Archive Predator media campaign including failed and unusable outputs

2026-10-21 07:36:12 +02:00
1 parent 64c3b48e06
commit 3bd36b1074
@@ -0,0 +1,41 @@
+# Predator: images, video and music on a 6 GB GPU
+
+Real local-hardware observations from October 20–21, 2026, using a Windows computer with a GTX 1060 6 GB and approximately 28 GiB system RAM. Requests travelled through the real Packet Foundry Warehouse → Registry → Factory relay to a local ComfyUI runtime. These are exploratory deployment and media-generation trials, not a controlled model leaderboard.
+
+## Coverage
+
+The archive retains **123 output files**: **110 images, 11 videos and two music samples**. Of those, 122 were accepted and acknowledged by the Warehouse; one additional Qwen image finished locally but delivery was cancelled after network and lease failures. The image total includes the exact 100-image prompt-complexity suite and ten earlier image outputs. Two pairs of early accepted images have identical original file hashes, so the archive contains **121 unique original media byte sequences**, while preserving every delivery individually.
+
+Eleven video contact sheets and the image used as LTX video input are derived reference material, not additional model outputs. Two synthetic codec-preflight MP4 fixtures are excluded from the model gallery and generation counts.
+
+Seven submitted jobs ended without media: a cancelled Wan baseline repeat, two LTX storage-guard stops, a Cog attention allocation failure, an H3 legacy-loader failure, an H3 dynamic trial cancelled for impractical runtime, and an incomplete YuE2 CPU probe. Their prompts, available timings, settings and findings appear under `failures` in `dataset.json`. No files are invented for jobs that produced none. Three further installation/schema preflight failures are described separately under `diagnostics`; these were not generation jobs.
+
+## Read the results
+
+`dataset.json` is the authoritative public projection. Each output record links its media and operator-approved workflow, complete input brief, actual conditioning text, seed, settings, available timing and resource measurements, cache observations, technical verification and limited quality findings. Private application, attempt, trace, tenant and host identifiers are omitted. Per-run public workflow hashes and captured source workflow canonical hashes distinguish the privacy-cleaned graph from its original.
+
+The 100-image suite uses ten scene families, five prompt-complexity tiers and adjacent seeds 42/43 for each brief: 50 briefs and exactly 100 images. Briefs range from 16 to 463 whitespace-counted words. Seed and cache state change together between the two arms, and complexity also changes constraint count; the experiment does not isolate prompt length as the sole cause of latency or quality.
+
+The exact ComfyUI and GGUF-node commits, selected model revisions and weight hashes, local conversion identities, recorded Packet Foundry wheel hashes and private integration source hashes are in the dataset. A deployed Packet Foundry Git commit was not recorded in the retained proof, so the current checkout is not substituted for it. No weights are redistributed. Custom workflow node names require the corresponding private integration; the graph alone is not a standalone installer.
+
+## Measurement boundaries
+
+Factory trace duration covers the attempt and differs from native Comfy execution time. Observer node intervals include loading, preparation and event delivery; they are not isolated kernel time or exact weight-loading measurements. Missing timings remain null. Stage telemetry is included where retained. A completed stage in a failed request is not a delivered result.
+
+Quantization labels describe weight storage. Video generators used FP32 arithmetic; an entire pipeline labelled Q8 integer computation would be inaccurate. CUDA parameter checks do not establish physical VRAM residency or absence of Windows WDDM paging. Resource extrema are sampled, often device-wide, and are not continuous process peaks. Cog decoder reserved allocator memory exceeded the card's physical 6 GB.
+
+Cache fields have explicit semantics in the dataset. The second Wan Q8 request followed an approximately 50-minute idle gap: its conditioning was cached, but it is not an immediate warm benchmark. The LTX image-input pair began in a runtime already used for its text pair. Reused encoder probe times describe the earlier encoding, not extra encoding during the cached request. Padded token positions are not content-only token counts, and tokenizer-internal truncation was not independently established.
