@simulatte/doppler 0.1.9 → 0.2.0

package/src/config/presets/models/mixtral.json

@@ -1,37 +0,0 @@
-{
-  "id": "mixtral",
-  "name": "Mixtral (MoE)",
-  "extends": "transformer",
-  "modelType": "mixtral",
-
-  "inference": {
-    "attention": {
-      "slidingWindow": 4096
-    },
-    "normalization": {
-      "rmsNormWeightOffset": false,
-      "rmsNormEps": 1e-5
-    }
-  },
-
-  "tokenizer": {
-    "bosToken": "<s>",
-    "eosTokens": ["</s>"],
-    "addBosToken": true
-  },
-
-  "tensorPatterns": {
-    "ffn": {
-      "gate": ["layers.{layer}.block_sparse_moe.experts.{expert}.w1.weight"],
-      "up": ["layers.{layer}.block_sparse_moe.experts.{expert}.w3.weight"],
-      "down": ["layers.{layer}.block_sparse_moe.experts.{expert}.w2.weight"]
-    }
-  },
-
-  "detection": {
-    "architecturePatterns": ["mixtral", "MixtralForCausalLM"],
-    "configPatterns": {
-      "num_local_experts": 8
-    }
-  }
-}

package/src/config/presets/models/modernbert.json

@@ -1,32 +0,0 @@
-{
-  "id": "modernbert",
-  "name": "ModernBERT",
-  "extends": "transformer",
-  "modelType": "embedding",
-
-  "inference": {
-    "ffn": {
-      "activation": "gelu"
-    },
-    "chatTemplate": {
-      "enabled": false
-    },
-    "output": {
-      "tieWordEmbeddings": false
-    }
-  },
-
-  "tokenizer": {
-    "addBosToken": false
-  },
-
-  "detection": {
-    "architecturePatterns": [
-      "ModernBertModel",
-      "modernbert"
-    ],
-    "modelTypePatterns": [
-      "modernbert"
-    ]
-  }
-}

package/src/config/presets/models/qwen3.json

@@ -1,49 +0,0 @@
-{
-  "id": "qwen3",
-  "name": "Qwen 3",
-  "extends": "transformer",
-
-  "architecture": {
-    "ropeTheta": 10000000
-  },
-
-  "inference": {
-    "attention": {
-      "slidingWindow": null,
-      "queryKeyNorm": true,
-      "attentionOutputGate": true
-    },
-    "output": {
-      "scaleEmbeddings": false
-    },
-    "normalization": {
-      "rmsNormWeightOffset": false,
-      "rmsNormEps": 1e-6
-    },
-    "chatTemplate": {
-      "type": "qwen",
-      "enabled": true
-    },
-    "rope": {
-      "ropeTheta": 10000000,
-      "mropeInterleaved": true,
-      "mropeSection": [11, 11, 10],
-      "partialRotaryFactor": 0.25
-    }
-  },
-
-  "tokenizer": {
-    "bosToken": "<|im_start|>",
-    "eosTokens": ["<|im_end|>", "<|endoftext|>"],
-    "unkTokenId": 248044,
-    "addBosToken": false
-  },
-
-  "detection": {
-    "architecturePatterns": ["qwen3", "Qwen3ForCausalLM", "Qwen2ForCausalLM"],
-    "modelTypePatterns": ["qwen3", "qwen2"],
-    "configPatterns": {
-      "model_type": "qwen2"
-    }
-  }
-}

package/src/config/presets/models/qwen3_5.json

@@ -1,16 +0,0 @@
-{
-  "id": "qwen3_5",
-  "name": "Qwen 3.5",
-  "extends": "qwen3",
-
-  "inference": {
-    "normalization": {
-      "rmsNormWeightOffset": true
-    }
-  },
-
-  "detection": {
-    "architecturePatterns": ["qwen3_5", "Qwen3_5ForCausalLM", "Qwen3_5ForConditionalGeneration"],
-    "modelTypePatterns": ["qwen3_5", "qwen3_5_text"]
-  }
-}

