{"entries":[{"name":"__init__.py","mode":"0100644","size":219,"lastCommit":{"hash":"672a6f101839a150180fc28891ebed379c8f2353","oid":"672a6f101839a150180fc28891ebed379c8f2353","parents":["3807c3de04cde853418033c95e96642876545f3e"],"author":{"name":"Brian","email":"mofosyne@gmail.com","when":"2024-07-18T20:40:15+10:00"},"committer":{"name":"GitHub","email":"noreply@github.com","when":"2024-07-18T20:40:15+10:00"},"message":"convert-*.py: GGUF Naming Convention Refactor and Metadata Override Refactor (#7499)\n\nMain thing is that the default output filename will take this form\r\n\r\n{name}{parameters}{finetune}{version}{encoding}{kind}\r\n\r\nIn addition this add and remove some entries in the KV store and adds a metadata class with automatic heuristics capability to derive some values based on model card content\r\n\r\n* No Change:\r\n - Internal GGUF Spec\r\n - `general.architecture`\r\n - `general.quantization_version`\r\n - `general.alignment`\r\n - `general.file_type`\r\n - General Model Details\r\n - `general.name`\r\n - `general.author`\r\n - `general.version`\r\n - `general.description`\r\n - Licensing details\r\n - `general.license`\r\n - Typically represents the converted GGUF repo (Unless made from scratch)\r\n - `general.url`\r\n - Model Source during conversion\r\n - `general.source.url`\r\n\r\n* Removed:\r\n - Model Source during conversion\r\n - `general.source.huggingface.repository`\r\n\r\n* Added:\r\n - General Model Details\r\n - `general.organization`\r\n - `general.finetune`\r\n - `general.basename`\r\n - `general.quantized_by`\r\n - `general.size_label`\r\n - Licensing details\r\n - `general.license.name`\r\n - `general.license.link`\r\n - Typically represents the converted GGUF repo (Unless made from scratch)\r\n - `general.doi`\r\n - `general.uuid`\r\n - `general.repo_url`\r\n - Model Source during conversion\r\n - `general.source.doi`\r\n - `general.source.uuid`\r\n - `general.source.repo_url`\r\n - Base Model Source\r\n - `general.base_model.count`\r\n - `general.base_model.{id}.name`\r\n - `general.base_model.{id}.author`\r\n - `general.base_model.{id}.version`\r\n - `general.base_model.{id}.organization`\r\n - `general.base_model.{id}.url` (Model Website/Paper)\r\n - `general.base_model.{id}.doi`\r\n - `general.base_model.{id}.uuid`\r\n - `general.base_model.{id}.repo_url` (Model Source Repository (git/svn/etc...))\r\n - Array based KV stores\r\n - `general.tags`\r\n - `general.languages`\r\n - `general.datasets`\r\n\r\n---------\r\n\r\nCo-authored-by: compilade \r\nCo-authored-by: Xuan Son Nguyen ","tree":"0fe1866907aae3695e938839f383826b6f7f36ba"}},{"name":"constants.py","mode":"0100644","size":225378,"lastCommit":{"hash":"279df66449f597f9b91d1976865486ba33e2b984","oid":"279df66449f597f9b91d1976865486ba33e2b984","parents":["9a5c96acb158e92e4c7910b0986c3927025777f0"],"author":{"name":"Chirag Patnaik","email":"chirag.patnaik@gmail.com","when":"2026-09-25T23:52:35+05:30"},"committer":{"name":"GitHub","email":"noreply@github.com","when":"2026-09-25T11:22:35-07:00"},"message":"spec : add DFlash2 support (local convolution + candidate selector) (#27342) (#27816) (#261)\n\n* spec : add DFlash2 support (local convolution + candidate selector) (#27342)\n\n* support DFlash2\n\n* Add p_min in DFlash2\n\nAssisted-by: Claude Opus 5\n\n* Revert unnecessary changes\n\nAssisted-by: Claude Opus 5\n\n* Revert draft sampling in rejection sampling\n\nAssisted-by: Claude Opus 5\n\n* Refactor code structure\n\nAssisted-by: