lm1b,1024,conditional_generative_perplexity,72.0438298362223
+duo-lm1b-steps-1024,duo,lm1b,1024,mauve_suffix,0.415894589443493
+duo-lm1b-steps-1024,duo,lm1b,1024,sample_entropy,3.810042092869218
+duo-lm1b-steps-1024,duo,lm1b,1024,reference_entropy,3.8171525992110147
+duo-lm1b-steps-1024,duo,lm1b,1024,sample_entropy_delta,-0.007110506341796885
+duo-lm1b-steps-1024,duo,lm1b,1024,self_bleu,0.1648459488962082
+duo-lm1b-steps-1024,duo,lm1b,1024,prefix_exact_match,1.0
+duo-lm1b-steps-1024,duo,lm1b,1024,seconds_per_sample,7.377475616100128
+duo_dcd-lm1b-steps-1,duo_dcd,lm1b,1,conditional_generative_perplexity,972.0891391616341
+duo_dcd-lm1b-steps-1,duo_dcd,lm1b,1,mauve_suffix,0.005170955824428433
+duo_dcd-lm1b-steps-1,duo_dcd,lm1b,1,sample_entropy,3.766506824476198
+duo_dcd-lm1b-steps-1,duo_dcd,lm1b,1,reference_entropy,3.8171525992110147
+duo_dcd-lm1b-steps-1,duo_dcd,lm1b,1,sample_entropy_delta,-0.05064577473481657
+duo_dcd-lm1b-steps-1,duo_dcd,lm1b,1,self_bleu,0.10595979154131205
+duo_dcd-lm1b-steps-1,duo_dcd,lm1b,1,prefix_exact_match,1.0
+duo_dcd-lm1b-steps-1,duo_dcd,lm1b,1,seconds_per_sample,0.02398861467372626
+duo_dcd-lm1b-steps-2,duo_dcd,lm1b,2,conditional_generative_perplexity,384.35445941799287
+duo_dcd-lm1b-steps-2,duo_dcd,lm1b,2,mauve_suffix,0.011389631189173955
+duo_dcd-lm1b-steps-2,duo_dcd,lm1b,2,sample_entropy,3.72297918807914
+duo_dcd-lm1b-steps-2,duo_dcd,lm1b,2,reference_entropy,3.8171525992110147
+duo_dcd-lm1b-steps-2,duo_dcd,lm1b,2,sample_entropy_delta,-0.09417341113187483
+duo_dcd-lm1b-steps-2,duo_dcd,lm1b,2,self_bleu,0.13281386091909597
+duo_dcd-lm1b-steps-2,duo_dcd,lm1b,2,prefix_exact_match,1.0
+duo_dcd-lm1b-steps-2,duo_dcd,lm1b,2,seconds_per_sample,0.018906686367699876
+duo_dcd-lm1b-steps-4,duo_dcd,lm1b,4,conditional_generative_perplexity,209.32037280649857
+duo_dcd-lm1b-steps-4,duo_dcd,lm1b,4,mauve_suffix,0.03587051089962172
+duo_dcd-lm1b-steps-4,duo_dcd,lm1b,4,sample_entropy,3.7426513131748633
+duo_dcd-lm1b-steps-4,duo_dcd,lm1b,4,reference_entropy,3.8171525992110147
+duo_dcd-lm1b-steps-4,duo_dcd,lm1b,4,sample_entropy_delta,-0.07450128603615136
+duo_dcd-lm1b-steps-4,duo_dcd,lm1b,4,self_bleu,0.14268675432242428
+duo_dcd-lm1b-steps-4,duo_dcd,lm1b,4,prefix_exact_match,1.0
+duo_dcd-lm1b-steps-4,duo_dcd,lm1b,4,seconds_per_sample,0.04598313111637253
+duo_dcd-lm1b-steps-8,duo_dcd,lm1b,8,conditional_generative_perplexity,155.4102272275335
+duo_dcd-lm1b-steps-8,duo_dcd,lm1b,8,mauve_suffix,0.08763089060698975
+duo_dcd-lm1b-steps-8,duo_dcd,lm1b,8,sample_entropy,3.7630187834968547
+duo_dcd-lm1b-steps-8,duo_dcd,lm1b,8,reference_entropy,3.8171525992110147
+duo_dcd-lm1b-steps-8,duo_dcd,lm1b,8,sample_entropy_delta,-0.05413381571415998
+duo_dcd-lm1b-steps-8,duo_dcd,lm1b,8,self_bleu,0.14312298725073697
+duo_dcd-lm1b-steps-8,duo_dcd,lm1b,8,prefix_exact_match,1.0
+duo_dcd-lm1b-steps-8,duo_dcd,lm1b,8,seconds_per_sample,0.06934818447916768
+duo_dcd-lm1b-steps-16,duo_dcd,lm1b,16,conditional_generative_perplexity,134.93143203086112
+duo_dcd-lm1b-steps-16,duo_dcd,lm1b,16,mauve_suffix,0.1321953620227347
+duo_dcd-lm1b-steps-16,duo_dcd,lm1b,16,sample_entropy,3.782079306471349
+duo_dcd-lm1b-steps-16,duo_dcd,lm1b,16,reference_entropy,3.8171525992110147
+duo_dcd-lm1b-steps-16,duo_dcd,lm1b,16,sample_entropy_delta,-0.03507329273966553
+duo_dcd-lm1b-steps-16,duo_dcd,lm1b,16,self_bleu,0.14340835127347715
+duo_dcd-lm1b-steps-16,duo_dcd,lm1b,16,prefix_exact_match,1.0
+duo_dcd-lm1b-steps-16,duo_dcd,lm1b,16,seconds_per_sample,0.13762991552357562
+duo_dcd-lm1b-steps-32,duo_dcd,lm1b,32,conditional_generative_perplexity,125.81750589142253
+duo_dcd-lm1b-steps-32,duo_dcd,lm1b,32,mauve_suffix,0.16729614820201782
+duo_dcd-lm1b-steps-32,duo_dcd,lm1b,32,sample_entropy,3.788740378490953
+duo_dcd-lm1b-steps-32,duo_dcd,lm1b,32,reference_entropy,3.8171525992110147
+duo_dcd-lm1b-steps-32,duo_dcd,lm1b,32,sample_entropy_delta,-0.028412220720061665
+duo_dcd-lm1b-steps-32,duo_dcd,lm1b,32,self_bleu,0.14381145161400047
+duo_dcd-lm1b-steps-32,duo_dcd,lm1b,32,prefix_exact_match,1.0
+duo_dcd-lm1b-steps-32,duo_dcd,lm1b,32,seconds_per_sample,0.2393840821168851
+flm-lm1b-steps-1,flm,lm1b,1,conditional_generative_perplexity,1.9548877062739092
+flm-lm1b-steps-1,flm,lm1b,1,mauve_suffix,0.0040720962619612555
+flm-lm1b-steps-1,flm,lm1b,1,sample_entropy,0.8335246118012893
+flm-lm1b-steps-1,flm,lm1b,1,reference_entropy,3.8171525992110147
+flm-lm1b-steps-1,flm,lm1b,1,sample_entropy_delta,-2.9836279874097253
+flm-lm1b-steps-1,flm,lm1b,1,self_bleu,0.9254259123738018
+flm-lm1b-steps-1,flm,lm1b,1,prefix_exact_match,1.0
+flm-lm1b-steps-1,flm,lm1b,1,seconds_per_sample,0.007126594660803676
+flm-lm1b-steps-2,flm,lm1b,2,conditional_generative_perplexity,2.928663405751989
+flm-lm1b-steps-2,flm,lm1b,2,mauve_suffix,0.0040720962619612555
+flm-lm1b-steps-2,flm,lm1b,2,sample_entropy,1.2976684172732178
+flm-lm1b-steps-2,flm,lm1b,2,reference_entropy,3.8171525992110147
+flm-lm1b-steps-2,flm,lm1b,2,sample_entropy_delta,-2.5194841819377967
+flm-lm1b-steps-2,flm,lm1b,2,self_bleu,0.8571350413199361
+flm-lm1b-steps-2,flm,lm1b,2,prefix_exact_match,1.0
+flm-lm1b-steps-2,flm,lm1b,2,seconds_per_sample,0.013325479361810721
+flm-lm1b-steps-4,flm,lm1b,4,conditional_generative_perplexity,4.447919515570954
+flm-lm1b-steps-4,flm,lm1b,4,mauve_suffix,0.004072096261961256
+flm-lm1b-steps-4,flm,lm1b,4,sample_entropy,1.6413411603948338
+flm-lm1b-steps-4,flm,lm1b,4,reference_entropy,3.8171525992110147
+flm-lm1b-steps-4,flm,lm1b,4,sample_entropy_delta,-2.175811438816181
+flm-lm1b-steps-4,flm,lm1b,4,self_bleu,0.7991751544839317
+flm-lm1b-steps-4,flm,lm1b,4,prefix_exact_match,1.0
+flm-lm1b-steps-4,flm,lm1b,4,seconds_per_sample,0.028052487556124106
+flm-lm1b-steps-8,flm,lm1b,8,conditional_generative_perplexity,6.00491602998456
+flm-lm1b-steps-8,flm,lm1b,8,mauve_suffix,0.004179943930085514
+flm-lm1b-steps-8,flm,lm1b,8,sample_entropy,1.8097444920868058
+flm-lm1b-steps-8,flm,lm1b,8,reference_entropy,3.8171525992110147
+flm-lm1b-steps-8,flm,lm1b,8,sample_entropy_delta,-2.007408107124209
+flm-lm1b-steps-8,flm,lm1b,8,self_bleu,0.7593250173826942
+flm-lm1b-steps-8,flm,lm1b,8,prefix_exact_match,1.0
+flm-lm1b-steps-8,flm,lm1b,8,seconds_per_sample,0.04932963251485489
+flm-lm1b-steps-16,flm,lm1b,16,conditional_generative_perplexity,8.335617731774379
+flm-lm1b-steps-16,flm,lm1b,16,mauve_suffix,0.004103136176217619
+flm-lm1b-steps-16,flm,lm1b,16,sample_entropy,1.9668683860632878
+flm-lm1b-steps-16,flm,lm1b,16,reference_entropy,3.8171525992110147
+flm-lm1b-steps-16,flm,lm1b,16,sample_entropy_delta,-1.8502842131477268
+flm-lm1b-steps-16,flm,lm1b,16,self_bleu,0.7145817168578211
+flm-lm1b-steps-16,flm,lm1b,16,prefix_exact_match,1.0
+flm-lm1b-steps-16,flm,lm1b,16,seconds_per_sample,0.10677115371800028
+flm-lm1b-steps-32,flm,lm1b,32,conditional_generative_perplexity,12.092915980161044
+flm-lm1b-steps-32,flm,lm1b,32,mauve_suffix,0.0040720962619612555
+flm-lm1b-steps-32,flm,lm1b,32,sample_entropy,2.0806292940402793
+flm-lm1b-steps-32,flm,lm1b,32,reference_entropy,3.8171525992110147
+flm-lm1b-steps-32,flm,lm1b,32,sample_entropy_delta,-1.7365233051707354
+flm-lm1b-steps-32,flm,lm1b,32,self_bleu,0.639138127621003
+flm-lm1b-steps-32,flm,lm1b,32,prefix_exact_match,1.0
+flm-lm1b-steps-32,flm,lm1b,32,seconds_per_sample,0.21033303574949969
+flm-lm1b-steps-1024,flm,lm1b,1024,conditional_generative_perplexity,9.472139926294247
+flm-lm1b-steps-1024,flm,lm1b,1024,mauve_suffix,0.004274281639634798
+flm-lm1b-steps-1024,flm,lm1b,1024,sample_entropy,2.190467673326771
+flm-lm1b-steps-1024,flm,lm1b,1024,reference_entropy,3.8171525992110147
+flm-lm1b-steps-1024,flm,lm1b,1024,sample_entropy_delta,-1.6266849258842435
+flm-lm1b-steps-1024,flm,lm1b,1024,self_bleu,0.6605278236354867
+flm-lm1b-steps-1024,flm,lm1b,1024,prefix_exact_match,1.0
+flm-lm1b-steps-1024,flm,lm1b,1024,seconds_per_sample,7.670933374945889
+fmlm-lm1b-steps-1,fmlm,lm1b,1,conditional_generative_perplexity,162.1960481565661
+fmlm-lm1b-steps-1,fmlm,lm1b,1,mauve_suffix,0.006534227811277743
+fmlm-lm1b-steps-1,fmlm,lm1b,1,sample_entropy,3.7822500659339635
+fmlm-lm1b-steps-1,fmlm,lm1b,1,reference_entropy,3.8171525992110147
+fmlm-lm1b-steps-1,fmlm,lm1b,1,sample_entropy_delta,-0.03490253327705117
+fmlm-lm1b-steps-1,fmlm,lm1b,1,self_bleu,0.16909926500809713
+fmlm-lm1b-steps-1,fmlm,lm1b,1,prefix_exact_match,1.0
+fmlm-lm1b-steps-1,fmlm,lm1b,1,seconds_per_sample,0.005986617907183245
+fmlm-lm1b-steps-2,fmlm,lm1b,2,conditional_generative_perplexity,143.61810773625584
+fmlm-lm1b-steps-2,fmlm,lm1b,2,mauve_suffix,0.012325578721461068
+fmlm-lm1b-steps-2,fmlm,lm1b,2,sample_entropy,3.8188728002550287
+fmlm-lm1b-steps-2,fmlm,lm1b,2,reference_entropy,3.8171525992110147
+fmlm-lm1b-steps-2,fmlm,lm1b,2,sample_entropy_delta,0.001720201044014047
+fmlm-lm1b-steps-2,fmlm,lm1b,2,self_bleu,0.16136826366939322
+fmlm-lm1b-steps-2,fmlm,lm1b,2,prefix_exact_match,1.0
+fmlm-lm1b-steps-2,fmlm,lm1b,2,seconds_per_sample,0.015742484785732813
+fmlm-lm1b-steps-4,fmlm,lm1b,4,conditional_generative_perplexity,124.74229462167105
+fmlm-lm1b-steps-4,fmlm,lm1b,4,mauve_suffix,0.016640859917227913
+fmlm-lm1b-steps-4,fmlm,lm1b,4,sample_entropy,3.8223502962571443
+fmlm-lm1b-steps-4,fmlm,lm1b,4,reference_entropy,3.8171525992110147
+fmlm-lm1b-steps-4,fmlm,lm1b,4,sample_entropy_delta,0.005197697046129601
+fmlm-lm1b-steps-4,fmlm,lm1b,4,self_bleu,0.1633891153212988
+fmlm-lm1b-steps-4,fmlm,lm1b,4,prefix_exact_match,1.0
+fmlm-lm1b-steps-4,fmlm,lm1b,4,seconds_per_sample,0.026481049208086915
+fmlm-lm1b-steps-8,fmlm,lm1b,8,conditional_generative_perplexity,108.29128070279461
+fmlm-lm1b-steps-8,fmlm,lm1b,8,mauve_suffix,0.023481378724795915
+fmlm-lm1b-steps-8,fmlm,lm1b,8,sample_entropy,3.81743248327292
+fmlm-lm1b-steps-8,fmlm,lm1b,8,reference_entropy,3.8171525992110147
+fmlm-lm1b-steps-8,fmlm,lm1b,8,sample_entropy_delta,0.00027988406190537773
+fmlm-lm1b-steps-8,fmlm,lm1b,8,self_bleu,0.16226303438688486
+fmlm-lm1b-steps-8,fmlm,lm1b,8,prefix_exact_match,1.0
+fmlm-lm1b-steps-8,fmlm,lm1b,8,seconds_per_sample,0.052056993954465725
+fmlm-lm1b-steps-16,fmlm,lm1b,16,conditional_generative_perplexity,93.34730625374856
+fmlm-lm1b-steps-16,fmlm,lm1b,16,mauve_suffix,0.023395364878960975
+fmlm-lm1b-steps-16,fmlm,lm1b,16,sample_entropy,3.810194860246701
+fmlm-lm1b-steps-16,fmlm,lm1b,16,reference_entropy,3.8171525992110147
+fmlm-lm1b-steps-16,fmlm,lm1b,16,sample_entropy_delta,-0.00695773896431362
+fmlm-lm1b-steps-16,fmlm,lm1b,16,self_bleu,0.16688286813259537
+fmlm-lm1b-steps-16,fmlm,lm1b,16,prefix_exact_match,1.0
+fmlm-lm1b-steps-16,fmlm,lm1b,16,seconds_per_sample,0.10537018949980848
+fmlm-lm1b-steps-32,fmlm,lm1b,32,conditional_generative_perplexity,81.08235208356075
+fmlm-lm1b-steps-32,fmlm,lm1b,32,mauve_suffix,0.025989271366052474
+fmlm-lm1b-steps-32,fmlm,lm1b,32,sample_entropy,3.8008536082446445
+fmlm-lm1b-steps-32,fmlm,lm1b,32,reference_entropy,3.8171525992110147
+fmlm-lm1b-steps-32,fmlm,lm1b,32,sample_entropy_delta,-0.01629899096637022
+fmlm-lm1b-steps-32,fmlm,lm1b,32,self_bleu,0.17032629629472096
+fmlm-lm1b-steps-32,fmlm,lm1b,32,prefix_exact_match,1.0
+fmlm-lm1b-steps-32,fmlm,lm1b,32,seconds_per_sample,0.21115368003665935
+mdlm-lm1b-steps-1,mdlm,lm1b,1,conditional_generative_perplexity,1064.2631849936188
+mdlm-lm1b-steps-1,mdlm,lm1b,1,mauve_suffix,0.00483854366589726
+mdlm-lm1b-steps-1,mdlm,lm1b,1,sample_entropy,3.8323178554140154
+mdlm-lm1b-steps-1,mdlm,lm1b,1,reference_entropy,3.8171525992110147
+mdlm-lm1b-steps-1,mdlm,lm1b,1,sample_entropy_delta,0.015165256203000688
+mdlm-lm1b-steps-1,mdlm,lm1b,1,self_bleu,0.11177542653795658
+mdlm-lm1b-steps-1,mdlm,lm1b,1,prefix_exact_match,1.0
+mdlm-lm1b-steps-1,mdlm,lm1b,1,seconds_per_sample,0.019062782506807707
+mdlm-lm1b-steps-2,mdlm,lm1b,2,conditional_generative_perplexity,545.8042529875876
+mdlm-lm1b-steps-2,mdlm,lm1b,2,mauve_suffix,0.007774155419971392
+mdlm-lm1b-steps-2,mdlm,lm1b,2,sample_entropy,3.8237285049053167
+mdlm-lm1b-steps-2,mdlm,lm1b,2,reference_entropy,3.8171525992110147
+mdlm-lm1b-steps-2,mdlm,lm1b,2,sample_entropy_delta,0.006575905694301998
+mdlm-lm1b-steps-2,mdlm,lm1b,2,self_bleu,0.11723423591524713
+mdlm-lm1b-steps-2,mdlm,lm1b,2,prefix_exact_match,1.0
+mdlm-lm1b-steps-2,mdlm,lm1b,2,seconds_per_sample,0.018605535791721195
+mdlm-lm1b-steps-4,mdlm,lm1b,4,conditional_generative_perplexity,262.1196216839746
+mdlm-lm1b-steps-4,mdlm,lm1b,4,mauve_suffix,0.050296948084236015
+mdlm-lm1b-steps-4,mdlm,lm1b,4,sample_entropy,3.8223798230691886
+mdlm-lm1b-steps-4,mdlm,lm1b,4,reference_entropy,3.8171525992110147
+mdlm-lm1b-steps-4,mdlm,lm1b,4,sample_entropy_delta,0.005227223858173868
+mdlm-lm1b-steps-4,mdlm,lm1b,4,self_bleu,0.13613229516483286
+mdlm-lm1b-steps-4,mdlm,lm1b,4,prefix_exact_match,1.0
+mdlm-lm1b-steps-4,mdlm,lm1b,4,seconds_per_sample,0.03258685475157108
+mdlm-lm1b-steps-8,mdlm,lm1b,8,conditional_generative_perplexity,147.58164874422377
+mdlm-lm1b-steps-8,mdlm,lm1b,8,mauve_suffix,0.15113775719337685
+mdlm-lm1b-steps-8,mdlm,lm1b,8,sample_entropy,3.8204394581725745
+mdlm-lm1b-steps-8,mdlm,lm1b,8,reference_entropy,3.8171525992110147
+mdlm-lm1b-steps-8,mdlm,lm1b,8,sample_entropy_delta,0.00328685896155978
+mdlm-lm1b-steps-8,mdlm,lm1b,8,self_bleu,0.15094632587601328
+mdlm-lm1b-steps-8,mdlm,lm1b,8,prefix_exact_match,1.0
+mdlm-lm1b-steps-8,mdlm,lm1b,8,seconds_per_sample,0.061980252954526804
+mdlm-lm1b-steps-16,mdlm,lm1b,16,conditional_generative_perplexity,107.01657472710622
+mdlm-lm1b-steps-16,mdlm,lm1b,16,mauve_suffix,0.22652224603977716
+mdlm-lm1b-steps-16,mdlm,lm1b,16,sample_entropy,3.819660797312084
+mdlm-lm1b-steps-16,mdlm,lm1b,16,reference_entropy,3.8171525992110147
+mdlm-lm1b-steps-16,mdlm,lm1b,16,sample_entropy_delta,0.002508198101069503
+mdlm-lm1b-steps-16,mdlm,lm1b,16,self_bleu,0.15697349769846364
+mdlm-lm1b-steps-16,mdlm,lm1b,16,prefix_exact_match,1.0
+mdlm-lm1b-steps-16,mdlm,lm1b,16,seconds_per_sample,0.12651682189607527
+mdlm-lm1b-steps-32,mdlm,lm1b,32,conditional_generative_perplexity,87.9809105575203
+mdlm-lm1b-steps-32,mdlm,lm1b,32,mauve_suffix,0.3395893682806676
+mdlm-lm1b-steps-32,mdlm,lm1b,32,sample_entropy,3.819441376044174
+mdlm-lm1b-steps-32,mdlm,lm1b,32,reference_entropy,3.8171525992110147
+mdlm-lm1b-steps-32,mdlm,lm1b,32,sample_entropy_delta,0.0022887768331592007
+mdlm-lm1b-steps-32,mdlm,lm1b,32,self_bleu,0.16143168286861326
+mdlm-lm1b-steps-32,mdlm,lm1b,32,prefix_exact_match,1.0
+mdlm-lm1b-steps-32,mdlm,lm1b,32,seconds_per_sample,0.2489504896075232
+mdlm-lm1b-steps-1024,mdlm,lm1b,1024,conditional_generative_perplexity,58.6389090098016
+mdlm-lm1b-steps-1024,mdlm,lm1b,1024,mauve_suffix,0.3842042386037965
+mdlm-lm1b-steps-1024,mdlm,lm1b,1024,sample_entropy,3.801667650020746
+mdlm-lm1b-steps-1024,mdlm,lm1b,1024,reference_entropy,3.8171525992110147
+mdlm-lm1b-steps-1024,mdlm,lm1b,1024,sample_entropy_delta,-0.015484949190268882
+mdlm-lm1b-steps-1024,mdlm,lm1b,1024,self_bleu,0.1680921892163879
+mdlm-lm1b-steps-1024,mdlm,lm1b,1024,prefix_exact_match,1.0
+mdlm-lm1b-steps-1024,mdlm,lm1b,1024,seconds_per_sample,7.402176677293028
+mdlm_di4c-owt-steps-1,mdlm_di4c,owt,1,conditional_generative_perplexity,2791.4371557642853
+mdlm_di4c-owt-steps-1,mdlm_di4c,owt,1,mauve_suffix,0.008204959866683138
+mdlm_di4c-owt-steps-1,mdlm_di4c,owt,1,sample_entropy,3.7424604010263938
+mdlm_di4c-owt-steps-1,mdlm_di4c,owt,1,reference_entropy,3.8090485583459235
+mdlm_di4c-owt-steps-1,mdlm_di4c,owt,1,sample_entropy_delta,-0.06658815731952972
+mdlm_di4c-owt-steps-1,mdlm_di4c,owt,1,self_bleu,0.017216425389052357
+mdlm_di4c-owt-steps-1,mdlm_di4c,owt,1,prefix_exact_match,1.0
+mdlm_di4c-owt-steps-1,mdlm_di4c,owt,1,seconds_per_sample,0.009990812352043577
+mdlm_di4c-owt-steps-2,mdlm_di4c,owt,2,conditional_generative_perplexity,944.5300897856437
+mdlm_di4c-owt-steps-2,mdlm_di4c,owt,2,mauve_suffix,0.02081879824149785
+mdlm_di4c-owt-steps-2,mdlm_di4c,owt,2,sample_entropy,3.750602440233099
+mdlm_di4c-owt-steps-2,mdlm_di4c,owt,2,reference_entropy,3.8090485583459235
+mdlm_di4c-owt-steps-2,mdlm_di4c,owt,2,sample_entropy_delta,-0.05844611811282441
+mdlm_di4c-owt-steps-2,mdlm_di4c,owt,2,self_bleu,0.018530838581378082
+mdlm_di4c-owt-steps-2,mdlm_di4c,owt,2,prefix_exact_match,1.0
+mdlm_di4c-owt-steps-2,mdlm_di4c,owt,2,seconds_per_sample,0.02093727396277245
+mdlm_di4c-owt-steps-4,mdlm_di4c,owt,4,conditional_generative_perplexity,226.33609985099912
+mdlm_di4c-owt-steps-4,mdlm_di4c,owt,4,mauve_suffix,0.271091930870634
+mdlm_di4c-owt-steps-4,mdlm_di4c,owt,4,sample_entropy,3.7514206599349937
+mdlm_di4c-owt-steps-4,mdlm_di4c,owt,4,reference_entropy,3.8090485583459235
+mdlm_di4c-owt-steps-4,mdlm_di4c,owt,4,sample_entropy_delta,-0.057627898410929745
+mdlm_di4c-owt-steps-4,mdlm_di4c,owt,4,self_bleu,0.02569586406877412
+mdlm_di4c-owt-steps-4,mdlm_di4c,owt,4,prefix_exact_match,1.0
+mdlm_di4c-owt-steps-4,mdlm_di4c,owt,4,seconds_per_sample,0.03869468237098772
+mdlm_di4c-owt-steps-8,mdlm_di4c,owt,8,conditional_generative_perplexity,79.87288537474217
+mdlm_di4c-owt-steps-8,mdlm_di4c,owt,8,mauve_suffix,0.446706965617256
+mdlm_di4c-owt-steps-8,mdlm_di4c,owt,8,sample_entropy,3.749191915840413
+mdlm_di4c-owt-steps-8,mdlm_di4c,owt,8,reference_entropy,3.8090485583459235
+mdlm_di4c-owt-steps-8,mdlm_di4c,owt,8,sample_entropy_delta,-0.05985664250551048
+mdlm_di4c-owt-steps-8,mdlm_di4c,owt,8,self_bleu,0.04232682549524337
+mdlm_di4c-owt-steps-8,mdlm_di4c,owt,8,prefix_exact_match,1.0
+mdlm_di4c-owt-steps-8,mdlm_di4c,owt,8,seconds_per_sample,0.08375158975832164
+mdlm_di4c-owt-steps-16,mdlm_di4c,owt,16,conditional_generative_perplexity,47.07727671992287
+mdlm_di4c-owt-steps-16,mdlm_di4c,owt,16,mauve_suffix,0.5557158118486929
+mdlm_di4c-owt-steps-16,mdlm_di4c,owt,16,sample_entropy,3.737626109038477
+mdlm_di4c-owt-steps-16,mdlm_di4c,owt,16,reference_entropy,3.8090485583459235
+mdlm_di4c-owt-steps-16,mdlm_di4c,owt,16,sample_entropy_delta,-0.07142244930744646
+mdlm_di4c-owt-steps-16,mdlm_di4c,owt,16,self_bleu,0.053125032999678364
+mdlm_di4c-owt-steps-16,mdlm_di4c,owt,16,prefix_exact_match,1.0
+mdlm_di4c-owt-steps-16,mdlm_di4c,owt,16,seconds_per_sample,0.17833757211337797
+mdlm_di4c-owt-steps-32,mdlm_di4c,owt,32,conditional_generative_perplexity,36.06460066053778
+mdlm_di4c-owt-steps-32,mdlm_di4c,owt,32,mauve_suffix,0.6089450357777276
+mdlm_di4c-owt-steps-32,mdlm_di4c,owt,32,sample_entropy,3.7288697853124337
+mdlm_di4c-owt-steps-32,mdlm_di4c,owt,32,reference_entropy,3.8090485583459235
+mdlm_di4c-owt-steps-32,mdlm_di4c,owt,32,sample_entropy_delta,-0.08017877303348975
+mdlm_di4c-owt-steps-32,mdlm_di4c,owt,32,self_bleu,0.059542885163979035
+mdlm_di4c-owt-steps-32,mdlm_di4c,owt,32,prefix_exact_match,1.0
+mdlm_di4c-owt-steps-32,mdlm_di4c,owt,32,seconds_per_sample,0.3458745952375466
+mdlm_sdtt-lm1b-steps-1,mdlm_sdtt,lm1b,1,conditional_generative_perplexity,912.046764668509
+mdlm_sdtt-lm1b-steps-1,mdlm_sdtt,lm1b,1,mauve_suffix,0.005058213509265277
+mdlm_sdtt-lm1b-steps-1,mdlm_sdtt,lm1b,1,sample_entropy,3.748283703983959
+mdlm_sdtt-lm1b-steps-1,mdlm_sdtt,lm1b,1,reference_entropy,3.8171525992110147
+mdlm_sdtt-lm1b-steps-1,mdlm_sdtt,lm1b,1,sample_entropy_delta,-0.06886889522705575
+mdlm_sdtt-lm1b-steps-1,mdlm_sdtt,lm1b,1,self_bleu,0.10331830201607839
+mdlm_sdtt-lm1b-steps-1,mdlm_sdtt,lm1b,1,prefix_exact_match,1.0
+mdlm_sdtt-lm1b-steps-1,mdlm_sdtt,lm1b,1,seconds_per_sample,0.010952344644465484
+mdlm_sdtt-lm1b-steps-2,mdlm_sdtt,lm1b,2,conditional_generative_perplexity,333.40456056383493
+mdlm_sdtt-lm1b-steps-2,mdlm_sdtt,lm1b,2,mauve_suffix,0.01318880671815744
+mdlm_sdtt-lm1b-steps-2,mdlm_sdtt,lm1b,2,sample_entropy,3.738391732910667
+mdlm_sdtt-lm1b-steps-2,mdlm_sdtt,lm1b,2,reference_entropy,3.8171525992110147
+mdlm_sdtt-lm1b-steps-2,mdlm_sdtt,lm1b,2,sample_entropy_delta,-0.07876086630034784
+mdlm_sdtt-lm1b-steps-2,mdlm_sdtt,lm1b,2,self_bleu,0.1368546232901211
+mdlm_sdtt-lm1b-steps-2,mdlm_sdtt,lm1b,2,prefix_exact_match,1.0
+mdlm_sdtt-lm1b-steps-2,mdlm_sdtt,lm1b,2,seconds_per_sample,0.017312765296082944
+mdlm_sdtt-lm1b-steps-4,mdlm_sdtt,lm1b,4,conditional_generative_perplexity,139.73297897434944
+mdlm_sdtt-lm1b-steps-4,mdlm_sdtt,lm1b,4,mauve_suffix,0.09093492679475954
+mdlm_sdtt-lm1b-steps-4,mdlm_sdtt,lm1b,4,sample_entropy,3.7415236711251048
+mdlm_sdtt-lm1b-steps-4,mdlm_sdtt,lm1b,4,reference_entropy,3.8171525992110147
+mdlm_sdtt-lm1b-steps-4,mdlm_sdtt,lm1b,4,sample_entropy_delta,-0.07562892808590993
+mdlm_sdtt-lm1b-steps-4,mdlm_sdtt,lm1b,4,self_bleu,0.16643315236358716
+mdlm_sdtt-lm1b-steps-4,mdlm_sdtt,lm1b,4,prefix_exact_match,1.0
+mdlm_sdtt-lm1b-steps-4,mdlm_sdtt,lm1b,4,seconds_per_sample,0.03484965949610341
+mdlm_sdtt-lm1b-steps-8,mdlm_sdtt,lm1b,8,conditional_generative_perplexity,82.75605033216789
+mdlm_sdtt-lm1b-steps-8,mdlm_sdtt,lm1b,8,mauve_suffix,0.21942479998092493
+mdlm_sdtt-lm1b-steps-8,mdlm_sdtt,lm1b,8,sample_entropy,3.7548982290978086
+mdlm_sdtt-lm1b-steps-8,mdlm_sdtt,lm1b,8,reference_entropy,3.8171525992110147
+mdlm_sdtt-lm1b-steps-8,mdlm_sdtt,lm1b,8,sample_entropy_delta,-0.062254370113206114
+mdlm_sdtt-lm1b-steps-8,mdlm_sdtt,lm1b,8,self_bleu,0.17550843768016466
+mdlm_sdtt-lm1b-steps-8,mdlm_sdtt,lm1b,8,prefix_exact_match,1.0
+mdlm_sdtt-lm1b-steps-8,mdlm_sdtt,lm1b,8,seconds_per_sample,0.07436269868048839
+mdlm_sdtt-lm1b-steps-16,mdlm_sdtt,lm1b,16,conditional_generative_perplexity,62.240678306789874
+mdlm_sdtt-lm1b-steps-16,mdlm_sdtt,lm1b,16,mauve_suffix,0.32770595251175666
+mdlm_sdtt-lm1b-steps-16,mdlm_sdtt,lm1b,16,sample_entropy,3.759916998261912
+mdlm_sdtt-lm1b-steps-16,mdlm_sdtt,lm1b,16,reference_entropy,3.8171525992110147
+mdlm_sdtt-lm1b-steps-16,mdlm_sdtt,lm1b,16,sample_entropy_delta,-0.05723560094910285
+mdlm_sdtt-lm1b-steps-16,mdlm_sdtt,lm1b,16,self_bleu,0.18156787232012347
+mdlm_sdtt-lm1b-steps-16,mdlm_sdtt,lm1b,16,prefix_exact_match,1.0
+mdlm_sdtt-lm1b-steps-16,mdlm_sdtt,lm1b,16,seconds_per_sample,0.14275758047006093
+mdlm_sdtt-lm1b-steps-32,mdlm_sdtt,lm1b,32,conditional_generative_perplexity,53.03848223781679
+mdlm_sdtt-lm1b-steps-32,mdlm_sdtt,lm1b,32,mauve_suffix,0.3901764342204239
+mdlm_sdtt-lm1b-steps-32,mdlm_sdtt,lm1b,32,sample_entropy,3.7595116431018822
+mdlm_sdtt-lm1b-steps-32,mdlm_sdtt,lm1b,32,reference_entropy,3.8171525992110147
+mdlm_sdtt-lm1b-steps-32,mdlm_sdtt,lm1b,32,sample_entropy_delta,-0.057640956109132446
+mdlm_sdtt-lm1b-steps-32,mdlm_sdtt,lm1b,32,self_bleu,0.17993570679965792
+mdlm_sdtt-lm1b-steps-32,mdlm_sdtt,lm1b,32,prefix_exact_match,1.0
+mdlm_sdtt-lm1b-steps-32,mdlm_sdtt,lm1b,32,seconds_per_sample,0.23791447999246884
+mdlm_sdtt-owt-steps-1,mdlm_sdtt,owt,1,conditional_generative_perplexity,2770.0619448869115
+mdlm_sdtt-owt-steps-1,mdlm_sdtt,owt,1,mauve_suffix,0.0073559087882152305
+mdlm_sdtt-owt-steps-1,mdlm_sdtt,owt,1,sample_entropy,3.737069846035564
+mdlm_sdtt-owt-steps-1,mdlm_sdtt,owt,1,reference_entropy,3.8090485583459235
+mdlm_sdtt-owt-steps-1,mdlm_sdtt,owt,1,sample_entropy_delta,-0.07197871231035968
+mdlm_sdtt-owt-steps-1,mdlm_sdtt,owt,1,self_bleu,0.01637466504973282
+mdlm_sdtt-owt-steps-1,mdlm_sdtt,owt,1,prefix_exact_match,1.0
+mdlm_sdtt-owt-steps-1,mdlm_sdtt,owt,1,seconds_per_sample,0.008862308925017715
+mdlm_sdtt-owt-steps-2,mdlm_sdtt,owt,2,conditional_generative_perplexity,1077.2700142678389
+mdlm_sdtt-owt-steps-2,mdlm_sdtt,owt,2,mauve_suffix,0.013707248537549552
+mdlm_sdtt-owt-steps-2,mdlm_sdtt,owt,2,sample_entropy,3.745571857497006
+mdlm_sdtt-owt-steps-2,mdlm_sdtt,owt,2,reference_entropy,3.8090485583459235
+mdlm_sdtt-owt-steps-2,mdlm_sdtt,owt,2,sample_entropy_delta,-0.06347670084891766
+mdlm_sdtt-owt-steps-2,mdlm_sdtt,owt,2,self_bleu,0.016799458482196977
+mdlm_sdtt-owt-steps-2,mdlm_sdtt,owt,2,prefix_exact_match,1.0
+mdlm_sdtt-owt-steps-2,mdlm_sdtt,owt,2,seconds_per_sample,0.018550986525951885
+mdlm_sdtt-owt-steps-4,mdlm_sdtt,owt,4,conditional_generative_perplexity,298.5037909872003
+mdlm_sdtt-owt-steps-4,mdlm_sdtt,owt,4,mauve_suffix,0.1671905044000403
+mdlm_sdtt-owt-steps-4,mdlm_sdtt,owt,4,sample_entropy,3.748584941150153
+mdlm_sdtt-owt-steps-4,mdlm_sdtt,owt,4,reference_entropy,3.8090485583459235
+mdlm_sdtt-owt-steps-4,mdlm_sdtt,owt,4,sample_entropy_delta,-0.060463617195770425
+mdlm_sdtt-owt-steps-4,mdlm_sdtt,owt,4,self_bleu,0.02383661208349576
+mdlm_sdtt-owt-steps-4,mdlm_sdtt,owt,4,prefix_exact_match,1.0
+mdlm_sdtt-owt-steps-4,mdlm_sdtt,owt,4,seconds_per_sample,0.0400318524625618
+mdlm_sdtt-owt-steps-8,mdlm_sdtt,owt,8,conditional_generative_perplexity,102.52068892439462
+mdlm_sdtt-owt-steps-8,mdlm_sdtt,owt,8,mauve_suffix,0.4434890512136005
+mdlm_sdtt-owt-steps-8,mdlm_sdtt,owt,8,sample_entropy,3.752913234833009
+mdlm_sdtt-owt-steps-8,mdlm_sdtt,owt,8,reference_entropy,3.8090485583459235
+mdlm_sdtt-owt-steps-8,mdlm_sdtt,owt,8,sample_entropy_delta,-0.0561353235129145
+mdlm_sdtt-owt-steps-8,mdlm_sdtt,owt,8,self_bleu,0.03526571529695085
+mdlm_sdtt-owt-steps-8,mdlm_sdtt,owt,8,prefix_exact_match,1.0
+mdlm_sdtt-owt-steps-8,mdlm_sdtt,owt,8,seconds_per_sample,0.07965370659076143
+mdlm_sdtt-owt-steps-16,mdlm_sdtt,owt,16,conditional_generative_perplexity,59.174016716989655
+mdlm_sdtt-owt-steps-16,mdlm_sdtt,owt,16,mauve_suffix,0.5158102383713354
+mdlm_sdtt-owt-steps-16,mdlm_sdtt,owt,16,sample_entropy,3.7507356855181246
+mdlm_sdtt-owt-steps-16,mdlm_sdtt,owt,16,reference_entropy,3.8090485583459235
+mdlm_sdtt-owt-steps-16,mdlm_sdtt,owt,16,sample_entropy_delta,-0.058312872827798845
+mdlm_sdtt-owt-steps-16,mdlm_sdtt,owt,16,self_bleu,0.04471545915510127
+mdlm_sdtt-owt-steps-16,mdlm_sdtt,owt,16,prefix_exact_match,1.0
+mdlm_sdtt-owt-steps-16,mdlm_sdtt,owt,16,seconds_per_sample,0.16996464456315152
+mdlm_sdtt-owt-steps-32,mdlm_sdtt,owt,32,conditional_generative_perplexity,43.943780683780034
+mdlm_sdtt-owt-steps-32,mdlm_sdtt,owt,32,mauve_suffix,0.5207367362678341
+mdlm_sdtt-owt-steps-32,mdlm_sdtt,owt,32,sample_entropy,3.737161209473734
+mdlm_sdtt-owt-steps-32,mdlm_sdtt,owt,32,reference_entropy,3.8090485583459235
+mdlm_sdtt-owt-steps-32,mdlm_sdtt,owt,32,sample_entropy_delta,-0.07188734887218962
+mdlm_sdtt-owt-steps-32,mdlm_sdtt,owt,32,self_bleu,0.05084042935920519
+mdlm_sdtt-owt-steps-32,mdlm_sdtt,owt,32,prefix_exact_match,1.0
+mdlm_sdtt-owt-steps-32,mdlm_sdtt,owt,32,seconds_per_sample,0.3209660451102536