package/src/config/presets/models/qwen3_vl.json

@@ -1,40 +0,0 @@
-{
-  "id": "qwen3_vl",
-  "name": "Qwen 3 VL",
-  "extends": "qwen3",
-
-  "architecture": {
-    "ropeTheta": 5000000
-  },
-
-  "inference": {
-    "normalization": {
-      "rmsNormWeightOffset": false
-    },
-    "rope": {
-      "ropeTheta": 5000000,
-      "mropeInterleaved": true,
-      "mropeSection": [24, 20, 20],
-      "partialRotaryFactor": null
-    }
-  },
-
-  "vision": {
-    "patchSize": 16,
-    "spatialMergeSize": 2,
-    "temporalPatchSize": 2,
-    "eps": 1e-6,
-    "minPixels": 3136,
-    "maxPixels": 1003520,
-    "projectorType": "spatial_merge",
-    "normalization": {
-      "mean": [0.48145466, 0.4578275, 0.40821073],
-      "std": [0.26862954, 0.26130258, 0.27577711]
-    }
-  },
-
-  "detection": {
-    "architecturePatterns": ["qwen3_vl", "Qwen3VLForConditionalGeneration"],
-    "modelTypePatterns": ["qwen3_vl"]
-  }
-}

package/src/config/presets/models/transformer.json

@@ -1,78 +0,0 @@
-{
-  "id": "transformer",
-  "name": "Base Transformer",
-  "modelType": "transformer",
-
-  "inference": {
-    "attention": {
-      "slidingWindow": null,
-      "attnLogitSoftcapping": null,
-      "queryKeyNorm": false,
-      "ropeScalingType": null
-    },
-    "normalization": {
-      "rmsNormWeightOffset": false,
-      "rmsNormEps": 1e-5,
-      "postAttentionNorm": false,
-      "preFeedforwardNorm": false,
-      "postFeedforwardNorm": false
-    },
-    "ffn": {
-      "activation": "silu",
-      "gatedActivation": true
-    },
-    "rope": {
-      "mropeInterleaved": false,
-      "mropeSection": null,
-      "partialRotaryFactor": null
-    },
-    "output": {
-      "finalLogitSoftcapping": null,
-      "tieWordEmbeddings": false
-    },
-    "layerPattern": {
-      "type": "all_attention"
-    }
-  },
-
-  "tokenizer": {
-    "addBosToken": true,
-    "addEosToken": false,
-    "allowArchFallback": false
-  },
-
-  "sampling": {
-    "temperature": 1.0,
-    "topK": 50,
-    "topP": 0.95,
-    "repetitionPenalty": 1.0
-  },
-
-  "tensorPatterns": {
-    "embedding": [
-      "model.embed_tokens.weight",
-      "embeddings.word_embeddings.weight",
-      "token_embd.weight"
-    ],
-    "lmHead": [
-      "lm_head.weight",
-      "output.weight"
-    ],
-    "attention": {
-      "qProj": ["layers.{layer}.self_attn.q_proj.weight"],
-      "kProj": ["layers.{layer}.self_attn.k_proj.weight"],
-      "vProj": ["layers.{layer}.self_attn.v_proj.weight"],
-      "oProj": ["layers.{layer}.self_attn.o_proj.weight"]
-    },
-    "ffn": {
-      "gate": ["layers.{layer}.mlp.gate_proj.weight"],
-      "up": ["layers.{layer}.mlp.up_proj.weight"],
-      "down": ["layers.{layer}.mlp.down_proj.weight"]
-    },
-    "norm": {
-      "input": ["layers.{layer}.input_layernorm.weight"],
-      "postAttn": ["layers.{layer}.post_attention_layernorm.weight"],
-      "final": ["model.norm.weight", "norm.weight"]
-    }
-  }
-}

package/src/config/presets/models/translategemma.json

@@ -1,30 +0,0 @@
-{
-  "id": "translategemma",
-  "name": "TranslateGemma",
-  "extends": "gemma3",
-
-  "sampling": {
-    "temperature": 1.0,
-    "topK": 64,
-    "topP": 0.95,
-    "repetitionPenalty": 1.0
-  },
-  "inference": {
-    "chatTemplate": {
-      "type": "translategemma",
-      "enabled": true
-    }
-  },
-
-  "detection": {
-    "architecturePatterns": [
-      "translategemma",
-      "TranslateGemma",
-      "Gemma3ForConditionalGeneration"
-    ],
-    "modelTypePatterns": [
-      "translategemma",
-      "translate_gemma"
-    ]
-  }
-}