Claude Opus 5\n\n* Delete embedding scaling\n\nAssisted-by: Claude Opus 5\n\n* Gate output transforms on DFlash2\n\nAssisted-by: Claude Opus 5\n\n* Optimize Dflash 2 cost\n\nAssisted-by: Claude Opus 5\n\n* Avoid using atoi\n\nAssisted-by: Claude Opus 5\n\n* Modify comments\n\nAssisted-by: Claude Opus 5\n\n* Move llama_model_dflash_selector_top_k to llama-ext.h\n\nAssisted-by: Claude Opus 5\n\n* Formatting\n\nAssisted-by: Claude Opus 5\n\n* Apply patch to fix the mrope bug\n\nAssisted-by: Claude Opus 5\n\n* fix ci\n\nAssisted-by: Claude Opus 5\n\n* Fix graph number calculation\n\nAssisted-by: Claude Opus 5\n\n* rename hid and unary\n\nAssisted-by: Claude Opus 5\n\n---------\n\n\n\n\n* revert top-k.cu changes\n\n---------\n\n\n\n(cherry picked from commit b10f9ca58c89ccfc3653ac01e979dd085d582b76)\n\nCo-authored-by: Xuan-Son Nguyen \nCo-authored-by: Zihan Zhang \nCo-authored-by: Jian Chen ","tree":"8ac2a17f083de363d9ffe06d35e10c1e1d4eff58"}},{"name":"gguf.py","mode":"0100644","size":478},{"name":"gguf_reader.py","mode":"0100644","size":16321,"lastCommit":{"hash":"0329fcdac8c2477c2dda1d5e43fd2e3616b99655","oid":"0329fcdac8c2477c2dda1d5e43fd2e3616b99655","parents":["6d05498314db1b57f81c271080018aa2d0b89be9"],"author":{"name":"Oğuzhan Akkaya","email":"oakkaya@ymail.com","when":"2026-08-19T02:35:27-04:00"},"committer":{"name":"GitHub","email":"noreply@github.com","when":"2026-08-19T09:35:27+03:00"},"message":"gguf-py : add size guards to GGUFReader (#27188)\n\n* gguf-py : add size guards to GGUFReader\n\nGuard kv_count, tensor_count, string length, and array length\nagainst crafted values that cause unbounded allocation or hangs.\n\nAssisted-by: opencode\n\n* gguf : validate tensor data section fits within file\n\nWhen no_alloc=true, gguf_init_from_reader accepted files where the\ntensor data section (computed from header claims) exceeded the remaining\nfile size. This allowed crafted GGUF files to pass validation while\nhaving insufficient data, leading to OOB reads when the loader later\nmapped tensor data from the file.\n\nAssisted-by: opencode\n\n* gguf-py : move size limits into gguf_reader.py\n\nPer review feedback, the limits are not part of gguf.h but are\narbitrary limits defined in gguf.cpp, so define them locally in\nthe reader instead of exporting them from constants.\n\nAssisted-by: opencode\n\n* remove gguf.ccp changes\n\n---------\n\nCo-authored-by: Sigbjørn Skjæret ","tree":"0580dc85ccba2d400a070c09a3ad8ecc0c6b79ef"}},{"name":"gguf_writer.py","mode":"0100644","size":68407,"lastCommit":{"hash":"279df66449f597f9b91d1976865486ba33e2b984","oid":"279df66449f597f9b91d1976865486ba33e2b984","parents":["9a5c96acb158e92e4c7910b0986c3927025777f0"],"author":{"name":"Chirag Patnaik","email":"chirag.patnaik@gmail.com","when":"2026-09-25T23:52:35+05:30"},"committer":{"name":"GitHub","email":"noreply@github.com","when":"2026-09-25T11:22:35-07:00"},"message":"spec : add DFlash2 support (local convolution + candidate selector) (#27342) (#27816) (#261)\n\n* spec : add DFlash2 support (local convolution + candidate selector) (#27342)\n\n* support DFlash2\n\n* Add p_min in DFlash2\n\nAssisted-by: Claude Opus 5\n\n* Revert unnecessary changes\n\nAssisted-by: Claude Opus 5\n\n* Revert draft sampling in rejection sampling\n\nAssisted-by: Claude Opus 5\n\n* Refactor code structure\n\nAssisted-by: Claude Opus 5\n\n* Delete embedding scaling\n\nAssisted-by: Claude Opus 5\n\n* Gate output transforms on DFlash2\n\nAssisted-by: Claude Opus 5\n\n* Optimize