c931ef3163940227c9b6291e04216245cce21429	a43741244a646d3f8fa3ff01bb9b7d059223d0a5	2026-05-18T17:34:37+08:00	wangbomeng		calrt: fix prefill_by_ids.fix create_buf
eeeef3f6d6a5aad9296323ca15d9d929745b8932	431ad6f1c787f1b47e9e7be14b2e1a7d8565485c	2026-04-28T16:16:19+08:00	wangbomeng		add unknown time info: prefill and decode
99962021f2dd1994dbbf90d5712f0ca2633f20f8	9af6682ef101c6d006c743c703cf150ac25f466f	2026-04-17T14:48:33+08:00	wangbomeng	origin/dev-ot	calrt: add onetoken slice and untile_prefill
a206460e610aef995e36e50b33e8bd41fe477541	d8b2aaa00049978cebac15041f960d75b552f97b	2026-04-14T10:23:43+08:00	wangbomeng		calrt: support mixture of prefill_by_ids and prefill_by_embeds
b8153b6b8f244b223c9f4c2940ae1f212634519e	82842cfc3fa3d5c004c4540aeff8a81f38509bee	2026-04-09T15:07:02+08:00	wangbomeng		server: fix limit tokens to max_seq_len
82842cfc3fa3d5c004c4540aeff8a81f38509bee	622a22991089c6debd5f4b57738f3df3a0bda8b5	2026-04-09T15:06:26+08:00	wangbomeng		server: fix limit tokens to max_seq_len
1e8a16cd0e3ad3c9a0f8746852a0a38c6e52322e	893e52bda0fee99bbda1521ea733a760bc4201f1	2026-04-01T10:14:22+08:00	wangbomeng		calrt: fix slice and add infer/copy time
e4eaab700333490e80ebefc8d023f6c1adfcc312	aeacc23a883400db0ebffb50e23355ada054da66	2026-03-25T09:03:55+08:00	wangbomeng		calrt: prefil to prefill
aeacc23a883400db0ebffb50e23355ada054da66	cbaa7492663630132adeddccd9fec5e44ffa3336	2026-03-25T08:49:22+08:00	wangbomeng		calrt: add slice only on decode
3c08ebf0e442cd730ddffd959ec676f9caf31c82	613bc2b9a9d7af92efbe42d876ffcdb244f7a00e	2026-03-24T02:43:03+08:00	wangbomeng		calrt: comment LOG_ERROR in untile_decode_cpy
16cbf81a731f5839a2f6b0eb9d8539826353748b	8b4e17d990c23025148c4abadefe1010a0dd0734	2026-03-23T09:38:21+08:00	wangbomeng		calrt: fix model_name of untile_cpy
b2c45d8ddefb7ec09435231c8ec5b736fabf9c91	c6ea3f0d2468e3b8eab5de80bacee7fe4c9dd3d7	2026-03-09T15:03:26+08:00	Bomeng Wang		clart: past_kv_len = n_tokens
583c387efaf7bc030574e554088de79d09a27fbd	0edc14a60a2d5596f71d728c655e3523da924d89	2026-02-27T02:24:56+08:00	wangbomeng		caserver: lim generated tokens to n_ctx_per_seq
0edc14a60a2d5596f71d728c655e3523da924d89	aa0a17d719326a63e0d6fea60aa0ced44e7abc33	2026-02-27T01:57:39+08:00	wangbomeng		caserver: limit generated tokens to n_ctx_per_seq
84f2c0b5f334a7ab77347c6779b6027a951d0ff0	0bd95719fed2f0247c8d0199dd897d24157dfae6	2026-02-09T01:46:36+08:00	wangbomeng		add bf16_to_fp32 in untile_decode/prefill_cpy
9328f21378275f2354a19c3af907332b216ab492	49fea70a20f550bfdb59d0ffe041b54919c22447	2026-01-28T15:32:36+08:00	Yunzhe Jia		add prefill-only mode
5d2387931528fbe04aa8d66de935147efb65ca4d	e6285a748657add8d974b78ca920c5b41753ee12	2025-12-08T16:40:34+08:00	Yunzhe Jia		add timestamp for one decode process
e492f5dc8cec17529f53ef2bbdf7b502107f8148	2e05afd22b3d77c7013b11eee88e88b17188f21a	2025-11-04T14:09:59+08:00	Yunzhe Jia		redo untile logic
070ff4d5356083d60b807bb34d36b31c3653a29e	bf7b0c9725ed0c406c5debaea022ec7258430634	2025-11-03T11:11:18+01:00	Xuan-Son Nguyen		mtmd: add --image-min/max-tokens (#16921)
cd5e3b57541ecc52421130742f4d89acbcf77cd4	87c9efc3b297b8a498716b1db3d061842e6fc85b	2025-11-02T18:14:04+02:00	Georgi Gerganov		server : support unified cache across slots (#16736)
cf659bbb8ef9eb048e5153a27cf787fd83c05560	d8b860a219c2415faac8cc0e50b48b4aa11e3b64	2025-11-01T15:51:36+01:00	Xuan-Son Nguyen		mtmd: refactor preprocessing + support max/min pixels (#16878)
b52edd25586fabb70f0c21b274473b307cf14499	517b7170e1a4d733583c4b07c5b7a49acc05911c	2025-10-30T18:42:57+02:00	Georgi Gerganov		server : remove n_past (#16818)
5a91109a5d7dab5d7adc40bedb397ede99a705b1	f8f071faddf32ea09f4234edb6e809b380a9ee26	2025-10-24T12:02:02+02:00	Daniel Bevenius		model-conversion : add trust_remote_code for orig model run [no ci] (#16751)
84bf3c677857279037adf67cdcfd89eaa4ca9281	c9c1972e2c2cc6a771fcc145bfa138700179f961	2025-10-20T21:38:20+02:00	Sigbjørn Skjæret		model : add BailingMoeV2 support (#16063)
17304cbcc1dd24de7741cbe57925d58e90a98ac1	3e3cb19f6449f9168a128eaeae01f8f41b049acc	2025-10-15T16:53:12+03:00	Georgi Gerganov		server : fix img token logs (#16595)
bc07349a7f87ba6eb31ed4b0ea9d9a7352185213	e60f241eacec42d3bd7c9edd37d236ebf35132a8	2025-10-14T08:48:50+03:00	Georgi Gerganov		server : dynamic token limit for prompt cache (#16560)
477a66b03501cf3bd067f8968b77ca4d053ff1bd	e60f01d941bc5b7fae62dd57fee4cec76ec0ea6e	2025-10-11T11:33:41+03:00	amirai21		convert : correctly handle LLaMA tokenizer for Jamba (#16470)
81086cd6a3ca1252f0dc0f938171648399179c53	68ee98ae181a5c83a5cc6261daeee69a1f588c15	2025-10-10T17:17:31+03:00	Georgi Gerganov		vocab : mark EOT token for Granite models (#16499)
1faa13a1187051af66b0fd9f0d6effe4c77f0b3e	1deee0f8d494981c32597dca8b5f8696d399b0f2	2025-10-09T22:54:57+02:00	Pascal		webui: updated the chat service to only include max_tokens in the req… (#16489)
c08002a1988348403a5fd59b1fa3de3a10a6f92f	3a002afafa8e06555f11aed88b02d055d8a166f3	2025-10-06T10:59:40-06:00	Gabe Goodhart		chat : Granite Docling stopping (#16438)
ca71fb9b368e3db96e028f80c4c9df6b6b370edd	35266573b968e1c947b367782fb4b3eddbb4f3c0	2025-10-05T06:57:47-06:00	Gabe Goodhart		model : Granite docling + Idefics3 preprocessing (SmolVLM) (#16206)
77233277c912d495d1069543ca540c21a2f9e5b4	e308efda8e54e3f07578937a45a5d83624eb7970	2025-10-03T11:30:39+02:00	Aleksander Grygier		Capture model name only after first token (streaming) or completed request (#16405)
e74c92e84236b2bab3f3c77bee4ead94928be360	b2ba81dbe07b6dbea9c96b13346c66973dede32c	2025-09-30T16:24:36-04:00	Bartowski		model : support GLM 4.6 (make a few NextN/MTP tensors not required) (#16359)
2f61c0f5bf8a620ca4c3872408803ab38cfb9613	3ffd0fae473c954bb3e67526b31262048fb508d4	2025-09-29T12:33:12+05:30	Vinkal		llama-cli: prevent spurious assistant token (#16202)
624207e676ab5eb3ce7af631902bb45fb73a8359	807e8c6d310952f2f5656afc63dab9d7083dcb5c	2025-09-27T02:03:33+08:00	Aaron Teo		devops: add s390x & ppc64le CI (#15925)
077c94d0caf87fbd3cf3288dbb5c0fd9670294cf	aa3ee0eb0b80efca126cedf9bcb4fb5864b46ce3	2025-09-25T22:35:05+08:00	Aman Gupta		CUDA: add a fused top-K MoE kernel (#16130)
aa3ee0eb0b80efca126cedf9bcb4fb5864b46ce3	d0991da39d3c39b3980c24cdfccb9a4c95a46870	2025-09-25T12:02:36+02:00	Daniel Bevenius		model-conversion : add embedding prompt file support (#15871)
351f3da39c85f59d581fc184f09283da7f099a3b	3ecb2f671a2f49d56357f99d135a94e841759178	2025-09-23T01:57:46+08:00	Haiyue Wang		clang-tidy : disable warning about performance enum size (#16127)
745cbcf2fe1eb88f8db615ac622f0b944d924ad6	1cbd80f8cf80a817715b1ccc5680fe2a3c5172c8	2025-09-17T15:30:55+08:00	Jie Fu (傅杰)		llama-quant : fix the verification of attention layers for encoder-decoder models (#16023)
1cbd80f8cf80a817715b1ccc5680fe2a3c5172c8	85286f354813056f6c835046c0acfa3bf6ba9432	2025-09-17T15:29:00+08:00	Jie Fu (傅杰)		examples : support encoder-decoder models in the simple example (#16002)
c0389dba43d50695f9d3f57dd1f1a14cbefc100c	00681dfc16ba4cebb9c7fbd2cf2656e06a0692a4	2025-09-11T15:59:37+08:00	hipudding		CANN: Disable acl_graph for prefill stage (#15933)
4f658855fa8f2e42b7ed9a5b298fa39a2e39b096	6ab397e12ba8e9f776341cdae68f7ffb2f8d2cde	2025-09-11T02:51:51+08:00	Jie Fu (傅杰)		llama : support T5 models with unequal number of encoder-decoder layers (#15909)
70cd37dbbebdb7a2f84f08207f18eabb0b291a55	acc1b008cfd95e63d5f99a370dbffb98e5a99d2c	2025-09-09T06:06:52+02:00	Daniel Bevenius		requirements : update transformers/torch for Embedding Gemma (#15828)
88021565f08e0b7c4e07ac089a15ec16fae9166c	56920f56651908d5cce7a310dabf54ac4f6fbb7f	2025-09-08T10:59:48-04:00	Jesse		chat : Deepseek V3.1 reasoning and tool calling support (OpenAI Style) (#15533)
56920f56651908d5cce7a310dabf54ac4f6fbb7f	b0d52998b962bd2681c34bf52af993af79f178b8	2025-09-08T21:50:05+07:00	Xuan-Son Nguyen		server : bring back timings_per_token (#15879)
fd621880f3fd908424e675a41715a2dc760247a2	4281c7b315f8a904549fe527039b976e08098d1a	2025-09-05T17:32:39-06:00	Gabe Goodhart		aLoRA Support (#15327)
5fac79cbc77b6d12c9feb5f34fc63586b35fd561	408ff524b40baf4f51a81d42a9828200dd4fcb6b	2025-09-05T14:31:24-06:00	Gabe Goodhart		Thinking model disabled assistant prefill (#15404)
3dc7397a2799bdc07bccf637ab7ae5a1e786d1a4	e92d53b29e393fc4c0f9f1f7c3fe651be8d36faa	2025-09-01T08:57:00+08:00	hipudding		CANN: fix RoPE cache issue on multi-device (#15629)
e8d99dd0b67f2ecc1e45fca8074a3a18c3e036d2	a8bca68f727844e7dcf24a956003b3c2039ea563	2025-08-28T18:39:31-06:00	Gabe Goodhart		nvidia nemotron nano v2 (nemotronh) (#15507)
84ab83cc0b4b7e769451ee48e4c7d1acef91ef25	55042b3692cb1467c9ee15c62c4a9fbf180f89e3	2025-08-28T15:49:50+02:00	Sigbjørn Skjæret		model : jina-embeddings-v3 support (#13693)
1e7489745a74996fc36e8fd05b73aa16bc184e0c	1cf123a343ab7ca5586aacb9e0a1d2de7fe33be4	2025-08-27T17:21:41+08:00	Chenguang Li		CANN: refactor mask handling and improve performance in FA (#15561)
a6a58d64785cb458ed9de52f391aa38142d38d64	0373486dbc0dccbdcb3b5fdd65759d88cec06196	2025-08-26T21:05:25+05:30	shalinib-ibm		llamafile: PowerPC Sgemm Optimization (#15558)
62cef26ac5b6b7acb635d3dc963813b43952dc2b	8f5afa94c4f929da71f560db7c9f38ef6a783d95	2025-08-26T16:12:29+02:00	Daniel Bevenius		model-conversion : add qat-q4 quantization targets (#15588)
79a546220c719e6a70627b243a478ab8d84dc9e1	85cc1ae998e4898d9fa992cb9b8620338cee97bf	2025-08-26T12:54:19+02:00	Xuan-Son Nguyen		mtmd : support Kimi VL model (#15458)
0115cfb7c1dec0753aecb34f3af4338bf9de8eeb	3a617cb18121a85eed17ac5585788810c5d9c1e0 0d5a470223fc90b6b6807921d68011ff06ae7f9e	2025-08-26T15:19:56+08:00	Yunzhe Jia		Merge branch 'master' into dev
6b64f74b55628e4193f4fb00313f07dbd8556528	0d5a470223fc90b6b6807921d68011ff06ae7f9e	2025-08-25T13:56:43+03:00	Georgi Gerganov		batched-bench : fix unified KV cache handling + pp timing (#15562)
4afb0a746f22abaa545b3ebdb76a400d7da3a713	e288693669cf9d0a71e2f2b8bd57305f06340257	2025-08-22T08:10:14Z	65a		server : Support multimodal completion and embeddings prompts in JSON format (#15108)
e288693669cf9d0a71e2f2b8bd57305f06340257	a0f98dd604d34826eb5ea2560d1e23fe726921df	2025-08-22T09:29:08+02:00	Tarek Dakhran		readme : model : mtdm : lfm2 improvements (#15476)
657b8a77bd01854f99d37a47318fa24f2e7e298f	ec5ab1a36c11dd3efcf4ec8d1ac89a13a8117bc3	2025-08-20T14:26:01+02:00	Daniel Bevenius		chat: handle gpt-oss return/end token inconsistency (#15421)
f0d3c7405c323784a60f14ddddfbac3f7404d417	f08c4c0d8d0cb6caaf8b7ad316039232b9fa059c	2025-08-19T08:45:12+03:00	Georgi Gerganov		batched-bench : use rand tokens (#15398)
f08c4c0d8d0cb6caaf8b7ad316039232b9fa059c	6d7f1117e3e3285d0c5c11b5ebb0439e27920082	2025-08-18T22:53:52+02:00	Xuan-Son Nguyen		mtmd : clean up clip_n_output_tokens (#15391)
5e6229a8409ac786e62cb133d09f1679a9aec13e	e2c1bfff5305c661ac53e9d57cb732ff626a2242	2025-08-15T19:50:52+02:00	Daniel Bevenius		common : fix double bos, use common_chat_templates for add_bos and add_eos (#15326)
b204a5a234f4cf0b3f449476da639a5510c6d157	646944cfa8961afd914dd6637739b3cda9a72e11	2025-08-14T09:23:11-05:00	Aldehir Rojas		gpt-oss: implement harmony parsing (#15181)
3a617cb18121a85eed17ac5585788810c5d9c1e0	c42293848dd9f03eba9a6c1b85a600b398f06541	2025-08-13T10:58:59+08:00	Yunzhe Jia		add calrt_decode
1ebbaddff2a44b0599df659175d3274bd5bbeb81	a3a7874272e5a060079658eb5cca4617b7f99062	2025-08-11T10:21:24+02:00	Daniel Bevenius		perplexity : update comments/error msg to use decode [no ci] (#15227)
99acbc9921b119aa7ed929eb5780a66a8f06e6d9	7ad67ba9fe2b909e271dd31b99c5fce3aba35899	2025-08-08T00:20:40+08:00	RunningLeon		llama : Support intern-s1 (#14875)
1cbf442f50a154aac0bf35d7b8b3436a2cdfd6f5	ba67d6c03fed3193987a4ef13050e6e2a4451df4	2025-07-24T15:50:19+08:00	Yunzhe Jia		1.sync with calrt prefill,decode,kv_update op 2.add hw pattern untile function
fd1234cb468935ea087d6929b2487926c3afff4b	f324a3b715d5c1081c110ce459f8a8486fb1ee89	2025-08-05T22:10:36+03:00	Georgi Gerganov		llama : add gpt-oss (#15091)
ef0144c087b33e5b8da42d529ac71aaf05cb49df	2721257e3e2c4c944ac8a08221113ee7cb503f1b	2025-08-05T04:29:25+10:00	Sam		model: support GLM 4.5 family of models (#14939)
97366dc6abdd0bdc74260bd3c42bd06f0feb7428	83bc2f288c0e08e676d9beca9c4669197e920593	2025-08-03T12:38:18-07:00	Csaba Kecskemeti		vocab : JetBrains Mellum pre-tokenizer (#15045)