+
+## Quality and failures remain visible
+
+All 100 complexity-suite thumbnails contained recognizable scenes; exact lettering, fine anatomy and full-resolution trait scores were not systematically assessed. The user listened to both YuE2 files and described them as usable music. Lyrics, intelligibility and comparative music quality were not scored.
+
+Videos include nearly static fox scenes, anatomy and direction errors, two unusable colored-noise clips from the original custom Cog bridge, and corrected Cog outputs with recognizable scenery but poor framing, continuity and requested-action adherence. The original bridge had concrete positional and sampling mismatches. Its noise is not a clean test of Cog quality or Q8 quantization. Five sampled frames and valid decoding do not establish natural full-rate motion or production suitability.
+
+Delivery, checksums, full decoding and actual LangSmith media attachment retrieval were verified separately from visual or listening quality. Comparable native experiments were checked for the completed pairs and 100-image arms. LangSmith's preview/comparison UI was not inspected.
+
+## Public-copy verification and attribution
+
+Original media remain retained privately. PNG copies were losslessly re-encoded with ancillary metadata removed, with decoded pixel bytes compared against each source. Video and FLAC containers were remuxed without metadata and without re-encoding; encoded stream hashes were compared against each original. Source and public-copy SHA-256 hashes are both recorded. Every public output was fully decoded; `publication-audit.json` records the 123-file checks. Video contact sheets are derived JPEGs, not replacement original clips.
+
+Prompts, measurements and observations are supplied as benchmark evidence. Model and companion terms remain separate from Packet Foundry's Apache-2.0 code license. In particular, FLUX.1 dev uses its non-commercial model license; YuE2 has CC-BY-NC-4.0 model terms; Qwen Image 2.1 uses its research terms; FLUX.3 components link their upstream FLUX Kommunity and companion terms. The model-source and license links in the dataset identify the relevant upstream material. This archive does not grant a new blanket license to model weights or override upstream conditions on their outputs.
@@ -0,0 +1,370 @@
+[
+  {
+    "id": "wan-failure-01",
+    "label": "Wan2.1 T2V 1.3B · cancelled",
+    "group": "wan21-small-v1",
+    "kind": "video",
+    "model_id": "wan",
+    "model": "Wan-AI/Wan2.1-T2V-1.3B",
+    "seed": 43,
+    "status": "cancelled",
+    "accepted_result": false,
+    "prompt": null,
+    "prompt_details": {
+      "last_known_brief_description": "A blue ceramic fox figurine on a desk beside a plant, with requested camera movement.",
+      "exact_request_text_verified": false,
+      "source": "Retained baseline discussion/approved template; exact submitted manifest not available in the local evidence projection."
+    },
+    "parameters": {
+      "width": 512,
+      "height": 320,
+      "frames": 81,
+      "fps": 16,
+      "template": "wan21-13b-text-v1",
+      "steps": 20,
+      "cfg": 5,
+      "sampler": "uni_pc",
+      "scheduler": "simple",
+      "math": "FP32",
+      "generator_storage": "BF16 source checkpoint",
+      "placement": "low-VRAM generator, CPU encoder and decoder"
+    },
+    "timing": {
+      "factory_trace_seconds": null,
+      "stages_seconds": {},
+      "limits": "No completed output or completed duration; completed stage observations do not establish delivered video."
+    },
+    "cache": {
+      "immediate_warm_claim": false
+    },
+    "resources": {
+      "measurement_gap": "No complete per-request RAM/VRAM extrema summary retained in this public projection; failure-specific allocation/guard limits appear in findings."
+    },
+    "verification": {
+      "full_decode": false,
+      "checksum": false,
+      "acknowledged": false,
+      "trace_attachment": false,
+      "native_experiment": false,
+      "quality_scope": "No accepted media."
+    },
+    "findings": [
+      "User cancelled the slow FP32 baseline repeat and requested Q8 next."