Dflash 2 cost\n\nAssisted-by: Claude Opus 5\n\n* Avoid using atoi\n\nAssisted-by: Claude Opus 5\n\n* Modify comments\n\nAssisted-by: Claude Opus 5\n\n* Move llama_model_dflash_selector_top_k to llama-ext.h\n\nAssisted-by: Claude Opus 5\n\n* Formatting\n\nAssisted-by: Claude Opus 5\n\n* Apply patch to fix the mrope bug\n\nAssisted-by: Claude Opus 5\n\n* fix ci\n\nAssisted-by: Claude Opus 5\n\n* Fix graph number calculation\n\nAssisted-by: Claude Opus 5\n\n* rename hid and unary\n\nAssisted-by: Claude Opus 5\n\n---------\n\n\n\n\n* revert top-k.cu changes\n\n---------\n\n\n\n(cherry picked from commit b10f9ca58c89ccfc3653ac01e979dd085d582b76)\n\nCo-authored-by: Xuan-Son Nguyen \nCo-authored-by: Zihan Zhang \nCo-authored-by: Jian Chen ","tree":"8ac2a17f083de363d9ffe06d35e10c1e1d4eff58"}},{"name":"lazy.py","mode":"0100644","size":9471,"lastCommit":{"hash":"29b28a98242692e352df38254c9a2bb9fa79c884","oid":"29b28a98242692e352df38254c9a2bb9fa79c884","parents":["cea560f483f0f03e828a6c76e78821debdecbe06"],"author":{"name":"Sigbjørn Skjæret","email":"sigbjorn.skjaeret@scala.com","when":"2026-03-21T08:54:34+01:00"},"committer":{"name":"GitHub","email":"noreply@github.com","when":"2026-03-21T08:54:34+01:00"},"message":"ci : switch from pyright to ty (#20826)\n\n* type fixes\n\n* switch to ty\n\n* tweak rules\n\n* tweak more rules\n\n* more tweaks\n\n* final tweak\n\n* use common import-not-found rule","tree":"84fa36d90684ab3c05dad5072f600129a1705610"}},{"name":"metadata.py","mode":"0100644","size":39318,"lastCommit":{"hash":"b3c3b96a139d4ef1bdec926ac17aa040981cfc5d","oid":"b3c3b96a139d4ef1bdec926ac17aa040981cfc5d","parents":["7584430716ee229751771ed0d6bbcb780d105eeb"],"author":{"name":"Tarek Dakhran","email":"tarek@liquid.ai","when":"2026-08-24T16:01:35+02:00"},"committer":{"name":"GitHub","email":"noreply@github.com","when":"2026-08-24T17:01:35+03:00"},"message":"misc : read repetition_penalty from generation_config.json (#27659)\n\n`repetition_penalty` is standard HF key for repetion penalty.\n\nCurrently, only `penalty_repeat` is mapped, read `repetition_penalty`\nand map it to `metadata.sampling_penalty_repeat`.","tree":"352a4e1511264a6f90a1d09c0d2f9c1212591a91"}},{"name":"py.typed","mode":"0100644","size":0},{"name":"quants.py","mode":"0100644","size":64946,"lastCommit":{"hash":"5d246a792ddf839e87fb777681dd2a9ac7f7eb5e","oid":"5d246a792ddf839e87fb777681dd2a9ac7f7eb5e","parents":["63248fc3e33e6f3b579dce6a743fd6ce8939af9c"],"author":{"name":"Sigbjørn Skjæret","email":"sigbjorn.skjaeret@scala.com","when":"2026-05-24T09:51:31+02:00"},"committer":{"name":"GitHub","email":"noreply@github.com","when":"2026-05-24T09:51:31+02:00"},"message":"convert : minor fixes for numpy 2.x (#23571)","tree":"bab313af1b865a906226eeac033295c2846e76c5"}},{"name":"scripts","mode":"0040000","size":0,"lastCommit":{"hash":"18d5846d31513f088a9e4d5be6822b1d57ccebc4","oid":"18d5846d31513f088a9e4d5be6822b1d57ccebc4","parents":["aeeacb3e8c4272c20d408b7dd6ce05daa7e92841"],"author":{"name":"bri-prism","email":"288398250+bri-prism@users.noreply.github.com","when":"2026-08-27T15:36:54-07:00"},"committer":{"name":"GitHub","email":"noreply@github.com","when":"2026-08-27T15:36:54-07:00"},"message":"dflash: ship the legacy dspark re-export script, and fail on a vocab-less drafter (#127)\n\n* dflash: ship the legacy dspark re-export script, and fail on a vocab-less drafter\n\nTwo changes, one purpose: let people move their own legacy drafters onto this\nbranch, and make a botched move obvious.