83bc2f288c0e08e676d9beca9c4669197e920593	6c7a441161080551ce8a52ba32563b6295067192	2025-08-03T09:56:25-05:00	Gabriel Larson		model : add text-only support for Kimi-VL (and find special tokens in text_config)  (#15051)
2bf3fbf0b54f97aef2b388b76d222789e1c170f1	711d5e6fe66eb6cd7a10d71cec4567321848be08	2025-08-02T14:39:01+02:00	Sigbjørn Skjæret		ci : check that pre-tokenizer hashes are up-to-date (#15032)
711d5e6fe66eb6cd7a10d71cec4567321848be08	f738989dcb9ccbe468c945553eafbeef7b869675	2025-08-02T05:51:02-05:00	Douglas Hanley		convert : fix Qwen3-Embedding pre-tokenizer hash (#15030)
f906275537d14c8fc7c6976d944233771fd6672c	a9f7541ec25c4c8547daf5ff48700ad2836e2b7d	2025-08-02T10:12:41+02:00	Johannes Gäßler		server: enable token array inputs for OAI API (#15001)
0f5ccd6fd1a1f709010312933db0316867cc30b6	1c872f71fb8a25589efa3ee9b6bf8b517cb8caa4	2025-08-01T21:31:12+08:00	stevenkuang		model : add hunyuan dense (#14878)
94933c8c2eeaa9a7983e3f6c08af76bd86724094	c1dacaa99b4ead6edbac928cd2da59436573f6b0	2025-07-31T05:25:23-07:00	g2mt		server : implement universal assisted decoding (#12635)
8a4a85627702b569d7d2810f2de06a4321656e9d	11490b36723d511d75fb601995c79b5c363ba3a2	2025-07-31T19:49:09+08:00	Aman Gupta		Add LLaDA 8b Diffusion model (#14771)
92b8810ec7aa6d778bc287cc918443cf67b962e2	00131d6eaf4df029e1ec84de868c2c5957503007	2025-07-30T15:46:13+02:00	Johannes Gäßler		CUDA: skip masked KV slices for all FA kernels (#14924)
00fa15fedc79263fa0285e6a3bbb0cfb3e3878a2	946b1f685909c8c9c044f145bce819c02f327eaa	2025-07-28T15:01:48+02:00	Xuan-Son Nguyen		mtmd : add support for Voxtral (#14862)
ca0ef2dddb022cb1337d775cd05cd27d7808aff4	89d1029559bd2968f76db854f9f113d73e34527c	2025-07-27T12:10:51+02:00	Daniel Bevenius		llama : clarify comment about pp and tg graphs [no ci] (#14895)
793c0d7f46384001738c337d7afa46b45ae32745	ce111d39d666f4a3b6a561ad020f6feb8cc67790	2025-07-25T10:47:39-06:00	Gabe Goodhart		metal: SSM_SCAN performance (#14743)
b4efd77f8ab407836ca73a5176f041650c5b2411	2be60cbc2707359241c2784f9d2e30d8fc7cdabb	2025-07-21T09:24:51+02:00	IsaacDynamo		server : add parse_special option to /tokenize endpoint (#14783)
09651d09ffc1e941bd1be23163abf5495c416547	349ea79fcebc75b0c55bf61594a47736966d4f95	2025-07-18T06:25:54+02:00	Nexes the Elder		graph : Pass the graph placeholder message in debug mode (#14748)
cb887f1bc1001c92f7b4a595b9014f3a454a07ab	d6fb3f6b49b27ef1c0f4cf5128e041f7e7dc03af	2025-07-17T23:15:32+02:00	Piotr Wilkin (ilintar)		model: add Ernie 4.5 MoE support (#14658)
19e5943d9e976b59ccd5a0a87ae3d2ff3da44390	496957e1cbcb522abc63aa18521036e40efce985	2025-07-16T23:17:43+02:00	Sigbjørn Skjæret		convert : make hf token optional (#14717)
4b91d6f71f14040979bcdb7b6729b3bca93ec1c1	cf91f217f1c8b98b6db8e4ba6b480f017f81d206	2025-07-16T08:52:04+02:00	Sigbjørn Skjæret		convert : only check for tokenizer folder if we need it (#14704)
68e37a61a7b24863f67541db65b4f6195962a268	cbc68be51d88b1d5531643b926a4b359c3cff131	2025-07-16T01:11:42+09:00	Shunta Saito		model : add PLaMo-2 support (#14560)
576c82eda210ca0111c04f5256bf77897a4d4cc4	0aedae00e6fb48680324a5ac5da9cba0e35de6b5	2025-07-11T16:36:04+09:00	Dowon		vocab : add midm-2.0 model pre-tokenizer (#14626)
0aedae00e6fb48680324a5ac5da9cba0e35de6b5	6bdda13981d6c8189b7dc4f9fb8ecb91c21529f8	2025-07-10T18:20:13-06:00	Gabe Goodhart		model : Granite Four (#13550)
4bb625b713fd9b294b4f7af87eaa752b710c7cc1	11ee0fea2a24da1d3206eeba8fc52b759d9dfb24	2025-07-10T13:41:00-04:00	Ryan Mangeno		Smoldocling support (#14597)
8f22dc0a53338c629c1ef8fa878d8e39bfe627c9	53903ae6fa5f1caf187889c839cdd1ad25da4018	2025-07-08T10:24:06+02:00	Xuan-Son Nguyen		model : add hunyuan moe (#14425)
ddef99522d1ba74193b7394e803fab8db5c78bae	668168814601d2aef6161ce49ab186bc74177ae7	2025-07-05T09:17:14+02:00	Sigbjørn Skjæret		server : fix assistant prefilling when content is an array (#14360)
5d46babdc2d4675d96ebcf23cac098a02f0d30cc	e17991c466ac835b2c71bd813c1ca7ff8dd97b94	2025-07-02T13:10:24-04:00	compilade		llama : initial Mamba-2 support (#9126)
caf5681fcb47dfe9bafee94ef9aa8f669ac986c7	83790b0e7e09ab17238b16452a33053a71dbdfad	2025-06-29T20:02:53+02:00	matteo		server : support jinja extra template kwargs (Qwen3 enable_thinking feature), from command line and from client (#13196)
8f52c6f8abb2ed9ad967cc0bfb90acb073011193	d2f325130c4ed77a3dcff2c23a926587fb08e2cc	2025-06-26T15:10:45+08:00	Yunzhe Jia		Squashed commit of the following:
66aba7aca9a245d71f1ddf02c7a97223529752a8	f1f5e82df6222dcbaca6396c0de44df259a1694f	2025-06-23T01:28:06+08:00	Ruikai Peng		run : avoid double tokenization (#14327)
f1f5e82df6222dcbaca6396c0de44df259a1694f	af3373f1adfca56119f3e4de0e6a0a8df8edf3d9	2025-06-22T20:10:07+03:00	Georgi Gerganov		examples : fix is_first logic for tokenization (#14329)
aa0ef5c578eef4c2adc7be1282f21bab5f3e8d26	bb16041caef45cd4348cd6f84906b5dfec7a1f6a	2025-06-21T18:12:05+02:00	Sigbjørn Skjæret		gguf-py : fix Qwen3-Embedding eos token (#14314)
dd6e6d0b6a4bbe3ebfc931d1eb14db2f2b1d70af	8308f98c7fb778e54bf75538f5234d8bd20915e9	2025-06-20T22:13:06+08:00	Ruikai Peng		vocab : prevent tokenizer overflow (#14301)
88fc854b4bd2e3caf10e705e6afcbbca136f0a3c	e28c1b93fd7d3f8faf9551d962e8a65fe2122e38	2025-06-20T14:04:09+02:00	Sigbjørn Skjæret		llama : improve sep token handling (#14272)
812939a9e90f99d1bd5bb1bc6b99d12600671d50	4c9fdfbe1580a66fd7d77c77418ce2c606a29fdd	2025-06-20T10:50:27+03:00	Georgi Gerganov		model : more uniform output id handling (#14275)
d2f325130c4ed77a3dcff2c23a926587fb08e2cc	6391ee7ffd97e723dfbdcc32f001e1b95a63cba6	2025-06-17T11:53:33+08:00	Yunzhe Jia		calrt: add input_token in graph; input_token will be copy to calrt_in_buffer
f6e1a7aa8787b5c00acba6370cb70a0beff48b1e	c3ee46fab49a765d2e32e171e9ed7a5fa121dd9c	2025-06-12T11:50:01+03:00	Georgi Gerganov		context : simplify output counting logic during decode (#14142)
a20b2b05bce6622c585459ebf46f142f113d021c	2e89f76b7af2c0b827be785e445f2e2b3e52e1ca	2025-06-12T02:56:04-04:00	compilade		context : round n_tokens to next multiple of n_seqs when reserving (#14140)
cc66a7f78f95dcb5208420f4dd1abc3fc6aec0cc	bd248d4dc7265437e1918dc53ae3f49a0c592e5f	2025-06-11T17:16:32+02:00	Sigbjørn Skjæret		tests : add test-tokenizers-repo (#14017)
dad5c44398b78467943ed0a603ea427fe9f6fc62	55f6b9fa6563f6ae49113f9abdc980c12348cc1c	2025-06-10T18:20:14-04:00	compilade		kv-cache : avoid modifying recurrent cells when setting inputs (#13834)
9f47fa5792bae5312615439e68dbf1826913f7ac	9e31bec4fd53634c9e5b04650488a09a055f5dab	2025-06-05T09:29:18+02:00	Sigbjørn Skjæret		vocab : warn about missing mask token (#14022)
9e31bec4fd53634c9e5b04650488a09a055f5dab	5a8ae3053ced350ed300ba91600519fcad1c6ba7	2025-06-05T09:06:29+03:00	Georgi Gerganov		context : fix pos_min initialization upon error decode (#14008)
5e1c3aed4074480f63e914d6c44c93536ed1452a	c496fe0b1da28618dd17bda8b0a6bf5004554080	2025-06-01T18:07:21+02:00	Sigbjørn Skjæret		convert : fix nomic-bert-moe mask token (#13757)
12d0188c0dc6146ffde6d277a93f232ccbe699f8	eb3949938e82a128855bc0676220bb2ce6e4228d	2025-05-31T10:24:04+03:00	Georgi Gerganov		kv-cache : refactor + add llama_memory_state_i (#13746)
db38704f0133be7832123495fa8fc2601ea999d4	07e4351ce663a7802b31f92fd7bc0e555c2044b6	2025-05-30T14:50:43+02:00	Sigbjørn Skjæret		convert : fix rwkv bos/eos token (#13844)
07e4351ce663a7802b31f92fd7bc0e555c2044b6	291f2b6913c7ef8350dbf0e77da38f7af131a08e	2025-05-30T12:24:37+02:00	Xuan-Son Nguyen		convert : allow partial update to the chkhsh pre-tokenizer list (#13847)
bc583e3c63c04a11d287c108ea9e6a515ead0423	72b090da2c50e540143fd312a2f9aa5f151e6136	2025-05-27T14:06:10+02:00	Xuan-Son Nguyen		mtmd : support Qwen 2.5 Omni (input audio+vision, no audio output) (#13784)
f9cd68398baf2ba8af4725ca9ed00bef205e6706	4f81b33e324b1669b282eb81104e4a1131be7dce	2025-05-27T12:07:52+03:00	Georgi Gerganov		sampling : make sure samplers return at least 1 token (#13822)
79c137f77677b3c8ee3c60a7da033721b938399a	22229314fc46b2f741bb21b12cde71f6c6a60b52	2025-05-26T14:03:54+03:00	Georgi Gerganov		examples : allow extracting embeddings from decoder contexts (#13797)
aa50ba462f63759e1731227f20f5009cfc2a0f16	de2ef53a4b2c0d703749a309d19fe68fd8f1b9ac	2025-05-25T16:22:29+02:00	Sigbjørn Skjæret		tests : improve UGM tokenizer test coverage (#13773)
c3a2624339187e89c4f65fd72a5fe7103968b5ad	ffd0eae60b7642e942c8245b89642527e36099e6	2025-05-24T12:29:09+02:00	Sigbjørn Skjæret		vocab : fix ugm tokenizer precision (#13743)
8a1d206f1d2b4e45918b589f3165b4be232f7ba8	797990c4bca0dca5be295c63e3fb2800dc0a69c2	2025-05-22T22:21:07+03:00	Georgi Gerganov		tts : fix n_ubatch + make WavTokenizer cache-less (#13713)
b7a17463ec190aeee7b9077c606c910fb4688b84	be0239693c1530a18496086331fc18d8a9adbad1	2025-05-21T00:55:30+08:00	l3utterfly		mtmd-helper : bug fix to token batching in mtmd (#13650)
e298d2fbd082a52c0f6ed02729f94e9bf630cf17	f0adb80bf7c2c0d80abb04f4533b5513622d9964	2025-05-20T08:05:46+03:00	Georgi Gerganov		kv-cache : add SWA support (#13194)
92ecdcc06a4c405a415bcaa0cb772bc560aa23b1	f71f40a2847d4c9f57b86cd206e0a27b2bfb6d1c	2025-05-19T13:04:14+02:00	Xuan-Son Nguyen		mtmd : add vision support for llama 4 (#13282)
6a2bc8bfb7cd502e5ebc72e36c97a6f848c21c2c	e3a7cf6c5bf6a0a24217f88607b06e4405a2b5d9	2025-05-17T17:59:48-04:00	Isaac McFadyen		server : added --no-prefill-assistant flag (#13608)
b2838049ccf50859cea4c390e81405c2fb01e820	aa48e373f256df9608395ee6881e7010840d6202	2025-05-15T05:57:02+03:00	Georgi Gerganov		bench : handle decode errors (#13548)
aa48e373f256df9608395ee6881e7010840d6202	e3a9421b78da5c810d812e6348a405f5acc39f34	2025-05-15T02:39:51+01:00	Olivier Chafik		`server`: inject date_string in llama 3.x template + fix date for firefunction v2 (#12802)
360a9c98e13d35f322b4c5b1309aab0cc90ed82b	09d13d94fb169093095416ac6323551c37b75339	2025-05-14T13:35:07+02:00	Xuan-Son Nguyen		server : fix cache_tokens bug with no cache_prompt (#13533)
43dfd741a5da77982e0a94d1e522a2481dfb914d	b064a51a4e7246fa43a3307f58e59f65e6be170a	2025-05-10T17:19:52+02:00	Sigbjørn Skjæret		llguidance : set tokenizer slices to default (#13424)
33eff4024084d1f0c8441b79f7208a52fad79858	17512a94d636c4b6c1332370acb3e5af3ca70918	2025-05-09T19:29:37+02:00	Xuan-Son Nguyen		 server : vision support via libmtmd (#12898)
efb8b47eda78ea8ae570d4fece3953aae499289e	0527771dd80bd18479dfaaa0a98be297fc3592bf	2025-05-09T05:53:58-04:00	Bartowski		imatrix : Add --parse-special for enabling parsing of special tokens in imatrix calculation (#13389)
f05a6d71a0f3dbf0730b56a1abbad41c0f42e63d	ee01d71e585f4a17ae83ec55d77acfdcf0bfa798	2025-05-08T14:25:39-04:00	Matt Clayton		mtmd : Expose helper_decode_image_chunk (#13366)
0ccc1213549e39ef4c1affb1bf5f49651ef4ce48	6562e5a4d6c58326dcd79002ea396d4141f1b18e	2025-05-08T22:03:53+09:00	welix		mtmd : fix the calculation of n_tokens for smolvlm (#13381)
6562e5a4d6c58326dcd79002ea396d4141f1b18e	51fb96b1ff2e1cc98b2492a012b7d93531a6a9a8	2025-05-08T14:28:33+03:00	Georgi Gerganov		context : allow cache-less context for embeddings (#13108)
27aa2595321c4d9cc4086a8e67bdea204b8309b0	9fdfcdaeddd1ef57c6d041b89cd8fb7048a0f028	2025-05-04T23:43:42+02:00	Xuan-Son Nguyen		mtmd : add C public API (#13184)
36667c8edcded08063ed51c7d57e9e086bbfc903	3bf785f3efa89ed28294fbf73054558a2b034bfb	2025-05-03T20:07:54+02:00	Xuan-Son Nguyen		clip : revert the change of BOI/EOI token for GLM-edge (⚠️ breaking change) (#13259)
3f3769ba76061a511f02f2a48da2ad2d93fce511	2f567611c0234bbca0a4009762acb47b56866095	2025-05-02T22:23:12+05:30	shalinib-ibm		ggml : Enable MMA for BF16 in llamafile_sgemm (#13148)
2f567611c0234bbca0a4009762acb47b56866095	7d2123484e6ba8fcd90fff8c01661b9dbbefdc9d	2025-05-02T11:42:30-04:00	Jared Van Bortel		llama-model : support Qwen2 embedding models and pooling_mode_lasttoken (#13245)
e2e1ddb93a01ce282e304431b37e60b3cddb6114	d9d398f84f96d16c308d4976f5e90222ecc2a492	2025-04-29T20:33:10+02:00	matteo		server : Prefilling assistant message in openai compatible API (#13174)
00e3e5a194e88e604e7c91391b9e90332888fd72	e98b3692be4cd8fbbd9a56fbacc2f2bf0bf26a68	2025-04-29T11:47:04+02:00	Xuan-Son Nguyen		mtmd : add qwen2vl and qwen2.5vl (#13141)
5f5e39e1ba5dbea814e41f2a15e035d749a520bc	eaea3253244dc4bbe07f6cd81325847ccc6cf93e	2025-04-28T15:52:15-04:00	AT		model : Nomic Embed Text V2 with Mixture-of-Experts (MoE) architecture (#12466)
4e87962e34a4b257ec374c4baf6b1568554b81a9	fb0471d1753824e75474c24f82fbdd54c94dceda	2025-04-28T16:12:56+02:00	Xuan-Son Nguyen		mtmd : fix glm-edge redundant token count (#13139)