+    ],
+    "media": []
+  },
+  {
+    "id": "ltx-failure-01",
+    "label": "LTX-Video 2B 0.9.8 distilled Q8_0 · cancelled",
+    "group": "ltxv2b-q8-text-v1",
+    "kind": "video",
+    "model_id": "ltx",
+    "model": "Lightricks/LTX-Video:2B-0.9.8-distilled-Q8_0",
+    "seed": 42,
+    "status": "cancelled",
+    "accepted_result": false,
+    "prompt": "A realistic red fox walks steadily across a snowy clearing from the left side of the frame to the right. Show the fox in full body from the side. Its four legs visibly alternate with each walking step, its body advances across the ground, and its bushy tail swings gently. Near the end the fox stops walking and turns its head toward the camera. The camera stays completely fixed in one wide shot. Continuous visible animal movement throughout the walk. Natural daylight, realistic anatomy, one continuous five-second shot, no cuts, no text.",
+    "prompt_details": {
+      "source": "Saved submitted-run ledger."
+    },
+    "parameters": {
+      "width": 512,
+      "height": 320,
+      "frames": 121,
+      "fps": 24,
+      "template": "ltxv2b-q8-text-v1",
+      "steps": 8,
+      "cfg": 1,
+      "sampler": "euler",
+      "sigmas": "1.0,0.9937,0.9875,0.9812,0.975,0.9094,0.725,0.4219,0.0",
+      "math": "FP32"
+    },
+    "timing": {
+      "factory_trace_seconds": null,
+      "stages_seconds": {
+        "generation_stage_seconds": 79.98499999998603,
+        "decoder_stage_seconds": 46.48399999999674,
+        "text_encoder": 44.578
+      },
+      "limits": "No completed output or completed duration; completed stage observations do not establish delivered video."
+    },
+    "cache": {
+      "immediate_warm_claim": null
+    },
+    "resources": {
+      "measurement_gap": "No complete per-request RAM/VRAM extrema summary retained in this public projection; failure-specific allocation/guard limits appear in findings."
+    },
+    "verification": {
+      "full_decode": false,
+      "checksum": false,
+      "acknowledged": false,
+      "trace_attachment": false,
+      "native_experiment": false,
+      "quality_scope": "No accepted media."
+    },
+    "findings": [
+      "system drive free below6GiB; overlappingCPUCogloaderpreflight also active; avoid overlapping setup in retry"
+    ],
+    "media": []
+  },
+  {
+    "id": "ltx-failure-02",
+    "label": "LTX-Video 2B 0.9.8 distilled Q8_0 · cancelled",
+    "group": "ltxv2b-q8-text-v2",
+    "kind": "video",
+    "model_id": "ltx",
+    "model": "Lightricks/LTX-Video:2B-0.9.8-distilled-Q8_0",
+    "seed": 42,
+    "status": "cancelled",
+    "accepted_result": false,
+    "prompt": "A realistic red fox walks steadily across a snowy clearing from the left side of the frame to the right. Show the fox in full body from the side. Its four legs visibly alternate with each walking step, its body advances across the ground, and its bushy tail swings gently. Near the end the fox stops walking and turns its head toward the camera. The camera stays completely fixed in one wide shot. Continuous visible animal movement throughout the walk. Natural daylight, realistic anatomy, one continuous five-second shot, no cuts, no text.",
+    "prompt_details": {
+      "source": "Saved submitted-run ledger."
+    },
+    "parameters": {
+      "width": 512,
+      "height": 320,
+      "frames": 121,
+      "fps": 24,
+      "template": "ltxv2b-q8-text-v1",
+      "steps": 8,
+      "cfg": 1,
+      "sampler": "euler",
+      "sigmas": "1.0,0.9937,0.9875,0.9812,0.975,0.9094,0.725,0.4219,0.0",
+      "math": "FP32"
+    },
+    "timing": {
+      "factory_trace_seconds": null,
+      "stages_seconds": {
+        "generation_stage_seconds": 79.71900000004098,
+        "decoder_stage_seconds": 44.46899999998277,
+        "text_encoder": 34.719
+      },
+      "limits": "No completed output or completed duration; completed stage observations do not establish delivered video."