\n\nLegacy drafters declare general.architecture = dspark and do not load here at\nall: \"unknown model architecture: 'dspark'\". They also namespace every head\ndifferently, dspark.markov_head_a/b instead of markov_w1/w2,\ndspark.confidence_head instead of conf_proj, dspark.log_snr_fc1/fc2 without the\nprefix, dspark.hidden_norm instead of enc.output_norm. They carry no vocab, only\na tokenizer.ggml.model = \"none\" stub, and their target_layers are one lower than\nthis branch wants because the runtime taps a layer's input.\n\ngguf_dspark_to_dflash.py does that conversion. Tensor data is copied byte for\nbyte, so nothing is requantized and a quantized drafter stays quantized. The\nvocab is taken from a donor GGUF, normally the target model.\n\nThe transform was derived from two shipped files that already exist in both\nformats, then checked against them: regenerating one reproduces its metadata\nexactly, all 31 tensors match byte for byte under the rename, and it drafts at\nthe same 41.243% acceptance with the same mean length 2.62.\n\nThe guard is the other half. Without a mask token, llama_vocab_mask returns\nLLAMA_TOKEN_NULL and every masked slot in the block is built with token id -1.\nThe drafter loads, runs, and reports timings, but accepts nothing. That is what\na conversion which drops the vocab looks like, and it reads as a bad drafter\nrather than a bad file. It cost a debugging cycle to find, so it now asserts\nwith a message naming the metadata key to check.\n\nVerified on this branch:\n- 14 legacy drafters converted, all load, arch dflash, vocab present, mask token\n 248319, target_layers shifted\n- against a low-bit target, three drafters from different families: 65.891% mean\n 3.58, 63.158% mean 3.47, and 41.243% mean 2.62\n- a file with the vocab stub and no mask token now aborts on load with the new\n message instead of drafting nothing\n\n* gguf-dspark-to-dflash: optional --drop-shared-tensors (#129)\n\n* gguf-dspark-to-dflash: name the real command in the usage line\n\n* speculative: name both tokenizer keys in the mask-token assert, condense comment\n\n---------\n\nCo-authored-by: Pasha Khosravi \nCo-authored-by: Pasha Khosravi ","tree":"3ac913c9586440620b4cf8aa24c43cb423c26224"}},{"name":"tensor_mapping.py","mode":"0100644","size":130488,"lastCommit":{"hash":"279df66449f597f9b91d1976865486ba33e2b984","oid":"279df66449f597f9b91d1976865486ba33e2b984","parents":["9a5c96acb158e92e4c7910b0986c3927025777f0"],"author":{"name":"Chirag Patnaik","email":"chirag.patnaik@gmail.com","when":"2026-09-25T23:52:35+05:30"},"committer":{"name":"GitHub","email":"noreply@github.com","when":"2026-09-25T11:22:35-07:00"},"message":"spec : add DFlash2 support (local convolution + candidate selector) (#27342) (#27816) (#261)\n\n* spec : add DFlash2 support (local convolution + candidate selector) (#27342)\n\n* support DFlash2\n\n* Add p_min in DFlash2\n\nAssisted-by: Claude Opus 5\n\n* Revert unnecessary changes\n\nAssisted-by: Claude Opus 5\n\n* Revert draft sampling in rejection sampling\n\nAssisted-by: Claude Opus 5\n\n* Refactor code structure\n\nAssisted-by: Claude Opus 5\n\n* Delete embedding scaling\n\nAssisted-by: Claude Opus 5\n\n* Gate output transforms on DFlash2\n\nAssisted-by: Claude Opus 5\n\n* Optimize Dflash 2 cost\n\nAssisted-by: Claude Opus 5\n\n* Avoid using atoi\n\nAssisted-by: Claude Opus 5\n\n* Modify comments\n\nAssisted-by: Claude Opus 5\n\n* Move llama_model_dflash_selector_top_k to