d2b2031e5f11b826dcc718138642f147a2009665	5fa9e63be82225fb3249c76f39ddda3e5bdec0a3	2025-04-28T14:20:56+02:00	Xuan-Son Nguyen		llama : (mrope) allow using normal 1D position for text token (#13138)
ced44be34290fab450f8344efa047d8a08e723b4	e291450b7602d7a36239e4ceeece37625f838373	2025-04-27T21:57:32+02:00	matteo		llama-chat : fix wrong template in GLM4-0414 (#13140)
553a5c3a9fdf771be2101bc3529937963f817457	13be08daf992c89d5169518229b3740041c0f419	2025-04-25T10:08:08+03:00	Radoslav Gerganov		rpc : do not wait for response when sending RPC_CMD_SET_TENSOR (#12943)
ecda2ec4b347031a9b8a89ee2efc664ce63f599c	eb1776b15a32d832f1266deeeab75b9d255c5849	2025-04-23T20:21:59+02:00	Xuan-Son Nguyen		mtmd : Support Pixtral 12B (#13065)
b9154ecff93ff54dc554411eb844a2a654be49f2	2db9ba1464f3de0aceb2b5289963e69fc369cb66	2025-04-18T10:04:51+02:00	Xuan-Son Nguyen		mtmd : add methods to access `mtmd_image_tokens` (#12906)
971f245b3b5f3f55991bb779cb541b00f82eea1d	12b17501e6015ffe568ac54fdf08e6580833bf1b	2025-04-17T01:37:05-07:00	Mikko Juola		llama : recognize IBM Granite 3.3 FIM tokens (#12988)
8b9cc7cdd8a0dcf0176c60c755322c95b5965299	64eda5deb9859e87a020e56bab5d2f9ca956f1de	2025-04-10T22:57:16+02:00	Xuan-Son Nguyen		llava : introduce libmtmd (#12849)
64eda5deb9859e87a020e56bab5d2f9ca956f1de	fe5b78c89670b2f37ecb216306bed3e677b49d9f	2025-04-10T17:24:44+02:00	Xuan-Son Nguyen		convert : ability to lazy-load safetensors remotely without downloading to disk (#12820)
0090950f679475c5ecaac2f7bca5049cca96492b	7ecd780b1a1d5214b8d04c25ebfc194d310816ed	2025-04-09T00:25:08-05:00	Jeff Bolz		vulkan: In coopmat2 mmq, load q4_k/q5_k scales through shared memory (#12833)
1466621e738779eefe1bb672e17dc55d63d166bb	82974011f312057b446c27267105bd7ad3810599	2025-04-07T23:06:44+02:00	Xuan-Son Nguyen		llama : Support llama 4 text-only (#12791)
a10b36c91a091f4606710fba4e9327fd71e0e738	83a88bd6affbe148a622ac730952ac5b8b585979	2025-04-02T14:32:59+03:00	Georgi Gerganov		llama : refactor kv cache guard (#12695)
c80a7759dab10657b9b6c3e87eef988a133b9b6a	250d7953e829ff0e16074510fef0e7cec696461b	2025-03-31T18:40:56+02:00	Daniel Bevenius		vocab : add special infill tokens for CodeLlama (#11850)
403fbacbbcd3650abcc53e84f5e6bf88e51dd6f5	a8a1f3356786cbf8bcc3422e3c8737fc33b453e7	2025-03-31T16:36:25+02:00	Sigbjørn Skjæret		convert : Qwerky : use lora_rank_tokenshift and lora_rank_decay if present (#12667)
52de2e594979be52f0da92601747aa44a13e50e4	2c3f8b850a4a6cff0f5dda2135c03fc81d33ed8b	2025-03-31T10:20:30+02:00	marcoStocchi		tts : remove printfs (#12640)
5d01670266859444366e4f333ade5e0e5e2ae63d	ef03229ff423dd1991f4f44ef1352f03334d86eb	2025-03-28T01:05:44-07:00	Benson Wong		server : include speculative decoding stats when timings_per_token is enabled (#12603)
13731766db91ec927c1b61bf502ac7a9be2b11b9	ab6ab8f809bd514d88e02a63869b4d619f13fa86	2025-03-28T13:13:22+05:30	amritahs-ibm		llamafile : ppc64le GEMV forwarding for FP32. (#12594)
c7b43ab60855f752ae79937fb93d561bc30b69a4	24feaec05792b972d5ff3e2b12d9237ebd50d1ac	2025-03-27T12:21:47+05:30	amritahs-ibm		llamafile : ppc64le MMA implementation for Q4_0. (#12489)
00d53800e00bb22a26bf710fa6bd1150e412cc1d	7ea75035b67f44c22ed7039967f718011fd35ce5	2025-03-24T06:47:24-04:00	compilade		llama-vocab : add SuperBPE pre-tokenizer (#12532)
732b5fbf5e7f9cf069942f0c5850ee959ef321ba	568013d0cd3d5add37c376b3d5e959809b711fc7	2025-03-20T02:36:37-04:00	Bartowski		convert : avoid calls to tokenizer.added_tokens_decoder (#12473)
2048b5913d51beab82dfe29955f9008130b936c0	f08f4b3187b691bb08a8884ed39ebaa94e956707	2025-03-13T06:10:05-04:00	Ishaan Gandhi		server : fix crash when using verbose output with input tokens that are not in printable range (#12178) (#12338)
2b3a25c212f8c4d8a49cec23e03343a7719d51c9	8352cdc87b207a735d34c431a36c425728cb4586	2025-03-10T09:44:42Z	Olivier Chafik		`sampler`: fixes trigger tokens + lazy grammars (fix typo cast from token to string) (#12291)
06a92a193a07afe445929607be9d5e4d033956fb	a057897ad4e48e3df0354256e0cc80dadcb57595	2025-03-05T15:25:45+08:00	Clauszy		server : fix cache reuse logic (#12161)
7b69003af74924ac04d97313c2e57c45ba56b616	ece9745bb8079b9f4af45df29b67ad0c6e50584d	2025-03-03T11:42:45+01:00	Xuan-Son Nguyen		webui : add ?m=... and ?q=... params (#12148)
34a846b5847a18d133b360074f1fb485b2632b2d	7a2c913e66353362d7f28d612fd3c9d51a831eda	2025-02-24T13:47:07-08:00	lhez		opencl: fix for small models (#11950)
af7747c95ae71a5db4184947f837179e82c70b77	a28e0d5eb18c18e6a4598286158f427269b1444e	2025-02-23T05:39:24+08:00	Aaron Teo		ggml-cpu: Support s390x SIMD Instruction Set (#12019)
36c258ee921dbb5c96bdc57c0872e4a9a129bef6	f3e64859edb0d55d4223ead78672597cd1a218df	2025-02-22T22:28:28+08:00	Ting Lou		llava: build clip image from pixels (#11999)
abd4d0bc4f1a9a0e429bc8ee0d5ece2a394a0a39	9626d9351a6dfb665400d9fccbda876a0a96ef67	2025-02-19T13:29:42+02:00	Georgi Gerganov		speculative : update default params (#11954)
63e489c025d61c7ca5ec06c5d10f36e2b76aaa1d	63ac12856303108ee46635e6c9e751f81415ee64	2025-02-18T18:03:23Z	Olivier Chafik		tool-call: refactor common chat / tool-call api (+ tests / fixes) (#11900)
3e693197724c31d53a9b69018c2f1bd0b93ebab2	a394039db004c6ee00098250d160b5aa018c2314	2025-02-13T07:07:51+01:00	Daniel Bevenius		llama : update llama_decode_internal ref [no ci] (#11840)
a18f481f99962638092d6f1c98b1d34d3e3256de	b9ab0a4d0b2ed19effec130921d05fb5c30b68c5	2025-02-11T14:06:45+01:00	Daniel Bevenius		server : use common_token_to_piece instead of common_detokenize (#11740)
2d219b389e8c8c40bce547b08c8aa7add60fde1f	333820d7491cd31c707a340ff23b984a84e40154	2025-02-07T08:55:47-05:00	Christian Fillion		vocab : ignore invalid UTF-8 input in the BPE tokenizer (#11729)
c3db0480bb820e120249014b380e1f4badf2a1c1	d774ab3acc4fee41fbed6dbfc192b57d5f79f34b	2025-02-06T01:55:25+01:00	Matvey Soloviev		readme : add link to Autopen under UIs (#11684)
cde383323959544abe10a4d79e1d3e1ee479933c	b3451785aca16fbf4214eeba7e4612676cdf8ccb	2025-02-03T23:49:27Z	Olivier Chafik		`tool-call`: allow `--chat-template chatml` w/ `--jinja`, default to chatml upon parsing issue, avoid double bos (#11616)
bfcce4d693617ec843d0b2510f6ee16e6bc6720d	69804487e0b10f2c5c06316f0ac0eb6ada68433f	2025-02-02T09:25:38Z	Olivier Chafik		`tool-call`: support Command R7B (+ return tool_plan "thoughts" in API) (#11585)
ff227703d6d6e1888bdc7af6138514092ffcdb96	0cec062a638700495673f5494d200b74340538be	2025-02-01T23:55:32-08:00	Michał Moskal		sampling : support for llguidance grammars (#10224)
0cec062a638700495673f5494d200b74340538be	53debe6f3c9cca87e9520a83ee8c14d88977afa4	2025-02-02T15:48:46+08:00	piDack		llama : add support for GLM-Edge and GLM-Edge-V series models (#10573)
ffd0821c57edc7e5d04338ab0c6b1461198df15f	4314e56c4f8c5091f45732f39bd94c0c6c323798	2025-01-30T11:10:59+01:00	mgroeber9110		vocab : correctly identify LF token for GPT-2 style BPE tokenizer (#11496)
df984e014714cba4c99ef894b20b51cbcef31b16	acd38efee316f3a5ed2e6afcbc5814807c347053	2025-01-27T12:07:12+01:00	Johannes Gäßler		llama: refactor llama_decode_impl (#11381)
6171c9d25820ccf676b243c172868819d882848f	e28245f35f2faaf249dd352998b3693a8cc28c51	2025-01-21T13:18:51Z	Olivier Chafik		Add Jinja template support (#11016)
6390a998bfc63241fde8509022b0768a71bf20bb	44e18ef93995f3040660750b527e5becf85899d0	2025-01-18T18:20:57+08:00	LostRuins Concedo		tts : add guide tokens support (#11186)
3edfa7d3753c29e44b964c0ff424d2ea8d5fdee6	667d72846c06b2cf4f7c8a4265e210991a49706b	2025-01-17T20:57:56+08:00	codezjx		llama.android: add field formatChat to control whether to parse special tokens when send message (#11270)
bbf3e55e352d309573bdafee01a014b0a2492155	c5bf0d1bd7eb8762edca47e0a95f64e6fbfaf5b1	2025-01-14T12:54:58+02:00	Georgi Gerganov		vocab : add dummy tokens for "no_vocab" type (#11231)
8f70fc3d1b1d3c17b61842330dd106d391cc1227	1244cdcf14900dd199907b13f25d9c91a507f578	2025-01-13T13:38:20+01:00	Daniel Bevenius		llama : remove 'd' from bad special token log (#11212)
08f10f69c38288e9e8bb1f933af63a3fc9013d40	afa8a9ec9b520137bbd1ca6838cda93ee39baf20	2025-01-12T12:15:53+02:00	Georgi Gerganov		llama : remove notion of CLS token (#11064)
afa8a9ec9b520137bbd1ca6838cda93ee39baf20	c05e8c9934f94fde49bc1bc9dc51eed282605150	2025-01-12T11:32:42+02:00	Georgi Gerganov		llama : add `llama_vocab`, functions -> methods, naming (#11110)
2739a71e4b88474833b64aa974ca4515574fd3c4	ba8a1f9c5b675459c55a83e3f97f10df3a66c788	2025-01-11T05:50:33+01:00	Daniel Bevenius		convert : sort print supported models [no ci] (#11179)
ff3fcabc727b2dd0c477d23a258217b27cc639fb	c3f9d25706ac84297067aeaa662c1f1af42ed443	2025-01-10T11:30:53+01:00	Daniel Bevenius		convert : add --print-supported-models option (#11172)
8cef75c743ba13ebbd6d380c531200c768a8b8aa	0d52a69e4bf0d6181beec7853307bdcdeec9905b	2025-01-08T16:24:19+05:30	amritahs-ibm		llamafile : ppc64le MMA INT8 implementation (#10912)
3e6e7a6bc2c4b980a0cf0fcb5cb3b79a965b5f14	ae2f606bb598b287f5fb69c9fdfc98b86598c6cc	2025-01-06T10:54:25+02:00	Georgi Gerganov		tokenize : escape the prompt (#11058)
727368c60f2ebf2d6a7473a4a9f80957ab063a8e	5047dd3546951dea3d65c02257d06c46c8662338	2025-01-06T10:52:15+02:00	Georgi Gerganov		llama : use LLAMA_TOKEN_NULL (#11062)
9394bbd484f802ce80d2858033583af3ef700d25	f922a9c542ee117550a168395c63ea79261f5c99	2025-01-04T21:06:11+01:00	fairydreaming		llama : Add support for DeepSeek V3 (#11049)
2f0ee84b9b02d2a98742308026f060ebdc2423f1	0da5d860266c6928b8c9408efbd264ae59fedda6	2025-01-02T18:06:12+01:00	Pierrick Hymbert		server: bench: minor fixes (#10765)
16cdce7b68218959e0658e2f95b4572573d5008e	d79d8f39b4da6deca4aea8bf130c6034c482b320	2024-12-28T15:08:54Z	Alexey Parfenov		server : fix token duplication when streaming with stop strings (#10997)
30caac3a68a54de8396b21e20ba972554c587230	60cfa728e27c28537657d4e627ed432508eb9537	2024-12-24T09:44:20+02:00	Georgi Gerganov		llama : the WPM vocabs use the CLS token as BOS (#10930)
b92a14a841fb4dfaf27b29d982ec8ba5289a3bff	6f0c9e034bb398915a6617ee4acc62adb87d387d	2024-12-23T08:35:44+08:00	Yun Dou		llama : support InfiniAI Megrez 3b (#10893)
7ae33a616f44ecc081f3dcb589be20962d1d4a92	ebdee9478ca7ba65497b9b96f7457698c6ee5115	2024-12-23T01:09:58+03:00	Billel Mokeddem		llama : add Falcon3 support (#10883)
57bb2c40cd94c5a09f5210ed8264cc93b21c4b7e	a3c33b1dce2d4f25040b75f66629104bd1e40128	2024-12-19T15:40:08+01:00	Xuan Son Nguyen		server : fix logprobs, make it OAI-compatible (#10783)
2fffc52b50992ac5bc64db19d33c39cbc06f52cf	7585edbdebd02861e0994dae67c9338731fb3fc5	2024-12-19T06:04:51-07:00	Sukriti Sharma		llama : fix Roberta embeddings (#10856)
0bf2d10c5514ff61b99897a4a5054f846e384e1e	7bbb5acf125d1d2840cac7d31b9aaa72210dd5ec	2024-12-18T19:27:21+02:00	Georgi Gerganov		tts : add OuteTTS support (#10784)
152610eda91217ac409342cd976d05f5114ad39f	0e70ba686e6c717a0aa41d88284e2a392c2bd0cd	2024-12-18T13:01:41+02:00	Georgi Gerganov		server : output embeddings for all tokens when pooling = none (#10861)
0e70ba686e6c717a0aa41d88284e2a392c2bd0cd	46828872c31b25df16169cbbf5c2225fa9cb0675	2024-12-18T11:05:29+02:00	Georgi Gerganov		server : add "tokens" output (#10853)
644fd71b44c4cdbfc6482fbf0353d289c3bc29e6	4ddd199f6f6b980e0a7ed9f9b44efeae2fbdf5c4	2024-12-16T12:31:14+02:00	Georgi Gerganov		sampling : refactor + optimize penalties sampler (#10803)
ba1cb19cdd0d92e012e0f6e009e0620f854b6afd	56eea0781cbd2608ed8ff524955495569b9264be	2024-12-14T20:43:46+08:00	HimariO		llama : add Qwen2VL support + multimodal RoPE (#10361)
64ed2091b24b2f9747148fdf49a34ed5938762c3	991f8aabeec89d801300bb179e52013fb0eb0584	2024-12-02T21:45:54+08:00	haopeng		server: Add "tokens per second" information in the backend (#10548)
8fc393f246c550d2481e53323a47644a94e8d01f	ec450d3bbf9fdb3cd06b27c00c684fd1861cb0cf	2024-11-09T15:06:54+08:00	haopeng		scripts : fix pattern and get n_tokens in one go (#10221)
2319126a70b541f8670225a04a38202bbdccbedb	3bcd40b3c593d14261fb2abfabad3c0fb5b9e318	2024-11-07T02:02:08-06:00	snadampal		fix q4_0_8_8 format for corrupted tokens issue (#10198)
d865d1478cd4e403f82d793c2afcd0f943412f05	1804adb0cfee4811eaf633741503d683a46e4c77	2024-11-01T13:33:14Z	sasha0552		server : fix smart selection of available slot (#10120)
958367bf530d943a902afa1ce1c342476098576b	40f2555797f97314de749873cdc29dc102be66e2	2024-10-24T21:51:22+02:00	Xuan Son Nguyen		server : refactor slot input data, move tokenizer to HTTP thread (#10023)