results/conditional/timing/lm1b/candi/steps_1/timing.json

@@ -0,0 +1 @@
+{"metadata":{"adapter_identity":"dlb.adapters.candi:v1|dlb.adapters.candi:CANDIAdapter|e63560f999a54ea08143cab77049eaa8bbe2965321470716a2fe3a94f2b7ed19","attempt_id":"1e36cbac338b4d0e9d8c88c7c4b97b0c","checkpoint_lock_id":"flm_lm1b_reproductions:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152:lm1b_CANDI.ckpt","checkpoint_selection":{"path":"lm1b_CANDI.ckpt","resource":"flm_lm1b_reproductions","teacher_family":"hybrid_candi"},"checkpoint_sha256":"2b301a6875a6ba9efa52a2a66773a5bcc49d5397f7c861a94495252e647df185","checkpoint_teacher_family":"hybrid_candi","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","cuda_runtime_version":"12.4","dataset":"lm1b","diversity_prompt_count":256,"environment":"dlb-candi","evaluation_continuation_length":64,"generation_mode":"conditional_prefix","gpu_compute_capability":"9.0","gpu_index":0,"gpu_name":"NVIDIA H200 NVL","model":"candi","nvidia_driver_version":"575.57.08","parameter_precision":"fp32","precision":"bf16-mixed","precision_evidence":"static_policy_bound_to_checkpoint_and_runtime_code_not_runtime_autocast_observation","precision_policy":"candi:pinned_internal_bf16_autocast_with_fp32_sensitive_ops","precision_policy_binding":{"adapter_identity":"dlb.adapters.candi:v1|dlb.adapters.candi:CANDIAdapter|e63560f999a54ea08143cab77049eaa8bbe2965321470716a2fe3a94f2b7ed19","checkpoint_lock_id":"flm_lm1b_reproductions:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152:lm1b_CANDI.ckpt","checkpoint_repository":null,"checkpoint_resource":"flm_lm1b_reproductions","checkpoint_revision":null,"checkpoint_selection":{"path":"lm1b_CANDI.ckpt","resource":"flm_lm1b_reproductions","teacher_family":"hybrid_candi"},"checkpoint_sha256":"2b301a6875a6ba9efa52a2a66773a5bcc49d5397f7c861a94495252e647df185","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","dataset":"lm1b","diversity_prompt_count":256,"evaluation_continuation_length":64,"experiment_config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","generation_mode":"conditional_prefix","model":"candi","prefix_length":64,"prompt_count":1024,"source_commit":"cd57ae9eec98d6ac71cd52bdc50eeec8dfd70f91","upstream":"candi"},"prefix_length":64,"prompt_count":1024,"pytorch_compiled_cuda_toolkit":"12.4","requested_precision":"author","runtime_code_binding":[{"kind":"sampling_callable","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/capture.py","sha256":"bb7d05734503f57d02187a8a94d91394b5a3b394b45a6c3556ed51e612591713","symbol":"_capture_teacher.<locals>.capture.<locals>.<lambda>"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/candi/algo.py","sha256":"131ffaeb45418710294a70e1258afdaaa213e8a51bd19ef6227eda8d7fd036e9","symbol":"CANDI"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/candi/models/dit_cont.py","sha256":"3ee29045168f0ea4c0faecc35c053cdd2b872b7ac11eb16def52c6dae084552a","symbol":"ContDIT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/candi/models/dit_cont.py","sha256":"3ee29045168f0ea4c0faecc35c053cdd2b872b7ac11eb16def52c6dae084552a","symbol":"DDiTBlock"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/candi/models/dit_cont.py","sha256":"3ee29045168f0ea4c0faecc35c053cdd2b872b7ac11eb16def52c6dae084552a","symbol":"DDiTFinalLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/candi/models/dit_cont.py","sha256":"3ee29045168f0ea4c0faecc35c053cdd2b872b7ac11eb16def52c6dae084552a","symbol":"EmbeddingLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/candi/models/dit_cont.py","sha256":"3ee29045168f0ea4c0faecc35c053cdd2b872b7ac11eb16def52c6dae084552a","symbol":"LayerNorm"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/candi/models/dit_cont.py","sha256":"3ee29045168f0ea4c0faecc35c053cdd2b872b7ac11eb16def52c6dae084552a","symbol":"Rotary"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/candi/models/dit_cont.py","sha256":"3ee29045168f0ea4c0faecc35c053cdd2b872b7ac11eb16def52c6dae084552a","symbol":"TimestepEmbedder"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/candi/trainer_base.py","sha256":"953ab0e52957472772150b32a342fe826dcc1a57ad3a4c48654de4985f0c18a4","symbol":"LogLinear"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-candi/lib/python3.11/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"GELU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-candi/lib/python3.11/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"SiLU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-candi/lib/python3.11/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"Softplus"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-candi/lib/python3.11/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"ModuleList"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-candi/lib/python3.11/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"Sequential"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-candi/lib/python3.11/site-packages/torch/nn/modules/dropout.py","sha256":"1fc9974c3cf2078dd4bc257f73afb6a6714adfec2e9d099d47cf923e766d3490","symbol":"Dropout"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-candi/lib/python3.11/site-packages/torch/nn/modules/linear.py","sha256":"d7f5cadf773deb5ab4bcc9f57414e8239b2859c70d2aa6f141576de1ec918823","symbol":"Linear"}],"sample_count":2048,"seed":42,"source_commit":"cd57ae9eec98d6ac71cd52bdc50eeec8dfd70f91","steps":1,"synchronization_policy":"torch.cuda.synchronize_before_start_and_after_generate"},"schema":"dlb-generation-timing-v1","timing":{"batch_size":1,"exclusions":["model_and_checkpoint_loading","first_compilation","token_decoding","metrics","file_io"],"median_seconds":0.00752258999273181,"mode":"primary_latency","num_timed_samples":32,"raw_durations_seconds":[0.007759853731840849,0.008103722240775824,0.007878145668655634,0.007913640234619379,0.007955094333738089,0.00796306598931551,0.007905327714979649,0.00789955910295248,0.007996948901563883,0.007830223068594933,0.007754056248813868,0.00765029713511467,0.007571546360850334,0.007522420026361942,0.007444891147315502,0.007402245886623859,0.007472493685781956,0.007584906183183193,0.007522759959101677,0.0075397370383143425,0.0073171048425138,0.007275390904396772,0.007241099141538143,0.007262291386723518,0.007286819163709879,0.007242861203849316,0.007243432104587555,0.007273267954587936,0.007271504960954189,0.007263233419507742,0.0072575840167701244,0.007269741967320442],"repeats":32,"seconds_per_sample":0.007558602053904906,"standard_deviation_convention":"population","standard_deviation_seconds":0.0002800327204288113,"warmups":5}}

results/conditional/timing/lm1b/candi/steps_32/timing.json

@@ -0,0 +1 @@
+{"metadata":{"adapter_identity":"dlb.adapters.candi:v1|dlb.adapters.candi:CANDIAdapter|e63560f999a54ea08143cab77049eaa8bbe2965321470716a2fe3a94f2b7ed19","attempt_id":"e8a901176bee4d118c73b9fe63510ac0","checkpoint_lock_id":"flm_lm1b_reproductions:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152:lm1b_CANDI.ckpt","checkpoint_selection":{"path":"lm1b_CANDI.ckpt","resource":"flm_lm1b_reproductions","teacher_family":"hybrid_candi"},"checkpoint_sha256":"2b301a6875a6ba9efa52a2a66773a5bcc49d5397f7c861a94495252e647df185","checkpoint_teacher_family":"hybrid_candi","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","cuda_runtime_version":"12.4","dataset":"lm1b","diversity_prompt_count":256,"environment":"dlb-candi","evaluation_continuation_length":64,"generation_mode":"conditional_prefix","gpu_compute_capability":"9.0","gpu_index":0,"gpu_name":"NVIDIA H200 NVL","model":"candi","nvidia_driver_version":"575.57.08","parameter_precision":"fp32","precision":"bf16-mixed","precision_evidence":"static_policy_bound_to_checkpoint_and_runtime_code_not_runtime_autocast_observation","precision_policy":"candi:pinned_internal_bf16_autocast_with_fp32_sensitive_ops","precision_policy_binding":{"adapter_identity":"dlb.adapters.candi:v1|dlb.adapters.candi:CANDIAdapter|e63560f999a54ea08143cab77049eaa8bbe2965321470716a2fe3a94f2b7ed19","checkpoint_lock_id":"flm_lm1b_reproductions:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152:lm1b_CANDI.ckpt","checkpoint_repository":null,"checkpoint_resource":"flm_lm1b_reproductions","checkpoint_revision":null,"checkpoint_selection":{"path":"lm1b_CANDI.ckpt","resource":"flm_lm1b_reproductions","teacher_family":"hybrid_candi"},"checkpoint_sha256":"2b301a6875a6ba9efa52a2a66773a5bcc49d5397f7c861a94495252e647df185","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","dataset":"lm1b","diversity_prompt_count":256,"evaluation_continuation_length":64,"experiment_config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","generation_mode":"conditional_prefix","model":"candi","prefix_length":64,"prompt_count":1024,"source_commit":"cd57ae9eec98d6ac71cd52bdc50eeec8dfd70f91","upstream":"candi"},"prefix_length":64,"prompt_count":1024,"pytorch_compiled_cuda_toolkit":"12.4","requested_precision":"author","runtime_code_binding":[{"kind":"sampling_callable","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/capture.py","sha256":"bb7d05734503f57d02187a8a94d91394b5a3b394b45a6c3556ed51e612591713","symbol":"_capture_teacher.<locals>.capture.<locals>.<lambda>"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/candi/algo.py","sha256":"131ffaeb45418710294a70e1258afdaaa213e8a51bd19ef6227eda8d7fd036e9","symbol":"CANDI"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/candi/models/dit_cont.py","sha256":"3ee29045168f0ea4c0faecc35c053cdd2b872b7ac11eb16def52c6dae084552a","symbol":"ContDIT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/candi/models/dit_cont.py","sha256":"3ee29045168f0ea4c0faecc35c053cdd2b872b7ac11eb16def52c6dae084552a","symbol":"DDiTBlock"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/candi/models/dit_cont.py","sha256":"3ee29045168f0ea4c0faecc35c053cdd2b872b7ac11eb16def52c6dae084552a","symbol":"DDiTFinalLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/candi/models/dit_cont.py","sha256":"3ee29045168f0ea4c0faecc35c053cdd2b872b7ac11eb16def52c6dae084552a","symbol":"EmbeddingLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/candi/models/dit_cont.py","sha256":"3ee29045168f0ea4c0faecc35c053cdd2b872b7ac11eb16def52c6dae084552a","symbol":"LayerNorm"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/candi/models/dit_cont.py","sha256":"3ee29045168f0ea4c0faecc35c053cdd2b872b7ac11eb16def52c6dae084552a","symbol":"Rotary"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/candi/models/dit_cont.py","sha256":"3ee29045168f0ea4c0faecc35c053cdd2b872b7ac11eb16def52c6dae084552a","symbol":"TimestepEmbedder"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/candi/trainer_base.py","sha256":"953ab0e52957472772150b32a342fe826dcc1a57ad3a4c48654de4985f0c18a4","symbol":"LogLinear"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-candi/lib/python3.11/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"GELU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-candi/lib/python3.11/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"SiLU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-candi/lib/python3.11/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"Softplus"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-candi/lib/python3.11/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"ModuleList"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-candi/lib/python3.11/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"Sequential"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-candi/lib/python3.11/site-packages/torch/nn/modules/dropout.py","sha256":"1fc9974c3cf2078dd4bc257f73afb6a6714adfec2e9d099d47cf923e766d3490","symbol":"Dropout"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-candi/lib/python3.11/site-packages/torch/nn/modules/linear.py","sha256":"d7f5cadf773deb5ab4bcc9f57414e8239b2859c70d2aa6f141576de1ec918823","symbol":"Linear"}],"sample_count":2048,"seed":42,"source_commit":"cd57ae9eec98d6ac71cd52bdc50eeec8dfd70f91","steps":32,"synchronization_policy":"torch.cuda.synchronize_before_start_and_after_generate"},"schema":"dlb-generation-timing-v1","timing":{"batch_size":1,"exclusions":["model_and_checkpoint_loading","first_compilation","token_decoding","metrics","file_io"],"median_seconds":0.20533425151370466,"mode":"primary_latency","num_timed_samples":32,"raw_durations_seconds":[0.2046427447348833,0.20466438820585608,0.20506100729107857,0.20522249536588788,0.20509565016254783,0.20502952905371785,0.20526133477687836,0.20508236018940806,0.2088904120028019,0.2058593207038939,0.20582817401736975,0.20555429439991713,0.20551946014165878,0.20463323080912232,0.20827291393652558,0.24337409297004342,0.22553377505391836,0.20509061310440302,0.2055131709203124,0.20510776760056615,0.20896028773859143,0.20560870692133904,0.20581622608006,0.20540716825053096,0.20498577086254954,0.20511473016813397,0.20801089191809297,0.20569194527342916,0.20478689623996615,0.20498869568109512,0.20774794928729534,0.20511452900245786],"repeats":32,"seconds_per_sample":0.20754595415201038,"standard_deviation_convention":"population","standard_deviation_seconds":0.007394012668292917,"warmups":5}}