package/src/config/presets/platforms/nvidia-gb200-8gpu.json

@@ -1,45 +0,0 @@
-{
-  "name": "nvidia-gb200-8gpu",
-  "description": "NVIDIA GB200 8-GPU Pod (DGX Pod Building Block)",
-  "emulation": {
-    "enabled": true,
-    "targetChip": "gb200-8gpu",
-    "timingMode": "functional",
-    "gpuSpec": {
-      "name": "B200",
-      "vramBytes": 206158430208,
-      "hbmBandwidthBytesPerSec": 8000000000000,
-      "fp16Tflops": 4500,
-      "fp8Tflops": 9000
-    },
-    "cpuSpec": {
-      "name": "Grace",
-      "cores": 72,
-      "memoryBytes": 515396075520,
-      "memoryBandwidthBytesPerSec": 546000000000
-    },
-    "topology": {
-      "gpuCount": 8,
-      "gpusPerNode": 8,
-      "nodeCount": 1,
-      "cpuCount": 2
-    },
-    "nvlink": {
-      "bandwidthBytesPerSec": 1800000000000,
-      "latencyUs": 0.8
-    },
-    "nvlinkC2C": {
-      "bandwidthBytesPerSec": 900000000000,
-      "latencyUs": 0.5,
-      "coherent": true
-    },
-    "parallelism": {
-      "tensorParallel": { "enabled": true, "degree": 8 },
-      "pipelineParallel": { "enabled": false, "stages": 1 },
-      "dataParallel": { "enabled": false, "degree": 1 },
-      "expertParallel": { "enabled": false, "degree": 1 }
-    },
-    "statsEnabled": true,
-    "logOperations": false
-  }
-}

package/src/config/presets/platforms/nvidia-gb200-nvl72.json

@@ -1,45 +0,0 @@
-{
-  "name": "nvidia-gb200-nvl72",
-  "description": "NVIDIA GB200 NVL72 (72 GPUs across 9 nodes)",
-  "emulation": {
-    "enabled": true,
-    "targetChip": "gb200-nvl72",
-    "timingMode": "functional",
-    "gpuSpec": {
-      "name": "B200",
-      "vramBytes": 206158430208,
-      "hbmBandwidthBytesPerSec": 8000000000000,
-      "fp16Tflops": 4500,
-      "fp8Tflops": 9000
-    },
-    "cpuSpec": {
-      "name": "Grace",
-      "cores": 72,
-      "memoryBytes": 515396075520,
-      "memoryBandwidthBytesPerSec": 546000000000
-    },
-    "topology": {
-      "gpuCount": 72,
-      "gpusPerNode": 8,
-      "nodeCount": 9,
-      "cpuCount": 18
-    },
-    "nvlink": {
-      "bandwidthBytesPerSec": 1800000000000,
-      "latencyUs": 0.8
-    },
-    "nvlinkC2C": {
-      "bandwidthBytesPerSec": 900000000000,
-      "latencyUs": 0.5,
-      "coherent": true
-    },
-    "parallelism": {
-      "tensorParallel": { "enabled": true, "degree": 8 },
-      "pipelineParallel": { "enabled": true, "stages": 9 },
-      "dataParallel": { "enabled": false, "degree": 1 },
-      "expertParallel": { "enabled": false, "degree": 1 }
-    },
-    "statsEnabled": true,
-    "logOperations": false
-  }
-}