c8c07d658a6cefc5a50cfdf6be7d726503612303	19d900a7565b8f6b0a708836a57d26966cb9efe2	2024-10-22T16:59:02+02:00	Xuan Son Nguyen		llama : fix empty batch causing llama_batch_allocr to crash (#9966)
c421ac072d46172ab18924e1e8be53680b54ed3b	4ff7fe1fb36b04ddd158b2de881c348c5f0ff5e4	2024-10-22T13:08:41+02:00	Xuan Son Nguyen		lora : warn user if new token is added in the adapter (#9948)
4ff7fe1fb36b04ddd158b2de881c348c5f0ff5e4	6b8447352df3d662b56280c8fc38d7f092885787	2024-10-22T18:33:37+08:00	Molly Sophia		llama : add chat template for RWKV-World + fix EOT (#9968)
cda0e4b648dde8fac162b3430b14a99597d3d74f	afd9909a6481402844aecefa8a8908afdd7f52f1	2024-10-18T23:18:01+02:00	Xuan Son Nguyen		llama : remove all_pos_0, all_pos_1, all_seq_id from llama_batch (#9745)
6f55bccbb8835d42147add4ee48807450f5ff535	17bb9280807cfbb6611b853aa1ef05114bd9efe9	2024-10-18T01:41:51+02:00	Daniel Bevenius		llama : rename batch_all to batch (#8881)
99bd4ac28c32cd17c0e337ff5601393b033dc5fc	3752217ed5a6a11864682fbf009bcb36afffd6bc	2024-10-17T22:32:47+03:00	Georgi Gerganov		llama : infill sampling handle very long tokens (#9924)
9e041024481f6b249ab8918e18b9477f873b5a5e	dbf18e4de9e7aa496871f1555f9f0c8d84567108	2024-10-16T19:34:28+02:00	Daniel Bevenius		llama : suppress conversion from 'size_t' to 'int' (#9046)
fbc98b748e7b075e327bcf13237057f647678049	dcdd535302fc9702a4709be25f56540d65163a44	2024-10-15T15:54:55+05:00	MaggotHATE		sampling : add XTC sampler (#9742)
11ac9800aff532715a5bc7991062c68ba3472e6e	943d20b4111c746bcd9dbc7e4771de313b08b50c	2024-10-12T08:21:51+03:00	Georgi Gerganov		llama : improve infill support and special token detection (#9798)
8c475b97b8ba7d678d4c9904b1161bd8811a9b44	58b16695e146628481c6b9b8a3b101c0c9bac00f	2024-10-05T15:55:04+03:00	Georgi Gerganov		rerank : use [SEP] token instead of [BOS] (#9737)
e3c355ba654d4164c1c09e5d0dcacecb8b214af8	841713e1e487bdb82fd106a52ad998c5f87b59e9	2024-10-03T10:22:15-04:00	compilade		convert : handle tokenizer merges format from transformers 4.45 (#9696)
1927378bcce20ba72b6c89d5977b854a4bcaeb5d	6f1d9d71f4c568778a7637ff6582e6f6ba5fb9d3	2024-10-01T02:31:36-04:00	compilade		convert : refactor rope_freqs generation (#9396)
faac0bae265449fd988c57bf894018edc36fbe1e	f99d3f8367174f7aba73c07fd87de687d4a0ece1	2024-09-29T05:25:00-07:00	matiaslin		common : ensure llama_batch size does not exceed max size (#9668)
f4d2b8846a6b34419ff9e9491aee6cd95e444bfc	1b2f992cd2cff0b69e5abe78bb8888d51ed19d67	2024-09-28T17:42:03+03:00	Georgi Gerganov		llama : add reranking support (#9510)
6102037bbb55521880ae78a6ee6c2a0c00c901df	9a913110cf471a8287ac06c43cbe307d3cf6df99	2024-09-28T20:10:58+08:00	Zhenwei Jin		vocab : refactor tokenizer to reduce init overhead (#9449)
9a913110cf471a8287ac06c43cbe307d3cf6df99	43bcdd9703ec19af7d2a519640b5ed6f4aac3d53	2024-09-28T12:08:43Z	nopperl		llama : add support for Chameleon (#8543)
31ac5834fe75c296476658a124c06c84772aa641	cea1486ecf34a1c7e014a9e290eb458f5a11f876	2024-09-24T10:16:06+03:00	Georgi Gerganov		llama : keep track of all EOG tokens in the vocab (#9609)
6443ddd98576a9da904ef9f07df4e4398bb6a01a	8a308354f6520df6bea851b435bd8054ee5617b4	2024-09-18T13:42:36+02:00	Daniel Bevenius		llama : use reserve/emplace_back in sampler_sample (#9534)
8344ef58f8cdb8ebd6faf4463ca32ae91c374c81	0226613853133c081b55bb892a41bb5eacc0bc94	2024-09-17T12:18:22+02:00	Michael Podvitskiy		llama : fix n_vocab init for 'no_vocab' case (#9511)
0abc6a2c25272d5cf01384dda8ee8bfec4ba8745	bd35cb0ae357185c173345f10dc89a4ff925fc25	2024-09-13T09:53:38+03:00	Georgi Gerganov		llama : llama_perf + option to disable timings during decode (#9355)
78203641fee3b1f82abaff0c7f667e1b4a286390	e6b7801bd189d102d901d3e72035611a25456ef1	2024-09-12T22:30:11+02:00	Mathijs Henquet		server : Add option to return token pieces in /tokenize endpoint (#9108)
c837981bba7cf6839b69d32b25552ce685936b14	3c26a1644dacfa6b5d58af550210524efd7b93fc	2024-09-12T20:28:20+09:00	daminho		py : add Phi-1.5/Phi-2 tokenizer (#9361)
39f852f44039b058fdd0611ee127c6efa7ba4a04	2b00fa799773cc75d53b841c03d21d7468a1e3a1	2024-09-12T19:25:16+08:00	Molly Sophia		py : add special tokens in hf_converter for RWKV v6 (#9428)
1b28061400eb9832603c9f1dfbec4d339a8490a2	8db003a19d7055b5bd248ce2afff9324e5b8da95	2024-09-11T17:52:13+02:00	slaren		llama : skip token bounds check when evaluating embeddings (#9437)
83008b7cfe90ad89d0c0ed2c2424fd75edc25ac1	0b4ac75772b744bb0a0d674927587621d1057884	2024-09-10T09:03:21+02:00	Daniel Bevenius		llama : update llm_build_copy_mask_state comment [no ci] (#9385)
eae597182cb61bbde0b26e7cec5999d28b9327fe	00b02bb249406ec0123757a67a5b714c3f33a699	2024-09-08T12:41:51+02:00	slaren		llama : sanitize tokens in the upper bound (#9359)
fbb7fcffbcfc50009c275e75982b1603a6cb6b80	a5b5d9a1014248f385939e6739e9db9de7147e55	2024-09-07T22:51:00-07:00	Kevin Gibbons		llama : set attrs of mislabelled EOT/EOM tokens (#9348)
faf69d4237c9ae4d7f572b4674d1002463e8acd3	e536426ded3fb4a8cd13626e53508cd92928d6c2	2024-09-08T00:33:13+03:00	Georgi Gerganov		llama : sanitize invalid tokens (#9357)
9b2c24c0993487d3b34a873980e292da571481f3	134bc38ecf3e2c5460581badce289a1ffa680453	2024-09-06T23:21:29+02:00	Xuan Son Nguyen		server : simplify state machine for slot (#9283)
9bc6db28d011d47a5f318dc4aebbe7927fac4629	32b2ec88bc44b086f3807c739daf28a1613abde1	2024-09-05T21:48:47-04:00	compilade		ggml-quants : ternary packing for TriLMs and BitNet b1.58 (#8151)
9c1ba557335c3cab46807e6478eb7d17a63361f1	c6d4cb46559b359d2682cf2a002e7fe01bb7a767	2024-09-02T16:51:01+05:30	Tushar		build(nix): Package gguf-py (#5664)
8f1d81a0b6f50b9bad72db0b6fcd299ad9ecd48c	a47667cff41f5a198eb791974e0afcc1cddd3229	2024-09-01T22:38:17+08:00	Molly Sophia		llama : support RWKV v6 models (#8980)
c679e0cb5c378bfb63643c44a453345ba8b90126	fb487bb5671b37267f35ff43fe8849ddccd925cd	2024-08-16T15:35:18+09:00	Minsoo Cheong		llama : add EXAONE model support (#9025)
4af8420afba39de4968adf2695ce12fd42422a13	6bda7ce6c3a9284fcbb70c1ace4107db8eb63e5c	2024-08-15T15:23:23+08:00	Zhenwei Jin		common : remove duplicate function llama_should_add_bos_token (#8778)
6bda7ce6c3a9284fcbb70c1ace4107db8eb63e5c	d5492f0525fa533817a67e93a4bde9d71d81cf58	2024-08-15T10:17:12+03:00	Esko Toivonen		llama : add pre-tokenizer regexes for BLOOM and gpt3-finnish (#8850)
5ef07e25ac39e62297a67208c5bcced50835a2dd	4134999e01f31256b15342b41c4de9e2477c4a6c	2024-08-12T10:21:50+03:00	Georgi Gerganov		server : handle models with missing EOS token (#8997)
7c3f55c10051c634546247387c5c359c9d499360	911b437f228e75aa3d235acec21bfddd23ecce2f	2024-08-10T11:43:26+02:00	fairydreaming		Add support for encoder-only T5 models (#8900)
cdd1889de62a7140c8c016405ec917464bde8bd3	c21a896405de4cdf4207eb8130555ceaac0ab110	2024-08-06T02:20:54-05:00	Douglas Hanley		convert : add support for XLMRoberta embedding models (#8658)
d3f0c7166adfa952237e0f437a5344362d8256d4	e31a4f679779220312c165b0f5994c680a610e38	2024-08-05T09:38:01+02:00	fairydreaming		Stop the generation when <|eom_id|> token is encountered - needed for Llama 3.1 tool call support (#8858)
978ba3d83d17b10fdf9807006048432b5b3769fc	ecf6b7f23e664afd7ff856ec39034240ce438daa	2024-08-04T18:16:23Z	ardfork		Server: Don't ignore llama.cpp params (#8754)
9d03d085dd6cb275c078690bb64073b9b043e95f	bfb4c74981f0a40d757b450b596a9fe4ca983d26	2024-07-27T12:45:02+02:00	Daniel Bevenius		common : add --no-warmup option for main/llama-cli (#8712)
938943cdbf4dd79005a394d732bc226f9e34e0ff	751fcfc6c33ea5f43cadd4d976f8fb176871df5e	2024-07-23T13:10:17+03:00	Georgi Gerganov		llama : move vocab, grammar and sampling into separate files (#8508)
d94c6e0ccbd29ee1ba4f44e9caa8682ad94df9fa	566daa5a5b38018b2727950bbd280239adb981b6	2024-07-22T10:43:01-04:00	Jason Stillerman		llama : add support for SmolLm pre-tokenizer (#8609)
6f11a83e4e7700fdf353ed4a29599cb662c792f6	e093dd238282a980d248db0786063f8174400f70	2024-07-22T13:33:22+03:00	Georgi Gerganov		llama : allow overrides for tokenizer flags (#8614)
e093dd238282a980d248db0786063f8174400f70	50e05353e88d50b644688caa91f5955e8bdb9eb9	2024-07-22T13:32:49+03:00	Georgi Gerganov		tests : re-enable tokenizer tests (#8611)
628154492a0b2dcc954a3af99e5c267097568eae	04bab6b7da0ed2b0a57174a57784d602aabb6c10	2024-07-22T16:02:09+08:00	Jan Boon		server : update doc to clarify n_keep when there is bos token (#8619)
c69c63039cd75f8f33f253ab7485d82a8b4cd403	69c487f4ed57bb4d4514a1b7ff12608d5a8e7ef0	2024-07-20T21:53:01-04:00	compilade		convert_hf : fix Gemma v1 conversion (#8597)
940362224d20e35f13aa5fd34a0d937ae57bdf7d	69b9945b44c3057ec17cb556994cd36060455d44	2024-07-20T09:43:51-04:00	Michael Coppola		llama : add support for Tekken pre-tokenizer (#8579)
69b9945b44c3057ec17cb556994cd36060455d44	c3776cacabce2ee35f172fb72be7a519752125fa	2024-07-20T09:09:37-04:00	Huifeng Ou		llama.swiftui: fix end of generation bug (#8268)
fa79495bb4897953a75607addd9d2cdd2ec63222	17eb6aa8a992cda37ee65cf848d9289bd6cad860	2024-07-13T23:35:10-04:00	compilade		llama : fix pre-tokenization of non-special added tokens (#8228)
c3ebcfa148e867a68e78fd5c4f0c23e8f84c788b	8a4441ea1a2564578134404f31158c318e9c0bf3	2024-07-12T03:14:12-05:00	Douglas Hanley		server : ensure batches are either all embed or all completion (#8420)
5aefbce27a66473d4b1263ba3f7bdd3d14245975	71c1121d11f1437be9421fd0cbaa011b9ef49098	2024-07-12T10:06:33+02:00	Jiří Podivín		convert : remove fsep token from GPTRefactForCausalLM (#8237)
9a55ffe6fbf7f19c865f8f277ca7cd585cc0b094	7a221b672e49dfae459b1af27210ba3f2b5419b6	2024-07-11T03:41:48-04:00	compilade		tokenize : add --no-parse-special option (#8423)
470939d483d1c89b7292f78bac1fd27c42c171ce	6f0dbf6ab087bcd286fb78560099ca0458316735	2024-07-08T03:26:53-04:00	Kevin Wang		common : preallocate sampling token data vector (#8363)
6f0dbf6ab087bcd286fb78560099ca0458316735	ffd00797d81ef7db1528b9e10adbdc333ade6495	2024-07-08T09:34:35+03:00	Georgi Gerganov		infill : assert prefix/suffix tokens + remove old space logic (#8351)
3fd62a6b1c9ca7b7c0093e984cc9c133c6f2726d	a8db2a9ce64cd4417f6a312ab61858f17f0f8584	2024-07-07T15:04:39-04:00	compilade		py : type-check all Python scripts with Pyright (#8341)
905942abdba5ba0b28a1b0805e51e4f818c54bc9	b5040086d436e7345e4fa33a5b9558060c75603f	2024-07-07T20:52:10+08:00	toyer		llama : support glm3 and glm4 (#8031)
86e7299ef5dff0f388922dc6fcbce009e99d8005	60d83a0149849e9217e4b8ae26e277a41aea906e	2024-07-06T15:32:04-05:00	Derrick T. Woolworth		added support for Authorization Bearer tokens when downloading model (#8307)
213701b51a17175d0d326b566efc03f30ec7fbe6	be20e7f49d9e5c6d9e8d9b4871eeba3df7a1639d	2024-07-05T19:01:35+02:00	jaime-m-p		Detokenizer fixes (#8039)
6f63d646c1a06a6e09f721009a2676864ae04e31	51d2ebadbbf365b894f3888361df42dbacb12b7a	2024-07-04T18:38:58+02:00	Daniel Bevenius		tokenize : add --show-count (token) option (#8299)
807b0c49ff7071094f97ebc3a0a8e2b9e274f503	f8c4c0738d72d2162736edd72dd5db8b269adca1	2024-07-04T15:46:11+02:00	fairydreaming		Inference support for T5 and FLAN-T5 model families (#5763)
20fc3804bfb727074bc270b6eacb60af8d0bf7d4	f619024764e72261f14d7c31d892b8fb976603b4	2024-07-04T10:41:03+03:00	Georgi Gerganov		convert : fix gemma v1 tokenizer convert (#8248)
5fac350b9cc49d0446fc291b9c4ad53666c77591	cb5fad4c6c2cbef92e9b8b63449e1cb7664e4846	2024-07-02T01:07:23+02:00	Xuan Son Nguyen		Fix gemma2 tokenizer convert (#8244)
e57dc62057d41211ac018056c19c02cd544694df	a27aa50ab7e07fe46aae619076b6e31d5663e914	2024-06-28T00:00:43-04:00	pculliton		llama: Add support for Gemma2ForCausalLM (#8156)
f675b20a3b7f878bf3be766b9a737e2c8321ff0d	911e35bb8bb2fd1c7d3f40f27e96ff432eae7e14	2024-06-27T11:58:54+03:00	kustaaya		Added support for Viking pre-tokenizer (#8135)
911e35bb8bb2fd1c7d3f40f27e96ff432eae7e14	ac146628e47451c531a3c7e62e6a973a2bb467a0	2024-06-27T09:46:41+02:00	Sigbjørn Skjæret		llama : fix CodeLlama FIM token checks (#8144)
6777c544bdd8c5d9de3220d6e2557957bbbf2a4f	163d50adaf8897d8b734d701ff332de6be63d484	2024-06-26T01:45:58+01:00	Olivier Chafik		`json`: fix additionalProperties, allow space after enum/const (#7840)
6fcbf6823553efabe52ed83e3c2a3329aa3387d1	e6bf007744eb06336a231ef39cf08146dd16d2ce	2024-06-25T21:14:35+02:00	fairydreaming		llama : implement Unigram tokenizer needed by T5 and FLAN-T5 model families (#5763)
d62e4aaa02540c89be8b59426340b909d02bbc9e	9a590c82262dd518137f85406e65e452fdf2aca3	2024-06-24T14:13:39+02:00	fairydreaming		gguf-py : fix tensor groups for encoder-decoder models in gguf-dump.py (#8090)