results/conditional/timing/lm1b/candi/steps_8/timing.json

@@ -0,0 +1 @@
+{"metadata":{"adapter_identity":"dlb.adapters.candi:v1|dlb.adapters.candi:CANDIAdapter|e63560f999a54ea08143cab77049eaa8bbe2965321470716a2fe3a94f2b7ed19","attempt_id":"f1dbef8c86b247928f432c4ae63b194e","checkpoint_lock_id":"flm_lm1b_reproductions:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152:lm1b_CANDI.ckpt","checkpoint_selection":{"path":"lm1b_CANDI.ckpt","resource":"flm_lm1b_reproductions","teacher_family":"hybrid_candi"},"checkpoint_sha256":"2b301a6875a6ba9efa52a2a66773a5bcc49d5397f7c861a94495252e647df185","checkpoint_teacher_family":"hybrid_candi","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","cuda_runtime_version":"12.4","dataset":"lm1b","diversity_prompt_count":256,"environment":"dlb-candi","evaluation_continuation_length":64,"generation_mode":"conditional_prefix","gpu_compute_capability":"9.0","gpu_index":0,"gpu_name":"NVIDIA H200 NVL","model":"candi","nvidia_driver_version":"575.57.08","parameter_precision":"fp32","precision":"bf16-mixed","precision_evidence":"static_policy_bound_to_checkpoint_and_runtime_code_not_runtime_autocast_observation","precision_policy":"candi:pinned_internal_bf16_autocast_with_fp32_sensitive_ops","precision_policy_binding":{"adapter_identity":"dlb.adapters.candi:v1|dlb.adapters.candi:CANDIAdapter|e63560f999a54ea08143cab77049eaa8bbe2965321470716a2fe3a94f2b7ed19","checkpoint_lock_id":"flm_lm1b_reproductions:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152:lm1b_CANDI.ckpt","checkpoint_repository":null,"checkpoint_resource":"flm_lm1b_reproductions","checkpoint_revision":null,"checkpoint_selection":{"path":"lm1b_CANDI.ckpt","resource":"flm_lm1b_reproductions","teacher_family":"hybrid_candi"},"checkpoint_sha256":"2b301a6875a6ba9efa52a2a66773a5bcc49d5397f7c861a94495252e647df185","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","dataset":"lm1b","diversity_prompt_count":256,"evaluation_continuation_length":64,"experiment_config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","generation_mode":"conditional_prefix","model":"candi","prefix_length":64,"prompt_count":1024,"source_commit":"cd57ae9eec98d6ac71cd52bdc50eeec8dfd70f91","upstream":"candi"},"prefix_length":64,"prompt_count":1024,"pytorch_compiled_cuda_toolkit":"12.4","requested_precision":"author","runtime_code_binding":[{"kind":"sampling_callable","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/capture.py","sha256":"bb7d05734503f57d02187a8a94d91394b5a3b394b45a6c3556ed51e612591713","symbol":"_capture_teacher.<locals>.capture.<locals>.<lambda>"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/candi/algo.py","sha256":"131ffaeb45418710294a70e1258afdaaa213e8a51bd19ef6227eda8d7fd036e9","symbol":"CANDI"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/candi/models/dit_cont.py","sha256":"3ee29045168f0ea4c0faecc35c053cdd2b872b7ac11eb16def52c6dae084552a","symbol":"ContDIT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/candi/models/dit_cont.py","sha256":"3ee29045168f0ea4c0faecc35c053cdd2b872b7ac11eb16def52c6dae084552a","symbol":"DDiTBlock"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/candi/models/dit_cont.py","sha256":"3ee29045168f0ea4c0faecc35c053cdd2b872b7ac11eb16def52c6dae084552a","symbol":"DDiTFinalLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/candi/models/dit_cont.py","sha256":"3ee29045168f0ea4c0faecc35c053cdd2b872b7ac11eb16def52c6dae084552a","symbol":"EmbeddingLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/candi/models/dit_cont.py","sha256":"3ee29045168f0ea4c0faecc35c053cdd2b872b7ac11eb16def52c6dae084552a","symbol":"LayerNorm"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/candi/models/dit_cont.py","sha256":"3ee29045168f0ea4c0faecc35c053cdd2b872b7ac11eb16def52c6dae084552a","symbol":"Rotary"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/candi/models/dit_cont.py","sha256":"3ee29045168f0ea4c0faecc35c053cdd2b872b7ac11eb16def52c6dae084552a","symbol":"TimestepEmbedder"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/candi/trainer_base.py","sha256":"953ab0e52957472772150b32a342fe826dcc1a57ad3a4c48654de4985f0c18a4","symbol":"LogLinear"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-candi/lib/python3.11/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"GELU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-candi/lib/python3.11/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"SiLU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-candi/lib/python3.11/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"Softplus"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-candi/lib/python3.11/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"ModuleList"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-candi/lib/python3.11/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"Sequential"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-candi/lib/python3.11/site-packages/torch/nn/modules/dropout.py","sha256":"1fc9974c3cf2078dd4bc257f73afb6a6714adfec2e9d099d47cf923e766d3490","symbol":"Dropout"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-candi/lib/python3.11/site-packages/torch/nn/modules/linear.py","sha256":"d7f5cadf773deb5ab4bcc9f57414e8239b2859c70d2aa6f141576de1ec918823","symbol":"Linear"}],"sample_count":2048,"seed":42,"source_commit":"cd57ae9eec98d6ac71cd52bdc50eeec8dfd70f91","steps":8,"synchronization_policy":"torch.cuda.synchronize_before_start_and_after_generate"},"schema":"dlb-generation-timing-v1","timing":{"batch_size":1,"exclusions":["model_and_checkpoint_loading","first_compilation","token_decoding","metrics","file_io"],"median_seconds":0.050887257093563676,"mode":"primary_latency","num_timed_samples":32,"raw_durations_seconds":[0.08097852999344468,0.0782812531106174,0.11627397779375315,0.05082677397876978,0.05084044532850385,0.051176529843360186,0.050719338934868574,0.05068989424034953,0.050903500989079475,0.050909189973026514,0.05084601417183876,0.05087933503091335,0.0508608459495008,0.05081158131361008,0.05089281592518091,0.05083485599607229,0.0509342891164124,0.05088169826194644,0.050855187233537436,0.05058915028348565,0.05042401608079672,0.05390882212668657,0.05078524025157094,0.0509467083029449,0.05090848868712783,0.050869989674538374,0.050856729969382286,0.05096474476158619,0.051205425057560205,0.05110044404864311,0.050956022925674915,0.05143340490758419],"repeats":32,"seconds_per_sample":0.054823288883198984,"standard_deviation_convention":"population","standard_deviation_seconds":0.013045186100231889,"warmups":5}}

results/conditional/timing/lm1b/duo/steps_1024/timing.json

@@ -0,0 +1 @@
+{"metadata":{"adapter_identity":"dlb.adapters.duo:v1|dlb.adapters.duo:DuoAdapter|8878182eb10d7d1e3f18219cb62c2153084842fdb0ab532b1a3c7df0e10da3c6","attempt_id":"f86c30fd76a841bdab6482dda17e7044","checkpoint_lock_id":"flm_lm1b_reproductions:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152:lm1b_Duo.ckpt","checkpoint_selection":{"path":"lm1b_Duo.ckpt","resource":"flm_lm1b_reproductions","teacher_family":"uniform_duo"},"checkpoint_sha256":"42694ec4a5d7aa1ca9d1f5e377a99a75dd7239b047236cdc644ee770b157245f","checkpoint_teacher_family":"uniform_duo","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","cuda_runtime_version":"12.4","dataset":"lm1b","diversity_prompt_count":256,"environment":"dlb-duo","evaluation_continuation_length":64,"generation_mode":"conditional_prefix","gpu_compute_capability":"9.0","gpu_index":0,"gpu_name":"NVIDIA H200 NVL","model":"duo","nvidia_driver_version":"575.57.08","parameter_precision":"fp32","precision":"bf16-mixed","precision_evidence":"static_policy_bound_to_checkpoint_and_runtime_code_not_runtime_autocast_observation","precision_policy":"duo:pinned_internal_bf16_autocast_with_fp32_sensitive_ops","precision_policy_binding":{"adapter_identity":"dlb.adapters.duo:v1|dlb.adapters.duo:DuoAdapter|8878182eb10d7d1e3f18219cb62c2153084842fdb0ab532b1a3c7df0e10da3c6","checkpoint_lock_id":"flm_lm1b_reproductions:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152:lm1b_Duo.ckpt","checkpoint_repository":null,"checkpoint_resource":"flm_lm1b_reproductions","checkpoint_revision":null,"checkpoint_selection":{"path":"lm1b_Duo.ckpt","resource":"flm_lm1b_reproductions","teacher_family":"uniform_duo"},"checkpoint_sha256":"42694ec4a5d7aa1ca9d1f5e377a99a75dd7239b047236cdc644ee770b157245f","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","dataset":"lm1b","diversity_prompt_count":256,"evaluation_continuation_length":64,"experiment_config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","generation_mode":"conditional_prefix","model":"duo","prefix_length":64,"prompt_count":1024,"source_commit":"7c9b498f5b717de064d6fad7e2509c866e6cb620","upstream":"duo"},"prefix_length":64,"prompt_count":1024,"pytorch_compiled_cuda_toolkit":"12.4","requested_precision":"author","runtime_code_binding":[{"kind":"sampling_callable","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/capture.py","sha256":"92c69119e85195e099a41668e8f75dd5a908c4e341fb220f714853db555dca0d","symbol":"_capture_teacher.<locals>.capture.<locals>.<lambda>"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/algo.py","sha256":"d6cf18959f77f5002f5a9dde3df3fc109809afc87e6d735da88d0cb28dfa9d46","symbol":"DUO_BASE"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/models/dit.py","sha256":"32ee71d182ca69645730b4c4624e9cba92bbed9afe3c8ace3aac74e8f1089f97","symbol":"DDiTBlock"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/models/dit.py","sha256":"32ee71d182ca69645730b4c4624e9cba92bbed9afe3c8ace3aac74e8f1089f97","symbol":"DDiTFinalLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/models/dit.py","sha256":"32ee71d182ca69645730b4c4624e9cba92bbed9afe3c8ace3aac74e8f1089f97","symbol":"DIT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/models/dit.py","sha256":"32ee71d182ca69645730b4c4624e9cba92bbed9afe3c8ace3aac74e8f1089f97","symbol":"EmbeddingLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/models/dit.py","sha256":"32ee71d182ca69645730b4c4624e9cba92bbed9afe3c8ace3aac74e8f1089f97","symbol":"LayerNorm"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/models/dit.py","sha256":"32ee71d182ca69645730b4c4624e9cba92bbed9afe3c8ace3aac74e8f1089f97","symbol":"Rotary"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/models/dit.py","sha256":"32ee71d182ca69645730b4c4624e9cba92bbed9afe3c8ace3aac74e8f1089f97","symbol":"TimestepEmbedder"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/trainer_base.py","sha256":"8e990b13034edde83563ec39a00eaaffdab44738229446db21a3a2e367855ed4","symbol":"LogLinear"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"GELU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"SiLU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"ModuleList"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"Sequential"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/dropout.py","sha256":"1fc9974c3cf2078dd4bc257f73afb6a6714adfec2e9d099d47cf923e766d3490","symbol":"Dropout"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/linear.py","sha256":"d7f5cadf773deb5ab4bcc9f57414e8239b2859c70d2aa6f141576de1ec918823","symbol":"Linear"}],"sample_count":2048,"seed":42,"source_commit":"7c9b498f5b717de064d6fad7e2509c866e6cb620","steps":1024,"synchronization_policy":"torch.cuda.synchronize_before_start_and_after_generate"},"schema":"dlb-generation-timing-v1","timing":{"batch_size":1,"exclusions":["model_and_checkpoint_loading","first_compilation","token_decoding","metrics","file_io"],"median_seconds":7.376055216416717,"mode":"primary_latency","num_timed_samples":32,"raw_durations_seconds":[7.7428727759979665,7.705725627951324,7.469200691673905,7.305480025708675,7.344188129063696,7.3166166502051055,7.075356550980359,7.555273668374866,7.302082043141127,7.293280880898237,7.462853233795613,7.744695962872356,7.203486165031791,7.402441346086562,7.10174246923998,7.175323430914432,7.141051292885095,7.533302010037005,7.655593199189752,7.416771720163524,7.6547971670515835,7.029395690653473,7.405532468110323,7.132406680844724,7.480991721153259,7.449837496038526,7.7260882360860705,7.170982911717147,7.556377111002803,7.14677893044427,7.3496690867468715,7.029024341143668],"repeats":32,"seconds_per_sample":7.377475616100128,"standard_deviation_convention":"population","standard_deviation_seconds":0.2158404851119753,"warmups":5}}

results/conditional/timing/lm1b/duo/steps_8/timing.json

@@ -0,0 +1 @@
+{"metadata":{"adapter_identity":"dlb.adapters.duo:v1|dlb.adapters.duo:DuoAdapter|8878182eb10d7d1e3f18219cb62c2153084842fdb0ab532b1a3c7df0e10da3c6","attempt_id":"f1c073c6f76c498295782060732e8421","checkpoint_lock_id":"flm_lm1b_reproductions:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152:lm1b_Duo.ckpt","checkpoint_selection":{"path":"lm1b_Duo.ckpt","resource":"flm_lm1b_reproductions","teacher_family":"uniform_duo"},"checkpoint_sha256":"42694ec4a5d7aa1ca9d1f5e377a99a75dd7239b047236cdc644ee770b157245f","checkpoint_teacher_family":"uniform_duo","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","cuda_runtime_version":"12.4","dataset":"lm1b","diversity_prompt_count":256,"environment":"dlb-duo","evaluation_continuation_length":64,"generation_mode":"conditional_prefix","gpu_compute_capability":"9.0","gpu_index":0,"gpu_name":"NVIDIA H200 NVL","model":"duo","nvidia_driver_version":"575.57.08","parameter_precision":"fp32","precision":"bf16-mixed","precision_evidence":"static_policy_bound_to_checkpoint_and_runtime_code_not_runtime_autocast_observation","precision_policy":"duo:pinned_internal_bf16_autocast_with_fp32_sensitive_ops","precision_policy_binding":{"adapter_identity":"dlb.adapters.duo:v1|dlb.adapters.duo:DuoAdapter|8878182eb10d7d1e3f18219cb62c2153084842fdb0ab532b1a3c7df0e10da3c6","checkpoint_lock_id":"flm_lm1b_reproductions:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152:lm1b_Duo.ckpt","checkpoint_repository":null,"checkpoint_resource":"flm_lm1b_reproductions","checkpoint_revision":null,"checkpoint_selection":{"path":"lm1b_Duo.ckpt","resource":"flm_lm1b_reproductions","teacher_family":"uniform_duo"},"checkpoint_sha256":"42694ec4a5d7aa1ca9d1f5e377a99a75dd7239b047236cdc644ee770b157245f","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","dataset":"lm1b","diversity_prompt_count":256,"evaluation_continuation_length":64,"experiment_config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","generation_mode":"conditional_prefix","model":"duo","prefix_length":64,"prompt_count":1024,"source_commit":"7c9b498f5b717de064d6fad7e2509c866e6cb620","upstream":"duo"},"prefix_length":64,"prompt_count":1024,"pytorch_compiled_cuda_toolkit":"12.4","requested_precision":"author","runtime_code_binding":[{"kind":"sampling_callable","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/capture.py","sha256":"92c69119e85195e099a41668e8f75dd5a908c4e341fb220f714853db555dca0d","symbol":"_capture_teacher.<locals>.capture.<locals>.<lambda>"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/algo.py","sha256":"d6cf18959f77f5002f5a9dde3df3fc109809afc87e6d735da88d0cb28dfa9d46","symbol":"DUO_BASE"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/models/dit.py","sha256":"32ee71d182ca69645730b4c4624e9cba92bbed9afe3c8ace3aac74e8f1089f97","symbol":"DDiTBlock"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/models/dit.py","sha256":"32ee71d182ca69645730b4c4624e9cba92bbed9afe3c8ace3aac74e8f1089f97","symbol":"DDiTFinalLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/models/dit.py","sha256":"32ee71d182ca69645730b4c4624e9cba92bbed9afe3c8ace3aac74e8f1089f97","symbol":"DIT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/models/dit.py","sha256":"32ee71d182ca69645730b4c4624e9cba92bbed9afe3c8ace3aac74e8f1089f97","symbol":"EmbeddingLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/models/dit.py","sha256":"32ee71d182ca69645730b4c4624e9cba92bbed9afe3c8ace3aac74e8f1089f97","symbol":"LayerNorm"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/models/dit.py","sha256":"32ee71d182ca69645730b4c4624e9cba92bbed9afe3c8ace3aac74e8f1089f97","symbol":"Rotary"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/models/dit.py","sha256":"32ee71d182ca69645730b4c4624e9cba92bbed9afe3c8ace3aac74e8f1089f97","symbol":"TimestepEmbedder"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/trainer_base.py","sha256":"8e990b13034edde83563ec39a00eaaffdab44738229446db21a3a2e367855ed4","symbol":"LogLinear"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"GELU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"SiLU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"ModuleList"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"Sequential"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/dropout.py","sha256":"1fc9974c3cf2078dd4bc257f73afb6a6714adfec2e9d099d47cf923e766d3490","symbol":"Dropout"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/linear.py","sha256":"d7f5cadf773deb5ab4bcc9f57414e8239b2859c70d2aa6f141576de1ec918823","symbol":"Linear"}],"sample_count":2048,"seed":42,"source_commit":"7c9b498f5b717de064d6fad7e2509c866e6cb620","steps":8,"synchronization_policy":"torch.cuda.synchronize_before_start_and_after_generate"},"schema":"dlb-generation-timing-v1","timing":{"batch_size":1,"exclusions":["model_and_checkpoint_loading","first_compilation","token_decoding","metrics","file_io"],"median_seconds":0.05401190044358373,"mode":"primary_latency","num_timed_samples":32,"raw_durations_seconds":[0.04980362392961979,0.04959653690457344,0.04951209668070078,0.049555703997612,0.04954571882262826,0.049675868824124336,0.059781030751764774,0.08761341916397214,0.15157404216006398,0.0870958250015974,0.09109791414812207,0.08751302491873503,0.0850060279481113,0.049421658739447594,0.049444292671978474,0.04939434677362442,0.04932577069848776,0.04938193690031767,0.04941212385892868,0.04935241211205721,0.05838718172162771,0.09301127959042788,0.08809782331809402,0.08776462124660611,0.08801177982240915,0.08872325113043189,0.08879941795021296,0.08937081508338451,0.058220176957547665,0.04942265013232827,0.049231166020035744,0.04932830622419715],"repeats":32,"seconds_per_sample":0.06817099513136782,"standard_deviation_convention":"population","standard_deviation_seconds":0.023604729213166663,"warmups":5}}

results/conditional/timing/lm1b/duo_dcd/steps_32/timing.json

@@ -0,0 +1 @@
+{"metadata":{"adapter_identity":"dlb.adapters.duo:v1|dlb.adapters.duo:DuoAdapter|8878182eb10d7d1e3f18219cb62c2153084842fdb0ab532b1a3c7df0e10da3c6","attempt_id":"d7a24c03819543b398e5e0ed05c50c20","checkpoint_lock_id":"recipe:duo_dcd_lm1b:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/reference_reproduction/duo_dcd/lm1b","recipe":"duo_dcd_lm1b","sampling_checkpoint":"model.ckpt","sampling_config":"config.yaml","source":"duo","source_commit":"7c9b498f5b717de064d6fad7e2509c866e6cb620","teacher_adapter":null,"teacher_family":"uniform_duo"},"checkpoint_sha256":"ac3d61fd6e9e412be79328c48b7c9a5529349e5468c49609b23572d0cc8e086b","checkpoint_teacher_family":"uniform_duo","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","cuda_runtime_version":"12.4","dataset":"lm1b","diversity_prompt_count":256,"environment":"dlb-duo","evaluation_continuation_length":64,"generation_mode":"conditional_prefix","gpu_compute_capability":"9.0","gpu_index":0,"gpu_name":"NVIDIA H200 NVL","model":"duo_dcd","nvidia_driver_version":"575.57.08","parameter_precision":"fp32","precision":"bf16-mixed","precision_evidence":"static_policy_bound_to_checkpoint_and_runtime_code_not_runtime_autocast_observation","precision_policy":"duo:pinned_internal_bf16_autocast_with_fp32_sensitive_ops","precision_policy_binding":{"adapter_identity":"dlb.adapters.duo:v1|dlb.adapters.duo:DuoAdapter|8878182eb10d7d1e3f18219cb62c2153084842fdb0ab532b1a3c7df0e10da3c6","checkpoint_lock_id":"recipe:duo_dcd_lm1b:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/reference_reproduction/duo_dcd/lm1b","recipe":"duo_dcd_lm1b","sampling_checkpoint":"model.ckpt","sampling_config":"config.yaml","source":"duo","source_commit":"7c9b498f5b717de064d6fad7e2509c866e6cb620","teacher_adapter":null,"teacher_family":"uniform_duo"},"checkpoint_sha256":"ac3d61fd6e9e412be79328c48b7c9a5529349e5468c49609b23572d0cc8e086b","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","dataset":"lm1b","diversity_prompt_count":256,"evaluation_continuation_length":64,"experiment_config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","generation_mode":"conditional_prefix","model":"duo_dcd","prefix_length":64,"prompt_count":1024,"source_commit":"7c9b498f5b717de064d6fad7e2509c866e6cb620","upstream":"duo"},"prefix_length":64,"prompt_count":1024,"pytorch_compiled_cuda_toolkit":"12.4","requested_precision":"author","runtime_code_binding":[{"kind":"sampling_callable","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/capture.py","sha256":"92c69119e85195e099a41668e8f75dd5a908c4e341fb220f714853db555dca0d","symbol":"_capture_teacher.<locals>.capture.<locals>.<lambda>"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/algo.py","sha256":"d6cf18959f77f5002f5a9dde3df3fc109809afc87e6d735da88d0cb28dfa9d46","symbol":"DUO_BASE"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/models/dit.py","sha256":"32ee71d182ca69645730b4c4624e9cba92bbed9afe3c8ace3aac74e8f1089f97","symbol":"DDiTBlock"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/models/dit.py","sha256":"32ee71d182ca69645730b4c4624e9cba92bbed9afe3c8ace3aac74e8f1089f97","symbol":"DDiTFinalLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/models/dit.py","sha256":"32ee71d182ca69645730b4c4624e9cba92bbed9afe3c8ace3aac74e8f1089f97","symbol":"DIT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/models/dit.py","sha256":"32ee71d182ca69645730b4c4624e9cba92bbed9afe3c8ace3aac74e8f1089f97","symbol":"EmbeddingLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/models/dit.py","sha256":"32ee71d182ca69645730b4c4624e9cba92bbed9afe3c8ace3aac74e8f1089f97","symbol":"LayerNorm"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/models/dit.py","sha256":"32ee71d182ca69645730b4c4624e9cba92bbed9afe3c8ace3aac74e8f1089f97","symbol":"Rotary"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/models/dit.py","sha256":"32ee71d182ca69645730b4c4624e9cba92bbed9afe3c8ace3aac74e8f1089f97","symbol":"TimestepEmbedder"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/trainer_base.py","sha256":"8e990b13034edde83563ec39a00eaaffdab44738229446db21a3a2e367855ed4","symbol":"LogLinear"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"GELU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"SiLU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"ModuleList"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"Sequential"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/dropout.py","sha256":"1fc9974c3cf2078dd4bc257f73afb6a6714adfec2e9d099d47cf923e766d3490","symbol":"Dropout"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/linear.py","sha256":"d7f5cadf773deb5ab4bcc9f57414e8239b2859c70d2aa6f141576de1ec918823","symbol":"Linear"}],"sample_count":2048,"seed":42,"source_commit":"7c9b498f5b717de064d6fad7e2509c866e6cb620","steps":32,"synchronization_policy":"torch.cuda.synchronize_before_start_and_after_generate"},"schema":"dlb-generation-timing-v1","timing":{"batch_size":1,"exclusions":["model_and_checkpoint_loading","first_compilation","token_decoding","metrics","file_io"],"median_seconds":0.21202572202309966,"mode":"primary_latency","num_timed_samples":32,"raw_durations_seconds":[0.1923399311490357,0.2232366008684039,0.2857207078486681,0.19028408685699105,0.21890654787421227,0.2941506402567029,0.19217837369069457,0.20187336718663573,0.3025941839441657,0.1902306261472404,0.21012444514781237,0.31504873000085354,0.29749768506735563,0.18973145866766572,0.21236895816400647,0.30094730807468295,0.18939576530829072,0.21184454392641783,0.3018624121323228,0.1888269418850541,0.2122069001197815,0.3060990599915385,0.18924848875030875,0.20165757508948445,0.30944866966456175,0.18907296983525157,0.20552125293761492,0.30568956211209297,0.18925349600613117,0.20323675824329257,0.3197149089537561,0.3199776718392968],"repeats":32,"seconds_per_sample":0.2393840821168851,"standard_deviation_convention":"population","standard_deviation_seconds":0.051851893608489914,"warmups":5}}

