Skip to content

Get up and running with Kimi, GLM, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models.

Commits on main

5c0b663
ollama-debug.c: change 'ld' to 'PRIi64'
shane.xb.qian committed 1 year ago
30d7a59
sample: do all sorting in topK
ParthSareen committed 1 year ago
4aeb67e
sample: simplify top_k=0 sorting
ParthSareen committed 1 year ago
3ba9163
sample: use container/heap for top_k
ParthSareen committed 1 year ago
1b7433b
models/gemma3: remove final logit softcap (#9692)
Bruce MacDonald committed 1 year ago
a70820d
6b45b1d
ollama-debug.c: correct mistype
shane.xb.qian committed 1 year ago
85ab552
Adding Gemma 3 to readme (#9671)
Michael committed 1 year ago
ad4e0bf
Merge pull request #9661 from ollama/gemma
Michael Yang committed 1 year ago
aee2850
all: address linter errors
jmorganca committed 1 year ago
83f0ec8
kvcache: fix tests by adding AvgPool2D stub
jmorganca committed 1 year ago
c6b6938
model: add more spm tokenizer tests
jmorganca committed 1 year ago
fb4664f
20e3593
use 2d pooling
Michael Yang committed 1 year ago
63a3940
ab39e08
f63e62e
Revert "Allow models to force a new batch"
jmorganca committed 1 year ago
65b0f32
Allow models to force a new batch
Jesse Gross committed 1 year ago
06007c0
Disable causal attention based on batch index
Jesse Gross committed 1 year ago
a8e83a7
Restrict Gemma to a single image per request
Jesse Gross committed 1 year ago
4750055
2c40c4d
use non-causal mask only for image positions
Michael Yang committed 1 year ago
e952789
use non-causal mask for inputs with images
Michael Yang committed 1 year ago
9d2a20a
fix gemma3 1b conversion
Patrick Devine committed 1 year ago
2e54d72
compat with upstream gguf
Michael Yang committed 1 year ago
6b32a2d
fallback to cpu
Michael Yang committed 1 year ago
c5cbe4f