de0d6a68ac99f307fe889c48e21124bc3b7ca29a	95f57bb5d5b18ef0beb2702a0d6c06e46804075c	2024-06-24T07:06:05+02:00	fairydreaming		gguf-py, convert-hf : model conversion support for T5 and FLAN-T5 model variants (#5763)
a927b0f3dd9a86ee042cd2bdcc8c9da4a855926b	80ea089d771f0c2d97afa8bead80ded412f600d7	2024-06-21T08:51:28+03:00	Georgi Gerganov		llama : optimize long word tokenization with WPM (#8034)
80ea089d771f0c2d97afa8bead80ded412f600d7	0e64591e8290037db6412665a56354b789a0597e	2024-06-21T00:38:22-05:00	Douglas Hanley		llama : allow pooled embeddings on any model (#7477)
37bef8943312d91183ff06d8f1214082a17344a5	91c188d6c296bd3384f2a02a83b71187aa3d18b3	2024-06-18T18:40:52+02:00	jaime-m-p		tokenizer : BPE fixes (#7530)
91c188d6c296bd3384f2a02a83b71187aa3d18b3	84f6de17f6a8602e7ff7f7c7bda36a73f510a2dd	2024-06-18T14:19:45+02:00	Sigbjørn Skjæret		Only use FIM middle token if it exists (#7648)
0c7b3595b9e5ad2355818e259f06b0dc3f0065b3	7b2f4a7d193ef2475259bbe7656fcccfab4b1217	2024-06-15T18:53:40+02:00	Xuan Son Nguyen		Add `cvector-generator` example (#7514)
6fcd1331efbfbb89c8c96eba2321bb7b4d0c40e4	41b9260f18eb7f325c952006ac46afc1d0d8ad2f	2024-06-14T12:20:04+02:00	Sigbjørn Skjæret		llama : more checks before assuming FIM tokens (#7644)
41b9260f18eb7f325c952006ac46afc1d0d8ad2f	172c8256840ffd882ab9992ecedbb587d9b21f15	2024-06-14T13:16:49+03:00	Elaine		convert : add Poro-34B-chat tokenizer support (#7713)
ad675e1c67a05b16e4e12abe30dbecfc808e7b7e	a143c04375828b1f72eb1a326115791b63e79345	2024-06-06T06:08:52-07:00	Clint Herron		Added support for . (any character) token in grammar engine. (#6467)
c90dbe026b456a233f8f0fbe752212e6a0503ca2	b90dc566c1c615289b05b50d61680f23744a21e7	2024-06-05T01:26:14+02:00	jaime-m-p		Fix per token atrributes bits (#7749)
3b38d48609280aa5f8ab7ea135a4351b2a5ee240	6d1616944d9efd342ed2a4fd318722adfc9febcd	2024-06-04T09:17:17+02:00	jaime-m-p		Per token attributes (#7685)
549279d8049d78620a2b081e26edb654f83c3bbd	9e405b6e2ecb888e860f7b92720b4809e21b3915	2024-06-03T08:34:43+03:00	Georgi Gerganov		llama : avoid double token-to-piece cache (#7654)
2e666832e6ac78194edf030bd1c295e21bdb022c	2ac95c9d5678d05e253691fb1f26471675bff5ad	2024-06-01T21:31:48+02:00	Yazan Agha-Schrader		server : new UI (#7633)
2ac95c9d5678d05e253691fb1f26471675bff5ad	750f60c03e4d3f53fa51910551ce87a3d508d2d7	2024-06-01T21:50:18+05:30	HanishKVC		SimpleChat: Simple histogram/repeatMatching driven garbageTrimming, Settings UI, Streaming mode, OpenAi Compat (Model, Authorization Bearer), Save/Restore session, Auto Settings UI (#7548)
5921b8f089d3b7bda86aac5a66825df6a6c10603	5dcdf946764fae49a8e2a90bf2f0960bde1c44e8	2024-05-30T19:01:41+03:00	Georgi Gerganov		llama : cache llama_token_to_piece (#7587)
9c4c9cc83f7297a10bb3b2af54a22ac154fd5b20	59b0d077662fab430446b3119fa142f3291c45b2	2024-05-30T13:40:00+02:00	Galunid		Move convert.py to examples/convert-legacy-llama.py (#7430)
eb57fee51f7b4d78039f003249873c2eb46f12f6	55d62262a99cd8bc28a1492975791fe433c8cc0f	2024-05-30T02:10:40+02:00	Galunid		gguf-py : Add tokenizer.ggml.pre to gguf-new-metadata.py (#7627)
02c1ecad07f0e2d2febe8196271bcc64bdc9c006	6bd12ce409f949012935b7d1b15a21ffa473a565	2024-05-28T21:46:34+02:00	jaime-m-p		Tokenizer WPM fixes (#7500)
5442939fcc5e6ae41abf40612a95fd71377e487e	56411a950f255b523a9edd684fd1632752474399	2024-05-28T20:49:49+02:00	Giuseppe Scrivano		llama : support small Granite models (#7481)
edc29433fa08b4e5aeb67649a29fc7713af13d04	8b99e2aa66ba39e4e1114effea6ef7430881eca4	2024-05-28T15:04:09+03:00	Georgi Gerganov		tests : fix test-tokenizer-0.sh
852aafb163d32d5bad63c10bc323a02c28fec59d	0136966dafb452601c23f30395878d5a65ddc559	2024-05-28T01:40:47+02:00	Djip007		update HIP_UMA #7399 (#7414)
d298382ad977ec89c8de7b57459b9d7965d2c272	32a28217f475119926c603341e8273b26932b56a	2024-05-27T00:10:17+10:00	Brian		main: replace --no-special with --special (#7534)
b9adcbbf92fc7096bee23fe61496d25652ebf765	9588f196b1d7b21bdff013fcf958c249576b2619	2024-05-26T06:26:34+05:30	HanishKVC		SimpleChat Completion Mode flexibility and cleanup, Settings gMe, Optional sliding window (#7480)
00c63907931bb08a0ed2b7e38cf44dd290143cb9	faa0e6979a11dcb731e9d778ad42ceaa0302015e	2024-05-25T05:04:03-04:00	Justine Tunney		main : don't print special tokens with --grammar (#6923)
57684331fc2d685f7d1f5775af0b9e47d1829833	b83bab15a5d2a1e7807d09613a9b34309d86cfaa	2024-05-24T18:14:42-07:00	Mikko Juola		Make tokenize CLI tool have nicer command line arguments. (#6188)
fbca2f27fc7fa9aa4a8ad0357478fdb908472908	0df0aa8e43c3378975269a51f9b876c8692e70da	2024-05-24T14:31:13+02:00	fairydreaming		Add support for ArcticForCausalLM (#7020)
3015851c5ac7334fb544a23a70a284c117b87044	55ac3b7aeaf52f19786ed96e885d89521fc0f6c8	2024-05-23T14:29:26+02:00	Daniel Bevenius		llama : add getters for n_threads/n_threads_batch (#7464)
c3f8d583560b4f261fa21c976793e538c60cd66c	11474e756de3f56b760986e73086d40e787e52f8	2024-05-21T19:53:48+03:00	Georgi Gerganov		tests : test-tokenizer-0.sh print more info (#7402)
d7e852c1bc8e85bf62a6f1aede08cd2de723404a	917dc8cfa67a72fb7c8bf7392270da3bf4833af4	2024-05-21T14:39:48+02:00	jaime-m-p		Tokenizer SPM fixes for phi-3 and llama-spm (bugfix) (#7425)
917dc8cfa67a72fb7c8bf7392270da3bf4833af4	fabf30b4c4fca32e116009527180c252919ca922	2024-05-20T20:15:57+02:00	jaime-m-p		Tokenizer SPM fixes for phi-3 and llama-spm (#7375)
1b01f06db0cff5f5f600bb754fc39fde565ed56a	41858392e17abead21735309bf17cb55183d8c31	2024-05-19T16:26:02+02:00	Johannes Gäßler		server: add test for token probs (#7347)
6aade19ee74b896c59929676629340b36be3e22c	ab33f7a338593f6cf1ae98b10b6f8684f63bd72c	2024-05-19T14:46:46+02:00	Anas Ahouzi		Add StableLM2 pre-tokenizer (#7349)
0f98acfac6cc561dc57586bfff778405e42b576b	ca57e0f35e33f714b9a6c2c4482b87bfe059c819	2024-05-18T10:04:55+02:00	Steffen Röcker		llama : add support for larger Granite Code Models (20B, 34B) (#7324)
29c60d8cddcfd14fa8a6bf023a6c4eb8692c76ba	359cbe3f46c90ce6f5151005e411b8fb74f8139e	2024-05-17T09:59:57+02:00	Johannes Gäßler		tokenization: add warning for double BOS (#7332)
dc020985b8755dd6aa93a2f002f43c3ede808cce	344f9126cc0d15891fde9472fe40b8572628ad7d	2024-05-15T14:44:49+01:00	agray3		Avoid unnecessarily disabling CUDA graphs (#7302)
9aa672490c848e45eaa704a554e0f1f6df995fc8	b1f8af1886e8187db6bb2a9b87cfc1c0f175f629	2024-05-13T10:35:14+02:00	Joan Fontanals		llama : rename jina tokenizers to v2 (#7249)
f99e1e456eaf69cc38c1982a2693ce41c0f897ef	5ae3426b0b64672991563d4c28b2018b9f961467	2024-05-11T16:12:06+08:00	Haoxiang Fei		llama : lookup word in vocab before doing BPE merges (#7193)
5ae3426b0b64672991563d4c28b2018b9f961467	b83cc3f5b303ff30c52874b2d5864dc6385ebf9f	2024-05-11T10:11:28+02:00	Johannes Gäßler		server: fix reported top tokens for temperature 0 (#7203)
b83cc3f5b303ff30c52874b2d5864dc6385ebf9f	9cb317f77e53067f7a138cc89ef7657148eae8e6	2024-05-11T09:46:09+02:00	Joan Fontanals		llama : add Jina Embeddings architecture (#6826)
f89fe2732c5709f6e86d5f4aee2e6d2a561f2eb2	d11afd665241c1b3910ab5f040d0216403019d87	2024-05-10T15:51:58+05:30	HanishKVC		Main+: optionally allow special tokens from user in interactive mode (#7097)
43248e559472556f368988575d9fba906b3eb139	a743d76a01f23038b2c85af1e9048ee836767b44	2024-05-09T15:30:44+02:00	jaime-m-p		llama3 custom regex split (#6965)
f31ec120bc36c6270e4948e6a065a7c4cfa0c404	fd9f92b154850014146f61717cd292a59a5cee5a	2024-05-09T14:13:05+02:00	Galunid		Add warning if token is invalid (#7173)
fd9f92b154850014146f61717cd292a59a5cee5a	22842164bcae3251b81ad9e497a16ef66833cb9e	2024-05-09T13:03:29+02:00	Daniel Bevenius		llama : update llama_timings.n_p_eval setting (#7160)
22842164bcae3251b81ad9e497a16ef66833cb9e	47345248827f426038098d016ed11975021b4919	2024-05-09T12:56:00+02:00	Sigbjørn Skjæret		gguf-py : add special token modification capability (#7166)
f98eb31c517c95960df1d0abc48002787f145f3b	bc4bba364fb96d908f2698e908648df5e6f55e02	2024-05-08T18:16:38-04:00	compilade		convert-hf : save memory with lazy evaluation (#7075)
911b3900dded9a1cfe0f0e41b82c7a29baf3a217	ad211edef5db1f1fb955874b7ca6a67bd0c88708	2024-05-08T14:27:58+02:00	Johan		server : add_special option for tokenize endpoint (#7059)
229ffff872f8ad0d21c997d18ee7a23692ae60a0	1fd9c1741d864d01cd7ec6d67227b92d7bfabf22	2024-05-08T20:06:43+08:00	Ren Xuancheng		llama : add BPE pre-tokenization for Qwen2 (#7114)
4cd621c26de2095cd7c4464bdec5fe2e696ef3f3	7e0b6a7b3ba94ff624dc27c1e0e735fded8819b8	2024-05-08T06:43:23-04:00	DAN™		convert : add BPE pre-tokenization for DBRX (#7132)
af0a5b616359809ce886ea433acedebb39b12969	b6aa6702030320a3d5fbc2508307af0d7c947e40	2024-05-07T23:07:58+02:00	Johannes Gäßler		server: fix incorrectly reported token probabilities (#7125)
3af34c1d1b0da47f85b95f60922abeded1cb5d33	04976db7a819fcf8bfefbfc09a3344210b79dd27	2024-05-07T19:51:31+02:00	RhinoDevel		main : update log text (EOS to EOG) (#7104)
889bdd76866ea31a7625ec2dcea63ff469f3e981	6fbd43221167bf96112f899daf22c127b282cbcf	2024-05-05T01:19:30-04:00	DAN™		command-r : add BPE pre-tokenization (#7063)
03fb8a002df2e96104f9e06de9c78d2a8ed91e92	92139b90af4841d7fd060b526bdd443b621770ff	2024-05-04T12:06:40+03:00	maor-ps		If first token generated from the server is the stop word the server will crash (#7038)
92139b90af4841d7fd060b526bdd443b621770ff	a2ac89d6efb41b535778bfeaecaae8fe295b6ed3	2024-05-04T08:32:32+03:00	Georgi Gerganov		tests : add test-tokenizer-0.sh + fix some tokenizers (#7036)
f4ab2a41476600a98067a9474ea8f9e6db41bcfa	3f167476b11efa7ab08f6cacdeb8cab0935c1249	2024-04-29T16:58:41+03:00	Georgi Gerganov		llama : fix BPE pre-tokenization (#6920)
7f5ff558eed0f732af8f25c2ab0645610bdec80c	9e4e077ec50fde6049b128662c72d37a3c28e34b	2024-04-26T12:15:30+02:00	Pierrick Hymbert		server: stop generation at `n_ctx_train` if `n_predict` is not set (#6638)
37246b1031b1680c0dcaf20aef736d6b446203fa	28103f4832e301a9c84d44ff0df9d75d46ab6c76	2024-04-24T05:15:29-05:00	Kyle Mistele		common : revert showing control tokens by default for server (#6860)
3fec68be4e9577fc53158366d3b3af039c17bb1f	c8297c6af5693555652c40b95974b95d49d2674d	2024-04-24T15:16:21+08:00	Junyang Lin		convert : add support of codeqwen due to tokenizer (#6707)
c8297c6af5693555652c40b95974b95d49d2674d	4e96a812b3ce7322a29a3008db2ed73d9087b176	2024-04-24T15:00:37+08:00	liuwei-git		llama : add phi3 support (#6852)
8960fe86ae075c846c5df8848230d1904ba8877f	c0956b09ba845a7cd787d5580d7c8b96e80f40f5	2024-04-22T15:41:11+03:00	Georgi Gerganov		llama : fix typo in <|im_end|> token text (#6745)
40f74e4d739e9250431cf339ae7588b28d8d0663	b9cc76d87e3d7ae5900f19d4fe8f8976d0a35888	2024-04-21T18:36:45+03:00	Georgi Gerganov		llama : add option to render special/control tokens (#6807)
7dbdba5690ca61b3ee8c92cfac8e7e251042e787	c1386c936e9fbc38eb2816c711ab28f13355708e	2024-04-21T15:03:39+02:00	Wouter		llama : add llama-3 chat template (#6751)
b97bc3966e852adb626c90be64fd48282800f504	b8109bc0139f15a5b321909f47510b89dca47ffc	2024-04-21T13:50:41+02:00	Pedro Cuenca		llama : support Llama 3 HF conversion (#6745)
03c0946d73c63ea73e1d85015b7088298443d438	e11b2e6e1e18522ca7cf129600875a0f6fb9307d	2024-04-18T13:49:01+02:00	Sigbjørn Skjæret		convert : support models with multiple chat templates (#6588)
8cc91dc63c0df397d644a581b2cbeea74eb51ae0	dbceec87c0221ec952e69448df6a71f1372a7487	2024-04-16T14:55:30-04:00	Justine Tunney		ggml : add llamafile sgemm (#6414)
4fbd8098e63670c6ae11a8adc350f5ba191cfda3	7593639ce335e8d7f89aa9a54d616951f273af60	2024-04-16T08:13:13+02:00	Daniel Bevenius		gguf : add special tokens metadata for FIM/Infill (#6689)
1b67731e184e27a465b8c5476061294a4af668ea	c4a3a4ff47d62d2503ddf9bd91b58c21f04fe3c3	2024-04-09T13:44:08-04:00	Jared Van Bortel		BERT tokenizer fixes (#6498)
5dc9dd7152dedc6046b646855585bd070c91e8c8	e11a8999b5690f810c2c99c14347f0834e68c524	2024-04-09T09:16:13+01:00	Carolinabanana		llama : add Command R Plus support (#6491)
75cd4c77292034ecec587ecb401366f57338f7c0	a8bd14d55717754a1f48313a846a2b16fa998ad2	2024-04-06T05:40:47+02:00	Pierrick Hymbert		ci: bench: support sse and fix prompt processing time / server: add tokens usage in stream OAI response (#6495)
db214fa578e00b01e0884fc2725c9349608bdab5	1ff4d9f3d683f02ef8a12e04bfac84300c44bd3a	2024-04-03T21:12:52+05:30	Abhishek Gopinath K		Missing tokenizer.model error during gguf conversion (#6443)