results/conditional/timing/lm1b/duo_dcd/steps_4/timing.json

@@ -0,0 +1 @@
+{"metadata":{"adapter_identity":"dlb.adapters.duo:v1|dlb.adapters.duo:DuoAdapter|8878182eb10d7d1e3f18219cb62c2153084842fdb0ab532b1a3c7df0e10da3c6","attempt_id":"7f1e770273e04cd29bef2c176e0a1be7","checkpoint_lock_id":"recipe:duo_dcd_lm1b:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/reference_reproduction/duo_dcd/lm1b","recipe":"duo_dcd_lm1b","sampling_checkpoint":"model.ckpt","sampling_config":"config.yaml","source":"duo","source_commit":"7c9b498f5b717de064d6fad7e2509c866e6cb620","teacher_adapter":null,"teacher_family":"uniform_duo"},"checkpoint_sha256":"ac3d61fd6e9e412be79328c48b7c9a5529349e5468c49609b23572d0cc8e086b","checkpoint_teacher_family":"uniform_duo","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","cuda_runtime_version":"12.4","dataset":"lm1b","diversity_prompt_count":256,"environment":"dlb-duo","evaluation_continuation_length":64,"generation_mode":"conditional_prefix","gpu_compute_capability":"9.0","gpu_index":0,"gpu_name":"NVIDIA H200 NVL","model":"duo_dcd","nvidia_driver_version":"575.57.08","parameter_precision":"fp32","precision":"bf16-mixed","precision_evidence":"static_policy_bound_to_checkpoint_and_runtime_code_not_runtime_autocast_observation","precision_policy":"duo:pinned_internal_bf16_autocast_with_fp32_sensitive_ops","precision_policy_binding":{"adapter_identity":"dlb.adapters.duo:v1|dlb.adapters.duo:DuoAdapter|8878182eb10d7d1e3f18219cb62c2153084842fdb0ab532b1a3c7df0e10da3c6","checkpoint_lock_id":"recipe:duo_dcd_lm1b:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/reference_reproduction/duo_dcd/lm1b","recipe":"duo_dcd_lm1b","sampling_checkpoint":"model.ckpt","sampling_config":"config.yaml","source":"duo","source_commit":"7c9b498f5b717de064d6fad7e2509c866e6cb620","teacher_adapter":null,"teacher_family":"uniform_duo"},"checkpoint_sha256":"ac3d61fd6e9e412be79328c48b7c9a5529349e5468c49609b23572d0cc8e086b","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","dataset":"lm1b","diversity_prompt_count":256,"evaluation_continuation_length":64,"experiment_config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","generation_mode":"conditional_prefix","model":"duo_dcd","prefix_length":64,"prompt_count":1024,"source_commit":"7c9b498f5b717de064d6fad7e2509c866e6cb620","upstream":"duo"},"prefix_length":64,"prompt_count":1024,"pytorch_compiled_cuda_toolkit":"12.4","requested_precision":"author","runtime_code_binding":[{"kind":"sampling_callable","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/capture.py","sha256":"92c69119e85195e099a41668e8f75dd5a908c4e341fb220f714853db555dca0d","symbol":"_capture_teacher.<locals>.capture.<locals>.<lambda>"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/algo.py","sha256":"d6cf18959f77f5002f5a9dde3df3fc109809afc87e6d735da88d0cb28dfa9d46","symbol":"DUO_BASE"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/models/dit.py","sha256":"32ee71d182ca69645730b4c4624e9cba92bbed9afe3c8ace3aac74e8f1089f97","symbol":"DDiTBlock"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/models/dit.py","sha256":"32ee71d182ca69645730b4c4624e9cba92bbed9afe3c8ace3aac74e8f1089f97","symbol":"DDiTFinalLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/models/dit.py","sha256":"32ee71d182ca69645730b4c4624e9cba92bbed9afe3c8ace3aac74e8f1089f97","symbol":"DIT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/models/dit.py","sha256":"32ee71d182ca69645730b4c4624e9cba92bbed9afe3c8ace3aac74e8f1089f97","symbol":"EmbeddingLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/models/dit.py","sha256":"32ee71d182ca69645730b4c4624e9cba92bbed9afe3c8ace3aac74e8f1089f97","symbol":"LayerNorm"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/models/dit.py","sha256":"32ee71d182ca69645730b4c4624e9cba92bbed9afe3c8ace3aac74e8f1089f97","symbol":"Rotary"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/models/dit.py","sha256":"32ee71d182ca69645730b4c4624e9cba92bbed9afe3c8ace3aac74e8f1089f97","symbol":"TimestepEmbedder"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/trainer_base.py","sha256":"8e990b13034edde83563ec39a00eaaffdab44738229446db21a3a2e367855ed4","symbol":"LogLinear"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"GELU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"SiLU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"ModuleList"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"Sequential"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/dropout.py","sha256":"1fc9974c3cf2078dd4bc257f73afb6a6714adfec2e9d099d47cf923e766d3490","symbol":"Dropout"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/linear.py","sha256":"d7f5cadf773deb5ab4bcc9f57414e8239b2859c70d2aa6f141576de1ec918823","symbol":"Linear"}],"sample_count":2048,"seed":42,"source_commit":"7c9b498f5b717de064d6fad7e2509c866e6cb620","steps":4,"synchronization_policy":"torch.cuda.synchronize_before_start_and_after_generate"},"schema":"dlb-generation-timing-v1","timing":{"batch_size":1,"exclusions":["model_and_checkpoint_loading","first_compilation","token_decoding","metrics","file_io"],"median_seconds":0.05177516909316182,"mode":"primary_latency","num_timed_samples":32,"raw_durations_seconds":[0.05219804821535945,0.05207215389236808,0.0513911503367126,0.05136224580928683,0.052346235141158104,0.052050271071493626,0.05188492685556412,0.0512741100974381,0.05119654117152095,0.0516662523150444,0.05221118684858084,0.05234398180618882,0.0523640220053494,0.05231251288205385,0.052295498084276915,0.051889183931052685,0.05220106290653348,0.052046255208551884,0.05222456809133291,0.05179820489138365,0.05183086544275284,0.051752133294939995,0.05167419370263815,0.042818928603082895,0.030191608238965273,0.02959556318819523,0.02902281377464533,0.029093521647155285,0.029114114120602608,0.02903627511113882,0.029047871939837933,0.02915389509871602],"repeats":32,"seconds_per_sample":0.04598313111637253,"standard_deviation_convention":"population","standard_deviation_seconds":0.009777098533271881,"warmups":5}}

results/conditional/timing/lm1b/duo_di4c/steps_2/timing.json

@@ -0,0 +1 @@
+{"metadata":{"adapter_identity":"dlb.adapters.di4c:v3|dlb.adapters.di4c:Di4CAdapter|e1636b46628a6015944553de2d5d8a1b829bb6bf3f9c6b789d829cb6db222ec9","attempt_id":"ca2039645d744faa8566f947678e261b","checkpoint_lock_id":"recipe:duo_di4c_lm1b:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/reference_reproduction/duo_di4c/lm1b","recipe":"duo_di4c_lm1b","sampling_checkpoint":"student_checkpoints/20000.ckpt","sampling_config":"config.yaml","source":"di4c","source_commit":"ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3","teacher_adapter":"uniform_to_absorbing","teacher_family":"uniform_duo"},"checkpoint_sha256":"0828425487cc7d017c3ef345f6bd673c79a83355bc3ac0e2fb263011a1a22d4a","checkpoint_teacher_family":"uniform_duo","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","cuda_runtime_version":"12.4","dataset":"lm1b","diversity_prompt_count":256,"environment":"dlb-di4c","evaluation_continuation_length":64,"generation_mode":"conditional_prefix","gpu_compute_capability":"9.0","gpu_index":0,"gpu_name":"NVIDIA H200 NVL","model":"duo_di4c","nvidia_driver_version":"575.57.08","parameter_precision":"fp32","precision":"bf16-mixed","precision_evidence":"static_policy_bound_to_checkpoint_and_runtime_code_not_runtime_autocast_observation","precision_policy":"di4c:pinned_bf16_mixed_config_and_internal_autocast","precision_policy_binding":{"adapter_identity":"dlb.adapters.di4c:v3|dlb.adapters.di4c:Di4CAdapter|e1636b46628a6015944553de2d5d8a1b829bb6bf3f9c6b789d829cb6db222ec9","checkpoint_lock_id":"recipe:duo_di4c_lm1b:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/reference_reproduction/duo_di4c/lm1b","recipe":"duo_di4c_lm1b","sampling_checkpoint":"student_checkpoints/20000.ckpt","sampling_config":"config.yaml","source":"di4c","source_commit":"ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3","teacher_adapter":"uniform_to_absorbing","teacher_family":"uniform_duo"},"checkpoint_sha256":"0828425487cc7d017c3ef345f6bd673c79a83355bc3ac0e2fb263011a1a22d4a","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","dataset":"lm1b","diversity_prompt_count":256,"evaluation_continuation_length":64,"experiment_config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","generation_mode":"conditional_prefix","model":"duo_di4c","prefix_length":64,"prompt_count":1024,"source_commit":"ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3","upstream":"di4c"},"prefix_length":64,"prompt_count":1024,"pytorch_compiled_cuda_toolkit":"12.4","requested_precision":"author","runtime_code_binding":[{"kind":"sampling_callable","path":"/home/ruizzhao/diffusion_baseline/adapters/_distilled_runtime.py","sha256":"39268b31d066526a2f0e9bb13bb531f98aaedb1aed46b4f4f0fd722cf02a5683","symbol":"benchmark_model.<locals>.<lambda>"},{"kind":"sampling_closure","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/conditional_runtime.py","sha256":"eef6d1a1062787f949be4656b7a1cc2e81718b42affd9facce4159e2c02c89b8","symbol":"token_project_fn.<locals>.project"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/core/diffusion/noise_schedule.py","sha256":"b2c06e11a4947b306db1c19eac181b99bbc6b36520526ebb36639696c95dcb21","symbol":"LogLinearNoise"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/core/distill/multi_round_sdtt.py","sha256":"75c61fbd1d3e1b87e10b25d67c0143ba0995753f579e0282bc1e3559c145704c","symbol":"MultiRoundSDTT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"DDiTBlock"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"DDitFinalLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"DIT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"EmbeddingLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"LayerNorm"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"Rotary"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"TimestepEmbedder"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"GELU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"SiLU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"ModuleList"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"Sequential"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/dropout.py","sha256":"1fc9974c3cf2078dd4bc257f73afb6a6714adfec2e9d099d47cf923e766d3490","symbol":"Dropout"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/linear.py","sha256":"d7f5cadf773deb5ab4bcc9f57414e8239b2859c70d2aa6f141576de1ec918823","symbol":"Linear"}],"sample_count":2048,"seed":42,"source_commit":"ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3","steps":2,"synchronization_policy":"torch.cuda.synchronize_before_start_and_after_generate"},"schema":"dlb-generation-timing-v1","timing":{"batch_size":1,"exclusions":["model_and_checkpoint_loading","first_compilation","token_decoding","metrics","file_io"],"median_seconds":0.01446571690030396,"mode":"primary_latency","num_timed_samples":32,"raw_durations_seconds":[0.014552013948559761,0.01458512619137764,0.014505872968584299,0.014495227951556444,0.014495817944407463,0.01651829108595848,0.014688904862850904,0.014486102852970362,0.014525294303894043,0.014429687056690454,0.014493213966488838,0.014456677716225386,0.014462247025221586,0.014469186775386333,0.01445250166580081,0.014423367101699114,0.01446997793391347,0.014440684113651514,0.014479263219982386,0.014513354748487473,0.014477099757641554,0.014476869255304337,0.014431759715080261,0.014461795799434185,0.014449878130108118,0.014434362761676311,0.014446282293647528,0.014455336146056652,0.014439941849559546,0.01444912701845169,0.014378217980265617,0.014403315726667643],"repeats":32,"seconds_per_sample":0.014538962495862506,"standard_deviation_convention":"population","standard_deviation_seconds":0.00035982260406084495,"warmups":5}}

results/conditional/timing/lm1b/duo_di4c/steps_32/timing.json

@@ -0,0 +1 @@
+{"metadata":{"adapter_identity":"dlb.adapters.di4c:v3|dlb.adapters.di4c:Di4CAdapter|e1636b46628a6015944553de2d5d8a1b829bb6bf3f9c6b789d829cb6db222ec9","attempt_id":"21f065d617cb4deaa7e643a6ffd237a3","checkpoint_lock_id":"recipe:duo_di4c_lm1b:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/reference_reproduction/duo_di4c/lm1b","recipe":"duo_di4c_lm1b","sampling_checkpoint":"student_checkpoints/20000.ckpt","sampling_config":"config.yaml","source":"di4c","source_commit":"ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3","teacher_adapter":"uniform_to_absorbing","teacher_family":"uniform_duo"},"checkpoint_sha256":"0828425487cc7d017c3ef345f6bd673c79a83355bc3ac0e2fb263011a1a22d4a","checkpoint_teacher_family":"uniform_duo","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","cuda_runtime_version":"12.4","dataset":"lm1b","diversity_prompt_count":256,"environment":"dlb-di4c","evaluation_continuation_length":64,"generation_mode":"conditional_prefix","gpu_compute_capability":"9.0","gpu_index":0,"gpu_name":"NVIDIA H200 NVL","model":"duo_di4c","nvidia_driver_version":"575.57.08","parameter_precision":"fp32","precision":"bf16-mixed","precision_evidence":"static_policy_bound_to_checkpoint_and_runtime_code_not_runtime_autocast_observation","precision_policy":"di4c:pinned_bf16_mixed_config_and_internal_autocast","precision_policy_binding":{"adapter_identity":"dlb.adapters.di4c:v3|dlb.adapters.di4c:Di4CAdapter|e1636b46628a6015944553de2d5d8a1b829bb6bf3f9c6b789d829cb6db222ec9","checkpoint_lock_id":"recipe:duo_di4c_lm1b:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/reference_reproduction/duo_di4c/lm1b","recipe":"duo_di4c_lm1b","sampling_checkpoint":"student_checkpoints/20000.ckpt","sampling_config":"config.yaml","source":"di4c","source_commit":"ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3","teacher_adapter":"uniform_to_absorbing","teacher_family":"uniform_duo"},"checkpoint_sha256":"0828425487cc7d017c3ef345f6bd673c79a83355bc3ac0e2fb263011a1a22d4a","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","dataset":"lm1b","diversity_prompt_count":256,"evaluation_continuation_length":64,"experiment_config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","generation_mode":"conditional_prefix","model":"duo_di4c","prefix_length":64,"prompt_count":1024,"source_commit":"ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3","upstream":"di4c"},"prefix_length":64,"prompt_count":1024,"pytorch_compiled_cuda_toolkit":"12.4","requested_precision":"author","runtime_code_binding":[{"kind":"sampling_callable","path":"/home/ruizzhao/diffusion_baseline/adapters/_distilled_runtime.py","sha256":"39268b31d066526a2f0e9bb13bb531f98aaedb1aed46b4f4f0fd722cf02a5683","symbol":"benchmark_model.<locals>.<lambda>"},{"kind":"sampling_closure","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/conditional_runtime.py","sha256":"eef6d1a1062787f949be4656b7a1cc2e81718b42affd9facce4159e2c02c89b8","symbol":"token_project_fn.<locals>.project"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/core/diffusion/noise_schedule.py","sha256":"b2c06e11a4947b306db1c19eac181b99bbc6b36520526ebb36639696c95dcb21","symbol":"LogLinearNoise"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/core/distill/multi_round_sdtt.py","sha256":"75c61fbd1d3e1b87e10b25d67c0143ba0995753f579e0282bc1e3559c145704c","symbol":"MultiRoundSDTT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"DDiTBlock"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"DDitFinalLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"DIT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"EmbeddingLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"LayerNorm"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"Rotary"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"TimestepEmbedder"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"GELU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"SiLU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"ModuleList"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"Sequential"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/dropout.py","sha256":"1fc9974c3cf2078dd4bc257f73afb6a6714adfec2e9d099d47cf923e766d3490","symbol":"Dropout"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/linear.py","sha256":"d7f5cadf773deb5ab4bcc9f57414e8239b2859c70d2aa6f141576de1ec918823","symbol":"Linear"}],"sample_count":2048,"seed":42,"source_commit":"ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3","steps":32,"synchronization_policy":"torch.cuda.synchronize_before_start_and_after_generate"},"schema":"dlb-generation-timing-v1","timing":{"batch_size":1,"exclusions":["model_and_checkpoint_loading","first_compilation","token_decoding","metrics","file_io"],"median_seconds":0.49475194537080824,"mode":"primary_latency","num_timed_samples":32,"raw_durations_seconds":[0.42964181024581194,0.5886196698993444,0.41297974763438106,0.5885575548745692,0.4234071341343224,0.5785956610925496,0.39881486678496003,0.5887242997996509,0.4246125537902117,0.5684358449652791,0.4209926910698414,0.5688208760693669,0.43035711720585823,0.5578746828250587,0.4511202988214791,0.5374476662836969,0.4499878422357142,0.5378943211399019,0.4639759915880859,0.5272039608098567,0.4775777789764106,0.5121743944473565,0.4762080768123269,0.5133105758577585,0.5080664809793234,0.4814374097622931,0.5175916249863803,0.4712734669446945,0.5257845530286431,0.46244002506136894,0.556845513638109,0.4422859172336757],"repeats":32,"seconds_per_sample":0.4966581377811963,"standard_deviation_convention":"population","standard_deviation_seconds":0.0583371845172985,"warmups":5}}

results/conditional/timing/lm1b/duo_di4c/steps_4/timing.json

@@ -0,0 +1 @@
+{"metadata":{"adapter_identity":"dlb.adapters.di4c:v3|dlb.adapters.di4c:Di4CAdapter|e1636b46628a6015944553de2d5d8a1b829bb6bf3f9c6b789d829cb6db222ec9","attempt_id":"54c3ed9afd674bde9a1fd6949905df30","checkpoint_lock_id":"recipe:duo_di4c_lm1b:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/reference_reproduction/duo_di4c/lm1b","recipe":"duo_di4c_lm1b","sampling_checkpoint":"student_checkpoints/20000.ckpt","sampling_config":"config.yaml","source":"di4c","source_commit":"ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3","teacher_adapter":"uniform_to_absorbing","teacher_family":"uniform_duo"},"checkpoint_sha256":"0828425487cc7d017c3ef345f6bd673c79a83355bc3ac0e2fb263011a1a22d4a","checkpoint_teacher_family":"uniform_duo","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","cuda_runtime_version":"12.4","dataset":"lm1b","diversity_prompt_count":256,"environment":"dlb-di4c","evaluation_continuation_length":64,"generation_mode":"conditional_prefix","gpu_compute_capability":"9.0","gpu_index":0,"gpu_name":"NVIDIA H200 NVL","model":"duo_di4c","nvidia_driver_version":"575.57.08","parameter_precision":"fp32","precision":"bf16-mixed","precision_evidence":"static_policy_bound_to_checkpoint_and_runtime_code_not_runtime_autocast_observation","precision_policy":"di4c:pinned_bf16_mixed_config_and_internal_autocast","precision_policy_binding":{"adapter_identity":"dlb.adapters.di4c:v3|dlb.adapters.di4c:Di4CAdapter|e1636b46628a6015944553de2d5d8a1b829bb6bf3f9c6b789d829cb6db222ec9","checkpoint_lock_id":"recipe:duo_di4c_lm1b:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/reference_reproduction/duo_di4c/lm1b","recipe":"duo_di4c_lm1b","sampling_checkpoint":"student_checkpoints/20000.ckpt","sampling_config":"config.yaml","source":"di4c","source_commit":"ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3","teacher_adapter":"uniform_to_absorbing","teacher_family":"uniform_duo"},"checkpoint_sha256":"0828425487cc7d017c3ef345f6bd673c79a83355bc3ac0e2fb263011a1a22d4a","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","dataset":"lm1b","diversity_prompt_count":256,"evaluation_continuation_length":64,"experiment_config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","generation_mode":"conditional_prefix","model":"duo_di4c","prefix_length":64,"prompt_count":1024,"source_commit":"ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3","upstream":"di4c"},"prefix_length":64,"prompt_count":1024,"pytorch_compiled_cuda_toolkit":"12.4","requested_precision":"author","runtime_code_binding":[{"kind":"sampling_callable","path":"/home/ruizzhao/diffusion_baseline/adapters/_distilled_runtime.py","sha256":"39268b31d066526a2f0e9bb13bb531f98aaedb1aed46b4f4f0fd722cf02a5683","symbol":"benchmark_model.<locals>.<lambda>"},{"kind":"sampling_closure","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/conditional_runtime.py","sha256":"eef6d1a1062787f949be4656b7a1cc2e81718b42affd9facce4159e2c02c89b8","symbol":"token_project_fn.<locals>.project"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/core/diffusion/noise_schedule.py","sha256":"b2c06e11a4947b306db1c19eac181b99bbc6b36520526ebb36639696c95dcb21","symbol":"LogLinearNoise"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/core/distill/multi_round_sdtt.py","sha256":"75c61fbd1d3e1b87e10b25d67c0143ba0995753f579e0282bc1e3559c145704c","symbol":"MultiRoundSDTT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"DDiTBlock"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"DDitFinalLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"DIT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"EmbeddingLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"LayerNorm"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"Rotary"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"TimestepEmbedder"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"GELU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"SiLU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"ModuleList"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"Sequential"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/dropout.py","sha256":"1fc9974c3cf2078dd4bc257f73afb6a6714adfec2e9d099d47cf923e766d3490","symbol":"Dropout"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/linear.py","sha256":"d7f5cadf773deb5ab4bcc9f57414e8239b2859c70d2aa6f141576de1ec918823","symbol":"Linear"}],"sample_count":2048,"seed":42,"source_commit":"ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3","steps":4,"synchronization_policy":"torch.cuda.synchronize_before_start_and_after_generate"},"schema":"dlb-generation-timing-v1","timing":{"batch_size":1,"exclusions":["model_and_checkpoint_loading","first_compilation","token_decoding","metrics","file_io"],"median_seconds":0.02952156006358564,"mode":"primary_latency","num_timed_samples":32,"raw_durations_seconds":[0.029989046044647694,0.02981955697759986,0.029793667141348124,0.029627752024680376,0.029709397815167904,0.029541379772126675,0.029616263695061207,0.029515069909393787,0.02960612904280424,0.029675244819372892,0.029529660940170288,0.029544003773480654,0.02952805021777749,0.02953487914055586,0.029606178868561983,0.02954008709639311,0.029498424381017685,0.029452945105731487,0.029354472644627094,0.02939276210963726,0.029466986190527678,0.02945350483059883,0.029528921004384756,0.029434535652399063,0.02937022689729929,0.029378720093518496,0.029449789319187403,0.029434957075864077,0.029493596870452166,0.02941132988780737,0.029496902134269476,0.029348914977163076],"repeats":32,"seconds_per_sample":0.029535729889175855,"standard_deviation_convention":"population","standard_deviation_seconds":0.00014078796599275553,"warmups":5}}

results/conditional/timing/lm1b/flm/steps_16/timing.json

@@ -0,0 +1 @@
+{"metadata":{"adapter_identity":"dlb.adapters.flm:v2|dlb.adapters.flm:FLMAdapter|c3a6a4573200fecf38a47fc6e7bc78254ca362fff3b96079dc2efa9f8e27aff0","attempt_id":"f6e563fc30f34c14a053f4f80c9411e2","checkpoint_lock_id":"recipe:flm_lm1b_official_ckpt:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/official/flm_ckpt/lm1b","recipe":"flm_lm1b_official_ckpt","sampling_checkpoint":"lm1b_flm.ckpt","sampling_config":null,"source":"flm","source_commit":"a1918d5164e5038e37d0b7a4fb2010ce75b863b3","teacher_adapter":null,"teacher_family":"continuous_flm"},"checkpoint_sha256":"99a97513471dc7232d4a0fc6b50d47301d97a6198e0cf13b77b3fe70c12c3665","checkpoint_teacher_family":"continuous_flm","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","cuda_runtime_version":"12.4","dataset":"lm1b","diversity_prompt_count":256,"environment":"dlb-flm","evaluation_continuation_length":64,"generation_mode":"conditional_prefix","gpu_compute_capability":"9.0","gpu_index":0,"gpu_name":"NVIDIA H200 NVL","model":"flm","nvidia_driver_version":"575.57.08","parameter_precision":"fp32","precision":"bf16-mixed-static-author-policy","precision_evidence":"static_policy_bound_to_checkpoint_and_runtime_code_not_runtime_autocast_observation","precision_policy":"flm:checkpoint_config_and_loaded_runtime_code_bound","precision_policy_binding":{"adapter_identity":"dlb.adapters.flm:v2|dlb.adapters.flm:FLMAdapter|c3a6a4573200fecf38a47fc6e7bc78254ca362fff3b96079dc2efa9f8e27aff0","checkpoint_lock_id":"recipe:flm_lm1b_official_ckpt:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/official/flm_ckpt/lm1b","recipe":"flm_lm1b_official_ckpt","sampling_checkpoint":"lm1b_flm.ckpt","sampling_config":null,"source":"flm","source_commit":"a1918d5164e5038e37d0b7a4fb2010ce75b863b3","teacher_adapter":null,"teacher_family":"continuous_flm"},"checkpoint_sha256":"99a97513471dc7232d4a0fc6b50d47301d97a6198e0cf13b77b3fe70c12c3665","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","dataset":"lm1b","diversity_prompt_count":256,"evaluation_continuation_length":64,"experiment_config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","generation_mode":"conditional_prefix","model":"flm","prefix_length":64,"prompt_count":1024,"source_commit":"a1918d5164e5038e37d0b7a4fb2010ce75b863b3","upstream":"flm"},"prefix_length":64,"prompt_count":1024,"pytorch_compiled_cuda_toolkit":"12.4","requested_precision":"author","runtime_code_binding":[{"kind":"sampling_callable","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/capture.py","sha256":"92c69119e85195e099a41668e8f75dd5a908c4e341fb220f714853db555dca0d","symbol":"_capture_teacher.<locals>.capture.<locals>.<lambda>"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/flm/algo.py","sha256":"65206adc72013bb9b09b68fd847239b3193dd65fd59669bbdee42e061583d76a","symbol":"FLM"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/flm/models/dit.py","sha256":"a5deb76947fa3c31d21e264e10591f504be1c987bc54b98553059f556648ba7a","symbol":"DDiTBlock"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/flm/models/dit.py","sha256":"a5deb76947fa3c31d21e264e10591f504be1c987bc54b98553059f556648ba7a","symbol":"DDiTFinalLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/flm/models/dit.py","sha256":"a5deb76947fa3c31d21e264e10591f504be1c987bc54b98553059f556648ba7a","symbol":"DIT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/flm/models/dit.py","sha256":"a5deb76947fa3c31d21e264e10591f504be1c987bc54b98553059f556648ba7a","symbol":"EmbeddingLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/flm/models/dit.py","sha256":"a5deb76947fa3c31d21e264e10591f504be1c987bc54b98553059f556648ba7a","symbol":"LayerNorm"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/flm/models/dit.py","sha256":"a5deb76947fa3c31d21e264e10591f504be1c987bc54b98553059f556648ba7a","symbol":"Rotary"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/flm/models/dit.py","sha256":"a5deb76947fa3c31d21e264e10591f504be1c987bc54b98553059f556648ba7a","symbol":"TimestepEmbedder"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/flm/trainer_base.py","sha256":"4a3cfa69d511023553290aa3ce975b9ced9cb73743f6a06e141de4ebfeec3d36","symbol":"LogLinear"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-flm/lib/python3.11/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"GELU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-flm/lib/python3.11/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"SiLU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-flm/lib/python3.11/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"Softplus"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-flm/lib/python3.11/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"ModuleList"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-flm/lib/python3.11/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"Sequential"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-flm/lib/python3.11/site-packages/torch/nn/modules/dropout.py","sha256":"1fc9974c3cf2078dd4bc257f73afb6a6714adfec2e9d099d47cf923e766d3490","symbol":"Dropout"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-flm/lib/python3.11/site-packages/torch/nn/modules/linear.py","sha256":"d7f5cadf773deb5ab4bcc9f57414e8239b2859c70d2aa6f141576de1ec918823","symbol":"Linear"}],"sample_count":2048,"seed":42,"source_commit":"a1918d5164e5038e37d0b7a4fb2010ce75b863b3","steps":16,"synchronization_policy":"torch.cuda.synchronize_before_start_and_after_generate"},"schema":"dlb-generation-timing-v1","timing":{"batch_size":1,"exclusions":["model_and_checkpoint_loading","first_compilation","token_decoding","metrics","file_io"],"median_seconds":0.10296167503111064,"mode":"primary_latency","num_timed_samples":32,"raw_durations_seconds":[0.12182492297142744,0.09483949467539787,0.09460793901234865,0.09502481808885932,0.0952430828474462,0.12008022703230381,0.12033671978861094,0.11076488578692079,0.09420921793207526,0.09458544524386525,0.09430577512830496,0.10423451801761985,0.12058247672393918,0.12180363992229104,0.12079978128895164,0.12319815997034311,0.11369691416621208,0.09508471097797155,0.09499550424516201,0.09503009682521224,0.10647934162989259,0.12242541508749127,0.12121089082211256,0.10168883204460144,0.09494538698345423,0.09537027915939689,0.09485046193003654,0.11726001976057887,0.12651124224066734,0.11711665987968445,0.09552960377186537,0.09804045502096415],"repeats":32,"seconds_per_sample":0.10677115371800028,"standard_deviation_convention":"population","standard_deviation_seconds":0.012091027960880197,"warmups":5}}

