Exact source and repack lineage of qwen_3_4b.safetensors
Could you identify the exact upstream Qwen model repository/revision and source shard SHA-256s used to produce split_files/text_encoders/qwen_3_4b.safetensors, 8,044,982,048 bytes, SHA-256 6c671498573ac2f7a5501502ccce8d2b08ea6ca2f661c458e708f36b36edfc5a?
This exact artifact appears in Comfy-Org/z_image_turbo commit 429cce00bd1829a3dfd6c57c3204772542876215 and Comfy-Org/vae-text-encorder-for-flux-klein-4b revision 5f526678002e43af5551dadb73ce2e8c91b43afe.
Please identify the conversion/repack tool and commit/version, command/options, tensor renaming/removal or precision changes, tokenizer/config source revisions, and whether that process reproduces this exact output digest. The upload commit mentions huggingface_hub, but does not identify a conversion tool. The tensor keys, dimensions and BF16 byte total match the Qwen3-4B architecture; this alone does not establish the exact weight revision.
For the converted artifact's provenance record, please confirm the applicable upstream license, required attribution/NOTICE files and whether the repack introduced any additional terms. A publisher statement identifying the input and output hashes or a reproducible conversion record would resolve the lineage question.