1ff4d9f3d683f02ef8a12e04bfac84300c44bd3a	076b08649ecc3b0e1c0709c2a086a63eddd1bf32	2024-04-03T23:24:31+08:00	kaizau		Add OpenChat, Alpaca, Vicuna chat templates (#6397)
057400a3fd457f4f214684eeb171444663b47a23	b75c38166cf0c66793a32d5c3d6cb69929dd083d	2024-03-29T08:23:22+01:00	Daniel Bevenius		llama : remove redundant reshape in build_kv_store (#6369)
b75c38166cf0c66793a32d5c3d6cb69929dd083d	bfe7dafc9cf96b9a09ead347fed9a547930fc631	2024-03-29T08:15:00+01:00	Pedro Cuenca		convert : allow conversion of Mistral HF models (#6144)
557410b8f06380560155ac7fcb8316d71ddc9837	55c1b2a3bbd470e9e2a3a0618b92cf64a885f806	2024-03-26T10:46:41-04:00	compilade		llama : greatly reduce output buffer memory usage (#6122)
e097633f63fdd26d492844f7eff056e4083fd9eb	d25b1c31b07c3675443a55a828dd58cfef5a241c	2024-03-26T13:32:19+01:00	Pedro Cuenca		convert-hf : fix exception in sentencepiece with added tokens (#6320)
94d1b3b4119209efcdd08df0dceaecbd1fe7f85c	95562175f83a49755ff6fd3bad09409417c8e6f9	2024-03-23T18:48:02-04:00	Jared Van Bortel		use _wfopen instead of fopen on Windows (#6248)
1d0331c12a2f2a6296b471232bd4e66fbf06e6a1	dba1af612926cbd4ebe2d876277af1e3305177e0	2024-03-22T19:47:14+01:00	Kawrakow		quantize: options for output and token embedding tensors qtype (#6239)
bd60d82d0cc8b6852ec535495a5042dbdf05de24	6c0b287748327741b113d7d6018b68c63039b1c5	2024-03-20T01:33:49-04:00	Jared Van Bortel		server tests : more pythonic process management; fix bare `except:` (#6146)
b0bc9f4a9da7c19f4779106ea83b23feca747566	4755afd1cbd40d93c017e5b98c39796f52345314	2024-03-15T09:22:24+01:00	slaren		llama-bench : use random tokens to improve accuracy with mixtral (#6069)
044ec4b2a567f649459ccd20af2f387c784faa51	77178eedc83d49f31bf757d8e12315d76460be78	2024-03-14T15:14:14+02:00	Georgi Gerganov		embedding : add EOS token if not present (#899)
f30ea47a87ed4446ad55adb265755dc9102956a2	d8fd0ccf6ac8b07791ffd1575eed436930854ae3	2024-03-13T18:54:21+01:00	slaren		llama : add pipeline parallelism support (#6017)
7ab7b733bb48250b2df26c12b00256ef42c76932	d9f65c97c3dc3aa6fa27470b8c6e69b437ec1a27	2024-03-11T04:03:17+08:00	Dean		android : fix utf8 decoding error (#5935)
621e86b331f8b0e71f79fd82a4ae1cd54c3e4396	77d1ac7e00bf049b9f2bba1b5a310a78318c49c4	2024-03-09T23:41:49+01:00	Pierrick Hymbert		server: benchmark: chat/completions scenario and other llm servers comparison (#5941)
0db32beaf09d90b8959d3d0cc493ed1e45685353	8a3012a4ad08112bb3dc3f1399afec4e93780c44	2024-03-09T11:16:53Z	Alexey Parfenov		server : fix passing prompt as tokens (#5955)
c2101a2e909ac7c08976d414e64e96c90ee5fa9e	515f7d0d4fce41c752fc253acf30707c3be2531e	2024-03-08T17:31:00-05:00	compilade		llama : support Mamba Selective State Space Models (#5328)
515f7d0d4fce41c752fc253acf30707c3be2531e	76e868821a94072fbc87cb1fcca291694319eae8	2024-03-08T10:53:37-05:00	compilade		llama : fix quantization of shared token_embd (#5944)
76e868821a94072fbc87cb1fcca291694319eae8	e457fb3540e0aaec47cfde0abf784c213f9216ee	2024-03-08T12:25:04+01:00	Pierrick Hymbert		server: metrics: add llamacpp:prompt_seconds_total and llamacpp:tokens_predicted_seconds_total, reset bucket only on /metrics. Fix values cast to int. Add Process-Start-Time-Unix header. (#5937)
af37fd8b30e37ccbffdd82e6f48559e2fb7ce7dd	581ed5c4fe3a8909aaa8313633ac443f471ba755	2024-03-08T12:40:02+02:00	Georgi Gerganov		server : fix EOS token detection with disabled cache (#5938)
2002bc96bf2cbf5ab981a17d7e994d817c9801f5	ceca1aef0738b57951cd12c603c3477e75312dec	2024-03-07T11:41:53+02:00	Georgi Gerganov		server : refactor (#5882)
21b08674331e1ea1b599f17c5ca91f0ed173be31	6a87ac3a52668e117d97bcea07b529c93188b303	2024-03-05T16:08:35+08:00	Neo Zhang Jianyu		[SYCL] fix mul_mat fault in CI/unit-test (#5862)
29ae62d2ae163e2b68aa0ad3bf2ab4636de0c957	e0843afe1b37890b631bc7d3d2da2ed36c862b91	2024-03-04T22:31:20+02:00	Georgi Gerganov		llama : fix embeddings (#5796)
5a51cc1bb4592f0d71f9af89cd08b11a066ba447	67be2ce1015d070b3b2cd488bcb041eefb61de72	2024-03-04T02:57:20-05:00	DAN™		main : support special tokens as reverse/anti prompt (#5847)
9731134296af3a6839cd682e51d9c2109a871de5	4a6e2d6142ab815c964924896891e9ab3e050632	2024-03-02T22:00:14+01:00	Pierrick Hymbert		server: tests: passkey challenge /  self-extend with context shift demo (#5832)
4d4d2366fc9c54d4a275065cfe9299c6cf7c5b78	c7a0ad8ec9ebb5ddb1c1c80c82f2ee041c525d47	2024-03-02T12:27:26-05:00	Jared Van Bortel		convert : automatically fall back to HfVocab if tokenizer.model doesn't exist (#5821)
177628bfd85565070916ad66a5ac4071ee0527d8	6c4416868df2e5455da7d20547f62bcf9735ba8e	2024-02-28T02:51:11-06:00	Douglas Hanley		llama : improve BERT tokenization (#5740)
f7625019c51ca437a5840576d92362cfa710e4a2	abbabc5e51d0d4656b438aec10b7fae9479ef37d	2024-02-25T13:43:50-05:00	compilade		server : fix crash when system prompt is bigger than batch size (#5714)
525213d2f5da1eaf4b922b6b792cb52b2c613368	fd43d66f46ee3b5345fb8a74a252d86ccd34a409	2024-02-24T12:28:55+01:00	Pierrick Hymbert		server: init functional tests (#5566)
15499eb94227401bdc8875da6eb85c15d37068f7	96633eeca1265ed03e57230de54032041c58f9cd	2024-02-22T17:05:23-05:00	Jared Van Bortel		mpt : do not duplicate token_embd.weight on disk (#5670)
96633eeca1265ed03e57230de54032041c58f9cd	847eedbdb2d1ebf14ef56eb507d4b4b975510908	2024-02-22T23:23:46+02:00	Georgi Gerganov		gemma : use more bits for the token_embd.weight tensor (#5650)
73122473ffd73030146276dbb85da7c8021a3ee4	0d4177126b0556e202efb85bf3f768be81076400	2024-02-15T14:14:37+01:00	Michaël de Vries		fix(gguf-py): special tokens are no longer skipped when add_<token>_token is set to false (#5487)
c4e6dd59e45ef7b14f7763fb073b517395dc176c	037259be689353081e7bae3c1ab4ab18e7fbe8c9	2024-02-13T18:18:16+02:00	Aarni Koskela		llama : allow raw byte in SPM vocabs; don't crash on nl 404 (#5478)
cf45252a7cfcb998bade46a886e20477cecc538a	03bf161eb6dea6400ee49c6dc6b69bdcfa9fd3fc	2024-02-13T15:14:22+02:00	Georgi Gerganov		tests : multi-thread the tokenizer tests (#5474)
49cc1f7d67de2da99f3ac185f9ff1319b7bf35f8	99b8b43d7b185a6483f28cf798a2d968b2e16ca7	2024-02-13T13:01:29+02:00	Georgi Gerganov		bert : add tests + fix quantization (#5475)
2891c8aa9af17f4ff636ff3868bc34ff72b56e25	97a336507ed9b971d72262bec7e2b8b7016a054a	2024-02-11T10:21:38-06:00	Douglas Hanley		Add support for BERT embedding models (#5423)
684780141a08200ec98eba3e982dbafd1d0b5000	85910c5b30f6e268321be8df044f5528a6efac52	2024-02-11T13:38:14Z	Alexey Parfenov		server : allow to specify tokens as strings in logit_bias (#5003)
0ef46da632c32faa1a538e5dc180994e8bbb46e1	ee1628bdfea8b0079fed0140ac2f00ef1b465b57	2024-02-07T02:17:25-06:00	Xiao-Yong Jin		llava-cli : always tokenize special tokens (#5382)
316c7faf7740fa98ea68f1445f4505810f706b9e	f3e2b4fa3f81a410ecb7dec929c259ef8d8dbb7d	2024-02-07T14:15:56+08:00	runfuture		llama : add MiniCPM support (#5346)
906cff55c2848fda091d888a1585915ec0c9ea9e	098f6d737b65134cf220d12b9b706e8cfc5e4610	2024-02-06T07:47:22+02:00	Georgi Gerganov		py : handle byte tokens in `get_token_type` (#5341)
e1e721094d8169636d55f68efe37f222cd3f0677	128dcbd3c9c4b12f42b560a4430427d7b2828628	2024-02-01T23:20:13-08:00	Ian Bull		llama : fix memory leak in llama_batch_free (#5252)
e6f291d15844398f8326940fe5ad7f2e02b5aa56	4003be0e5feef320f3707786f22722b73cff9356	2024-01-30T20:17:30+02:00	Georgi Gerganov		server : fix context shift (#5195)
e76627bcce9f77adb6034cb127b7ec93d4287b69	fbe7dfa53caff0a7e830b676e6e949917a5c71b4	2024-01-29T18:24:19+09:00	Sang-Kil Park		py : improve BPE tokenizer support (#5189)
9241c3a2ace544aef708334e54bbdddb90208ee8	b2b2bf988c098851b4f3831f0cf38394bff75121	2024-01-28T09:59:49+01:00	Johannes Gäßler		Apply min_p to unsorted tokens (#5115)
44879ee885f48ecf4675dd216b373dce0a6f3690	9ecdd12e95aee20d6dfaf5f5a0f0ce5ac1fb2747	2024-01-23T15:17:20+02:00	Kawrakow		Additional KL-divergence statistics (#5081)
d6bd4d46ddb6926087c11e0f6633ab1c81da58c3	152d9d05e097e35f1cac21262946d57faec7542a	2024-01-22T06:21:52-05:00	compilade		llama : support StableLM 2 1.6B (#5052)
8b20858e5e9c44b99b4b31ae9c40b8f20d01d94f	57e2a7a52a819883f40dada8a2edc24ecf48186b	2024-01-19T10:45:06+02:00	Georgi Gerganov		perplexity : faster Winogrande via batching (#5024)
ad19812cda4062c9f154ef16315df41fbe6a770a	682986a08eb5cb04865d2e713449f17304d266d8	2024-01-18T15:33:01+02:00	Georgi Gerganov		perplexity : faster HellaSwag via batching (#5017)
4f4bf35f46600441dec2f941e667291eeb9a18d8	2b3a665d3917edf393761a24c4835447894df74a	2024-01-17T15:45:03+02:00	Georgi Gerganov		py : fix missing added_tokens_dict for SPM and BPE vocabs (#4971)
cec8a4847062fbd76253e3b085683f39d91e80d3	334a835a1ccc8106a5fa355683a965efb1bfa24b	2024-01-16T18:54:24+01:00	Daniel Bevenius		finetune : add training data file to log message (#4979)
f172de03f11465dc6c5a0fc3a22f8ec254c6832c	2d57de525541247132e354f561ff48775fba5d85	2024-01-13T18:47:38+02:00	Georgi Gerganov		llama : fix detokenization of non-special added-tokens (#4916)
df845cc982e7e2ea7b9900e29d55b15338faa78d	6b48ed089377330cdb362970a51c1c89b6d857a8	2024-01-13T17:29:43+01:00	David Friehs		llama : minimize size used for state save/load (#4820)
722d33f34ec74c6f7046109f936d0928ffe171bc	c30b1ef39aeba497a943416d2897d69fee055b96	2024-01-14T00:09:08+08:00	Yann Follet		main : add parameter --no-display-prompt (#4541)
ee8243adaa9a9f51ff449213383874e49efe368f	15ebe59210e7fd9817ff67f51fa1a5ee2d004294	2024-01-13T14:16:11Z	makomk		server : fix crash with multimodal models without BOS token (#4904)
3ca63b4538dfc78aaec88cd2c3e3f8417c1924e3	b0377875488b33f7114138687d828da1de61775d	2024-01-11T22:43:05+02:00	Georgi Gerganov		main : disable token count by default (#4874)
d8d90aa343c22fe01429d3540e47ded87e9dcb9d	43f76bf1c362c067fce46bb8dcda0b64af8a9533	2024-01-11T17:22:34Z	Someone		ci: nix-flake-update: new token with pr permissions (#4879)
43f76bf1c362c067fce46bb8dcda0b64af8a9533	2f043328e3116724d15b915b5c6078e2df860a69	2024-01-11T16:14:52Z	pudepiedj		main : print total token count and tokens consumed so far (#4874)
6efb8eb30e7025b168f3fda3ff83b9b386428ad6	36e5a08b203542dca53cca4eaf172c5dc4bbc991	2024-01-09T13:46:46-05:00	Austin		convert.py : fix vanilla LLaMA model conversion (#4818)
128de3585b0f58b1e562733448fc00109f23a95d	8c5833031857c9e9ada61948bae894ab9c785f86	2024-01-09T05:02:05-05:00	Behnam M		server : update readme about token probs (#4777)
dd5ae06405c5565b99889bdb3f168f4351252cfb	668b31fc7d86245435ad6574e0e1126e734049e2	2024-01-08T16:02:32+01:00	Kawrakow		SOTA 2-bit quants (#4773)
c75ca5d96f902564cbbbdd7f5cade80d53c288bb	96e80dabc6e73ff68b09b68947b1fc25883c5094	2024-01-06T16:12:59+01:00	Daniel Illescas Romero		llama.swiftui : use correct pointer for llama_token_eos (#4797)
012cf349aec8ffb47c9def5dc018240fa3721e8b	a91928014fcf51fe297823fcff0788de4f14206e	2024-01-04T19:56:33+02:00	Georgi Gerganov		server : send token probs for "stream == false" (#4714)
0ef3ca2ac62016c0c545de1c89dc2e3e130f4a99	540938f8904707dd74cb3be18495f853b312e72f	2024-01-02T15:48:49Z	Phil H		server : add token counts to html footer (#4738)
0040d42eeb237197054cc7790df5776eacfa608e	83e633c27efdf0eb0ba54249e784b0ea760b1007	2024-01-02T06:15:16-08:00	Marcus Dunn		llama : replace all API facing `int`'s with `int32_t` (#4577)
65e5f6dadbba4b496bba27f573e473c66b446496	ea5497df5d138c83b2b0ca70aefdc4b1175c1001	2023-12-28T11:20:00-08:00	Justine Tunney		Fix OpenAI server sampling w.r.t. temp and seed (#4668)
f56d6077d0c37e6606ac0a4fa3169de70593acfe	dc68f0054cd279cddddb0cae0c9ef4f9cbaa512a	2023-12-27T17:37:25+09:00	wonjun Jang		Add byte token type when tokenizer.model is not exists (#4641)
7082d24cec35e9ce9147535a2224dfc67ee0a78c	ba661751322a7c201fd3bef71af077c5aebfaa2a	2023-12-22T17:05:56+01:00	LeonEricsson		lookup : add prompt lookup decoding example (#4484)
880e352277fc017df4d5794f0c21c44e1eae2b84	66f35a2f48e1965a13835a523e677223dbf148be	2023-12-21T18:07:34+01:00	howlger		py : open merges file as 'utf-8' (#4566)
b9e74f9bca5fdf7d0a22ed25e7a9626335fdfa48	3c04bf6da89eaf4c7d317e0518f0687dfcbf2de7	2023-12-18T17:27:47Z	Ebey Abraham		llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490)
2994f0c5a2e8c96955b422dedc93ec2595d16b82	b1306c439490c7fa4ec33594500d980d1e9e15e6	2023-12-17T19:39:02-05:00	Jared Van Bortel		decode : fix logits_valid for legacy API (#4516)