results/conditional/timing/lm1b/fmlm/steps_2/timing.json

@@ -0,0 +1 @@
+{"metadata":{"adapter_identity":"dlb.adapters.flm:v2|dlb.adapters.flm:FLMAdapter|c3a6a4573200fecf38a47fc6e7bc78254ca362fff3b96079dc2efa9f8e27aff0","attempt_id":"501c7406ceeb483abcaa98af4548a2e3","checkpoint_lock_id":"recipe:fmlm_lm1b_official_ckpt:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/official/fmlm_ckpt/lm1b","recipe":"fmlm_lm1b_official_ckpt","sampling_checkpoint":"lm1b_fmlm.ckpt","sampling_config":null,"source":"flm","source_commit":"a1918d5164e5038e37d0b7a4fb2010ce75b863b3","teacher_adapter":null,"teacher_family":"continuous_flm"},"checkpoint_sha256":"cb096021fda9344c88b184d9b3ac8a0a3bc45e30be7bcc917e024971a9d67395","checkpoint_teacher_family":"continuous_flm","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","cuda_runtime_version":"12.4","dataset":"lm1b","diversity_prompt_count":256,"environment":"dlb-flm","evaluation_continuation_length":64,"generation_mode":"conditional_prefix","gpu_compute_capability":"9.0","gpu_index":0,"gpu_name":"NVIDIA H200 NVL","model":"fmlm","nvidia_driver_version":"575.57.08","parameter_precision":"fp32","precision":"bf16-mixed-static-author-policy","precision_evidence":"static_policy_bound_to_checkpoint_and_runtime_code_not_runtime_autocast_observation","precision_policy":"flm:checkpoint_config_and_loaded_runtime_code_bound","precision_policy_binding":{"adapter_identity":"dlb.adapters.flm:v2|dlb.adapters.flm:FLMAdapter|c3a6a4573200fecf38a47fc6e7bc78254ca362fff3b96079dc2efa9f8e27aff0","checkpoint_lock_id":"recipe:fmlm_lm1b_official_ckpt:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/official/fmlm_ckpt/lm1b","recipe":"fmlm_lm1b_official_ckpt","sampling_checkpoint":"lm1b_fmlm.ckpt","sampling_config":null,"source":"flm","source_commit":"a1918d5164e5038e37d0b7a4fb2010ce75b863b3","teacher_adapter":null,"teacher_family":"continuous_flm"},"checkpoint_sha256":"cb096021fda9344c88b184d9b3ac8a0a3bc45e30be7bcc917e024971a9d67395","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","dataset":"lm1b","diversity_prompt_count":256,"evaluation_continuation_length":64,"experiment_config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","generation_mode":"conditional_prefix","model":"fmlm","prefix_length":64,"prompt_count":1024,"source_commit":"a1918d5164e5038e37d0b7a4fb2010ce75b863b3","upstream":"flm"},"prefix_length":64,"prompt_count":1024,"pytorch_compiled_cuda_toolkit":"12.4","requested_precision":"author","runtime_code_binding":[{"kind":"sampling_callable","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/capture.py","sha256":"92c69119e85195e099a41668e8f75dd5a908c4e341fb220f714853db555dca0d","symbol":"_capture_teacher.<locals>.capture.<locals>.<lambda>"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/flm/algo.py","sha256":"65206adc72013bb9b09b68fd847239b3193dd65fd59669bbdee42e061583d76a","symbol":"FMLM"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/flm/models/dit.py","sha256":"a5deb76947fa3c31d21e264e10591f504be1c987bc54b98553059f556648ba7a","symbol":"DDiTBlock"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/flm/models/dit.py","sha256":"a5deb76947fa3c31d21e264e10591f504be1c987bc54b98553059f556648ba7a","symbol":"DDiTFinalLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/flm/models/dit.py","sha256":"a5deb76947fa3c31d21e264e10591f504be1c987bc54b98553059f556648ba7a","symbol":"DIT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/flm/models/dit.py","sha256":"a5deb76947fa3c31d21e264e10591f504be1c987bc54b98553059f556648ba7a","symbol":"EmbeddingLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/flm/models/dit.py","sha256":"a5deb76947fa3c31d21e264e10591f504be1c987bc54b98553059f556648ba7a","symbol":"LayerNorm"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/flm/models/dit.py","sha256":"a5deb76947fa3c31d21e264e10591f504be1c987bc54b98553059f556648ba7a","symbol":"Rotary"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/flm/models/dit.py","sha256":"a5deb76947fa3c31d21e264e10591f504be1c987bc54b98553059f556648ba7a","symbol":"TimestepEmbedder"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/flm/trainer_base.py","sha256":"4a3cfa69d511023553290aa3ce975b9ced9cb73743f6a06e141de4ebfeec3d36","symbol":"LogLinear"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-flm/lib/python3.11/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"GELU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-flm/lib/python3.11/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"SiLU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-flm/lib/python3.11/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"Softplus"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-flm/lib/python3.11/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"ModuleList"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-flm/lib/python3.11/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"Sequential"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-flm/lib/python3.11/site-packages/torch/nn/modules/dropout.py","sha256":"1fc9974c3cf2078dd4bc257f73afb6a6714adfec2e9d099d47cf923e766d3490","symbol":"Dropout"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-flm/lib/python3.11/site-packages/torch/nn/modules/linear.py","sha256":"d7f5cadf773deb5ab4bcc9f57414e8239b2859c70d2aa6f141576de1ec918823","symbol":"Linear"}],"sample_count":2048,"seed":42,"source_commit":"a1918d5164e5038e37d0b7a4fb2010ce75b863b3","steps":2,"synchronization_policy":"torch.cuda.synchronize_before_start_and_after_generate"},"schema":"dlb-generation-timing-v1","timing":{"batch_size":1,"exclusions":["model_and_checkpoint_loading","first_compilation","token_decoding","metrics","file_io"],"median_seconds":0.016555583337321877,"mode":"primary_latency","num_timed_samples":32,"raw_durations_seconds":[0.011666236910969019,0.011690034065395594,0.011649210937321186,0.011608447879552841,0.011593475937843323,0.0116512649692595,0.0116270468570292,0.011590661015361547,0.011565042193979025,0.011661710683256388,0.013877038843929768,0.019234127830713987,0.0194161469116807,0.019583933986723423,0.01979566877707839,0.019457069225609303,0.0197182297706604,0.01983427768573165,0.020395569037646055,0.019701262935996056,0.019628542009741068,0.019544694107025862,0.01953742280602455,0.01954900100827217,0.019499714951962233,0.01963816676288843,0.02041879389435053,0.011770077049732208,0.011885324027389288,0.011648059822618961,0.011631743982434273,0.011691516265273094],"repeats":32,"seconds_per_sample":0.015742484785732813,"standard_deviation_convention":"population","standard_deviation_seconds":0.003966569816944137,"warmups":5}}

results/conditional/timing/lm1b/fmlm/steps_8/timing.json

@@ -0,0 +1 @@
+{"metadata":{"adapter_identity":"dlb.adapters.flm:v2|dlb.adapters.flm:FLMAdapter|c3a6a4573200fecf38a47fc6e7bc78254ca362fff3b96079dc2efa9f8e27aff0","attempt_id":"fcee11848fa74d6c97e23025cb1ea704","checkpoint_lock_id":"recipe:fmlm_lm1b_official_ckpt:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/official/fmlm_ckpt/lm1b","recipe":"fmlm_lm1b_official_ckpt","sampling_checkpoint":"lm1b_fmlm.ckpt","sampling_config":null,"source":"flm","source_commit":"a1918d5164e5038e37d0b7a4fb2010ce75b863b3","teacher_adapter":null,"teacher_family":"continuous_flm"},"checkpoint_sha256":"cb096021fda9344c88b184d9b3ac8a0a3bc45e30be7bcc917e024971a9d67395","checkpoint_teacher_family":"continuous_flm","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","cuda_runtime_version":"12.4","dataset":"lm1b","diversity_prompt_count":256,"environment":"dlb-flm","evaluation_continuation_length":64,"generation_mode":"conditional_prefix","gpu_compute_capability":"9.0","gpu_index":0,"gpu_name":"NVIDIA H200 NVL","model":"fmlm","nvidia_driver_version":"575.57.08","parameter_precision":"fp32","precision":"bf16-mixed-static-author-policy","precision_evidence":"static_policy_bound_to_checkpoint_and_runtime_code_not_runtime_autocast_observation","precision_policy":"flm:checkpoint_config_and_loaded_runtime_code_bound","precision_policy_binding":{"adapter_identity":"dlb.adapters.flm:v2|dlb.adapters.flm:FLMAdapter|c3a6a4573200fecf38a47fc6e7bc78254ca362fff3b96079dc2efa9f8e27aff0","checkpoint_lock_id":"recipe:fmlm_lm1b_official_ckpt:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/official/fmlm_ckpt/lm1b","recipe":"fmlm_lm1b_official_ckpt","sampling_checkpoint":"lm1b_fmlm.ckpt","sampling_config":null,"source":"flm","source_commit":"a1918d5164e5038e37d0b7a4fb2010ce75b863b3","teacher_adapter":null,"teacher_family":"continuous_flm"},"checkpoint_sha256":"cb096021fda9344c88b184d9b3ac8a0a3bc45e30be7bcc917e024971a9d67395","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","dataset":"lm1b","diversity_prompt_count":256,"evaluation_continuation_length":64,"experiment_config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","generation_mode":"conditional_prefix","model":"fmlm","prefix_length":64,"prompt_count":1024,"source_commit":"a1918d5164e5038e37d0b7a4fb2010ce75b863b3","upstream":"flm"},"prefix_length":64,"prompt_count":1024,"pytorch_compiled_cuda_toolkit":"12.4","requested_precision":"author","runtime_code_binding":[{"kind":"sampling_callable","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/capture.py","sha256":"92c69119e85195e099a41668e8f75dd5a908c4e341fb220f714853db555dca0d","symbol":"_capture_teacher.<locals>.capture.<locals>.<lambda>"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/flm/algo.py","sha256":"65206adc72013bb9b09b68fd847239b3193dd65fd59669bbdee42e061583d76a","symbol":"FMLM"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/flm/models/dit.py","sha256":"a5deb76947fa3c31d21e264e10591f504be1c987bc54b98553059f556648ba7a","symbol":"DDiTBlock"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/flm/models/dit.py","sha256":"a5deb76947fa3c31d21e264e10591f504be1c987bc54b98553059f556648ba7a","symbol":"DDiTFinalLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/flm/models/dit.py","sha256":"a5deb76947fa3c31d21e264e10591f504be1c987bc54b98553059f556648ba7a","symbol":"DIT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/flm/models/dit.py","sha256":"a5deb76947fa3c31d21e264e10591f504be1c987bc54b98553059f556648ba7a","symbol":"EmbeddingLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/flm/models/dit.py","sha256":"a5deb76947fa3c31d21e264e10591f504be1c987bc54b98553059f556648ba7a","symbol":"LayerNorm"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/flm/models/dit.py","sha256":"a5deb76947fa3c31d21e264e10591f504be1c987bc54b98553059f556648ba7a","symbol":"Rotary"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/flm/models/dit.py","sha256":"a5deb76947fa3c31d21e264e10591f504be1c987bc54b98553059f556648ba7a","symbol":"TimestepEmbedder"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/flm/trainer_base.py","sha256":"4a3cfa69d511023553290aa3ce975b9ced9cb73743f6a06e141de4ebfeec3d36","symbol":"LogLinear"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-flm/lib/python3.11/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"GELU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-flm/lib/python3.11/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"SiLU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-flm/lib/python3.11/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"Softplus"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-flm/lib/python3.11/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"ModuleList"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-flm/lib/python3.11/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"Sequential"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-flm/lib/python3.11/site-packages/torch/nn/modules/dropout.py","sha256":"1fc9974c3cf2078dd4bc257f73afb6a6714adfec2e9d099d47cf923e766d3490","symbol":"Dropout"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-flm/lib/python3.11/site-packages/torch/nn/modules/linear.py","sha256":"d7f5cadf773deb5ab4bcc9f57414e8239b2859c70d2aa6f141576de1ec918823","symbol":"Linear"}],"sample_count":2048,"seed":42,"source_commit":"a1918d5164e5038e37d0b7a4fb2010ce75b863b3","steps":8,"synchronization_policy":"torch.cuda.synchronize_before_start_and_after_generate"},"schema":"dlb-generation-timing-v1","timing":{"batch_size":1,"exclusions":["model_and_checkpoint_loading","first_compilation","token_decoding","metrics","file_io"],"median_seconds":0.05434661451727152,"mode":"primary_latency","num_timed_samples":32,"raw_durations_seconds":[0.05026452103629708,0.047743663657456636,0.047476775012910366,0.05464558303356171,0.055712690111249685,0.05563131393864751,0.058512394316494465,0.055308067705482244,0.0582078667357564,0.05541905900463462,0.05518364766612649,0.05741242691874504,0.05526823736727238,0.057534193620085716,0.049681838136166334,0.04710687696933746,0.048970425967127085,0.047245478723198175,0.0470884689129889,0.048071777913719416,0.046951639000326395,0.047120457980781794,0.05404764600098133,0.05531309591606259,0.055714111775159836,0.05539539270102978,0.05555858323350549,0.0548581681214273,0.047017139848321676,0.04706696607172489,0.04720920417457819,0.047086094971746206],"repeats":32,"seconds_per_sample":0.052056993954465725,"standard_deviation_convention":"population","standard_deviation_seconds":0.004210522263795925,"warmups":5}}

results/conditional/timing/lm1b/langflow/steps_4/timing.json

@@ -0,0 +1 @@
+{"metadata":{"adapter_identity":"dlb.adapters.langflow:v1|dlb.adapters.langflow:LangFlowAdapter|7d80794cf6a132fc8d4c179d723e8e486c70476888b2ba51dd92cdc3f5284f84","attempt_id":"6a7dc4a2c882448b98e50f75323330ad","checkpoint_lock_id":"langflow_lm1b_hf:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152:all","checkpoint_selection":{"path":null,"resource":"langflow_lm1b_hf","teacher_family":null},"checkpoint_sha256":"2050f55b63e952941eb190711828176529f2f510031cc0e0fc1cd39f6998d7bb","checkpoint_teacher_family":"continuous_langflow","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","cuda_runtime_version":"12.4","dataset":"lm1b","diversity_prompt_count":256,"environment":"dlb-langflow","evaluation_continuation_length":64,"generation_mode":"conditional_prefix","gpu_compute_capability":"9.0","gpu_index":0,"gpu_name":"NVIDIA H200 NVL","model":"langflow","nvidia_driver_version":"575.57.08","parameter_precision":"fp32","precision":"bf16-mixed","precision_evidence":"static_policy_bound_to_checkpoint_and_runtime_code_not_runtime_autocast_observation","precision_policy":"langflow:pinned_internal_bf16_autocast_with_fp32_sensitive_ops","precision_policy_binding":{"adapter_identity":"dlb.adapters.langflow:v1|dlb.adapters.langflow:LangFlowAdapter|7d80794cf6a132fc8d4c179d723e8e486c70476888b2ba51dd92cdc3f5284f84","checkpoint_lock_id":"langflow_lm1b_hf:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152:all","checkpoint_repository":"Continuous-Rivals-Discrete/langflow-lm1b","checkpoint_resource":"langflow_lm1b_hf","checkpoint_revision":"d5d395d9bdf01f697ba8830110e3e78c63d77cb5","checkpoint_selection":{"path":null,"resource":"langflow_lm1b_hf","teacher_family":null},"checkpoint_sha256":"2050f55b63e952941eb190711828176529f2f510031cc0e0fc1cd39f6998d7bb","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","dataset":"lm1b","diversity_prompt_count":256,"evaluation_continuation_length":64,"experiment_config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","generation_mode":"conditional_prefix","model":"langflow","prefix_length":64,"prompt_count":1024,"source_commit":"a712b08570c56c787d6ef24f8e906a2dcdf768f5","upstream":"langflow"},"prefix_length":64,"prompt_count":1024,"pytorch_compiled_cuda_toolkit":"12.4","requested_precision":"author","runtime_code_binding":[{"kind":"sampling_callable","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/capture.py","sha256":"daaf59774f53f83ef7987b1432cd14769226e701310308480bbac28ef5217256","symbol":"_capture_langflow.<locals>.generate.<locals>.<lambda>"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/langflow/langflow/model.py","sha256":"f3d65b468b61bd8ce46fa94764f1248d3f06e79064b6cfcb97c377be9933c8c3","symbol":"DDiTBlock"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/langflow/langflow/model.py","sha256":"f3d65b468b61bd8ce46fa94764f1248d3f06e79064b6cfcb97c377be9933c8c3","symbol":"DDiTFinalLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/langflow/langflow/model.py","sha256":"f3d65b468b61bd8ce46fa94764f1248d3f06e79064b6cfcb97c377be9933c8c3","symbol":"EmbeddingLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/langflow/langflow/model.py","sha256":"f3d65b468b61bd8ce46fa94764f1248d3f06e79064b6cfcb97c377be9933c8c3","symbol":"GumbelProposal"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/langflow/langflow/model.py","sha256":"f3d65b468b61bd8ce46fa94764f1248d3f06e79064b6cfcb97c377be9933c8c3","symbol":"LangFlow"},{"kind":"sampling_closure","path":"/home/ruizzhao/diffusion_baseline/upstreams/langflow/langflow/model.py","sha256":"f3d65b468b61bd8ce46fa94764f1248d3f06e79064b6cfcb97c377be9933c8c3","symbol":"LangFlow.generate_samples"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/langflow/langflow/model.py","sha256":"f3d65b468b61bd8ce46fa94764f1248d3f06e79064b6cfcb97c377be9933c8c3","symbol":"LangFlowBackbone"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/langflow/langflow/model.py","sha256":"f3d65b468b61bd8ce46fa94764f1248d3f06e79064b6cfcb97c377be9933c8c3","symbol":"LayerNorm"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/langflow/langflow/model.py","sha256":"f3d65b468b61bd8ce46fa94764f1248d3f06e79064b6cfcb97c377be9933c8c3","symbol":"Rotary"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/langflow/langflow/model.py","sha256":"f3d65b468b61bd8ce46fa94764f1248d3f06e79064b6cfcb97c377be9933c8c3","symbol":"TimestepEmbedder"},{"kind":"bound_method_owner","path":"/opt/anaconda3/envs/dlb-langflow/lib/python3.12/site-packages/torch/autograd/grad_mode.py","sha256":"de61684a7f73eb361c0f47c22159a2bfc67df6f78ddfa5f198049b0654f1f4a7","symbol":"no_grad"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-langflow/lib/python3.12/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"GELU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-langflow/lib/python3.12/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"SiLU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-langflow/lib/python3.12/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"ModuleList"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-langflow/lib/python3.12/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"Sequential"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-langflow/lib/python3.12/site-packages/torch/nn/modules/linear.py","sha256":"d7f5cadf773deb5ab4bcc9f57414e8239b2859c70d2aa6f141576de1ec918823","symbol":"Linear"},{"kind":"sampling_closure","path":"/opt/anaconda3/envs/dlb-langflow/lib/python3.12/site-packages/torch/utils/_contextlib.py","sha256":"cf7aa5b08f44974ba8c1d08cd71ef70ffd13d1c48a4931576eb235306bfa46b5","symbol":"LangFlow.generate_samples"},{"kind":"sampling_closure","path":"/opt/anaconda3/envs/dlb-langflow/lib/python3.12/site-packages/torch/utils/_contextlib.py","sha256":"cf7aa5b08f44974ba8c1d08cd71ef70ffd13d1c48a4931576eb235306bfa46b5","symbol":"_DecoratorContextManager.clone"}],"sample_count":2048,"seed":42,"source_commit":"a712b08570c56c787d6ef24f8e906a2dcdf768f5","steps":4,"synchronization_policy":"torch.cuda.synchronize_before_start_and_after_generate"},"schema":"dlb-generation-timing-v1","timing":{"batch_size":1,"exclusions":["model_and_checkpoint_loading","first_compilation","token_decoding","metrics","file_io"],"median_seconds":0.05176334083080292,"mode":"primary_latency","num_timed_samples":32,"raw_durations_seconds":[0.024382154922932386,0.024320730939507484,0.02380808489397168,0.023703995160758495,0.02374735102057457,0.023818400222808123,0.02348446799442172,0.023698376026004553,0.024382936768233776,0.023426249157637358,0.04565664194524288,0.05207185400649905,0.05195063818246126,0.04383384622633457,0.051895533222705126,0.05173721071332693,0.051898899022489786,0.0519180279225111,0.05204197810962796,0.04664710210636258,0.05181003175675869,0.051945810206234455,0.05192228499799967,0.05190054094418883,0.05195284215733409,0.05188011983409524,0.051915072835981846,0.051789470948278904,0.05182166118174791,0.05186244333162904,0.038297141902148724,0.023149653803557158],"repeats":32,"seconds_per_sample":0.041208486014511436,"standard_deviation_convention":"population","standard_deviation_seconds":0.012917261706972848,"warmups":5}}

results/conditional/timing/lm1b/mdlm/steps_16/timing.json

@@ -0,0 +1 @@
+{"metadata":{"adapter_identity":"dlb.adapters.mdlm:v1|dlb.adapters.mdlm:MDLMAdapter|56a619efd4f7abcf587f94757a334f9d4481f7f76f19d56bc7a45b930aa9ebd3","attempt_id":"db30514ad63e444d83fdf87456a7aa4d","checkpoint_lock_id":"flm_lm1b_reproductions:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152:lm1b_MDLM_.ckpt","checkpoint_selection":{"path":"lm1b_MDLM_.ckpt","resource":"flm_lm1b_reproductions","teacher_family":"masked_mdlm"},"checkpoint_sha256":"51910b37ddabb1c716654d934526f483abb04837ddf9ca76c29ad41632807f01","checkpoint_teacher_family":"masked_mdlm","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","cuda_runtime_version":"12.4","dataset":"lm1b","diversity_prompt_count":256,"environment":"dlb-mdlm","evaluation_continuation_length":64,"generation_mode":"conditional_prefix","gpu_compute_capability":"9.0","gpu_index":0,"gpu_name":"NVIDIA H200 NVL","model":"mdlm","nvidia_driver_version":"575.57.08","parameter_precision":"fp32","precision":"bf16-mixed","precision_evidence":"static_policy_bound_to_checkpoint_and_runtime_code_not_runtime_autocast_observation","precision_policy":"mdlm:pinned_internal_bf16_autocast_with_fp32_sensitive_ops","precision_policy_binding":{"adapter_identity":"dlb.adapters.mdlm:v1|dlb.adapters.mdlm:MDLMAdapter|56a619efd4f7abcf587f94757a334f9d4481f7f76f19d56bc7a45b930aa9ebd3","checkpoint_lock_id":"flm_lm1b_reproductions:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152:lm1b_MDLM_.ckpt","checkpoint_repository":null,"checkpoint_resource":"flm_lm1b_reproductions","checkpoint_revision":null,"checkpoint_selection":{"path":"lm1b_MDLM_.ckpt","resource":"flm_lm1b_reproductions","teacher_family":"masked_mdlm"},"checkpoint_sha256":"51910b37ddabb1c716654d934526f483abb04837ddf9ca76c29ad41632807f01","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","dataset":"lm1b","diversity_prompt_count":256,"evaluation_continuation_length":64,"experiment_config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","generation_mode":"conditional_prefix","model":"mdlm","prefix_length":64,"prompt_count":1024,"source_commit":"c112c526d193436838c98d81455ee51f90309470","upstream":"mdlm"},"prefix_length":64,"prompt_count":1024,"pytorch_compiled_cuda_toolkit":"12.4","requested_precision":"author","runtime_code_binding":[{"kind":"sampling_callable","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/capture.py","sha256":"92c69119e85195e099a41668e8f75dd5a908c4e341fb220f714853db555dca0d","symbol":"_capture_teacher.<locals>.capture.<locals>.<lambda>"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/mdlm/diffusion.py","sha256":"f8b4f1fdea2ffc960c9852583bee58865aace44baf958f83d5f9b8e352752121","symbol":"BPD"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/mdlm/diffusion.py","sha256":"f8b4f1fdea2ffc960c9852583bee58865aace44baf958f83d5f9b8e352752121","symbol":"Diffusion"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/mdlm/diffusion.py","sha256":"f8b4f1fdea2ffc960c9852583bee58865aace44baf958f83d5f9b8e352752121","symbol":"NLL"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/mdlm/diffusion.py","sha256":"f8b4f1fdea2ffc960c9852583bee58865aace44baf958f83d5f9b8e352752121","symbol":"Perplexity"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/mdlm/models/dit.py","sha256":"58751fa6fe0ee555debc86b91995754bf34e818c076efbe0f0a102bee6899fa4","symbol":"DDiTBlock"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/mdlm/models/dit.py","sha256":"58751fa6fe0ee555debc86b91995754bf34e818c076efbe0f0a102bee6899fa4","symbol":"DDitFinalLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/mdlm/models/dit.py","sha256":"58751fa6fe0ee555debc86b91995754bf34e818c076efbe0f0a102bee6899fa4","symbol":"DIT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/mdlm/models/dit.py","sha256":"58751fa6fe0ee555debc86b91995754bf34e818c076efbe0f0a102bee6899fa4","symbol":"EmbeddingLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/mdlm/models/dit.py","sha256":"58751fa6fe0ee555debc86b91995754bf34e818c076efbe0f0a102bee6899fa4","symbol":"LayerNorm"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/mdlm/models/dit.py","sha256":"58751fa6fe0ee555debc86b91995754bf34e818c076efbe0f0a102bee6899fa4","symbol":"Rotary"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/mdlm/models/dit.py","sha256":"58751fa6fe0ee555debc86b91995754bf34e818c076efbe0f0a102bee6899fa4","symbol":"TimestepEmbedder"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/mdlm/noise_schedule.py","sha256":"da16320cebbc727c03e8e45297593505fb41a597c7a5ef29eacd7043207a763c","symbol":"LogLinearNoise"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-mdlm/lib/python3.9/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"GELU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-mdlm/lib/python3.9/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"SiLU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-mdlm/lib/python3.9/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"Softplus"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-mdlm/lib/python3.9/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"ModuleList"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-mdlm/lib/python3.9/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"Sequential"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-mdlm/lib/python3.9/site-packages/torch/nn/modules/dropout.py","sha256":"1fc9974c3cf2078dd4bc257f73afb6a6714adfec2e9d099d47cf923e766d3490","symbol":"Dropout"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-mdlm/lib/python3.9/site-packages/torch/nn/modules/linear.py","sha256":"d7f5cadf773deb5ab4bcc9f57414e8239b2859c70d2aa6f141576de1ec918823","symbol":"Linear"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-mdlm/lib/python3.9/site-packages/torchmetrics/collections.py","sha256":"b305613023a66917d5e79639e92ad9c3510dccc043ea5417616fc48612f3db88","symbol":"MetricCollection"}],"sample_count":2048,"seed":42,"source_commit":"c112c526d193436838c98d81455ee51f90309470","steps":16,"synchronization_policy":"torch.cuda.synchronize_before_start_and_after_generate"},"schema":"dlb-generation-timing-v1","timing":{"batch_size":1,"exclusions":["model_and_checkpoint_loading","first_compilation","token_decoding","metrics","file_io"],"median_seconds":0.11032765940763056,"mode":"primary_latency","num_timed_samples":32,"raw_durations_seconds":[0.097872877959162,0.09790594922378659,0.09785482008010149,0.14071327028796077,0.1623998200520873,0.11546386405825615,0.09784288192167878,0.09784227097406983,0.09776211809366941,0.1612861412577331,0.16518149711191654,0.16552864015102386,0.16605121176689863,0.16624461906030774,0.1665918012149632,0.16641690209507942,0.12576043140143156,0.09768231632187963,0.09782396303489804,0.09768722392618656,0.15115208504721522,0.16700677759945393,0.10519145475700498,0.09785812627524137,0.09775271313264966,0.10039108106866479,0.1726176254451275,0.154507412109524,0.0989297297783196,0.0977508807554841,0.0977553678676486,0.1257124268449843],"repeats":32,"seconds_per_sample":0.12651682189607527,"standard_deviation_convention":"population","standard_deviation_seconds":0.030594181123883853,"warmups":5}}