llama-ext.h\n\nAssisted-by: Claude Opus 5\n\n* Formatting\n\nAssisted-by: Claude Opus 5\n\n* Apply patch to fix the mrope bug\n\nAssisted-by: Claude Opus 5\n\n* fix ci\n\nAssisted-by: Claude Opus 5\n\n* Fix graph number calculation\n\nAssisted-by: Claude Opus 5\n\n* rename hid and unary\n\nAssisted-by: Claude Opus 5\n\n---------\n\n\n\n\n* revert top-k.cu changes\n\n---------\n\n\n\n(cherry picked from commit b10f9ca58c89ccfc3653ac01e979dd085d582b76)\n\nCo-authored-by: Xuan-Son Nguyen \nCo-authored-by: Zihan Zhang \nCo-authored-by: Jian Chen ","tree":"8ac2a17f083de363d9ffe06d35e10c1e1d4eff58"}},{"name":"utility.py","mode":"0100644","size":13534,"lastCommit":{"hash":"8f48807380305a5985df78f67e29862664c9afec","oid":"8f48807380305a5985df78f67e29862664c9afec","parents":["bf6bc3c1551491b5eb019f0fb56c743f25d3eb79"],"author":{"name":"compilade","email":"git@compilade.net","when":"2025-12-22T14:25:16-05:00"},"committer":{"name":"GitHub","email":"noreply@github.com","when":"2025-12-22T20:25:16+01:00"},"message":"gguf-py : do not align the data start offset (#18291)\n\nThe safetensors format doesn't require alignment.","tree":"251fc048a31ae80c93f37d524019675cd61718bc"}},{"name":"vocab.py","mode":"0100644","size":42431,"lastCommit":{"hash":"1bfbdb134e4b983f7cbbde252d004483e31206a2","oid":"1bfbdb134e4b983f7cbbde252d004483e31206a2","parents":["68f30663cf49102a85913ae6e278e6d1e9a1109e"],"author":{"name":"o7si","email":"32285332+o7si@users.noreply.github.com","when":"2026-06-11T15:37:23+08:00"},"committer":{"name":"GitHub","email":"noreply@github.com","when":"2026-06-11T10:37:23+03:00"},"message":"vocab : adopt leading TemplateProcessing special token as BOS (#24428)","tree":"3aa472b76e2b30845d6f96838f3103bd177f93f4"}}],"lastCommit":{"hash":"279df66449f597f9b91d1976865486ba33e2b984","oid":"279df66449f597f9b91d1976865486ba33e2b984","parents":["9a5c96acb158e92e4c7910b0986c3927025777f0"],"author":{"name":"Chirag Patnaik","email":"chirag.patnaik@gmail.com","when":"2026-09-25T23:52:35+05:30"},"committer":{"name":"GitHub","email":"noreply@github.com","when":"2026-09-25T11:22:35-07:00"},"message":"spec : add DFlash2 support (local convolution + candidate selector) (#27342) (#27816) (#261)\n\n* spec : add DFlash2 support (local convolution + candidate selector) (#27342)\n\n* support DFlash2\n\n* Add p_min in DFlash2\n\nAssisted-by: Claude Opus 5\n\n* Revert unnecessary changes\n\nAssisted-by: Claude Opus 5\n\n* Revert draft sampling in rejection sampling\n\nAssisted-by: Claude Opus 5\n\n* Refactor code structure\n\nAssisted-by: Claude Opus 5\n\n* Delete embedding scaling\n\nAssisted-by: Claude Opus 5\n\n* Gate output transforms on DFlash2\n\nAssisted-by: Claude Opus 5\n\n* Optimize Dflash 2 cost\n\nAssisted-by: Claude Opus 5\n\n* Avoid using atoi\n\nAssisted-by: Claude Opus 5\n\n* Modify comments\n\nAssisted-by: Claude Opus 5\n\n* Move llama_model_dflash_selector_top_k to llama-ext.h\n\nAssisted-by: Claude Opus 5\n\n* Formatting\n\nAssisted-by: Claude Opus 5\n\n* Apply patch to fix the mrope bug\n\nAssisted-by: Claude Opus 5\n\n* fix ci\n\nAssisted-by: Claude Opus 5\n\n* Fix graph number calculation\n\nAssisted-by: Claude Opus 5\n\n* rename hid and unary\n\nAssisted-by: Claude Opus 5\n\n---------\n\n\n\n\n* revert top-k.cu changes\n\n---------\n\n\n\n(cherry picked from commit b10f9ca58c89ccfc3653ac01e979dd085d582b76)\n\nCo-authored-by: Xuan-Son Nguyen \nCo-authored-by: Zihan Zhang \nCo-authored-by: Jian Chen ","tree":"8ac2a17f083de363d9ffe06d35e10c1e1d4eff58"}}