f7f468a97dceec2f8fe8b1ed7a2091083446ebc7	919c40660fd27157b391b5832d2a577d5afef4cb	2023-12-17T10:45:46-05:00	Jared Van Bortel		gguf-py : fail fast on nonsensical special token IDs (#4489)
873637afc7924f435ac44c067630a28e82eefa7b	0353a1840134b24b07ab61fd4490192f28c4db6b	2023-12-14T17:09:34+09:00	wonjun Jang		convert : support loading vocab from fast tokenizer config (#3633)
948ff137ec37f1ec74c02905917fa0afc9b97514	4d98d9a65665eee3838cef936641f640e3f5b649	2023-12-13T23:57:15+04:00	shibe2		server : fix handling of characters that span multiple tokens when streaming (#4446)
799a1cb13b0b1b560ab0ceff485caed68faa8f1f	fecac45658a99eddc4d6e36ba0310ca8f87a77f0	2023-12-13T13:04:25+01:00	slaren		llama : add Mixtral support (#4406)
6391817cd19a4507c6c941a1fd08756268662b2d	d9d4cfef64ea416dd66632173787d03ffb180cc7	2023-12-12T04:25:57-05:00	crasm		llama : document logits_all deprecation (#4418)
e18f7345a300920e234f732077bda660cc6cda9c	fe680e3d1080a765e5d3150ffd7bab189742898d	2023-12-09T16:29:27-05:00	Xiang (Kevin) Li		grammar : revert the replacement of llama_token_to_piece with id_to_token (#4396)
5f6e0c0dff1e7a89331e6b25eca9a9fd71324069	5aa365d88fdb8fdd430ef3fc141c7a5fd37c3502	2023-12-05T10:55:12-10:00	Marcus Dunn		grammar : pre-computed pieces + reserve mem + less string copies (#4330)
23b5e12eb5a76489b4c3ee22213a081da68b1809	d208995c6da66f252d4054c1c5a90eb8ccb7a2f7	2023-12-04T17:04:21+01:00	Daniel Bevenius		simple : update error message for KV cache check (#4324)
5c9f90cba1cc6b0a2a7d19ee5dcb73cad6331d30	4fa44e84adb4c78e1885694cc3513982d4af2b08	2023-12-04T22:43:45+09:00	Miwa / Ensan		swift : fix prompt tokenization logic (#4321)
b220222a64ce760bfbec9c770f11db3ec6a6abb6	511f52c334e37033f9c9de07b98fca4abc9470bd	2023-12-02T03:19:45+09:00	Miwa / Ensan		swift : fix token_to_piece implementation (#4278)
b18c66ca6eee4fe0465cff5042daf05005dc9ab2	f4d973cecb7368c985720ba9100ae6abba14806d	2023-11-30T22:43:08+01:00	Daniel Bevenius		llama : fix alignment of general.name in print meta (#4254)
8406b0924bf323f37d536dee8b8165c1f3d9d11d	b38a16dfcff88d547f78f52d1bea31b84a05aff7	2023-11-28T10:32:03+02:00	Georgi Gerganov		ggml : re-enable BLAS for CPU when src0 != F32 + remove redundant full offload checks in llama.cpp (#4240)
922754a8d60080e956891f6cee1fb03aa48d57c6	22da05536ff4ad963080773bef1fb839fdab95d3	2023-11-26T20:33:07+02:00	Georgi Gerganov		lookahead : add example for lookahead decoding (#4207)
f837c3a992b2b6146936cb120871a8cf9d0e3857	3014b5415d08e3dff961da6eea835b9760a701b8	2023-11-25T08:58:23-08:00	Marcus Dunn		llama : grammar `reserve` space in `decode_utf8` (#4210)
af19d3573481d409b3c4e55494810eb1f65a9aae	e9c13ff78114af6fc6a4f27cc8dcdda0f3d389fb	2023-11-25T11:29:06+02:00	Georgi Gerganov		server : OAI API compatibility (#4198)
2568a4bf548d7392e9c78c008b33b4c11d53fe95	b35f3d0def3efde92ed465d92a267430d957e87d	2023-11-24T18:25:10+09:00	eastriver		main.swift : fix eos checking (#4197)
40a34fe8d034bd484efd79ccbb95059ca6308dcb	dae06c06e5c6232ae2be4d567dd5101e1e96c814	2023-11-20T03:50:04-06:00	Branden Butler		speculative : fix prompt tokenization in speculative example (#4025)
35985acffaab1eb4ffcbc22c86063bc630f24a89	e937066420b79a757bf80e9836eb12b88420a218	2023-11-19T18:50:49+02:00	Georgi Gerganov		gitignore : tokenize
28a2e6e7d476717881be6eb9e2d3331342cec57b	0b5c3b04572a05f80163a365070fb377a837ac27	2023-11-18T14:48:17-07:00	Kerfuffle		tokenize example: Respect normal add BOS token behavior (#4126)
5ad387e994dde77a47ec547a4a65f7611dc325f4	2fa02b4b3d86182381311c98b75065ee1b7c2930	2023-11-17T18:01:38+02:00	Georgi Gerganov		tokenize : fix trailing whitespace
2fa02b4b3d86182381311c98b75065ee1b7c2930	2ab0707acbf3a3ca9e5bc5959c7920c22eba2257	2023-11-17T17:36:44+02:00	zakkor		examples : add tokenize (#4039)
91f6499393d2d999331fbfdba47a7f8b9f913f0d	8da46278e1a57107591653275f8e03a281de94f0	2023-11-16T19:14:37-07:00	Kerfuffle		Respect tokenizer.ggml.add_bos_token value when tokenizing (#4040)
a6fc554e268634494f33b0de76f9dde650dd292f	1cf2850d52bb027aa215e039ed9a0c61beeef8d3	2023-11-15T11:34:47-05:00	Jared Van Bortel		llama : restore prefix space in llama tokenizer (#4081)
34b0a082074b073eb14c2bd93c0c070e20ddcd16	4a4fd3eefad5bd17ab6bcd8e2181b4f62eae76cf	2023-11-10T22:04:50-07:00	Kerfuffle		gguf-py: Refactor and allow reading/modifying existing GGUF files (#3981)
238657db2364cfb728c694470a4a81702afea760	07178c98e1b61a5e2af39d347add12e7eb9e08e1	2023-10-31T14:44:49-05:00	kalomaze		samplers : Min-P sampler implementation [alternative to Top P/Top K] (#3841)
6e08281e588bbba1a5d180290a94a43f167f3a1a	2046eb4345e62c4575b3cdc0115a51db89f3fb70	2023-10-29T11:31:40-06:00	Kerfuffle		Extend llama_kv_cache_seq_rm to allow matching any sequence (#3843)
8a2f2fea2914aaa3f4b2f82800c7de15f15bdb09	bd6d9e205982b34e0ba2c3b22bbf31a1ef1a1bb5	2023-10-28T15:25:15+03:00	Georgi Gerganov		convert : ignore tokens if their IDs are within [0, vocab_size) (#3831)
ee1a0ec9cb367ba41d138134795cbbbe93d2bf1c	177461104b454163473dced2a5038f4e016cdb7e	2023-10-28T14:23:11+03:00	Georgi Gerganov		llama : add option for greedy sampling with probs (#3813)
daab3d7f45832e10773c99f3484b0d5b14d86c0c	469c9addef75893e6be12edda852d12e840bf064	2023-10-24T09:17:17+02:00	Galunid		Add more tokenizer tests (#3742)
69a6735087c3634963c642fd69f0851ac479cd78	5be6c803fa5378f62a1590f3ad8c6b64c7c0c2ce	2023-10-23T21:46:00+02:00	Galunid		Update special token handling in conversion scripts for gpt2 derived tokenizers (#3746)
5be6c803fa5378f62a1590f3ad8c6b64c7c0c2ce	6336701c9378c23c85d1c0e464b663ca2bbb8e60	2023-10-23T12:40:03-07:00	Marcus Dunn		llama : remove token functions with `context` args in favor of `model` (#3720)
438c2ca83045a00ef244093d27e9ed41a8cb4ea9	9e70cc03229df19ca2d28ce23cc817198f897278	2023-10-22T22:53:08+03:00	Georgi Gerganov		server : parallel decoding and multimodal (#3677)
9e70cc03229df19ca2d28ce23cc817198f897278	5a42a5f8e8a86da9ac88008d748cf232a83aa0e1	2023-10-22T21:21:42+02:00	goerch		Add test for MPT tokenization (#3728)
a5e7dbd6141128bfa3c40a19c2945a181df625d3	d3956aea53369455008159cc405ed4c496976692	2023-10-22T12:14:56-06:00	Kerfuffle		llama : validate special token ids are in range when loading GGUF model (#3635)
d3956aea53369455008159cc405ed4c496976692	22c69a27945e7acf9690dd3210d316f22182751c	2023-10-22T20:09:51+02:00	vvhg1		main : escape prompt for cfg_negative_prompt and consecutive inputs in main with interactive (#3623)
40e5ce054f4c4fa555e4510ea5f760bb29185332	a5e8c1d8c71f01d98ae2ec63a57c118664f9764d	2023-10-11T21:30:06+04:00	shibe2		CLBlast: Fix temporary buffer size for f16 conversion (wsize)
1a159553f921a9209fed8c714494e57b3649f232	281ef73c258cc1eebec8a64264240432d5878c4b	2023-10-17T17:11:01+02:00	staviq		tokenizer : special token handling (#3538)
940efa95fec0b8a98c226a889d2ad839dfeeae0d	11bff290458f12f020b588792707f76ec658a27a	2023-10-16T23:58:00+03:00	Georgi Gerganov		llava : fix tokenization to not add bos between image embeddings and user prompt (#3645)
2a4bcbacead886996f175f33479d1d874a3e577f	424b6381c4daeed62e6600e0402e72f39845b58d	2023-10-13T12:33:16+02:00	Daniel Bevenius		llama : remove n_threads from llama_decode_internal (#3614)
233fc1c69f6f415f35363e18a755f9610e89161b	c5b49360d0d9e49f32e05a9116e90bd0b39a282d	2023-10-10T18:59:52+02:00	goerch		Minor improvements in GPT2 tokenizer (#3567)
f5f9121de140eff558f13b5c5e78a3a3b6b94377	11ea5c7d96f2c28e1c99659e08ec0a44574056e2	2023-10-10T09:50:23+02:00	Jan Ploski		llm : add MPT support (#3417)
11ea5c7d96f2c28e1c99659e08ec0a44574056e2	95bd60a0a69f57e9a2ff1269667ea484a1a9bb40	2023-10-10T09:31:21+02:00	vvhg1		infill. : fix tokenization (#3508)
3a716b4dae545c3db307594fbc509a95d3e21b6e	1faaae8c2bdc4a21302e367e0754c3fe74a8113e	2023-10-07T06:57:01+02:00	goerch		Fix for #3454 (#3455)
9ca79d5cbbc8d43f2bff951404b6a40ff1ee3788	0c731ca4039ccff86ffab90eaae4ca98037c4496	2023-10-06T10:10:13-06:00	Kerfuffle		kv cache slot search improvements (#3493)
97af49fa395df77e4c18af0e1655b2fee67c9686	16820a5a0d885113f21021ce934f0b0027b9d69a	2023-10-06T07:44:24-05:00	Jhen-Jie Hong		server : reuse llama_sample_token common util (#3494)
e8b8d32e8663ffc55a02c9721af3a5190382cbb0	8f3a642ec1d878b2d0a0d15e3a4277f522790d4c	2023-10-05T09:02:55-05:00	Jhen-Jie Hong		server : fix incorrect num_tokens_predicted (#3480)
ac2219fef34eb5b713c286c34c6e4162c39c8f3b	48be797ffbd80b062f55778e09e97180eb25d2ab	2023-10-03T21:04:01+03:00	Georgi Gerganov		llama : fix session saving/loading (#3400)
ff5a3f0c09dfa0a8e0bf76d1748df5c6dee0e8ff	1c84003c08027f5d3a4cb876f51d6b6224a34d0e	2023-10-03T09:16:26+02:00	goerch		Work on the BPE tokenizer (#3252)
16bc66d9479edd5ee12ec734973554d4493c5dfa	0512d66670de3f650c579519833c085014b0f200	2023-09-28T21:42:38+02:00	slaren		llama.cpp : split llama_context_params into model and context params (#3301)
0e76a8992c8200237bbc6471a53fb8796b3872f7	2db94d98eda56982d80238840b0652b4137a2a84	2023-09-28T20:40:11+02:00	xaedes		train : finetune LORA (#2632)
ecf90b1a5114034bc0939b3968f549fe4d63cf6d	2619109ad57d7a75388a9cce51e5da645410d92e	2023-09-28T14:30:15-04:00	Cebtenzzre		gguf : make token scores and types optional (#3347)
ec893798b7a2a803466cc8f063051499ec3d96f7	45855b3f1c7bdd0320aa632334d0b3e8965c26c4	2023-09-28T19:04:36+03:00	Georgi Gerganov		llama : custom attention mask + parallel decoding + no context swaps (#3228)
b08e75baea294e366628b898e85c0bd359b58115	e6616cf0db2b63189fc34d0076f654af9adecdf8	2023-09-16T13:41:33+02:00	goerch		Fixing the last deviations from sentencepiece indicated by test-tokenizer-1 (#3170)
35f73049af6c676a106a5a990a819ae0bc3fcd7d	71ca2fad7d6c0ef95ef9944fb3a1a843e481f314	2023-09-14T09:14:44-07:00	Leng Yue		speculative : add heuristic algorithm (#3006)
71ca2fad7d6c0ef95ef9944fb3a1a843e481f314	1b6c650d16048d6427dd502a9627e72837265844	2023-09-13T15:19:44+02:00	goerch		whisper : tokenizer fix + re-enable tokenizer test for LLaMa (#3096)
15b67a66c2f2d6032415b28a699b5131962318f1	be8c9c245bd129ebabb80e0a7a8dd7daeb4d30af	2023-09-07T15:52:34+02:00	slaren		llama-bench : use two tokens in the warmup run for prompt evals (#3059)
c4f496648c1e32efeb714200e7eae7fc7cfbb223	fec2fb19e4229aac58c98171c46e77144b99f8a3	2023-09-07T15:49:09+03:00	Georgi Gerganov		metal : fix kernel_norm (fixes Falcon on Metal) (#3057)
921772104ba2219bfdc2b2980d05ebc0aa0c92a4	2ba85c8609309a59d49c45ab43c31800b7ba141c	2023-09-05T08:46:17+03:00	Georgi Gerganov		speculative : add grammar support (#2991)
37301347767d555d0a66c043ce4ef6ead8e61c55	d9151e6f570eb20bfd54427bd8a337d9b1a08018	2023-09-03T19:18:09+09:00	opparco		llama : fix bpe tokenize from byte (#2889)
afc43d5f82588d2ed71ea104e8262f5e5da13980	6460f758dbd472653296044d36bed8c4554988f5	2023-09-03T11:48:49+03:00	Alon		cov : add Code Coverage and codecov.io integration (#2928)
cff7b0bf07cb46e1ad4fd199f6bdeb538925c8c4	340af42f09a80e32f4998857b4f0543e41124525	2023-09-02T23:52:13-06:00	Kerfuffle		convert.py : BPE fixes (#2938)
950929442070874d45561d2a4b68b010457767de	35092fb54712d032860f3976a6fc1ae1f84a4a28	2023-08-30T12:42:51+03:00	alonfaraj		make : add test and update CI (#2897)
dc07dc492ef9640bbb82904d7c7679f7bdcf6d76	ad9ddcff6ef322db5cf13785bd7c856b610d242e	2023-08-30T02:25:50-06:00	Kerfuffle		convert : various script cleanups/fixes + merges and special token handling (#2842)
8341a25957b319a03d4a811176cd5ad7f2b0fbd4	849408957c687cde4ab32c147107f643fc55130b	2023-08-30T08:29:32+02:00	staviq		main : log file (#2748)
fa3582f509a2715e80a473e79f88dcd1ebff44c2	e37e69dcc3d52f21222a63cafed2a71b3f6b53c6	2023-08-29T23:55:45+03:00	Kawrakow		Tell users attmepting to run perplexity with too few tokens to use more (#2882)
e37e69dcc3d52f21222a63cafed2a71b3f6b53c6	53885d7256909ec3e2176cdc2477f3986c15ec69	2023-08-29T23:55:03+03:00	Kawrakow		10X faster BPE tokenizer (#2876)
44c117f41ee01c5ac8fb86bba041f08d8b87b46d	43033b7bb4858da4f591715b3babdf906c9b7cbc	2023-08-28T21:51:47+02:00	xaedes		train : mem usage and other improvements  (#2439)
230d46c723edf5999752e4cb67fd94edb19ef9c7	463173a6c0ff353055eb90665794884c888c790f	2023-08-27T15:13:31+01:00	Olivier Chafik		examples : update llama2.c converter to read vocab and write models in GGUF format (#2751)
463173a6c0ff353055eb90665794884c888c790f	eaa13a48ff4136f01c1cdb79cacd61b67ec53095	2023-08-27T16:50:33+03:00	Kawrakow		llama : speedup tokenization (#2831)
edd4c1481708fcd788b0e423268304fd26e2b125	1591e2e590762011b43b10a9b6e04f13f98f2aa5	2023-08-27T14:19:19+03:00	Georgi Gerganov		llama : more tokenizer fixes (#2810)