results/conditional/timing/lm1b/mdlm/steps_32/timing.json

@@ -0,0 +1 @@
+{"metadata":{"adapter_identity":"dlb.adapters.mdlm:v1|dlb.adapters.mdlm:MDLMAdapter|56a619efd4f7abcf587f94757a334f9d4481f7f76f19d56bc7a45b930aa9ebd3","attempt_id":"bef84e9482f546ac95696f9b397c31ab","checkpoint_lock_id":"flm_lm1b_reproductions:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152:lm1b_MDLM_.ckpt","checkpoint_selection":{"path":"lm1b_MDLM_.ckpt","resource":"flm_lm1b_reproductions","teacher_family":"masked_mdlm"},"checkpoint_sha256":"51910b37ddabb1c716654d934526f483abb04837ddf9ca76c29ad41632807f01","checkpoint_teacher_family":"masked_mdlm","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","cuda_runtime_version":"12.4","dataset":"lm1b","diversity_prompt_count":256,"environment":"dlb-mdlm","evaluation_continuation_length":64,"generation_mode":"conditional_prefix","gpu_compute_capability":"9.0","gpu_index":0,"gpu_name":"NVIDIA H200 NVL","model":"mdlm","nvidia_driver_version":"575.57.08","parameter_precision":"fp32","precision":"bf16-mixed","precision_evidence":"static_policy_bound_to_checkpoint_and_runtime_code_not_runtime_autocast_observation","precision_policy":"mdlm:pinned_internal_bf16_autocast_with_fp32_sensitive_ops","precision_policy_binding":{"adapter_identity":"dlb.adapters.mdlm:v1|dlb.adapters.mdlm:MDLMAdapter|56a619efd4f7abcf587f94757a334f9d4481f7f76f19d56bc7a45b930aa9ebd3","checkpoint_lock_id":"flm_lm1b_reproductions:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152:lm1b_MDLM_.ckpt","checkpoint_repository":null,"checkpoint_resource":"flm_lm1b_reproductions","checkpoint_revision":null,"checkpoint_selection":{"path":"lm1b_MDLM_.ckpt","resource":"flm_lm1b_reproductions","teacher_family":"masked_mdlm"},"checkpoint_sha256":"51910b37ddabb1c716654d934526f483abb04837ddf9ca76c29ad41632807f01","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","dataset":"lm1b","diversity_prompt_count":256,"evaluation_continuation_length":64,"experiment_config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","generation_mode":"conditional_prefix","model":"mdlm","prefix_length":64,"prompt_count":1024,"source_commit":"c112c526d193436838c98d81455ee51f90309470","upstream":"mdlm"},"prefix_length":64,"prompt_count":1024,"pytorch_compiled_cuda_toolkit":"12.4","requested_precision":"author","runtime_code_binding":[{"kind":"sampling_callable","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/capture.py","sha256":"92c69119e85195e099a41668e8f75dd5a908c4e341fb220f714853db555dca0d","symbol":"_capture_teacher.<locals>.capture.<locals>.<lambda>"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/mdlm/diffusion.py","sha256":"f8b4f1fdea2ffc960c9852583bee58865aace44baf958f83d5f9b8e352752121","symbol":"BPD"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/mdlm/diffusion.py","sha256":"f8b4f1fdea2ffc960c9852583bee58865aace44baf958f83d5f9b8e352752121","symbol":"Diffusion"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/mdlm/diffusion.py","sha256":"f8b4f1fdea2ffc960c9852583bee58865aace44baf958f83d5f9b8e352752121","symbol":"NLL"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/mdlm/diffusion.py","sha256":"f8b4f1fdea2ffc960c9852583bee58865aace44baf958f83d5f9b8e352752121","symbol":"Perplexity"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/mdlm/models/dit.py","sha256":"58751fa6fe0ee555debc86b91995754bf34e818c076efbe0f0a102bee6899fa4","symbol":"DDiTBlock"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/mdlm/models/dit.py","sha256":"58751fa6fe0ee555debc86b91995754bf34e818c076efbe0f0a102bee6899fa4","symbol":"DDitFinalLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/mdlm/models/dit.py","sha256":"58751fa6fe0ee555debc86b91995754bf34e818c076efbe0f0a102bee6899fa4","symbol":"DIT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/mdlm/models/dit.py","sha256":"58751fa6fe0ee555debc86b91995754bf34e818c076efbe0f0a102bee6899fa4","symbol":"EmbeddingLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/mdlm/models/dit.py","sha256":"58751fa6fe0ee555debc86b91995754bf34e818c076efbe0f0a102bee6899fa4","symbol":"LayerNorm"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/mdlm/models/dit.py","sha256":"58751fa6fe0ee555debc86b91995754bf34e818c076efbe0f0a102bee6899fa4","symbol":"Rotary"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/mdlm/models/dit.py","sha256":"58751fa6fe0ee555debc86b91995754bf34e818c076efbe0f0a102bee6899fa4","symbol":"TimestepEmbedder"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/mdlm/noise_schedule.py","sha256":"da16320cebbc727c03e8e45297593505fb41a597c7a5ef29eacd7043207a763c","symbol":"LogLinearNoise"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-mdlm/lib/python3.9/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"GELU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-mdlm/lib/python3.9/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"SiLU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-mdlm/lib/python3.9/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"Softplus"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-mdlm/lib/python3.9/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"ModuleList"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-mdlm/lib/python3.9/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"Sequential"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-mdlm/lib/python3.9/site-packages/torch/nn/modules/dropout.py","sha256":"1fc9974c3cf2078dd4bc257f73afb6a6714adfec2e9d099d47cf923e766d3490","symbol":"Dropout"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-mdlm/lib/python3.9/site-packages/torch/nn/modules/linear.py","sha256":"d7f5cadf773deb5ab4bcc9f57414e8239b2859c70d2aa6f141576de1ec918823","symbol":"Linear"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-mdlm/lib/python3.9/site-packages/torchmetrics/collections.py","sha256":"b305613023a66917d5e79639e92ad9c3510dccc043ea5417616fc48612f3db88","symbol":"MetricCollection"}],"sample_count":2048,"seed":42,"source_commit":"c112c526d193436838c98d81455ee51f90309470","steps":32,"synchronization_policy":"torch.cuda.synchronize_before_start_and_after_generate"},"schema":"dlb-generation-timing-v1","timing":{"batch_size":1,"exclusions":["model_and_checkpoint_loading","first_compilation","token_decoding","metrics","file_io"],"median_seconds":0.230269574560225,"mode":"primary_latency","num_timed_samples":32,"raw_durations_seconds":[0.31469431705772877,0.21296378038823605,0.1941940658725798,0.3127167532220483,0.3143009329214692,0.21486022975295782,0.19770192680880427,0.2934532412327826,0.2218902548775077,0.19456613715738058,0.29060949804261327,0.22349235322326422,0.19445876078680158,0.28969266079366207,0.3123256731778383,0.24105161102488637,0.19391134195029736,0.2782387100160122,0.3169549652375281,0.22610577195882797,0.19400205090641975,0.29058851627632976,0.31532581616193056,0.22307164780795574,0.19403140619397163,0.2945712758228183,0.22591610066592693,0.19406658969819546,0.2780809779651463,0.23443337716162205,0.1951922969892621,0.28895262628793716],"repeats":32,"seconds_per_sample":0.2489504896075232,"standard_deviation_convention":"population","standard_deviation_seconds":0.04695566437018485,"warmups":5}}

results/conditional/timing/lm1b/mdlm_di4c/steps_1/timing.json

@@ -0,0 +1 @@
+{"metadata":{"adapter_identity":"dlb.adapters.di4c:v3|dlb.adapters.di4c:Di4CAdapter|e1636b46628a6015944553de2d5d8a1b829bb6bf3f9c6b789d829cb6db222ec9","attempt_id":"0ff58465b2254724b7de3ab5578b8ad4","checkpoint_lock_id":"recipe:mdlm_di4c_lm1b:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/reference_reproduction/mdlm_di4c/lm1b","recipe":"mdlm_di4c_lm1b","sampling_checkpoint":"student_checkpoints/20000.ckpt","sampling_config":"config.yaml","source":"di4c","source_commit":"ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3","teacher_adapter":"masked_to_absorbing","teacher_family":"masked_mdlm"},"checkpoint_sha256":"569faa9af7f866644c622bf3360a246d2ed3f3ddd7572999bdaa7c8a8d3d7f44","checkpoint_teacher_family":"masked_mdlm","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","cuda_runtime_version":"12.4","dataset":"lm1b","diversity_prompt_count":256,"environment":"dlb-di4c","evaluation_continuation_length":64,"generation_mode":"conditional_prefix","gpu_compute_capability":"9.0","gpu_index":0,"gpu_name":"NVIDIA H200 NVL","model":"mdlm_di4c","nvidia_driver_version":"575.57.08","parameter_precision":"fp32","precision":"bf16-mixed","precision_evidence":"static_policy_bound_to_checkpoint_and_runtime_code_not_runtime_autocast_observation","precision_policy":"di4c:pinned_bf16_mixed_config_and_internal_autocast","precision_policy_binding":{"adapter_identity":"dlb.adapters.di4c:v3|dlb.adapters.di4c:Di4CAdapter|e1636b46628a6015944553de2d5d8a1b829bb6bf3f9c6b789d829cb6db222ec9","checkpoint_lock_id":"recipe:mdlm_di4c_lm1b:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/reference_reproduction/mdlm_di4c/lm1b","recipe":"mdlm_di4c_lm1b","sampling_checkpoint":"student_checkpoints/20000.ckpt","sampling_config":"config.yaml","source":"di4c","source_commit":"ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3","teacher_adapter":"masked_to_absorbing","teacher_family":"masked_mdlm"},"checkpoint_sha256":"569faa9af7f866644c622bf3360a246d2ed3f3ddd7572999bdaa7c8a8d3d7f44","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","dataset":"lm1b","diversity_prompt_count":256,"evaluation_continuation_length":64,"experiment_config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","generation_mode":"conditional_prefix","model":"mdlm_di4c","prefix_length":64,"prompt_count":1024,"source_commit":"ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3","upstream":"di4c"},"prefix_length":64,"prompt_count":1024,"pytorch_compiled_cuda_toolkit":"12.4","requested_precision":"author","runtime_code_binding":[{"kind":"sampling_callable","path":"/home/ruizzhao/diffusion_baseline/adapters/_distilled_runtime.py","sha256":"39268b31d066526a2f0e9bb13bb531f98aaedb1aed46b4f4f0fd722cf02a5683","symbol":"benchmark_model.<locals>.<lambda>"},{"kind":"sampling_closure","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/conditional_runtime.py","sha256":"eef6d1a1062787f949be4656b7a1cc2e81718b42affd9facce4159e2c02c89b8","symbol":"token_project_fn.<locals>.project"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/core/diffusion/noise_schedule.py","sha256":"b2c06e11a4947b306db1c19eac181b99bbc6b36520526ebb36639696c95dcb21","symbol":"LogLinearNoise"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/core/distill/multi_round_sdtt.py","sha256":"75c61fbd1d3e1b87e10b25d67c0143ba0995753f579e0282bc1e3559c145704c","symbol":"MultiRoundSDTT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"DDiTBlock"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"DDitFinalLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"DIT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"EmbeddingLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"LayerNorm"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"Rotary"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"TimestepEmbedder"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"GELU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"SiLU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"ModuleList"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"Sequential"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/dropout.py","sha256":"1fc9974c3cf2078dd4bc257f73afb6a6714adfec2e9d099d47cf923e766d3490","symbol":"Dropout"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/linear.py","sha256":"d7f5cadf773deb5ab4bcc9f57414e8239b2859c70d2aa6f141576de1ec918823","symbol":"Linear"}],"sample_count":2048,"seed":42,"source_commit":"ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3","steps":1,"synchronization_policy":"torch.cuda.synchronize_before_start_and_after_generate"},"schema":"dlb-generation-timing-v1","timing":{"batch_size":1,"exclusions":["model_and_checkpoint_loading","first_compilation","token_decoding","metrics","file_io"],"median_seconds":0.007162156980484724,"mode":"primary_latency","num_timed_samples":32,"raw_durations_seconds":[0.00724647706374526,0.007222580257803202,0.007242952007800341,0.0071796840056777,0.007192695047706366,0.007204842753708363,0.0071981120854616165,0.007204101886600256,0.007200636900961399,0.007167686242610216,0.007216351106762886,0.007166733965277672,0.007189129013568163,0.007150079123675823,0.0071546961553394794,0.007153934799134731,0.007157579995691776,0.007319288793951273,0.00715661933645606,0.007119212299585342,0.0071831196546554565,0.007295791991055012,0.007155688013881445,0.007129928097128868,0.007135716732591391,0.0071257008239626884,0.007118841167539358,0.007103638257831335,0.0071052503772079945,0.007154526188969612,0.0070968070067465305,0.007085619959980249],"repeats":32,"seconds_per_sample":0.007172938159783371,"standard_deviation_convention":"population","standard_deviation_seconds":5.35804255483238e-05,"warmups":5}}

results/conditional/timing/lm1b/mdlm_di4c/steps_2/timing.json

@@ -0,0 +1 @@
+{"metadata":{"adapter_identity":"dlb.adapters.di4c:v3|dlb.adapters.di4c:Di4CAdapter|e1636b46628a6015944553de2d5d8a1b829bb6bf3f9c6b789d829cb6db222ec9","attempt_id":"455cf3bce92a4540b9c77e538cdf7056","checkpoint_lock_id":"recipe:mdlm_di4c_lm1b:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/reference_reproduction/mdlm_di4c/lm1b","recipe":"mdlm_di4c_lm1b","sampling_checkpoint":"student_checkpoints/20000.ckpt","sampling_config":"config.yaml","source":"di4c","source_commit":"ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3","teacher_adapter":"masked_to_absorbing","teacher_family":"masked_mdlm"},"checkpoint_sha256":"569faa9af7f866644c622bf3360a246d2ed3f3ddd7572999bdaa7c8a8d3d7f44","checkpoint_teacher_family":"masked_mdlm","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","cuda_runtime_version":"12.4","dataset":"lm1b","diversity_prompt_count":256,"environment":"dlb-di4c","evaluation_continuation_length":64,"generation_mode":"conditional_prefix","gpu_compute_capability":"9.0","gpu_index":0,"gpu_name":"NVIDIA H200 NVL","model":"mdlm_di4c","nvidia_driver_version":"575.57.08","parameter_precision":"fp32","precision":"bf16-mixed","precision_evidence":"static_policy_bound_to_checkpoint_and_runtime_code_not_runtime_autocast_observation","precision_policy":"di4c:pinned_bf16_mixed_config_and_internal_autocast","precision_policy_binding":{"adapter_identity":"dlb.adapters.di4c:v3|dlb.adapters.di4c:Di4CAdapter|e1636b46628a6015944553de2d5d8a1b829bb6bf3f9c6b789d829cb6db222ec9","checkpoint_lock_id":"recipe:mdlm_di4c_lm1b:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/reference_reproduction/mdlm_di4c/lm1b","recipe":"mdlm_di4c_lm1b","sampling_checkpoint":"student_checkpoints/20000.ckpt","sampling_config":"config.yaml","source":"di4c","source_commit":"ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3","teacher_adapter":"masked_to_absorbing","teacher_family":"masked_mdlm"},"checkpoint_sha256":"569faa9af7f866644c622bf3360a246d2ed3f3ddd7572999bdaa7c8a8d3d7f44","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","dataset":"lm1b","diversity_prompt_count":256,"evaluation_continuation_length":64,"experiment_config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","generation_mode":"conditional_prefix","model":"mdlm_di4c","prefix_length":64,"prompt_count":1024,"source_commit":"ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3","upstream":"di4c"},"prefix_length":64,"prompt_count":1024,"pytorch_compiled_cuda_toolkit":"12.4","requested_precision":"author","runtime_code_binding":[{"kind":"sampling_callable","path":"/home/ruizzhao/diffusion_baseline/adapters/_distilled_runtime.py","sha256":"39268b31d066526a2f0e9bb13bb531f98aaedb1aed46b4f4f0fd722cf02a5683","symbol":"benchmark_model.<locals>.<lambda>"},{"kind":"sampling_closure","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/conditional_runtime.py","sha256":"eef6d1a1062787f949be4656b7a1cc2e81718b42affd9facce4159e2c02c89b8","symbol":"token_project_fn.<locals>.project"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/core/diffusion/noise_schedule.py","sha256":"b2c06e11a4947b306db1c19eac181b99bbc6b36520526ebb36639696c95dcb21","symbol":"LogLinearNoise"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/core/distill/multi_round_sdtt.py","sha256":"75c61fbd1d3e1b87e10b25d67c0143ba0995753f579e0282bc1e3559c145704c","symbol":"MultiRoundSDTT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"DDiTBlock"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"DDitFinalLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"DIT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"EmbeddingLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"LayerNorm"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"Rotary"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"TimestepEmbedder"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"GELU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"SiLU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"ModuleList"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"Sequential"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/dropout.py","sha256":"1fc9974c3cf2078dd4bc257f73afb6a6714adfec2e9d099d47cf923e766d3490","symbol":"Dropout"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/linear.py","sha256":"d7f5cadf773deb5ab4bcc9f57414e8239b2859c70d2aa6f141576de1ec918823","symbol":"Linear"}],"sample_count":2048,"seed":42,"source_commit":"ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3","steps":2,"synchronization_policy":"torch.cuda.synchronize_before_start_and_after_generate"},"schema":"dlb-generation-timing-v1","timing":{"batch_size":1,"exclusions":["model_and_checkpoint_loading","first_compilation","token_decoding","metrics","file_io"],"median_seconds":0.04133602394722402,"mode":"primary_latency","num_timed_samples":32,"raw_durations_seconds":[0.0333600677549839,0.03812663070857525,0.04154093936085701,0.04143990436568856,0.041349414736032486,0.041345377918332815,0.041430819779634476,0.04141563596203923,0.04141434375196695,0.04129849700257182,0.04132666997611523,0.04141081916168332,0.022354363929480314,0.01717715198174119,0.01711285300552845,0.017191693652421236,0.016922950744628906,0.016813183669000864,0.017005898989737034,0.036843220703303814,0.0415173233486712,0.03916452266275883,0.04276465578004718,0.04138324689120054,0.04141151998192072,0.04144517192617059,0.04140912601724267,0.04057982424274087,0.038145650178194046,0.04148272890597582,0.04148790705949068,0.04132316401228309],"repeats":32,"seconds_per_sample":0.035593602442531846,"standard_deviation_convention":"population","standard_deviation_seconds":0.009616989485969377,"warmups":5}}

results/conditional/timing/lm1b/mdlm_di4c/steps_8/timing.json

@@ -0,0 +1 @@
+{"metadata":{"adapter_identity":"dlb.adapters.di4c:v3|dlb.adapters.di4c:Di4CAdapter|e1636b46628a6015944553de2d5d8a1b829bb6bf3f9c6b789d829cb6db222ec9","attempt_id":"60368859fb984d9ea0fb87b17c5b54fa","checkpoint_lock_id":"recipe:mdlm_di4c_lm1b:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/reference_reproduction/mdlm_di4c/lm1b","recipe":"mdlm_di4c_lm1b","sampling_checkpoint":"student_checkpoints/20000.ckpt","sampling_config":"config.yaml","source":"di4c","source_commit":"ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3","teacher_adapter":"masked_to_absorbing","teacher_family":"masked_mdlm"},"checkpoint_sha256":"569faa9af7f866644c622bf3360a246d2ed3f3ddd7572999bdaa7c8a8d3d7f44","checkpoint_teacher_family":"masked_mdlm","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","cuda_runtime_version":"12.4","dataset":"lm1b","diversity_prompt_count":256,"environment":"dlb-di4c","evaluation_continuation_length":64,"generation_mode":"conditional_prefix","gpu_compute_capability":"9.0","gpu_index":0,"gpu_name":"NVIDIA H200 NVL","model":"mdlm_di4c","nvidia_driver_version":"575.57.08","parameter_precision":"fp32","precision":"bf16-mixed","precision_evidence":"static_policy_bound_to_checkpoint_and_runtime_code_not_runtime_autocast_observation","precision_policy":"di4c:pinned_bf16_mixed_config_and_internal_autocast","precision_policy_binding":{"adapter_identity":"dlb.adapters.di4c:v3|dlb.adapters.di4c:Di4CAdapter|e1636b46628a6015944553de2d5d8a1b829bb6bf3f9c6b789d829cb6db222ec9","checkpoint_lock_id":"recipe:mdlm_di4c_lm1b:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/reference_reproduction/mdlm_di4c/lm1b","recipe":"mdlm_di4c_lm1b","sampling_checkpoint":"student_checkpoints/20000.ckpt","sampling_config":"config.yaml","source":"di4c","source_commit":"ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3","teacher_adapter":"masked_to_absorbing","teacher_family":"masked_mdlm"},"checkpoint_sha256":"569faa9af7f866644c622bf3360a246d2ed3f3ddd7572999bdaa7c8a8d3d7f44","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","dataset":"lm1b","diversity_prompt_count":256,"evaluation_continuation_length":64,"experiment_config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","generation_mode":"conditional_prefix","model":"mdlm_di4c","prefix_length":64,"prompt_count":1024,"source_commit":"ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3","upstream":"di4c"},"prefix_length":64,"prompt_count":1024,"pytorch_compiled_cuda_toolkit":"12.4","requested_precision":"author","runtime_code_binding":[{"kind":"sampling_callable","path":"/home/ruizzhao/diffusion_baseline/adapters/_distilled_runtime.py","sha256":"39268b31d066526a2f0e9bb13bb531f98aaedb1aed46b4f4f0fd722cf02a5683","symbol":"benchmark_model.<locals>.<lambda>"},{"kind":"sampling_closure","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/conditional_runtime.py","sha256":"eef6d1a1062787f949be4656b7a1cc2e81718b42affd9facce4159e2c02c89b8","symbol":"token_project_fn.<locals>.project"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/core/diffusion/noise_schedule.py","sha256":"b2c06e11a4947b306db1c19eac181b99bbc6b36520526ebb36639696c95dcb21","symbol":"LogLinearNoise"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/core/distill/multi_round_sdtt.py","sha256":"75c61fbd1d3e1b87e10b25d67c0143ba0995753f579e0282bc1e3559c145704c","symbol":"MultiRoundSDTT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"DDiTBlock"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"DDitFinalLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"DIT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"EmbeddingLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"LayerNorm"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"Rotary"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/di4c/sdtt/src/sdtt/models/dit_orig.py","sha256":"c2e7ced4ccbb92bcf9e6fdc0fc2b41130447d8d15e5d696e7f3a927177578912","symbol":"TimestepEmbedder"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"GELU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"SiLU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"ModuleList"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"Sequential"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/dropout.py","sha256":"1fc9974c3cf2078dd4bc257f73afb6a6714adfec2e9d099d47cf923e766d3490","symbol":"Dropout"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-di4c/lib/python3.10/site-packages/torch/nn/modules/linear.py","sha256":"d7f5cadf773deb5ab4bcc9f57414e8239b2859c70d2aa6f141576de1ec918823","symbol":"Linear"}],"sample_count":2048,"seed":42,"source_commit":"ac61ff9fe8e85120f9e1d2a8c5a332f8b8353dd3","steps":8,"synchronization_policy":"torch.cuda.synchronize_before_start_and_after_generate"},"schema":"dlb-generation-timing-v1","timing":{"batch_size":1,"exclusions":["model_and_checkpoint_loading","first_compilation","token_decoding","metrics","file_io"],"median_seconds":0.14962840685620904,"mode":"primary_latency","num_timed_samples":32,"raw_durations_seconds":[0.15086489310488105,0.1508006751537323,0.15089436899870634,0.15082118567079306,0.12087079975754023,0.06817262480035424,0.10907449293881655,0.1431314698420465,0.15104216616600752,0.15130962617695332,0.15341841103509068,0.1523319249972701,0.0792751768603921,0.06813657889142632,0.13305954914540052,0.14887798577547073,0.15037882793694735,0.15086302114650607,0.1509096031077206,0.1309251538477838,0.06773078674450517,0.09734546905383468,0.14515784895047545,0.15103283198550344,0.1513690766878426,0.15085415821522474,0.15067644510418177,0.09151683701202273,0.06737793516367674,0.12767956499010324,0.14521750016137958,0.1506821527145803],"repeats":32,"seconds_per_sample":0.13005622319178656,"standard_deviation_convention":"population","standard_deviation_seconds":0.030236000252981617,"warmups":5}}

