{"payload":{"feedbackUrl":"https://github.com/orgs/community/discussions/53140","repo":{"id":746049876,"defaultBranch":"master","name":"llama.cpp","ownerLogin":"compilade","currentUserCanPush":false,"isFork":true,"isEmpty":false,"createdAt":"2024-01-20T22:53:26.000Z","ownerAvatar":"https://avatars.githubusercontent.com/u/113953597?v=4","public":true,"private":false,"isOrgOwned":false},"refInfo":{"name":"","listCacheKey":"v0:1725927522.0","currentOid":""},"activityList":{"items":[{"before":"141dd55e53079a15851e1c8ec87e360d1d07876c","after":"e83d2707d3edae76e2d00408b0e0cfa39468d509","ref":"refs/heads/compilade/convert-separate-extra-tensors","pushedAt":"2024-09-16T20:29:33.000Z","pushType":"push","commitsCount":67,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"convert : adapt MiniCPM3 to separate rope_freqs insertion\n\nMiniCPM3's tokenizer is treated as a SentencePiece tokenizer to avoid\nhaving to run its custom Python code which mixes tokenization\nin the same file as tool calls.\n\ngguf-py : add long and short RoPE factors to tensor mappings\n\nEmpty, but the key names are used to populate the mappings.","shortMessageHtmlLink":"convert : adapt MiniCPM3 to separate rope_freqs insertion"}},{"before":"b5349cb4eb3c13cc552ae81e1b98d6db796e221c","after":"94596be679df0eaaccfd5c1bdca97633184fddb0","ref":"refs/heads/compilade/convert-check-index","pushedAt":"2024-09-10T01:08:11.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"convert : identify missing model files","shortMessageHtmlLink":"convert : identify missing model files"}},{"before":null,"after":"141dd55e53079a15851e1c8ec87e360d1d07876c","ref":"refs/heads/compilade/convert-separate-extra-tensors","pushedAt":"2024-09-10T00:18:42.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"convert : refactor rope_freqs generation\n\nThis should also fix vocab-only conversion for Phi-3.","shortMessageHtmlLink":"convert : refactor rope_freqs generation"}},{"before":"e715417d7c9ddfb4beb3d15455d9c6dfd1637240","after":"b5349cb4eb3c13cc552ae81e1b98d6db796e221c","ref":"refs/heads/compilade/convert-check-index","pushedAt":"2024-09-10T00:16:10.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"convert : identify missing model files","shortMessageHtmlLink":"convert : identify missing model files"}},{"before":null,"after":"e715417d7c9ddfb4beb3d15455d9c6dfd1637240","ref":"refs/heads/compilade/convert-check-index","pushedAt":"2024-09-09T23:51:43.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"convert : identify missing model files","shortMessageHtmlLink":"convert : identify missing model files"}},{"before":"3ad0603c65a33bf403627bc0a188da3a093bb5d0","after":"d19101c9a0e38359a303127bb5ccde47395ee083","ref":"refs/heads/compilade/imatrix-batched-chunks","pushedAt":"2024-09-08T15:04:41.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"imatrix : use FMA and sort tensor names","shortMessageHtmlLink":"imatrix : use FMA and sort tensor names"}},{"before":"c8ab6a3ba356e902b94499baaf7ab0191c3b6afe","after":"3ad0603c65a33bf403627bc0a188da3a093bb5d0","ref":"refs/heads/compilade/imatrix-batched-chunks","pushedAt":"2024-09-08T14:05:28.000Z","pushType":"push","commitsCount":98,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"Merge branch 'master' into compilade/imatrix-batched-chunks","shortMessageHtmlLink":"Merge branch 'master' into compilade/imatrix-batched-chunks"}},{"before":"3de9300c3786d52fb709596a0c5ac1dc65c9f08d","after":"c8ab6a3ba356e902b94499baaf7ab0191c3b6afe","ref":"refs/heads/compilade/imatrix-batched-chunks","pushedAt":"2024-09-08T14:04:27.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"imatrix : fix conversion problems","shortMessageHtmlLink":"imatrix : fix conversion problems"}},{"before":"347247a24ec0db754216b7d466bac021bef9ae6a","after":"3de9300c3786d52fb709596a0c5ac1dc65c9f08d","ref":"refs/heads/compilade/imatrix-batched-chunks","pushedAt":"2024-09-06T21:17:41.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"imatrix : use GGUF to store imatrix data","shortMessageHtmlLink":"imatrix : use GGUF to store imatrix data"}},{"before":"03a138dfe936f96b508b02f97d313d150842dc04","after":"2f5e28f9459b4b081fd38981677229307fe12b80","ref":"refs/heads/compilade/bitnet-ternary-metal","pushedAt":"2024-08-23T03:33:36.