+    },
+    "cache": {
+      "immediate_warm_claim": null
+    },
+    "resources": {
+      "measurement_gap": "No complete per-request RAM/VRAM extrema summary retained in this public projection; failure-specific allocation/guard limits appear in findings."
+    },
+    "verification": {
+      "full_decode": false,
+      "checksum": false,
+      "acknowledged": false,
+      "trace_attachment": false,
+      "native_experiment": false,
+      "quality_scope": "No accepted media."
+    },
+    "findings": [
+      "C drive free below6GiB again, without CPU preflight overlap; D scratch and Windows paging capacity configured afterward"
+    ],
+    "media": []
+  },
+  {
+    "id": "cog-failure-01",
+    "label": "CogVideoX 2B local Q8_0 · failed",
+    "group": "cogvideox2b-q8-text-v1",
+    "kind": "video",
+    "model_id": "cog",
+    "model": "zai-org/CogVideoX-2b:Q8_0",
+    "seed": 42,
+    "status": "failed",
+    "accepted_result": false,
+    "prompt": "A realistic red fox walks steadily across a snowy clearing from the left side of the frame to the right. Show the fox in full body from the side. Its four legs visibly alternate with each walking step, its body advances across the ground, and its bushy tail swings gently. Near the end the fox stops walking and turns its head toward the camera. The camera stays completely fixed in one wide shot. Continuous visible animal movement throughout the walk. Natural daylight, realistic anatomy, one continuous six-second shot, no cuts, no text.",
+    "prompt_details": {
+      "source": "Saved submitted-run ledger."
+    },
+    "parameters": {
+      "width": 512,
+      "height": 320,
+      "frames": 49,
+      "fps": 8,
+      "template": "cogvideox2b-q8-text-v1",
+      "steps": 30,
+      "cfg": 6,
+      "attention": "native split-attention allocation path"
+    },
+    "timing": {
+      "factory_trace_seconds": null,
+      "stages_seconds": {},
+      "limits": "No completed output or completed duration; completed stage observations do not establish delivered video."
+    },
+    "cache": {
+      "immediate_warm_claim": null
+    },
+    "resources": {
+      "measurement_gap": "No complete per-request RAM/VRAM extrema summary retained in this public projection; failure-specific allocation/guard limits appear in findings."
+    },
+    "verification": {
+      "full_decode": false,
+      "checksum": false,
+      "acknowledged": false,
+      "trace_attachment": false,
+      "native_experiment": false,
+      "quality_scope": "No accepted media."
+    },
+    "findings": [
+      "native split-attention CUDA OOM allocating8.16GiB; no result",
+      "Requested8.16GiB attention allocation exceeded availableGPU memory; no result saved."
+    ],
+    "media": []
+  },
+  {
+    "id": "h3-legacy-failure",
+    "label": "MiniMax H3 · legacy loader failure",
+    "group": "h3-legacy",
+    "kind": "video",
+    "model_id": "h3",
+    "model": "Comfy-Org/MiniMax-H3:W6A8",
+    "seed": 42,
+    "status": "failed",
+    "accepted_result": false,
+    "prompt": "A small red ceramic fox figurine sits on a pale wooden desk beside a green potted plant against a plain cream wall. The camera slowly arcs from left to right around the figurine. Soft morning light enters from the left, with subtle room ambience. Realistic ceramic texture, smooth continuous camera movement, no text.",
+    "prompt_details": {
+      "actual_positive_text": "Create one video clip from the supplied scene description.\n\nSource generation_brief:\nA small red ceramic fox figurine sits on a pale wooden desk beside a green potted plant against a plain cream wall. The camera slowly arcs from left to right around the figurine. Soft morning light enters from the left, with subtle room ambience. Realistic ceramic texture, smooth continuous camera movement, no text."