results/conditional/timing/lm1b/mdlm_sdtt/steps_1/timing.json

@@ -0,0 +1 @@
+{"metadata":{"adapter_identity":"dlb.adapters.sdtt:v3|dlb.adapters.sdtt:SDTTAdapter|d4144d2cac65e5b953ee3a8a8f9451655ca58e442b3ef78210938af9899e1009","attempt_id":"0eb7e61750d54638a3a0b9df00f90561","checkpoint_lock_id":"recipe:mdlm_sdtt_lm1b:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/reference_reproduction/mdlm_sdtt/lm1b","recipe":"mdlm_sdtt_lm1b","sampling_checkpoint":"student_checkpoints/70000.ckpt","sampling_config":"config.yaml","source":"sdtt","source_commit":"1150985e90b8f2d5749e4469d5154eff9ec922c4","teacher_adapter":"masked_to_absorbing","teacher_family":"masked_mdlm"},"checkpoint_sha256":"1a1a7a258109c94982f7f45edcc70feb0cc33204f4a48cc40e92809761d5c7a0","checkpoint_teacher_family":"masked_mdlm","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","cuda_runtime_version":"12.4","dataset":"lm1b","diversity_prompt_count":256,"environment":"dlb-sdtt","evaluation_continuation_length":64,"generation_mode":"conditional_prefix","gpu_compute_capability":"9.0","gpu_index":0,"gpu_name":"NVIDIA H200 NVL","model":"mdlm_sdtt","nvidia_driver_version":"575.57.08","parameter_precision":"fp32","precision":"bf16-mixed","precision_evidence":"static_policy_bound_to_checkpoint_and_runtime_code_not_runtime_autocast_observation","precision_policy":"sdtt:pinned_bf16_mixed_config_and_internal_autocast","precision_policy_binding":{"adapter_identity":"dlb.adapters.sdtt:v3|dlb.adapters.sdtt:SDTTAdapter|d4144d2cac65e5b953ee3a8a8f9451655ca58e442b3ef78210938af9899e1009","checkpoint_lock_id":"recipe:mdlm_sdtt_lm1b:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/reference_reproduction/mdlm_sdtt/lm1b","recipe":"mdlm_sdtt_lm1b","sampling_checkpoint":"student_checkpoints/70000.ckpt","sampling_config":"config.yaml","source":"sdtt","source_commit":"1150985e90b8f2d5749e4469d5154eff9ec922c4","teacher_adapter":"masked_to_absorbing","teacher_family":"masked_mdlm"},"checkpoint_sha256":"1a1a7a258109c94982f7f45edcc70feb0cc33204f4a48cc40e92809761d5c7a0","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","dataset":"lm1b","diversity_prompt_count":256,"evaluation_continuation_length":64,"experiment_config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","generation_mode":"conditional_prefix","model":"mdlm_sdtt","prefix_length":64,"prompt_count":1024,"source_commit":"1150985e90b8f2d5749e4469d5154eff9ec922c4","upstream":"sdtt"},"prefix_length":64,"prompt_count":1024,"pytorch_compiled_cuda_toolkit":"12.4","requested_precision":"author","runtime_code_binding":[{"kind":"sampling_callable","path":"/home/ruizzhao/diffusion_baseline/adapters/_distilled_runtime.py","sha256":"39268b31d066526a2f0e9bb13bb531f98aaedb1aed46b4f4f0fd722cf02a5683","symbol":"benchmark_model.<locals>.<lambda>"},{"kind":"sampling_closure","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/conditional_runtime.py","sha256":"41190be05ea8be5968c41f9f7fc40e3d6c087b93b9908839f042df1b6d9a402b","symbol":"token_project_fn.<locals>.project"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/core/diffusion/noise_schedule.py","sha256":"b2c06e11a4947b306db1c19eac181b99bbc6b36520526ebb36639696c95dcb21","symbol":"LogLinearNoise"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/core/distill/multi_round_sdtt.py","sha256":"879e46ab826d70b5cc9d420c890df2dec67db1cca8f204881870ed5581ba266e","symbol":"MultiRoundSDTT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/models/dit_orig.py","sha256":"8f140471181a002cee9f8639e40006709364957cdade6f8f3c69ffc8a60a5e07","symbol":"DDiTBlock"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/models/dit_orig.py","sha256":"8f140471181a002cee9f8639e40006709364957cdade6f8f3c69ffc8a60a5e07","symbol":"DDitFinalLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/models/dit_orig.py","sha256":"8f140471181a002cee9f8639e40006709364957cdade6f8f3c69ffc8a60a5e07","symbol":"DIT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/models/dit_orig.py","sha256":"8f140471181a002cee9f8639e40006709364957cdade6f8f3c69ffc8a60a5e07","symbol":"EmbeddingLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/models/dit_orig.py","sha256":"8f140471181a002cee9f8639e40006709364957cdade6f8f3c69ffc8a60a5e07","symbol":"LayerNorm"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/models/dit_orig.py","sha256":"8f140471181a002cee9f8639e40006709364957cdade6f8f3c69ffc8a60a5e07","symbol":"Rotary"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/models/dit_orig.py","sha256":"8f140471181a002cee9f8639e40006709364957cdade6f8f3c69ffc8a60a5e07","symbol":"TimestepEmbedder"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-sdtt/lib/python3.10/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"GELU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-sdtt/lib/python3.10/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"SiLU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-sdtt/lib/python3.10/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"ModuleList"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-sdtt/lib/python3.10/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"Sequential"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-sdtt/lib/python3.10/site-packages/torch/nn/modules/dropout.py","sha256":"1fc9974c3cf2078dd4bc257f73afb6a6714adfec2e9d099d47cf923e766d3490","symbol":"Dropout"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-sdtt/lib/python3.10/site-packages/torch/nn/modules/linear.py","sha256":"d7f5cadf773deb5ab4bcc9f57414e8239b2859c70d2aa6f141576de1ec918823","symbol":"Linear"}],"sample_count":2048,"seed":42,"source_commit":"1150985e90b8f2d5749e4469d5154eff9ec922c4","steps":1,"synchronization_policy":"torch.cuda.synchronize_before_start_and_after_generate"},"schema":"dlb-generation-timing-v1","timing":{"batch_size":1,"exclusions":["model_and_checkpoint_loading","first_compilation","token_decoding","metrics","file_io"],"median_seconds":0.010955392150208354,"mode":"primary_latency","num_timed_samples":32,"raw_durations_seconds":[0.011056781746447086,0.011164287105202675,0.010855883825570345,0.011159810237586498,0.010968056041747332,0.010969537775963545,0.01096745627000928,0.010965472087264061,0.011021648067981005,0.010784745216369629,0.011125667952001095,0.010962668340653181,0.010985002852976322,0.011277670972049236,0.010782862082123756,0.010776433162391186,0.010773456655442715,0.010849544312804937,0.010809944011271,0.011162985116243362,0.010948115959763527,0.01094531174749136,0.010918690823018551,0.010917038191109896,0.011153440922498703,0.010760799050331116,0.010785515885800123,0.010765745770186186,0.010813017841428518,0.011157577391713858,0.010910707991570234,0.010979153215885162],"repeats":32,"seconds_per_sample":0.010952344644465484,"standard_deviation_convention":"population","standard_deviation_seconds":0.0001427582480093406,"warmups":5}}

results/conditional/timing/lm1b/mdlm_sdtt/steps_32/timing.json

@@ -0,0 +1 @@
+{"metadata":{"adapter_identity":"dlb.adapters.sdtt:v3|dlb.adapters.sdtt:SDTTAdapter|d4144d2cac65e5b953ee3a8a8f9451655ca58e442b3ef78210938af9899e1009","attempt_id":"94d9a9813c1048a48645f9509c90118d","checkpoint_lock_id":"recipe:mdlm_sdtt_lm1b:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/reference_reproduction/mdlm_sdtt/lm1b","recipe":"mdlm_sdtt_lm1b","sampling_checkpoint":"student_checkpoints/70000.ckpt","sampling_config":"config.yaml","source":"sdtt","source_commit":"1150985e90b8f2d5749e4469d5154eff9ec922c4","teacher_adapter":"masked_to_absorbing","teacher_family":"masked_mdlm"},"checkpoint_sha256":"1a1a7a258109c94982f7f45edcc70feb0cc33204f4a48cc40e92809761d5c7a0","checkpoint_teacher_family":"masked_mdlm","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","cuda_runtime_version":"12.4","dataset":"lm1b","diversity_prompt_count":256,"environment":"dlb-sdtt","evaluation_continuation_length":64,"generation_mode":"conditional_prefix","gpu_compute_capability":"9.0","gpu_index":0,"gpu_name":"NVIDIA H200 NVL","model":"mdlm_sdtt","nvidia_driver_version":"575.57.08","parameter_precision":"fp32","precision":"bf16-mixed","precision_evidence":"static_policy_bound_to_checkpoint_and_runtime_code_not_runtime_autocast_observation","precision_policy":"sdtt:pinned_bf16_mixed_config_and_internal_autocast","precision_policy_binding":{"adapter_identity":"dlb.adapters.sdtt:v3|dlb.adapters.sdtt:SDTTAdapter|d4144d2cac65e5b953ee3a8a8f9451655ca58e442b3ef78210938af9899e1009","checkpoint_lock_id":"recipe:mdlm_sdtt_lm1b:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/reference_reproduction/mdlm_sdtt/lm1b","recipe":"mdlm_sdtt_lm1b","sampling_checkpoint":"student_checkpoints/70000.ckpt","sampling_config":"config.yaml","source":"sdtt","source_commit":"1150985e90b8f2d5749e4469d5154eff9ec922c4","teacher_adapter":"masked_to_absorbing","teacher_family":"masked_mdlm"},"checkpoint_sha256":"1a1a7a258109c94982f7f45edcc70feb0cc33204f4a48cc40e92809761d5c7a0","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","dataset":"lm1b","diversity_prompt_count":256,"evaluation_continuation_length":64,"experiment_config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","generation_mode":"conditional_prefix","model":"mdlm_sdtt","prefix_length":64,"prompt_count":1024,"source_commit":"1150985e90b8f2d5749e4469d5154eff9ec922c4","upstream":"sdtt"},"prefix_length":64,"prompt_count":1024,"pytorch_compiled_cuda_toolkit":"12.4","requested_precision":"author","runtime_code_binding":[{"kind":"sampling_callable","path":"/home/ruizzhao/diffusion_baseline/adapters/_distilled_runtime.py","sha256":"39268b31d066526a2f0e9bb13bb531f98aaedb1aed46b4f4f0fd722cf02a5683","symbol":"benchmark_model.<locals>.<lambda>"},{"kind":"sampling_closure","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/conditional_runtime.py","sha256":"41190be05ea8be5968c41f9f7fc40e3d6c087b93b9908839f042df1b6d9a402b","symbol":"token_project_fn.<locals>.project"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/core/diffusion/noise_schedule.py","sha256":"b2c06e11a4947b306db1c19eac181b99bbc6b36520526ebb36639696c95dcb21","symbol":"LogLinearNoise"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/core/distill/multi_round_sdtt.py","sha256":"879e46ab826d70b5cc9d420c890df2dec67db1cca8f204881870ed5581ba266e","symbol":"MultiRoundSDTT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/models/dit_orig.py","sha256":"8f140471181a002cee9f8639e40006709364957cdade6f8f3c69ffc8a60a5e07","symbol":"DDiTBlock"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/models/dit_orig.py","sha256":"8f140471181a002cee9f8639e40006709364957cdade6f8f3c69ffc8a60a5e07","symbol":"DDitFinalLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/models/dit_orig.py","sha256":"8f140471181a002cee9f8639e40006709364957cdade6f8f3c69ffc8a60a5e07","symbol":"DIT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/models/dit_orig.py","sha256":"8f140471181a002cee9f8639e40006709364957cdade6f8f3c69ffc8a60a5e07","symbol":"EmbeddingLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/models/dit_orig.py","sha256":"8f140471181a002cee9f8639e40006709364957cdade6f8f3c69ffc8a60a5e07","symbol":"LayerNorm"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/models/dit_orig.py","sha256":"8f140471181a002cee9f8639e40006709364957cdade6f8f3c69ffc8a60a5e07","symbol":"Rotary"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/models/dit_orig.py","sha256":"8f140471181a002cee9f8639e40006709364957cdade6f8f3c69ffc8a60a5e07","symbol":"TimestepEmbedder"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-sdtt/lib/python3.10/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"GELU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-sdtt/lib/python3.10/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"SiLU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-sdtt/lib/python3.10/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"ModuleList"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-sdtt/lib/python3.10/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"Sequential"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-sdtt/lib/python3.10/site-packages/torch/nn/modules/dropout.py","sha256":"1fc9974c3cf2078dd4bc257f73afb6a6714adfec2e9d099d47cf923e766d3490","symbol":"Dropout"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-sdtt/lib/python3.10/site-packages/torch/nn/modules/linear.py","sha256":"d7f5cadf773deb5ab4bcc9f57414e8239b2859c70d2aa6f141576de1ec918823","symbol":"Linear"}],"sample_count":2048,"seed":42,"source_commit":"1150985e90b8f2d5749e4469d5154eff9ec922c4","steps":32,"synchronization_policy":"torch.cuda.synchronize_before_start_and_after_generate"},"schema":"dlb-generation-timing-v1","timing":{"batch_size":1,"exclusions":["model_and_checkpoint_loading","first_compilation","token_decoding","metrics","file_io"],"median_seconds":0.23672646563500166,"mode":"primary_latency","num_timed_samples":32,"raw_durations_seconds":[0.24717091768980026,0.2532607610337436,0.2147402260452509,0.2223890800960362,0.26066140411421657,0.22783574601635337,0.21436337660998106,0.2630653018131852,0.2335263560526073,0.21486339392140508,0.2538208202458918,0.26108289090916514,0.24654149310663342,0.21461068699136376,0.23646275093778968,0.259780652821064,0.2598665147088468,0.26178938522934914,0.260384239256382,0.23297512019053102,0.2146293157711625,0.24744744319468737,0.23587520979344845,0.21227786783128977,0.24616327183321118,0.23356198007240891,0.2121676690876484,0.2418259889818728,0.23637903202325106,0.21185001218691468,0.24490427086129785,0.23699018033221364],"repeats":32,"seconds_per_sample":0.23791447999246884,"standard_deviation_convention":"population","standard_deviation_seconds":0.017424741506248138,"warmups":5}}

results/conditional/timing/lm1b/mdlm_sdtt/steps_4/timing.json

@@ -0,0 +1 @@
+{"metadata":{"adapter_identity":"dlb.adapters.sdtt:v3|dlb.adapters.sdtt:SDTTAdapter|d4144d2cac65e5b953ee3a8a8f9451655ca58e442b3ef78210938af9899e1009","attempt_id":"f11de2c111f14652a62d10578dd801a2","checkpoint_lock_id":"recipe:mdlm_sdtt_lm1b:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/reference_reproduction/mdlm_sdtt/lm1b","recipe":"mdlm_sdtt_lm1b","sampling_checkpoint":"student_checkpoints/70000.ckpt","sampling_config":"config.yaml","source":"sdtt","source_commit":"1150985e90b8f2d5749e4469d5154eff9ec922c4","teacher_adapter":"masked_to_absorbing","teacher_family":"masked_mdlm"},"checkpoint_sha256":"1a1a7a258109c94982f7f45edcc70feb0cc33204f4a48cc40e92809761d5c7a0","checkpoint_teacher_family":"masked_mdlm","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","cuda_runtime_version":"12.4","dataset":"lm1b","diversity_prompt_count":256,"environment":"dlb-sdtt","evaluation_continuation_length":64,"generation_mode":"conditional_prefix","gpu_compute_capability":"9.0","gpu_index":0,"gpu_name":"NVIDIA H200 NVL","model":"mdlm_sdtt","nvidia_driver_version":"575.57.08","parameter_precision":"fp32","precision":"bf16-mixed","precision_evidence":"static_policy_bound_to_checkpoint_and_runtime_code_not_runtime_autocast_observation","precision_policy":"sdtt:pinned_bf16_mixed_config_and_internal_autocast","precision_policy_binding":{"adapter_identity":"dlb.adapters.sdtt:v3|dlb.adapters.sdtt:SDTTAdapter|d4144d2cac65e5b953ee3a8a8f9451655ca58e442b3ef78210938af9899e1009","checkpoint_lock_id":"recipe:mdlm_sdtt_lm1b:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/reference_reproduction/mdlm_sdtt/lm1b","recipe":"mdlm_sdtt_lm1b","sampling_checkpoint":"student_checkpoints/70000.ckpt","sampling_config":"config.yaml","source":"sdtt","source_commit":"1150985e90b8f2d5749e4469d5154eff9ec922c4","teacher_adapter":"masked_to_absorbing","teacher_family":"masked_mdlm"},"checkpoint_sha256":"1a1a7a258109c94982f7f45edcc70feb0cc33204f4a48cc40e92809761d5c7a0","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","dataset":"lm1b","diversity_prompt_count":256,"evaluation_continuation_length":64,"experiment_config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","generation_mode":"conditional_prefix","model":"mdlm_sdtt","prefix_length":64,"prompt_count":1024,"source_commit":"1150985e90b8f2d5749e4469d5154eff9ec922c4","upstream":"sdtt"},"prefix_length":64,"prompt_count":1024,"pytorch_compiled_cuda_toolkit":"12.4","requested_precision":"author","runtime_code_binding":[{"kind":"sampling_callable","path":"/home/ruizzhao/diffusion_baseline/adapters/_distilled_runtime.py","sha256":"39268b31d066526a2f0e9bb13bb531f98aaedb1aed46b4f4f0fd722cf02a5683","symbol":"benchmark_model.<locals>.<lambda>"},{"kind":"sampling_closure","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/conditional_runtime.py","sha256":"41190be05ea8be5968c41f9f7fc40e3d6c087b93b9908839f042df1b6d9a402b","symbol":"token_project_fn.<locals>.project"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/core/diffusion/noise_schedule.py","sha256":"b2c06e11a4947b306db1c19eac181b99bbc6b36520526ebb36639696c95dcb21","symbol":"LogLinearNoise"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/core/distill/multi_round_sdtt.py","sha256":"879e46ab826d70b5cc9d420c890df2dec67db1cca8f204881870ed5581ba266e","symbol":"MultiRoundSDTT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/models/dit_orig.py","sha256":"8f140471181a002cee9f8639e40006709364957cdade6f8f3c69ffc8a60a5e07","symbol":"DDiTBlock"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/models/dit_orig.py","sha256":"8f140471181a002cee9f8639e40006709364957cdade6f8f3c69ffc8a60a5e07","symbol":"DDitFinalLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/models/dit_orig.py","sha256":"8f140471181a002cee9f8639e40006709364957cdade6f8f3c69ffc8a60a5e07","symbol":"DIT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/models/dit_orig.py","sha256":"8f140471181a002cee9f8639e40006709364957cdade6f8f3c69ffc8a60a5e07","symbol":"EmbeddingLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/models/dit_orig.py","sha256":"8f140471181a002cee9f8639e40006709364957cdade6f8f3c69ffc8a60a5e07","symbol":"LayerNorm"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/models/dit_orig.py","sha256":"8f140471181a002cee9f8639e40006709364957cdade6f8f3c69ffc8a60a5e07","symbol":"Rotary"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/models/dit_orig.py","sha256":"8f140471181a002cee9f8639e40006709364957cdade6f8f3c69ffc8a60a5e07","symbol":"TimestepEmbedder"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-sdtt/lib/python3.10/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"GELU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-sdtt/lib/python3.10/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"SiLU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-sdtt/lib/python3.10/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"ModuleList"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-sdtt/lib/python3.10/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"Sequential"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-sdtt/lib/python3.10/site-packages/torch/nn/modules/dropout.py","sha256":"1fc9974c3cf2078dd4bc257f73afb6a6714adfec2e9d099d47cf923e766d3490","symbol":"Dropout"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-sdtt/lib/python3.10/site-packages/torch/nn/modules/linear.py","sha256":"d7f5cadf773deb5ab4bcc9f57414e8239b2859c70d2aa6f141576de1ec918823","symbol":"Linear"}],"sample_count":2048,"seed":42,"source_commit":"1150985e90b8f2d5749e4469d5154eff9ec922c4","steps":4,"synchronization_policy":"torch.cuda.synchronize_before_start_and_after_generate"},"schema":"dlb-generation-timing-v1","timing":{"batch_size":1,"exclusions":["model_and_checkpoint_loading","first_compilation","token_decoding","metrics","file_io"],"median_seconds":0.027368255890905857,"mode":"primary_latency","num_timed_samples":32,"raw_durations_seconds":[0.02724753087386489,0.027154087089002132,0.028145138174295425,0.04438226716592908,0.04537599207833409,0.04473684076219797,0.04455847758799791,0.04430241463705897,0.044202601071447134,0.04408028395846486,0.027488980907946825,0.02688243007287383,0.02674698317423463,0.026777029037475586,0.026833625975996256,0.026797129772603512,0.026767485309392214,0.026834918186068535,0.026759200729429722,0.026845694053918123,0.026825653854757547,0.02675496507436037,0.02681747078895569,0.026691037230193615,0.04413607995957136,0.04470649501308799,0.04473383678123355,0.055190909188240767,0.0438050520606339,0.04384199809283018,0.04164831293746829,0.02711818227544427],"repeats":32,"seconds_per_sample":0.03484965949610341,"standard_deviation_convention":"population","standard_deviation_seconds":0.009146808820655052,"warmups":5}}

results/conditional/timing/lm1b/mdlm_sdtt/steps_8/timing.json

@@ -0,0 +1 @@
+{"metadata":{"adapter_identity":"dlb.adapters.sdtt:v3|dlb.adapters.sdtt:SDTTAdapter|d4144d2cac65e5b953ee3a8a8f9451655ca58e442b3ef78210938af9899e1009","attempt_id":"4358b6f2a79842ae8aea2c1a8d165011","checkpoint_lock_id":"recipe:mdlm_sdtt_lm1b:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/reference_reproduction/mdlm_sdtt/lm1b","recipe":"mdlm_sdtt_lm1b","sampling_checkpoint":"student_checkpoints/70000.ckpt","sampling_config":"config.yaml","source":"sdtt","source_commit":"1150985e90b8f2d5749e4469d5154eff9ec922c4","teacher_adapter":"masked_to_absorbing","teacher_family":"masked_mdlm"},"checkpoint_sha256":"1a1a7a258109c94982f7f45edcc70feb0cc33204f4a48cc40e92809761d5c7a0","checkpoint_teacher_family":"masked_mdlm","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","cuda_runtime_version":"12.4","dataset":"lm1b","diversity_prompt_count":256,"environment":"dlb-sdtt","evaluation_continuation_length":64,"generation_mode":"conditional_prefix","gpu_compute_capability":"9.0","gpu_index":0,"gpu_name":"NVIDIA H200 NVL","model":"mdlm_sdtt","nvidia_driver_version":"575.57.08","parameter_precision":"fp32","precision":"bf16-mixed","precision_evidence":"static_policy_bound_to_checkpoint_and_runtime_code_not_runtime_autocast_observation","precision_policy":"sdtt:pinned_bf16_mixed_config_and_internal_autocast","precision_policy_binding":{"adapter_identity":"dlb.adapters.sdtt:v3|dlb.adapters.sdtt:SDTTAdapter|d4144d2cac65e5b953ee3a8a8f9451655ca58e442b3ef78210938af9899e1009","checkpoint_lock_id":"recipe:mdlm_sdtt_lm1b:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152","checkpoint_selection":{"output":"checkpoints/reference_reproduction/mdlm_sdtt/lm1b","recipe":"mdlm_sdtt_lm1b","sampling_checkpoint":"student_checkpoints/70000.ckpt","sampling_config":"config.yaml","source":"sdtt","source_commit":"1150985e90b8f2d5749e4469d5154eff9ec922c4","teacher_adapter":"masked_to_absorbing","teacher_family":"masked_mdlm"},"checkpoint_sha256":"1a1a7a258109c94982f7f45edcc70feb0cc33204f4a48cc40e92809761d5c7a0","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","dataset":"lm1b","diversity_prompt_count":256,"evaluation_continuation_length":64,"experiment_config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","generation_mode":"conditional_prefix","model":"mdlm_sdtt","prefix_length":64,"prompt_count":1024,"source_commit":"1150985e90b8f2d5749e4469d5154eff9ec922c4","upstream":"sdtt"},"prefix_length":64,"prompt_count":1024,"pytorch_compiled_cuda_toolkit":"12.4","requested_precision":"author","runtime_code_binding":[{"kind":"sampling_callable","path":"/home/ruizzhao/diffusion_baseline/adapters/_distilled_runtime.py","sha256":"39268b31d066526a2f0e9bb13bb531f98aaedb1aed46b4f4f0fd722cf02a5683","symbol":"benchmark_model.<locals>.<lambda>"},{"kind":"sampling_closure","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/conditional_runtime.py","sha256":"41190be05ea8be5968c41f9f7fc40e3d6c087b93b9908839f042df1b6d9a402b","symbol":"token_project_fn.<locals>.project"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/core/diffusion/noise_schedule.py","sha256":"b2c06e11a4947b306db1c19eac181b99bbc6b36520526ebb36639696c95dcb21","symbol":"LogLinearNoise"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/core/distill/multi_round_sdtt.py","sha256":"879e46ab826d70b5cc9d420c890df2dec67db1cca8f204881870ed5581ba266e","symbol":"MultiRoundSDTT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/models/dit_orig.py","sha256":"8f140471181a002cee9f8639e40006709364957cdade6f8f3c69ffc8a60a5e07","symbol":"DDiTBlock"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/models/dit_orig.py","sha256":"8f140471181a002cee9f8639e40006709364957cdade6f8f3c69ffc8a60a5e07","symbol":"DDitFinalLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/models/dit_orig.py","sha256":"8f140471181a002cee9f8639e40006709364957cdade6f8f3c69ffc8a60a5e07","symbol":"DIT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/models/dit_orig.py","sha256":"8f140471181a002cee9f8639e40006709364957cdade6f8f3c69ffc8a60a5e07","symbol":"EmbeddingLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/models/dit_orig.py","sha256":"8f140471181a002cee9f8639e40006709364957cdade6f8f3c69ffc8a60a5e07","symbol":"LayerNorm"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/models/dit_orig.py","sha256":"8f140471181a002cee9f8639e40006709364957cdade6f8f3c69ffc8a60a5e07","symbol":"Rotary"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/sdtt/src/sdtt/models/dit_orig.py","sha256":"8f140471181a002cee9f8639e40006709364957cdade6f8f3c69ffc8a60a5e07","symbol":"TimestepEmbedder"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-sdtt/lib/python3.10/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"GELU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-sdtt/lib/python3.10/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"SiLU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-sdtt/lib/python3.10/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"ModuleList"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-sdtt/lib/python3.10/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"Sequential"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-sdtt/lib/python3.10/site-packages/torch/nn/modules/dropout.py","sha256":"1fc9974c3cf2078dd4bc257f73afb6a6714adfec2e9d099d47cf923e766d3490","symbol":"Dropout"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-sdtt/lib/python3.10/site-packages/torch/nn/modules/linear.py","sha256":"d7f5cadf773deb5ab4bcc9f57414e8239b2859c70d2aa6f141576de1ec918823","symbol":"Linear"}],"sample_count":2048,"seed":42,"source_commit":"1150985e90b8f2d5749e4469d5154eff9ec922c4","steps":8,"synchronization_policy":"torch.cuda.synchronize_before_start_and_after_generate"},"schema":"dlb-generation-timing-v1","timing":{"batch_size":1,"exclusions":["model_and_checkpoint_loading","first_compilation","token_decoding","metrics","file_io"],"median_seconds":0.08791635348461568,"mode":"primary_latency","num_timed_samples":32,"raw_durations_seconds":[0.08803247194737196,0.08848789101466537,0.08868449181318283,0.057188794016838074,0.055363813880831,0.05497903423383832,0.05480921408161521,0.05476147076115012,0.05437490716576576,0.05478890240192413,0.08780023502185941,0.0881121838465333,0.08848389517515898,0.0739459814503789,0.054789504036307335,0.05468807788565755,0.05459192022681236,0.05791018018499017,0.05486958613619208,0.05471891490742564,0.07553209504112601,0.0889464640058577,0.08928536390885711,0.08859483385458589,0.08855027612298727,0.08880364522337914,0.08899381710216403,0.0888224639929831,0.09298731340095401,0.08898165682330728,0.08942262502387166,0.08930433308705688],"repeats":32,"seconds_per_sample":0.07436269868048839,"standard_deviation_convention":"population","standard_deviation_seconds":0.01622311604022807,"warmups":5}}

results/conditional/timing/lm1b/rdlm/steps_1000/.timing.json.007fab9a2054482e8c95ae127aab11f7.superseded.json