package/src/config/presets/platforms/nvidia-gh200-nvl2.json

@@ -1,44 +0,0 @@
-{
-  "name": "nvidia-gh200-nvl2",
-  "description": "NVIDIA GH200 NVL2 (2 Superchips connected via NVLink)",
-  "emulation": {
-    "enabled": true,
-    "targetChip": "gh200-nvl2",
-    "timingMode": "functional",
-    "gpuSpec": {
-      "name": "H200",
-      "vramBytes": 154618822656,
-      "hbmBandwidthBytesPerSec": 4800000000000,
-      "fp16Tflops": 1979
-    },
-    "cpuSpec": {
-      "name": "Grace",
-      "cores": 72,
-      "memoryBytes": 515396075520,
-      "memoryBandwidthBytesPerSec": 546000000000
-    },
-    "topology": {
-      "gpuCount": 2,
-      "gpusPerNode": 2,
-      "nodeCount": 1,
-      "cpuCount": 2
-    },
-    "nvlink": {
-      "bandwidthBytesPerSec": 900000000000,
-      "latencyUs": 1.0
-    },
-    "nvlinkC2C": {
-      "bandwidthBytesPerSec": 900000000000,
-      "latencyUs": 0.5,
-      "coherent": true
-    },
-    "parallelism": {
-      "tensorParallel": { "enabled": true, "degree": 2 },
-      "pipelineParallel": { "enabled": false, "stages": 1 },
-      "dataParallel": { "enabled": false, "degree": 1 },
-      "expertParallel": { "enabled": false, "degree": 1 }
-    },
-    "statsEnabled": true,
-    "logOperations": false
-  }
-}

package/src/config/presets/platforms/nvidia-gh200.json

@@ -1,44 +0,0 @@
-{
-  "name": "nvidia-gh200",
-  "description": "NVIDIA GH200 Grace Hopper Superchip (1 GPU + 1 CPU)",
-  "emulation": {
-    "enabled": true,
-    "targetChip": "gh200",
-    "timingMode": "functional",
-    "gpuSpec": {
-      "name": "H200",
-      "vramBytes": 154618822656,
-      "hbmBandwidthBytesPerSec": 4800000000000,
-      "fp16Tflops": 1979
-    },
-    "cpuSpec": {
-      "name": "Grace",
-      "cores": 72,
-      "memoryBytes": 515396075520,
-      "memoryBandwidthBytesPerSec": 546000000000
-    },
-    "topology": {
-      "gpuCount": 1,
-      "gpusPerNode": 1,
-      "nodeCount": 1,
-      "cpuCount": 1
-    },
-    "nvlink": {
-      "bandwidthBytesPerSec": 900000000000,
-      "latencyUs": 1.0
-    },
-    "nvlinkC2C": {
-      "bandwidthBytesPerSec": 900000000000,
-      "latencyUs": 0.5,
-      "coherent": true
-    },
-    "parallelism": {
-      "tensorParallel": { "enabled": false, "degree": 1 },
-      "pipelineParallel": { "enabled": false, "stages": 1 },
-      "dataParallel": { "enabled": false, "degree": 1 },
-      "expertParallel": { "enabled": false, "degree": 1 }
-    },
-    "statsEnabled": true,
-    "logOperations": false
-  }
-}

package/src/config/presets/runtime/compute/f16-activations.json

@@ -1,30 +0,0 @@
-{
-  "id": "compute/f16-activations",
-  "name": "f16-activations",
-  "description": "Experimental F16 activations for reduced memory bandwidth",
-  "intent": "investigate",
-  "stability": "canonical",
-  "owner": "doppler-core",
-  "createdAtUtc": "2026-02-25T00:00:00Z",
-  "extends": "default",
-  "runtime": {
-    "inference": {
-      "compute": {
-        "activationDtype": "f16"
-      },
-      "kvcache": {
-        "kvDtype": "f16"
-      },
-      "session": {
-        "compute": {
-          "defaults": {
-            "activationDtype": "f16"
-          }
-        },
-        "kvcache": {
-          "kvDtype": "f16"
-        }
-      }
-    }
-  }
-}

package/src/config/presets/runtime/compute/f16-batched.json

@@ -1,32 +0,0 @@
-{
-  "id": "compute/f16-batched",
-  "name": "f16-batched",
-  "description": "F16 activations with multi-token decode batching for lower submit/readback overhead.",
-  "intent": "investigate",
-  "stability": "canonical",
-  "owner": "doppler-core",
-  "createdAtUtc": "2026-02-25T00:00:00Z",
-  "extends": "compute/f16-activations",
-  "runtime": {
-    "inference": {
-      "batching": {
-        "batchSize": 16,
-        "stopCheckMode": "batch",
-        "readbackInterval": 8,
-        "ringTokens": 8,
-        "ringStop": 8,
-        "ringStaging": 8
-      },
-      "session": {
-        "decodeLoop": {
-          "batchSize": 16,
-          "stopCheckMode": "batch",
-          "readbackInterval": 8,
-          "ringTokens": 8,
-          "ringStop": 8,
-          "ringStaging": 8
-        }
-      }
-    }
-  }
-}