c1ac54b77aaba10d029084d152be786102010eb2	730d9c681e339b76407659344e5a2cd50af7d7d5	2023-08-26T16:11:45-07:00	Bruce MacDonald		server : add `/detokenize` endpoint (#2802)
72f895c923ba98b8f2af294440206f35915c0501	50526f37eba0b28336700890242ff282b949cd83	2023-08-26T14:12:56-04:00	Dr. Tom Murphy VII Ph.D		main : fix bug (penalize_nl=false doesn't work) + suppress warning on mingw (#1528)
2ba83c8685177faea3399db9564f9c52df75c366	bae5c5f679e043371bc2b4dffff8d4964d6cb953	2023-08-26T13:45:53+02:00	klosax		Fix spm whitespaces (#2806)
c82742ac9cd96fd34aa961978805c1d8a361d589	28b2c996ca0ab90a5669946084f13443ec98e241	2023-08-25T11:18:48-04:00	Matt Pulver		llama : add llama_beam_search() (#2267)
29674ab4e847fcaba60cc6558f0d46d5f74ae279	5439a0ab57c16b556ffa91a0953df5e46b1e7fb4	2023-08-25T18:32:45+08:00	Jhen-Jie Hong		server : display token probabilities in the UI (#2489)
2e5f70a25fc4576e9ed78603fe493eb7702c37a3	d0f77b1353fc820d1ff1e6b87bc6bedde315938d	2023-08-24T14:49:30-07:00	Marcus Dunn		Added `enum` to `llama_token_get_type` return type (#2774)
cf658adc832badaaa2ca119fe86070e5a830f8f6	a192860cfec89a38d59a943623bf595b1fe4495b	2023-08-23T23:08:04+03:00	Georgi Gerganov		llm : add Falcon support (#2717)
95385241a91a616788a3bb76d12c9b7b2379ca2d	335acd2ffd7b04501c6d8773ab9fcee6e7bf8639	2023-08-23T20:33:05+01:00	Olivier Chafik		examples : restore the functionality to import llama2.c models (#2685)
5290c38e6e9b66ee2b543e560e301c1a1a90929c	cc34dbda9681418a2b18382446b90cdcec398d82	2023-08-23T16:46:03+02:00	klosax		main : insert bos if no tokens (#2727)
b8ad1b66b23f9b2e6e4531e9a62753323036a556	f5fe98d11bdf9e7797bcfb05c0c3601ffc4b9d26	2023-08-23T02:12:12-05:00	Xiao-Yong Jin		server : allow json array in prompt or content for direct token input (#2306)
777f42ba18b29f25c71ff8de3ecf97b8017304c0	46ef5b5fcf4c366e1fb27726b6394adbbf8fd0ea	2023-08-22T17:39:39-06:00	Kerfuffle		Improve handling of special tokens in GGML to GGUF converter (#2725)
46ef5b5fcf4c366e1fb27726b6394adbbf8fd0ea	c63bb1d16a70c03440671b76954bb767513cead8	2023-08-22T23:10:42+02:00	goerch		llama : fix whitespace escaping in tokenizer (#2724)
6381d4e110bd0ec02843a60bbeb8b6fc37a9ace9	dadbed99e65252d79f81101a392d0d6497b86caa	2023-08-21T23:07:43+03:00	Georgi Gerganov		gguf : new file format with flexible meta data (beta) (#2398)
cb1c0727bd59803b439b6a3af121c99e6393ff3d	9e232f0234073358e7031c1b8d7aa45020469a3b	2023-08-21T11:11:31+03:00	Kawrakow		HellaSwag: split token evaluation into batches if needed (#2681)
220d9318647a8ce127dbf7c9de5400455f41e7d8	81844fbcfd93a162b7aeaea9e4f2ab1358f7f97e	2023-08-02T16:21:11+08:00	ldwang		readme : add Aquila-7B model series to supported models (#2487)
ee1b497c985f61d6ec519c39fcfed78a3c6f1d06	d73b8d48b45d6e2c0ae9bb8c39623c4024adc275	2023-07-29T02:10:05+08:00	eric8607242		llama : support more diverse tokenizers? (#2420)
fce48caf9a6b9930eee9e2a5971428cdff403ba8	875086bdb95ce1f3294439811536533199e3b579	2023-07-25T21:22:09+08:00	ldwang		convert.py : support bpe tokenizer (#2228)
84e09a7d8bc4ab6d658b5cd81295ac0add60be78	2f9cf974a066ac0e03fbb235d834b01b0164d743	2023-07-23T23:58:10-04:00	Evan Jones		llama : add grammar-based sampling (#1773)
355c80f49e32b0b15c0a457f3bad380e57f5b9ac	83a00ce69bef9124c0702424a012ea799128b77d	2023-07-23T06:16:48-05:00	AustinMroz		examples : simplify vim plugin (#2327)
1d1630996920f889cdc08de26cebf2415958540e	db4047ad5cd8eae04db3b2efe0245e69a376601a	2023-07-09T05:59:53-03:00	oobabooga		llama : remove "first token must be BOS" restriction (#2153)
f257fd255044decffad93dee2502875ce66ad80c	7ee76e45afae7f9a7a53e93393accfb5b36684e1	2023-07-05T03:06:12+09:00	jwj7140		Add an API example using server.cpp similar to OAI. (#2009)
cc45a7feb8412e84ff292207621412fffc0d3d51	55dbb915cc2a95048f56e667b09dfad38d840421	2023-07-04T02:43:55+08:00	Howard Su		Fix crash of test-tokenizer-0 under Debug build (#2064)
d7d2e6a0f0c74f7a570dae384dfff371ac744d2a	46088f72318981341a2d646f12f6eee6aec06d65	2023-07-03T05:38:44+08:00	WangHaoranRobin		server: add option to output probabilities for completion (#1962)
ca7c3f4da5d144d4cd1dd44903552e6ba49b8ec8	b97ca431db35ec96a339a721acb1219c1dd78bed	2023-06-19T18:14:09+03:00	Kawrakow		cuda : faster k-quants on older GPUs (#1930)
b24c3049d96557c24782e4d32feaae65f47277af	0ede372a51fd8160688e01b587582666c14e94e5	2023-06-18T17:41:26+02:00	Johannes Gäßler		Added tokens per second to info prints (#1928)
794db3e7b982fee37e3995db9c3a216a57ff65e3	5ddf7ea1fb42bac21026de2f77e0f9c069b92234	2023-06-17T07:53:04-04:00	Randall Fitzgerald		Server Example Refactor and Improvements (#1570)
e32089b2c20b1b87b22912f4a8b93fe01647d5b9	2347e45e7bdb09c9a7d74b2c0bc86c2b65f0c343	2023-06-13T21:04:40+02:00	xaedes		train : improved training-from-scratch example (#1652)
74a6d922f12ccfe16b0c265f43be8978c6f25e98	e4caa8da59c1c97dc23fa336f4d726984a20560f	2023-06-12T22:39:21+03:00	Kawrakow		Metal implementation for all k_quants (#1807)
fa84c4b3e80199a5683438f062009c031a06c4fa	12b063f0ecf280e98028e444fc492ee6222cdcdc	2023-06-11T08:19:17-06:00	Kerfuffle		Fix issue where interactive mode crashes when input exceeds ctx size (#1789)
e9b66ee9829039d4ab54550d6222e42a0b31e52a	4f0154b0bad775ac4651bf73b5c216eb43c45cdc	2023-06-10T11:28:11+03:00	Kawrakow		metal : add Q4_1 implementation (#1785)
72ff5282bf0388c60821f504c4c8cc2b1f491aa6	0bf7cf1b296fc9fca05411b37afdf08a531487d2	2023-06-08T22:28:21+03:00	Kawrakow		metal : add Q2_K implementation (#1762)
0f291e1f65c1d68201e71ce99c89562a36686b6d	8fc8179919a11738910db07a800f2b176f8adf09	2023-06-08T19:46:22+03:00	Kawrakow		metal : Q6_K implementation (#1752)
4161bdc04debb70bf5f275492b4d89fd9330087c	0035858273ebe0694926bf4414d279f3e1cd109d	2023-06-08T10:08:23+03:00	Kawrakow		metal : add Q4_K implementation (#1733)
99009e72f8072fa552eb02efee436be596c71cdd	5220a991a5e92bddad9542267ab445a2c033681c	2023-06-05T22:56:18+03:00	Kawrakow		ggml : add SOTA 2,3,4,5,6 bit k-quantizations (#1684)
2e6cd4b02549e343bef3768e6b946f999c82e823	7e4ea5beff567f53be92f75f9089e6f11fa5dabd	2023-05-22T23:33:24+02:00	0cc4m		OpenCL Token Generation Acceleration (#1459)
affc76edfdefa7b326f526e463cc65ff13fcfb92	ea600071cb005267e9e8f2629c1e406dd5fde083	2023-05-20T14:19:28+02:00	Johannes Gäßler		cuda : loading models directly into VRAM, norm calculation on GPU, broadcasting for ggml_mul (#1483)
7694b52b9a206b93d59139c3c7c9b55da0f5aa59	79e3efb0e97b65b6cc72cd9ee970fa8189ad79a4	2023-05-19T10:24:59-07:00	Jason McCartney		main : make reverse prompt option act as a stop token in non-interactive mode (#1032)
905d87b70aa189623d500a28602d7a3a755a4769	f954edda935a70a14cf0cc45ecc7fe7d60cf3e4b	2023-05-13T15:38:36+02:00	Johannes Gäßler		ggml : GPU-accelerated token generation (#1412)
f954edda935a70a14cf0cc45ecc7fe7d60cf3e4b	f048af0230ad5381bb20b9e3c1467ec6fc7debdf	2023-05-13T14:56:40+02:00	xaedes		ggml : implement backward pass for llama + small training-llama-from-scratch example (#1360)
b9fd7eee57df101d4a3e3eabc9fd6c2cb13c9ca1	b608b55a3ea8e4760c617418538465449175bdb8	2023-05-12T00:23:08+03:00	Georgi Gerganov		ggml : remove bit shuffling (#1405)
b608b55a3ea8e4760c617418538465449175bdb8	cf348a60e0af3905acd1d297cb064b918265b7ac	2023-05-11T10:10:19-05:00	CRD716		prompts : model agnostic DAN (#1304)
f9a6364912fd0463fddfdbc9ef9f79fdc281570d	95078cc554fe03d4512363c7e4dec963f0047c72	2023-05-08T17:41:54+03:00	Georgi Gerganov		llama : require first token to be BOS (#1303)
018f2279f5fe3ef743bd8254b23ea8f0efae7e73	9411288271ab548216902a029f42a0a38ebcedb7	2023-04-22T02:27:06+08:00	源文雨		cmake : link threads publicly to ggml (#1042)
9411288271ab548216902a029f42a0a38ebcedb7	8687c1f2581d059cd5b6a9502f89bd343566062a	2023-04-21T11:18:09-07:00	Alex Klinkhamer		main : evaluate tokens in batches after swapping context (#1014)
1bfc153e2f35ddd9d64b084e8d1a5e6fa57ad1c9	3d59769c3bb7e72c915646ddb1e239b1face19f5	2023-04-21T17:18:26+02:00	Kawrakow		ggml : a faster version for Q4_1 x Q8_0 dot products (#1083)
884e7d7a2bfd7325b107442d6758983f5886ed3d	7cd5c4a3e9106151d48f328bb3c94c298a211f18	2023-04-19T20:10:08+03:00	Georgi Gerganov		ggml : use 8-bit precision for Q4_1 intermediate results (#1047)
723dac55fa2ba7adc6e3fc8609781d1ad0378906	0f07cacb05f49704d35a39aa27cfd4b419eb6f8d	2023-04-14T00:03:03-07:00	comex		py : new conversion script (#545)
e986f94829bae0b9e66b326acbbba179931c84f1	c0bb1d3ce21005ab21d686626ba87261a6e3a660	2023-04-02T12:23:04+02:00	Christian Falch		Added api for getting/setting the kv_cache (#685)
cbef542879962fdc491656cd0c8cadd65a5f1356	9733104be5389ebb1ff05095eca2a70280cd875a	2023-03-29T21:31:24+02:00	Pavol Rusnak		py : cleanup the code
2e17dfd80a473099dacc0f41c9146d233c6a5972	20a1a4e09c522a80e2a0db51643d25fa38326065	2023-03-23T15:22:47-05:00	rabidcopy		Replace EOS with newline to prevent context/memory being flushed by EOS in interactive mode (#333)
a50e39c6fe36be3de0941b3c05aaf9c37912fd47	a140219e81cfb80356438112cd2290d701b282bb	2023-03-23T14:15:48Z	Stephan Walter		Revert "Delete SHA256SUMS for now" (#429)
8eea5ae0e5f31238a97c79ea9103c27647380e37	93208cfb929c2323e5d2ac6bf354e278040e70ed	2023-03-23T12:26:19+02:00	anzz1		Delete SHA256SUMS for now (#416)
305ba6f0e6daa3796aad9dd18053a1945dd4cc58	4122dffff958cd137175b58f1f27c0913528d7ba	2023-03-22T18:16:35+01:00	tjohnman		Don't force immediate interactive without `-i` (#354)
56e659a0b271436e24813a801640d015e7b05328	40ea807a972ec7b5a426f034ebfa593b5e7a06ed	2023-03-22T17:09:38+01:00	Erik Scholz		fix perplexity after c-api refactor (#390)
f5a77a629bd0f37ae1696747633ab42a5530ec15	da0e9fe90ccf6e73597eb19dd0cfc0a28363fb3b	2023-03-22T07:32:36+02:00	Georgi Gerganov		Introduce C-style API (#370)
353ec251a42491f5192c48561da4b444ef67f23c	89d5d90f3b6d25f134da7a8e252c3432bffcf674	2023-03-21T14:21:50-03:00	Fabio R. Sluzala		We could use std::unordered_map over std::map (#305)
486ae645fd3eda8b9d7413d5ff34fb65a3e337fb	3ab3e6582f7320c2b6568c892fdfc8215caf7e6c	2023-03-21T09:27:42-07:00	Gary Linscott		Compute perplexity over prompt (#270)
6a612959e1b6c37b68b6b141329751a2902b1030	d5f56a5e5a0069329a81f96460221e7afb1daddc	2023-03-21T17:05:06+01:00	tjohnman		Check for reverse prompt by characters instead of tokens (#292) (#330)
d5f56a5e5a0069329a81f96460221e7afb1daddc	3bfa3b43b7319b71853bfc7d3cf4e9767c24bbc8	2023-03-21T17:04:43+01:00	tjohnman		Check for reverse prompt by characters instead of tokens (#292) (#330)
e0ffc861fae5ac8b40ce973f822d03db02929d36	8f644a0a859938c787d329d27f98e03c58d7df27	2023-03-21T08:34:49-07:00	Kevin Kwok		Update IPFS links to quantized alpaca with new tokenizer format (#352)
eb34620aeceaf9d9df7fcb19acc17ad41b9f60f8	2e664f1ff413995506c9a54f3a8d5b8c64e37a91	2023-03-21T17:29:41+02:00	Georgi Gerganov		Add tokenizer test + revert to C++11 (#355)
6b6d5b5024faaf82019d08cde5e8a9d69c6ca316	a791a68b613b162c88a83f5f0225223bc167c762	2023-03-21T03:33:10+08:00	Qingyou Meng		Fixed tokenizer.model not found error when model dir is symlink (#325)
0f1b21cb90ac6b84a9af70cafb8e13b5389e3b32	074bea2eb1f1349a0118239c4152914aecaa1be4	2023-03-20T18:05:20+01:00	Bernat Vadell		Docker - Fix publish docker image in GitHub Registry (#235)
074bea2eb1f1349a0118239c4152914aecaa1be4	5cb63e2493c49bc2c3b9b355696e8dc26cdd0380	2023-03-20T03:17:23-07:00	Mack Straight		sentencepiece bpe compatible tokenizer (#252)
368d0c8a9ebae16a20e1c8971b21ee888bdefad5	50fae10d0339f2bd639f69dd679c0201d939a265	2023-03-19T19:31:17+01:00	tjohnman		Respect the maximum number of tokens in interactive. (#298)
d7def1a7524f712e5ebb7cd02bab0f13aa56a7f9	6f61c18ec9a30416e21ed5abfb1321bdb14979be	2023-03-18T17:10:47-07:00	Ronsor		Warn user if a context size greater than 2048 tokens is specified (#274)
a81d0c2a171a4446e6a21a3ec74a0c0768d71184	b2de7f18dfbb93463eeb5b4392117bbe82d5bd1b	2023-03-18T04:17:19-07:00	Gary Linscott		Fix n^2 loop in tokenization (#254)
c9f670a17755311aa28c411f5c7f3c8c05434770	4f546091102a418ffdc6230f872ac56e5cedb835	2023-03-17T21:05:58+01:00	thement		Implement non-greedy tokenizer that tries to maximize token lengths (#242)
956dfda8ad8cea7961e22e0384bbc315bf79aed2	113e685d18ac4edb20f647fd34b000941556f6a6	2023-03-15T12:37:50-07:00	Ronsor		Use `tokenizer.vocab_size()` instead of hardcoding 32000 in convert-pth-to-ggml.py (#142)
460c48254098b28d422382a2bbff6a0b3d7f7e17	c80e2a8f2adeda202cbffe76ef800f134e51f03f	2023-03-13T00:35:51+01:00	Matvey Soloviev		Fix token count accounting