Tags

Tags give the ability to mark specific points in history as being important

b1939

57744932 · ci : fix Windows CI by updating Intel SDE version (#5053) · Jan 22, 2024
b1892

cec8a484 · finetune : add training data file to log message (#4979) · Jan 16, 2024
b1891

334a835a · ggml : importance matrix support for legacy quants (#4969) · Jan 16, 2024
b1889

959ef0c0 · perplexity : fix kv cache handling for hellaswag (#4981) · Jan 16, 2024
b1887

158f8c9e · metal : localized logic in `ggml_metal_graph_compute` (#4924) · Jan 16, 2024
b1886

862f5e41 · android : introduce starter project example (#4926) · Jan 16, 2024
b1885

3a48d558 · metal : replace loop of dispatch_async with dispatch_apply (#4934) · Jan 16, 2024
b1884

7c8d3abd · metal : log `recommendedMaxWorkingSetSize` on iOS 16+ (#4936) · Jan 16, 2024
b1882

a0b3ac8c · ggml : introduce GGML_CALL function annotation (#4850) · Jan 16, 2024
b1881

d75c232e · finetune : use LLAMA_FILE_MAGIC_GGLA (#4961) · Jan 16, 2024
b1880

e0324285 · speculative : threading options (#4959) · Jan 16, 2024
b1879

3e5ca793 · pass cpu-architecture arguments only to host code (C;C++) (#4943) · Jan 15, 2024
b1878

44833967 · llama : apply classifier-free guidance to logits directly (#4951) · Jan 15, 2024
b1876

ddb008d8 · cuda : fix dequantize kernel names (#4938) · Jan 15, 2024
b1875

2faaef39 · llama : check for 256 divisibility for IQ2_XS, IQ2_XXS (#4950) · Jan 15, 2024
b1874

4a3156de · CUDA: faster dequantize kernels for Q4_0 and Q4_1 (#4938) · Jan 15, 2024
b1873

a836c8f5 · llama : fix missing quotes (#4937) · Jan 14, 2024
b1872

467a882f · Add ability to use importance matrix for all k-quants (#4930) · Jan 14, 2024
b1871

bb0c1392 · llama : check LLAMA_TRACE env for extra logging (#4929) · Jan 14, 2024
b1869

03c52674 · llama : use LLAMA_LOG_ macros for logging · Jan 14, 2024

Prev
1
…
32
33
34
35
36
37
38
39
40
…
99
Next