package/src/config/presets/runtime/default.json

@@ -1,101 +0,0 @@
-{
-  "id": "default",
-  "name": "default",
-  "description": "Default DOPPLER configuration with balanced settings",
-  "intent": "investigate",
-  "stability": "canonical",
-  "owner": "doppler-core",
-  "createdAtUtc": "2026-02-25T00:00:00Z",
-  "runtime": {
-    "shared": {
-      "debug": {
-        "logOutput": {
-          "stdout": true,
-          "file": null,
-          "append": true
-        },
-        "logHistory": {
-          "maxLogHistoryEntries": 1000
-        },
-        "logLevel": {
-          "defaultLogLevel": "info"
-        },
-        "trace": {
-          "enabled": false,
-          "categories": [
-            "all"
-          ],
-          "layers": null,
-          "maxDecodeSteps": 0,
-          "file": null
-        }
-      }
-    },
-    "loading": {
-      "storage": {
-        "alignment": {
-          "bufferAlignmentBytes": 256
-        }
-      },
-      "distribution": {
-        "cdnBasePath": "https://huggingface.co",
-        "progressUpdateIntervalMs": 100
-      },
-      "allowF32UpcastNonMatmul": false
-    },
-    "inference": {
-      "compute": {
-        "activationDtype": "f16"
-      },
-      "prompt": "Hello from Doppler.",
-      "batching": {
-        "maxTokens": 128
-      },
-      "sampling": {
-        "temperature": 1,
-        "topK": 32,
-        "topP": 0.95,
-        "repetitionPenalty": 1,
-        "greedyThreshold": 0.01
-      },
-      "kvcache": {
-        "maxSeqLen": 8192,
-        "layout": "contiguous",
-        "pageSize": 256,
-        "kvDtype": "f16"
-      },
-      "moe": {
-        "routing": {
-          "topK": 2,
-          "routerDtype": "f32",
-          "normalizeWeights": true
-        },
-        "cache": {
-          "dequantCacheMaxEntries": 8
-        }
-      },
-      "session": {
-        "compute": {
-          "defaults": {
-            "activationDtype": "f16"
-          }
-        },
-        "kvcache": {
-          "layout": "contiguous",
-          "kvDtype": "f16",
-          "pageSize": 256
-        }
-      },
-      "kernelPathPolicy": {
-        "mode": "capability-aware",
-        "sourceScope": [
-          "model",
-          "manifest",
-          "config",
-          "execution-v0"
-        ],
-        "onIncompatible": "remap"
-      }
-    }
-  }
-}

package/src/config/presets/runtime/diagnostics/debug-logits.json

@@ -1,53 +0,0 @@
-{
-  "id": "diagnostics/debug-logits",
-  "name": "debug-logits",
-  "description": "Trace sample + logits with a top-k probe for a single token.",
-  "intent": "investigate",
-  "stability": "canonical",
-  "owner": "doppler-core",
-  "createdAtUtc": "2026-02-25T00:00:00Z",
-  "extends": "modes/debug",
-  "runtime": {
-    "inference": {
-      "prompt": "The sky is",
-      "batching": {
-        "maxTokens": 1
-      },
-      "sampling": {
-        "temperature": 0
-      }
-    },
-    "shared": {
-      "debug": {
-        "logLevel": {
-          "defaultLogLevel": "debug"
-        },
-        "trace": {
-          "enabled": true,
-          "categories": [
-            "sample",
-            "logits"
-          ],
-          "maxDecodeSteps": 1
-        },
-        "probes": [
-          {
-            "id": "topk",
-            "stage": "logits_final",
-            "tokens": [
-              -1
-            ],
-            "dims": [
-              138,
-              496,
-              0,
-              1,
-              2,
-              3
-            ]
-          }
-        ]
-      }
-    }
-  }
-}