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"metal : support TQ2_0\n\nMostly adapted from the IQ2_TN kernels\nfrom https://github.com/ikawrakow/ik_llama.cpp/pull/13\nwhich were themselves adapted from the Q2_K kernels.","shortMessageHtmlLink":"metal : support TQ2_0"}},{"before":"732c48f726ed65205dec1445e0bd35975a9dbc7c","after":"03a138dfe936f96b508b02f97d313d150842dc04","ref":"refs/heads/compilade/bitnet-ternary-metal","pushedAt":"2024-08-22T21:45:49.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"metal : support TQ2_0\n\nMostly adapted from the IQ2_TN kernels\nfrom https://github.com/ikawrakow/ik_llama.cpp/pull/13","shortMessageHtmlLink":"metal : support TQ2_0"}},{"before":"ad466ea886dcb071e1508952c9c2ffa6453081b8","after":"732c48f726ed65205dec1445e0bd35975a9dbc7c","ref":"refs/heads/compilade/bitnet-ternary-metal","pushedAt":"2024-08-22T21:44:08.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"metal : support TQ2_0","shortMessageHtmlLink":"metal : support TQ2_0"}},{"before":"732c48f726ed65205dec1445e0bd35975a9dbc7c","after":"ad466ea886dcb071e1508952c9c2ffa6453081b8","ref":"refs/heads/compilade/bitnet-ternary-metal","pushedAt":"2024-08-22T21:43:07.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"metal : support TQ2_0\n\nCo-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>","shortMessageHtmlLink":"metal : support TQ2_0"}},{"before":"b7ab848bfd536c8113b2abaf768bb0784c2080b4","after":"732c48f726ed65205dec1445e0bd35975a9dbc7c","ref":"refs/heads/compilade/bitnet-ternary-metal","pushedAt":"2024-08-22T21:36:39.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"metal : support TQ2_0","shortMessageHtmlLink":"metal : support TQ2_0"}},{"before":null,"after":"b7ab848bfd536c8113b2abaf768bb0784c2080b4","ref":"refs/heads/compilade/bitnet-ternary-metal","pushedAt":"2024-08-22T21:23:30.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"metal : support TQ2_0","shortMessageHtmlLink":"metal : support TQ2_0"}},{"before":"bce54642c8ac6ff41a55140d4f477bee77048e21","after":"347247a24ec0db754216b7d466bac021bef9ae6a","ref":"refs/heads/compilade/imatrix-batched-chunks","pushedAt":"2024-08-20T19:36:06.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"imatrix : fix segfault when using a single chunk per batch","shortMessageHtmlLink":"imatrix : fix segfault when using a single chunk per batch"}},{"before":null,"after":"bce54642c8ac6ff41a55140d4f477bee77048e21","ref":"refs/heads/compilade/imatrix-batched-chunks","pushedAt":"2024-08-20T19:18:37.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"imatrix : allow processing multiple chunks per batch\n\n* perplexity : simplify filling the batch","shortMessageHtmlLink":"imatrix : allow processing multiple chunks per batch"}},{"before":"88142afd1d0c97d1b0a92bd98dec7f4cac152c0a","after":"a79974ff7ac2ff475370b9f5b7f9e19c37b9b638","ref":"refs/heads/compilade/mamba2","pushedAt":"2024-08-19T02:44:02.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"llama : support running Mamba-Codestral-7B-v0.1","shortMessageHtmlLink":"llama : support running Mamba-Codestral-7B-v0.1"}},{"before":"e584c9c4ab008df591c73cd1d119dab22d403616","after":"88142afd1d0c97d1b0a92bd98dec7f4cac152c0a","ref":"refs/heads/compilade/mamba2","pushedAt":"2024-08-19T01:51:09.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"ggml : SIMD ggml_ssm_scan for Mamba-2\n\n* ggml : improve ggml_mul speed when masking recurrent states","shortMessageHtmlLink":"ggml : SIMD ggml_ssm_scan for Mamba-2"}},{"before":"de5129cf187055c651b2919d3c8b476fcc044704","after":"e584c9c4ab008df591c73cd1d119dab22d403616","ref":"refs/heads/compilade/mamba2","pushedAt":"2024-08-18T22:17:40.000Z","pushType":"push","commitsCount":2,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"wip : Mamba1 works again, and Mamba2 works!!!","shortMessageHtmlLink":"wip : Mamba1 works again, and Mamba2 works!!!"}},{"before":null,"after":"de5129cf187055c651b2919d3c8b476fcc044704","ref":"refs/heads/compilade/mamba2","pushedAt":"2024-08-17T20:32:39.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"wip : TODO fix ssm scan dimensions","shortMessageHtmlLink":"wip : TODO fix ssm scan dimensions"}},{"before":"b216b3dba25b4974fac6cbd1e52631e223801fb7","after":"8956543c091c6851089ed7467fa44ac2b2b0ee37","ref":"refs/heads/xsn/fix_lora","pushedAt":"2024-07-15T06:51:32.