@@ -0,0 +1 @@
+{"metadata":{"adapter_identity":"dlb.adapters.rdlm:v1|dlb.adapters.rdlm:RDLMAdapter|4cd983ae0746373fcae3d6cc0698aa19a49f61d2cf4eb395289b425aef2bab3a","attempt_id":"40f1b2545fb645eda8c6f6f87a16c0a1","checkpoint_lock_id":"rdlm_lm1b_drive:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152:all","checkpoint_selection":{"path":null,"resource":"rdlm_lm1b_drive","teacher_family":null},"checkpoint_sha256":"dc37df729f77d369cf15bf7ee587437f9a1b877aa8dd4da08ac732c9956b282b","checkpoint_teacher_family":"continuous_rdlm","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","cuda_runtime_version":"12.1","dataset":"lm1b","diversity_prompt_count":256,"environment":"dlb-rdlm","evaluation_continuation_length":64,"generation_mode":"conditional_prefix","gpu_compute_capability":"9.0","gpu_index":0,"gpu_name":"NVIDIA H200 NVL","model":"rdlm","nvidia_driver_version":"575.57.08","parameter_precision":"fp32","precision":"bf16-mixed","precision_evidence":"static_policy_bound_to_checkpoint_and_runtime_code_not_runtime_autocast_observation","precision_policy":"rdlm:pinned_internal_bf16_autocast_with_fp32_sensitive_ops","precision_policy_binding":{"adapter_identity":"dlb.adapters.rdlm:v1|dlb.adapters.rdlm:RDLMAdapter|4cd983ae0746373fcae3d6cc0698aa19a49f61d2cf4eb395289b425aef2bab3a","checkpoint_lock_id":"rdlm_lm1b_drive:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152:all","checkpoint_repository":null,"checkpoint_resource":"rdlm_lm1b_drive","checkpoint_revision":null,"checkpoint_selection":{"path":null,"resource":"rdlm_lm1b_drive","teacher_family":null},"checkpoint_sha256":"dc37df729f77d369cf15bf7ee587437f9a1b877aa8dd4da08ac732c9956b282b","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","dataset":"lm1b","diversity_prompt_count":256,"evaluation_continuation_length":64,"experiment_config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","generation_mode":"conditional_prefix","model":"rdlm","prefix_length":64,"prompt_count":1024,"source_commit":"67443aa6a2d0fa981eb7c6105f9cbc563e59c5c1","upstream":"rdlm"},"prefix_length":64,"prompt_count":1024,"pytorch_compiled_cuda_toolkit":"12.1","requested_precision":"author","runtime_code_binding":[{"kind":"sampling_closure","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/capture.py","sha256":"92c69119e85195e099a41668e8f75dd5a908c4e341fb220f714853db555dca0d","symbol":"_capture_rdlm.<locals>.sampling_factory.<locals>.project"},{"kind":"sampling_callable","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/capture.py","sha256":"92c69119e85195e099a41668e8f75dd5a908c4e341fb220f714853db555dca0d","symbol":"_rdlm_benchmark_sampling_fn.<locals>.sample.<locals>.<lambda>"},{"kind":"sampling_closure","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/conditional_runtime.py","sha256":"41190be05ea8be5968c41f9f7fc40e3d6c087b93b9908839f042df1b6d9a402b","symbol":"rdlm_project_fn.<locals>.project"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/rdlm/model/dit.py","sha256":"6b59f156ce1ba2897f7ace4601f550bff2cf738dc68458e67efd630d46849d50","symbol":"DiT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/rdlm/model/dit.py","sha256":"6b59f156ce1ba2897f7ace4601f550bff2cf738dc68458e67efd630d46849d50","symbol":"DiTBlock"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/rdlm/model/dit.py","sha256":"6b59f156ce1ba2897f7ace4601f550bff2cf738dc68458e67efd630d46849d50","symbol":"DiTFinalLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/rdlm/model/dit.py","sha256":"6b59f156ce1ba2897f7ace4601f550bff2cf738dc68458e67efd630d46849d50","symbol":"EmbeddingLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/rdlm/model/dit.py","sha256":"6b59f156ce1ba2897f7ace4601f550bff2cf738dc68458e67efd630d46849d50","symbol":"LayerNorm"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/rdlm/model/dit.py","sha256":"6b59f156ce1ba2897f7ace4601f550bff2cf738dc68458e67efd630d46849d50","symbol":"TimestepEmbedder"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/rdlm/model/rotary.py","sha256":"0f65c0f8e306fa29f152f040c6d70ac641eeaeae5b48757df8a39885ede422fd","symbol":"Rotary"},{"kind":"sampling_closure","path":"/home/ruizzhao/diffusion_baseline/upstreams/rdlm/sampling.py","sha256":"287d370c55eb29f562279f96aae0b759d6deadd435a1324f4124b9849acfd359","symbol":"get_sde_sampler.<locals>.pc_sampler"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-rdlm/lib/python3.9/site-packages/torch/_dynamo/eval_frame.py","sha256":"96a28ad58cbfc01005269673e0a98665319dccd593d3182b862537dce7c2d7d7","symbol":"OptimizedModule"},{"kind":"bound_method_owner","path":"/opt/anaconda3/envs/dlb-rdlm/lib/python3.9/site-packages/torch/autograd/grad_mode.py","sha256":"4f5bc6ae9d40f2c4777dd969fe6397f921fb648f43326c2147ce45d73be5ec6f","symbol":"no_grad"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-rdlm/lib/python3.9/site-packages/torch/nn/modules/activation.py","sha256":"90a597b30623ca02e4d165da6202898bb324279d9c2eebd016188210118702a1","symbol":"GELU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-rdlm/lib/python3.9/site-packages/torch/nn/modules/activation.py","sha256":"90a597b30623ca02e4d165da6202898bb324279d9c2eebd016188210118702a1","symbol":"SiLU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-rdlm/lib/python3.9/site-packages/torch/nn/modules/container.py","sha256":"8e2a1829811af9845cb8bf9200ce66ebcac91efc7d6ba5d78b21cc70b34540f6","symbol":"ModuleList"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-rdlm/lib/python3.9/site-packages/torch/nn/modules/container.py","sha256":"8e2a1829811af9845cb8bf9200ce66ebcac91efc7d6ba5d78b21cc70b34540f6","symbol":"Sequential"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-rdlm/lib/python3.9/site-packages/torch/nn/modules/dropout.py","sha256":"da4f9c79d7289384fbaabd3c5be16dbe9950835705a11cd97119df292f9f63bd","symbol":"Dropout"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-rdlm/lib/python3.9/site-packages/torch/nn/modules/linear.py","sha256":"86ddc6010e944590eb39eb12bfcda6716be5da834a0f33c44c054e98ffad0c65","symbol":"Linear"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-rdlm/lib/python3.9/site-packages/torch/nn/parallel/distributed.py","sha256":"00ab15a4fb56b0b94446698572b00ff5fd87a603de4560a2cf71ee2c7c2ffd6a","symbol":"DistributedDataParallel"},{"kind":"sampling_closure","path":"/opt/anaconda3/envs/dlb-rdlm/lib/python3.9/site-packages/torch/utils/_contextlib.py","sha256":"7e96d18a2ba0cff0edfb440ef90699358cb0633cf58f06013a7be3cf0087539e","symbol":"_DecoratorContextManager.clone"},{"kind":"sampling_closure","path":"/opt/anaconda3/envs/dlb-rdlm/lib/python3.9/site-packages/torch/utils/_contextlib.py","sha256":"7e96d18a2ba0cff0edfb440ef90699358cb0633cf58f06013a7be3cf0087539e","symbol":"get_sde_sampler.<locals>.pc_sampler"}],"seed":42,"source_commit":"67443aa6a2d0fa981eb7c6105f9cbc563e59c5c1","steps":1000,"synchronization_policy":"torch.cuda.synchronize_before_start_and_after_generate"},"schema":"dlb-generation-timing-v1","timing":{"batch_size":1,"exclusions":["model_and_checkpoint_loading","first_compilation","token_decoding","metrics","file_io"],"median_seconds":4.7119392200838774,"mode":"primary_latency","num_timed_samples":32,"raw_durations_seconds":[4.4776998101733625,4.615873063914478,4.70581367239356,5.1244133920408785,4.727834053803235,4.853146298322827,4.974498535040766,4.984726836904883,4.872867461759597,4.5900008119642735,4.598749704193324,4.707280792761594,4.862007641233504,4.5945104896090925,4.855502102058381,4.47622617892921,4.697192898020148,4.863835144788027,4.857635545078665,4.65964715834707,4.686284052208066,4.716597647406161,4.602907411754131,5.002939217258245,4.789878958370537,4.692220168188214,4.985826483927667,4.599059068132192,4.498514471575618,4.6788383992388844,4.750663097016513,4.9561729822307825],"repeats":32,"seconds_per_sample":4.7518551108951215,"standard_deviation_convention":"population","standard_deviation_seconds":0.16388104832402361,"warmups":5}}

results/conditional/timing/lm1b/rdlm/steps_1024/timing.json

@@ -0,0 +1 @@
+{"metadata":{"adapter_identity":"dlb.adapters.rdlm:v1|dlb.adapters.rdlm:RDLMAdapter|4cd983ae0746373fcae3d6cc0698aa19a49f61d2cf4eb395289b425aef2bab3a","attempt_id":"7b9160da33af4724966eb8a78d037aaf","checkpoint_lock_id":"rdlm_lm1b_drive:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152:all","checkpoint_selection":{"path":null,"resource":"rdlm_lm1b_drive","teacher_family":null},"checkpoint_sha256":"dc37df729f77d369cf15bf7ee587437f9a1b877aa8dd4da08ac732c9956b282b","checkpoint_teacher_family":"continuous_rdlm","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","cuda_runtime_version":"12.1","dataset":"lm1b","diversity_prompt_count":256,"environment":"dlb-rdlm","evaluation_continuation_length":64,"generation_mode":"conditional_prefix","gpu_compute_capability":"9.0","gpu_index":0,"gpu_name":"NVIDIA H200 NVL","model":"rdlm","nvidia_driver_version":"575.57.08","parameter_precision":"fp32","precision":"bf16-mixed","precision_evidence":"static_policy_bound_to_checkpoint_and_runtime_code_not_runtime_autocast_observation","precision_policy":"rdlm:pinned_internal_bf16_autocast_with_fp32_sensitive_ops","precision_policy_binding":{"adapter_identity":"dlb.adapters.rdlm:v1|dlb.adapters.rdlm:RDLMAdapter|4cd983ae0746373fcae3d6cc0698aa19a49f61d2cf4eb395289b425aef2bab3a","checkpoint_lock_id":"rdlm_lm1b_drive:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152:all","checkpoint_repository":null,"checkpoint_resource":"rdlm_lm1b_drive","checkpoint_revision":null,"checkpoint_selection":{"path":null,"resource":"rdlm_lm1b_drive","teacher_family":null},"checkpoint_sha256":"dc37df729f77d369cf15bf7ee587437f9a1b877aa8dd4da08ac732c9956b282b","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-lm1b-c64.json","conditioning_manifest_sha256":"0e5c5bf5b8a1445b348a65045f680ac31713dbfa1e690bedb73e0d1e7472a039","dataset":"lm1b","diversity_prompt_count":256,"evaluation_continuation_length":64,"experiment_config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","generation_mode":"conditional_prefix","model":"rdlm","prefix_length":64,"prompt_count":1024,"source_commit":"67443aa6a2d0fa981eb7c6105f9cbc563e59c5c1","upstream":"rdlm"},"prefix_length":64,"prompt_count":1024,"pytorch_compiled_cuda_toolkit":"12.1","requested_precision":"author","runtime_code_binding":[{"kind":"sampling_closure","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/capture.py","sha256":"bb7d05734503f57d02187a8a94d91394b5a3b394b45a6c3556ed51e612591713","symbol":"_capture_rdlm.<locals>.sampling_factory.<locals>.project"},{"kind":"sampling_callable","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/capture.py","sha256":"bb7d05734503f57d02187a8a94d91394b5a3b394b45a6c3556ed51e612591713","symbol":"_rdlm_benchmark_sampling_fn.<locals>.sample.<locals>.<lambda>"},{"kind":"sampling_closure","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/conditional_runtime.py","sha256":"150d2a1640f5b7acf50afc3cf1d3e1a6e82e51c3c7d44cacb41fbdb5a70ac49c","symbol":"rdlm_project_fn.<locals>.project"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/rdlm/model/dit.py","sha256":"6b59f156ce1ba2897f7ace4601f550bff2cf738dc68458e67efd630d46849d50","symbol":"DiT"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/rdlm/model/dit.py","sha256":"6b59f156ce1ba2897f7ace4601f550bff2cf738dc68458e67efd630d46849d50","symbol":"DiTBlock"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/rdlm/model/dit.py","sha256":"6b59f156ce1ba2897f7ace4601f550bff2cf738dc68458e67efd630d46849d50","symbol":"DiTFinalLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/rdlm/model/dit.py","sha256":"6b59f156ce1ba2897f7ace4601f550bff2cf738dc68458e67efd630d46849d50","symbol":"EmbeddingLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/rdlm/model/dit.py","sha256":"6b59f156ce1ba2897f7ace4601f550bff2cf738dc68458e67efd630d46849d50","symbol":"LayerNorm"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/rdlm/model/dit.py","sha256":"6b59f156ce1ba2897f7ace4601f550bff2cf738dc68458e67efd630d46849d50","symbol":"TimestepEmbedder"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/rdlm/model/rotary.py","sha256":"0f65c0f8e306fa29f152f040c6d70ac641eeaeae5b48757df8a39885ede422fd","symbol":"Rotary"},{"kind":"sampling_closure","path":"/home/ruizzhao/diffusion_baseline/upstreams/rdlm/sampling.py","sha256":"287d370c55eb29f562279f96aae0b759d6deadd435a1324f4124b9849acfd359","symbol":"get_sde_sampler.<locals>.pc_sampler"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-rdlm/lib/python3.9/site-packages/torch/_dynamo/eval_frame.py","sha256":"96a28ad58cbfc01005269673e0a98665319dccd593d3182b862537dce7c2d7d7","symbol":"OptimizedModule"},{"kind":"bound_method_owner","path":"/opt/anaconda3/envs/dlb-rdlm/lib/python3.9/site-packages/torch/autograd/grad_mode.py","sha256":"4f5bc6ae9d40f2c4777dd969fe6397f921fb648f43326c2147ce45d73be5ec6f","symbol":"no_grad"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-rdlm/lib/python3.9/site-packages/torch/nn/modules/activation.py","sha256":"90a597b30623ca02e4d165da6202898bb324279d9c2eebd016188210118702a1","symbol":"GELU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-rdlm/lib/python3.9/site-packages/torch/nn/modules/activation.py","sha256":"90a597b30623ca02e4d165da6202898bb324279d9c2eebd016188210118702a1","symbol":"SiLU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-rdlm/lib/python3.9/site-packages/torch/nn/modules/container.py","sha256":"8e2a1829811af9845cb8bf9200ce66ebcac91efc7d6ba5d78b21cc70b34540f6","symbol":"ModuleList"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-rdlm/lib/python3.9/site-packages/torch/nn/modules/container.py","sha256":"8e2a1829811af9845cb8bf9200ce66ebcac91efc7d6ba5d78b21cc70b34540f6","symbol":"Sequential"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-rdlm/lib/python3.9/site-packages/torch/nn/modules/dropout.py","sha256":"da4f9c79d7289384fbaabd3c5be16dbe9950835705a11cd97119df292f9f63bd","symbol":"Dropout"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-rdlm/lib/python3.9/site-packages/torch/nn/modules/linear.py","sha256":"86ddc6010e944590eb39eb12bfcda6716be5da834a0f33c44c054e98ffad0c65","symbol":"Linear"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-rdlm/lib/python3.9/site-packages/torch/nn/parallel/distributed.py","sha256":"00ab15a4fb56b0b94446698572b00ff5fd87a603de4560a2cf71ee2c7c2ffd6a","symbol":"DistributedDataParallel"},{"kind":"sampling_closure","path":"/opt/anaconda3/envs/dlb-rdlm/lib/python3.9/site-packages/torch/utils/_contextlib.py","sha256":"7e96d18a2ba0cff0edfb440ef90699358cb0633cf58f06013a7be3cf0087539e","symbol":"_DecoratorContextManager.clone"},{"kind":"sampling_closure","path":"/opt/anaconda3/envs/dlb-rdlm/lib/python3.9/site-packages/torch/utils/_contextlib.py","sha256":"7e96d18a2ba0cff0edfb440ef90699358cb0633cf58f06013a7be3cf0087539e","symbol":"get_sde_sampler.<locals>.pc_sampler"}],"sample_count":2048,"seed":42,"source_commit":"67443aa6a2d0fa981eb7c6105f9cbc563e59c5c1","steps":1024,"synchronization_policy":"torch.cuda.synchronize_before_start_and_after_generate"},"schema":"dlb-generation-timing-v1","timing":{"batch_size":1,"exclusions":["model_and_checkpoint_loading","first_compilation","token_decoding","metrics","file_io"],"median_seconds":3.736564058577642,"mode":"primary_latency","num_timed_samples":32,"raw_durations_seconds":[3.7382240500301123,3.7370584020391107,3.7350788949988782,3.742235033772886,3.734225946944207,3.735549407079816,3.736579948104918,3.7411643899977207,3.738023371901363,3.741289373021573,3.7354830149561167,3.739131391979754,3.7346438681706786,3.7412878810428083,3.7377394665963948,3.7425505071878433,3.7357114059850574,3.737760099116713,3.7365481690503657,3.7380082989111543,3.7375176358036697,3.7410456580109894,3.7319227531552315,3.7338713631033897,3.7343638380989432,3.7316030631773174,3.7318821609951556,3.7352647301740944,3.728509906679392,3.738202416803688,3.732163241598755,3.7323257313109934],"repeats":32,"seconds_per_sample":3.7364676693687215,"standard_deviation_convention":"population","standard_deviation_seconds":0.0034101949956221057,"warmups":5}}

results/conditional/timing/owt/duo/steps_8/.timing.json.97ce5a2effed489cb1a8713b7d67cbd8.superseded.json

@@ -0,0 +1 @@
+{"metadata":{"adapter_identity":"dlb.adapters.duo:v1|dlb.adapters.duo:DuoAdapter|8878182eb10d7d1e3f18219cb62c2153084842fdb0ab532b1a3c7df0e10da3c6","attempt_id":"ead3fdb5286047058f1485c436437aa8","checkpoint_lock_id":"duo_owt_hf:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152:all","checkpoint_selection":{"path":null,"resource":"duo_owt_hf","teacher_family":null},"checkpoint_sha256":"89ba2aa1ec5c763b36abed7c1dbfdff3b41463d8dd44baefb9866a9500c666c1","checkpoint_teacher_family":"uniform_duo","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json","conditioning_manifest_sha256":"7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32","config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","cuda_runtime_version":"12.4","dataset":"owt","diversity_prompt_count":256,"environment":"dlb-duo","evaluation_continuation_length":64,"generation_mode":"conditional_prefix","gpu_compute_capability":"9.0","gpu_index":0,"gpu_name":"NVIDIA H200 NVL","model":"duo","nvidia_driver_version":"575.57.08","parameter_precision":"fp32","precision":"bf16-mixed","precision_evidence":"static_policy_bound_to_checkpoint_and_runtime_code_not_runtime_autocast_observation","precision_policy":"duo:pinned_internal_bf16_autocast_with_fp32_sensitive_ops","precision_policy_binding":{"adapter_identity":"dlb.adapters.duo:v1|dlb.adapters.duo:DuoAdapter|8878182eb10d7d1e3f18219cb62c2153084842fdb0ab532b1a3c7df0e10da3c6","checkpoint_lock_id":"duo_owt_hf:ce8ff6a327d9f1470fef9a7ac66813ac94d4cc50268bc2998627498dc6f3a152:all","checkpoint_repository":"s-sahoo/duo","checkpoint_resource":"duo_owt_hf","checkpoint_revision":"ef8ad7a25a4bf9624306230913dd7ae5cb2a00bd","checkpoint_selection":{"path":null,"resource":"duo_owt_hf","teacher_family":null},"checkpoint_sha256":"89ba2aa1ec5c763b36abed7c1dbfdff3b41463d8dd44baefb9866a9500c666c1","completion_schedule":"c0:p0-1023;c1-4:p0-255","completions_per_diversity_prompt":5,"conditioning_config_sha256":"89d70134c4fb1ce21c18e46cfb9753f6cdcd460b879d6e60e99b2c7b56ae5114","conditioning_manifest":"/home/ruizzhao/diffusion_baseline/data/manifests/conditional-owt-c64.json","conditioning_manifest_sha256":"7718eb8d4d3e75ec7628d2f15f3c73c84f6470af0303ad4017da5e0ddae5bd32","dataset":"owt","diversity_prompt_count":256,"evaluation_continuation_length":64,"experiment_config_sha256":"6ce963030186c79cab2bc15670737a77ebfad08a0608bc87ee5c54e489ce678f","generation_mode":"conditional_prefix","model":"duo","prefix_length":64,"prompt_count":1024,"source_commit":"7c9b498f5b717de064d6fad7e2509c866e6cb620","upstream":"duo"},"prefix_length":64,"prompt_count":1024,"pytorch_compiled_cuda_toolkit":"12.4","requested_precision":"author","runtime_code_binding":[{"kind":"loaded_module","path":"/home/ruizzhao/.cache/huggingface/modules/transformers_modules/owt/model.py","sha256":"dcfef1c27da31fce4d6a7449a2759461070f5ed825b5e13c38361b228dade58d","symbol":"DDiTBlock"},{"kind":"loaded_module","path":"/home/ruizzhao/.cache/huggingface/modules/transformers_modules/owt/model.py","sha256":"dcfef1c27da31fce4d6a7449a2759461070f5ed825b5e13c38361b228dade58d","symbol":"DDiTFinalLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/.cache/huggingface/modules/transformers_modules/owt/model.py","sha256":"dcfef1c27da31fce4d6a7449a2759461070f5ed825b5e13c38361b228dade58d","symbol":"DUO"},{"kind":"loaded_module","path":"/home/ruizzhao/.cache/huggingface/modules/transformers_modules/owt/model.py","sha256":"dcfef1c27da31fce4d6a7449a2759461070f5ed825b5e13c38361b228dade58d","symbol":"EmbeddingLayer"},{"kind":"loaded_module","path":"/home/ruizzhao/.cache/huggingface/modules/transformers_modules/owt/model.py","sha256":"dcfef1c27da31fce4d6a7449a2759461070f5ed825b5e13c38361b228dade58d","symbol":"HFDIT"},{"kind":"loaded_module","path":"/home/ruizzhao/.cache/huggingface/modules/transformers_modules/owt/model.py","sha256":"dcfef1c27da31fce4d6a7449a2759461070f5ed825b5e13c38361b228dade58d","symbol":"LayerNorm"},{"kind":"loaded_module","path":"/home/ruizzhao/.cache/huggingface/modules/transformers_modules/owt/model.py","sha256":"dcfef1c27da31fce4d6a7449a2759461070f5ed825b5e13c38361b228dade58d","symbol":"Rotary"},{"kind":"loaded_module","path":"/home/ruizzhao/.cache/huggingface/modules/transformers_modules/owt/model.py","sha256":"dcfef1c27da31fce4d6a7449a2759461070f5ed825b5e13c38361b228dade58d","symbol":"TimestepEmbedder"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/capture.py","sha256":"92c69119e85195e099a41668e8f75dd5a908c4e341fb220f714853db555dca0d","symbol":"_adapt_hf_masked_lm_backbone.<locals>.KeywordBackboneAdapter"},{"kind":"sampling_callable","path":"/home/ruizzhao/diffusion_baseline/src/dlb/adapters/capture.py","sha256":"92c69119e85195e099a41668e8f75dd5a908c4e341fb220f714853db555dca0d","symbol":"_capture_teacher.<locals>.capture.<locals>.<lambda>"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/algo.py","sha256":"d6cf18959f77f5002f5a9dde3df3fc109809afc87e6d735da88d0cb28dfa9d46","symbol":"DUO_BASE"},{"kind":"loaded_module","path":"/home/ruizzhao/diffusion_baseline/upstreams/duo/trainer_base.py","sha256":"8e990b13034edde83563ec39a00eaaffdab44738229446db21a3a2e367855ed4","symbol":"LogLinear"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"GELU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/activation.py","sha256":"1e748e0df0d70491fe82f0178f32126ee6ee138a53de7fc76463af10c75f5477","symbol":"SiLU"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"ModuleList"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/container.py","sha256":"33af8d1fe4fb542cbba5542c5079b1bd8ef8a6493abbef329411f71b2071d24f","symbol":"Sequential"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/dropout.py","sha256":"1fc9974c3cf2078dd4bc257f73afb6a6714adfec2e9d099d47cf923e766d3490","symbol":"Dropout"},{"kind":"loaded_module","path":"/opt/anaconda3/envs/dlb-duo/lib/python3.12/site-packages/torch/nn/modules/linear.py","sha256":"d7f5cadf773deb5ab4bcc9f57414e8239b2859c70d2aa6f141576de1ec918823","symbol":"Linear"}],"seed":42,"source_commit":"7c9b498f5b717de064d6fad7e2509c866e6cb620","steps":8,"synchronization_policy":"torch.cuda.synchronize_before_start_and_after_generate"},"schema":"dlb-generation-timing-v1","timing":{"batch_size":1,"exclusions":["model_and_checkpoint_loading","first_compilation","token_decoding","metrics","file_io"],"median_seconds":0.08972604013979435,"mode":"primary_latency","num_timed_samples":32,"raw_durations_seconds":[0.09621173003688455,0.09629351506009698,0.08913765707984567,0.08913589408621192,0.08912293380126357,0.08919790992513299,0.089867337141186,0.09811956668272614,0.09770194627344608,0.09577158465981483,0.08914576983079314,0.08913610503077507,0.08916006190702319,0.0895847431384027,0.09908444713801146,0.09829198103398085,0.10167816560715437,0.08918144507333636,0.08912876294925809,0.08912329468876123,0.08914890419691801,0.09454049589112401,0.09973069792613387,0.09977128077298403,0.09520338289439678,0.08917202986776829,0.08915087720379233,0.0891712186858058,0.08957474818453193,0.1016949419863522,0.1005092030391097,0.09750675596296787],"repeats":32,"seconds_per_sample":0.09341404336737469,"standard_deviation_convention":"population","standard_deviation_seconds":0.004688172891915897,"warmups":5}}