+    },
+    "parameters": {
+      "width": 384,
+      "height": 256,
+      "frames": 124,
+      "fps": 24,
+      "steps": 20,
+      "sampler": "res_multistep",
+      "scheduler": "simple",
+      "generator_storage": "W6A8",
+      "text_encoder": "Qwen3VL32B INT8ConvRot CPU",
+      "video_decoder": "INT8ConvRot CPU",
+      "audio_decoder": "FP32 CPU",
+      "math": "FP32",
+      "memory_mode": "legacy"
+    },
+    "timing": {
+      "factory_trace_seconds": 5.234758,
+      "stages_seconds": {},
+      "limits": "Legacy5.234758s is failedattempt trace, not generation time. Dynamic completiontime unknown because cancelled."
+    },
+    "cache": {
+      "second_request_submitted": false
+    },
+    "resources": {
+      "measurement_gap": "No complete per-request RAM/VRAM extrema summary retained in this public projection; failure-specific allocation/guard limits appear in findings."
+    },
+    "verification": {
+      "full_decode": false,
+      "checksum": false,
+      "acknowledged": false,
+      "trace_attachment": false,
+      "native_experiment": true,
+      "quality_scope": "No generated video/audio output."
+    },
+    "findings": [
+      "WindowsOSerror1455(pagingfile too small) while legacy CLIPLoader opened encoder; zero media outputs."
+    ],
+    "media": [],
+    "workflow_path": "workflows/h3-legacy-failure.json",
+    "workflow_sha256": "5c7fd74009dedf7cb5dc21d0a58b9f7ccf81c8bb08d27134601299659c3034a3"
+  },
+  {
+    "id": "h3-dynamic-failure",
+    "label": "MiniMax H3 · dynamic cancelled",
+    "group": "h3-dynamic",
+    "kind": "video",
+    "model_id": "h3",
+    "model": "Comfy-Org/MiniMax-H3:W6A8",
+    "seed": 42,
+    "status": "cancelled",
+    "accepted_result": false,
+    "prompt": "A small red ceramic fox figurine sits on a pale wooden desk beside a green potted plant against a plain cream wall. The camera slowly arcs from left to right around the figurine. Soft morning light enters from the left, with subtle room ambience. Realistic ceramic texture, smooth continuous camera movement, no text.",
+    "prompt_details": {
+      "actual_positive_text": "Create one video clip from the supplied scene description.\n\nSource generation_brief:\nA small red ceramic fox figurine sits on a pale wooden desk beside a green potted plant against a plain cream wall. The camera slowly arcs from left to right around the figurine. Soft morning light enters from the left, with subtle room ambience. Realistic ceramic texture, smooth continuous camera movement, no text."
+    },
+    "parameters": {
+      "width": 384,
+      "height": 256,
+      "frames": 124,
+      "fps": 24,
+      "steps": 20,
+      "sampler": "res_multistep",
+      "scheduler": "simple",
+      "generator_storage": "W6A8",
+      "text_encoder": "Qwen3VL32B INT8ConvRot CPU",
+      "video_decoder": "INT8ConvRot CPU",
+      "audio_decoder": "FP32 CPU",
+      "math": "FP32",
+      "memory_mode": "dynamic"
+    },
+    "timing": {
+      "factory_trace_seconds": null,
+      "stages_seconds": {},
+      "limits": "Legacy5.234758s is failedattempt trace, not generation time. Dynamic completiontime unknown because cancelled."
+    },
+    "cache": {
+      "second_request_submitted": false
+    },
+    "resources": {
+      "measurement_gap": "No complete per-request RAM/VRAM extrema summary retained in this public projection; failure-specific allocation/guard limits appear in findings."