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"convert_hf : simplify modify_tensors for InternLM2\n\n* convert_lora : lazy conversion\n\n* llama : load and use alpha from LoRA adapters","shortMessageHtmlLink":"convert_hf : simplify modify_tensors for InternLM2"}},{"before":"8db62494f43d86908c448001c66ae375ba3d009d","after":"b216b3dba25b4974fac6cbd1e52631e223801fb7","ref":"refs/heads/xsn/fix_lora","pushedAt":"2024-07-10T05:34:09.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"convert_lora : MoE LoRA conversion support\n\n* convert_lora : prefer safetensors, similarly to convert_hf","shortMessageHtmlLink":"convert_lora : MoE LoRA conversion support"}},{"before":"46cc8a40d686492c143e9c7bbb8d73c2b6c1d4e8","after":"8db62494f43d86908c448001c66ae375ba3d009d","ref":"refs/heads/xsn/fix_lora","pushedAt":"2024-07-10T04:36:18.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"convert_hf : MoE LoRA conversion support","shortMessageHtmlLink":"convert_hf : MoE LoRA conversion support"}},{"before":"91b9946fac3ba659f6735334b81c889702a3d530","after":"46cc8a40d686492c143e9c7bbb8d73c2b6c1d4e8","ref":"refs/heads/xsn/fix_lora","pushedAt":"2024-07-10T04:33:51.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"convert_hf : MoE LoRA conversion support","shortMessageHtmlLink":"convert_hf : MoE LoRA conversion support"}},{"before":"3026c32956f6686fc3e80695d58fd5c77aa98d3a","after":"91b9946fac3ba659f6735334b81c889702a3d530","ref":"refs/heads/xsn/fix_lora","pushedAt":"2024-07-10T04:12:18.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"convert_hf : MoE LoRA conversion support","shortMessageHtmlLink":"convert_hf : MoE LoRA conversion support"}},{"before":null,"after":"3026c32956f6686fc3e80695d58fd5c77aa98d3a","ref":"refs/heads/xsn/fix_lora","pushedAt":"2024-07-10T03:50:24.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"convert_hf : MoE LoRA conversion support","shortMessageHtmlLink":"convert_hf : MoE LoRA conversion support"}},{"before":"01a22dd496e88b4d96a6837e46a004e61ae1fe54","after":"e19cb3aeb728987aa0a58a119d800c06fdd6aad7","ref":"refs/heads/smaller-output-buffer","pushedAt":"2024-03-17T21:08:25.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"llama : fix wrong n_outputs in llama_set_inputs\n\nA mismatch happened when using a smaller n_ubatch than n_batch and then using\nllama_batch_get_one(). The decision of what n_outputs should be now almost\nfully depends on how lctx.n_outputs is set in llama_decode_internal.\nThe conditions are simpler this way.\n\n* llama : when saving the state, recalculate n_outputs\n\nThis ensures the correct number of outputs for the entire previous batch\nis stored in the session file, even when n_ubatch is smaller than n_batch.","shortMessageHtmlLink":"llama : fix wrong n_outputs in llama_set_inputs"}},{"before":"408fcb0f91dba87386d9bf02a12eb0cf71c08559","after":"01a22dd496e88b4d96a6837e46a004e61ae1fe54","ref":"refs/heads/smaller-output-buffer","pushedAt":"2024-03-17T20:41:10.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"llama : fix wrong n_outputs in llama_set_inputs\n\nA mismatch happened when using a smaller n_ubatch than n_batch and then using\nllama_batch_get_one(). The decision of what n_outputs should be now almost\nfully depends on how lctx.n_outputs is set in llama_decode_internal.\nThe conditions are simpler this way.\n\n* llama : when saving the state, recalculate n_outputs\n\nThis ensures the correct number of outputs for the entire previous batch\nis stored in the session file, even when n_ubatch is smaller than n_batch.","shortMessageHtmlLink":"llama : fix wrong n_outputs in llama_set_inputs"}},{"before":"dd6b7fb9ae2b5708d7e34338e5e4cc2bfeaa1023","after":"408fcb0f91dba87386d9bf02a12eb0cf71c08559","ref":"refs/heads/smaller-output-buffer","pushedAt":"2024-03-17T19:37:08.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"compilade","name":null,"path":"/compilade","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/113953597?s=80&v=4"},"commit":{"message":"llama : fix llama_get_embeddings_ith when the resulting id is 0","shortMessageHtmlLink":"llama : fix llama_get_embeddings_ith when the resulting id is 0"}}],"hasNextPage":true,"hasPreviousPage":false,"activityType":"all","actor":null,"timePeriod":"all","sort":"DESC","perPage":30,"startCursor":"Y3Vyc29yOnYyOpK7MjAyNC0wOS0xNlQyMDoyOTozMy4wMDAwMDBazwAAAAS32rVx","endCursor":"Y3Vyc29yOnYyOpK7MjAyNC0wMy0xN1QxOTozNzowOC4wMDAwMDBazwAAAAQX8q_D"}},"title":"Activity · compilade/llama.cpp"}