+    },
+    "verification": {
+      "full_decode": false,
+      "checksum": false,
+      "acknowledged": false,
+      "trace_attachment": false,
+      "native_experiment": false,
+      "quality_scope": "No generated video/audio output."
+    },
+    "findings": [
+      "Dynamic loader passed prior failure point, but user aborted during slow sampling because anticipated runtime was impractical. No output, no secondrequest."
+    ],
+    "media": [],
+    "workflow_path": "workflows/h3-dynamic-failure.json",
+    "workflow_sha256": "5c7fd74009dedf7cb5dc21d0a58b9f7ccf81c8bb08d27134601299659c3034a3"
+  },
+  {
+    "id": "yue2-cpu-cancelled",
+    "label": "YuE2 · cancelled CPU semanticgeneration probe",
+    "group": "yue2-cpu-probe",
+    "kind": "audio",
+    "model_id": "yue2",
+    "model": "Comfy-Org/YuE2:int8-convrot",
+    "seed": 42,
+    "status": "cancelled",
+    "accepted_result": false,
+    "prompt": "Gentle acoustic folk, warm vocals, soft guitar and piano, relaxed morning atmosphere.",
+    "prompt_details": {
+      "actual_positive_text": "Create one music clip from the supplied musical style description.\n\nSource generation_brief:\nGentle acoustic folk, warm vocals, soft guitar and piano, relaxed morning atmosphere.",
+      "lyrics": "[Verse]\nMorning sunlight fills the room\nLittle flowers start to bloom\n[Chorus]\nTake a breath and start again\nLet the daylight wander in"
+    },
+    "parameters": {
+      "max_duration_seconds": 15,
+      "nar_steps": 32,
+      "semantic_execution": "CPU"
+    },
+    "timing": {
+      "factory_trace_seconds": 698.101266,
+      "last_observer_elapsed_seconds": 834.844,
+      "stages_seconds": {},
+      "limits": "Closedcancelledattempttrace(11:35:19.587812–11:46:57.689078UTC), not completedmusicgenerationtime. Last native observer834.844s continued aftertraceclosure."
+    },
+    "cache": {
+      "second_request_submitted": false
+    },
+    "resources": {
+      "measurement_gap": "No complete per-request RAM/VRAM extrema summary retained in this public projection; failure-specific allocation/guard limits appear in findings."
+    },
+    "verification": {
+      "full_decode": false,
+      "checksum": false,
+      "acknowledged": false,
+      "trace_attachment": false,
+      "native_experiment": true,
+      "quality_scope": "No audio output."
+    },
+    "findings": [
+      "Cancelled incompleteCPUprobe at observed47/375 semantic tokens. Later successfulGPUAR pair is separate. Expected lease_lost after intentionalcancellation is not evidence of a freshnetworkfailure."
+    ],
+    "media": [],
+    "workflow_path": "workflows/yue2-cpu-cancelled.json",
+    "workflow_sha256": "d55b45fe9f441376981c34ff1890e3956cb945cd72c19a6a430df379cb4235fc"
+  }
+]
@@ -0,0 +1,6 @@
+{
+  "gpu": "NVIDIA GeForce GTX 1060 6 GB",
+  "ram_gib_approx": 28,
+  "os": "Windows 10 Pro 19045",
+  "driver": "582.28"
+}
@@ -0,0 +1,11 @@
+{
+  "schema_version": "media-campaign-1",
+  "archive_run_id": "3cc53336-2555-5e90-99c3-7b94dcb40acc",
+  "source": "Packet Foundry retained evidence",
+  "publication_url": "https://benchmarks.weeyuga.com/benchmarks/predator-media-20261021.html",
+  "article_url": "https://benchmarks.weeyuga.com/findings/images-video-music-on-6gb.html",
+  "recorded_attempts": 130,
+  "retained_outputs": 123,
+  "private_operator_logs_excluded": true,
+  "generated_media_terms": "See README.md and dataset.json model-source/terms links; generated media are excluded from any blanket archive data license."
+}