github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-24 06:31:50 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #14312] qwen2.5vl:7b uses significantly more RAM than qwen3-vl:8b despite smaller architecture and identical quantization #9313

New Issue
Open
opened 2026-04-12 22:10:36 -05:00 by GiteaMirror · 1 comment
GiteaMirror commented 2026-04-12 22:10:36 -05:00
Owner
Copy Link

Originally created by @brendensoares on GitHub (Feb 18, 2026).
Original GitHub issue: https://github.com/ollama/ollama/issues/14312

What is the issue?

What happened?

When running both models locally with Ollama using the same quantization (Q4_K_M) and comparable context sizes, qwen2.5vl:7b consistently consumes significantly more total memory than qwen3-vl:8b.

Observed behavior:

  • qwen2.5vl:7b
    • 8,192 context → ~15 GB
    • 16,384 context → ~17 GB
  • qwen3-vl:8b
    • 32,768 context → ~11 GB

This occurs even though:

  • qwen2.5vl:7b has ~8.3B parameters
  • qwen3-vl:8b has ~8.8B parameters
  • Both use Q4_K_M quantization

Additionally, doubling context for qwen2.5vl:7b (8k → 16k) increases memory by ~2 GB, which is significantly larger than what would be expected from KV cache growth alone based on published architecture parameters.

From the model configs:

Qwen2.5-VL-7B (text tower)

  • 28 layers
  • 4 KV heads
  • head dim = 128

Expected KV cache size at fp16:

  • ~0.44 GB @ 8k
  • ~0.88 GB @ 16k

However, the observed memory increase (~2 GB when doubling context) suggests a ~4–5× multiplier over the theoretical KV size.

This indicates that:

  • KV cache may be stored at higher precision (e.g., fp32),
  • KV may be duplicated or staged,
  • or there is unexpected additional runtime allocation for this model.

What did you expect to happen?

Given:

  • qwen3-vl:8b has more layers (36 vs 28),
  • more KV heads (8 vs 4),
  • larger hidden size (4096 vs 3584),

The theoretical KV cache size for qwen3-vl:8b should be significantly larger than for qwen2.5vl:7b at the same context length.

Therefore, total runtime memory usage for qwen3-vl:8b should be equal to or greater than qwen2.5vl:7b, not substantially smaller.

The observed behavior contradicts expectations based on architecture-derived KV cache size and parameter count.

Related issues that appear consistent with unexpected memory estimation or allocation behavior in Qwen 2.5 models:

Relevant log output

## Relevant log output

ollama ps output:


qwen2.5vl:7b    5ced39dfa4ba    15 GB    100% GPU     8192       4 minutes from now
qwen2.5vl:7b    5ced39dfa4ba    17 GB    12%/88% CPU/GPU    16384      4 minutes from now
qwen3-vl:8b     901cae732162    11 GB    100% GPU     32768      4 minutes from now


Model metadata:


ollama show qwen3-vl:8b
  parameters          8.8B
  embedding length    4096
  quantization        Q4_K_M

ollama show qwen2.5vl:7b
  parameters          8.3B
  embedding length    3584
  quantization        Q4_K_M

OS

Windows

GPU

Nvidia

CPU

Intel

Ollama version

0.16.2

Originally created by @brendensoares on GitHub (Feb 18, 2026). Original GitHub issue: https://github.com/ollama/ollama/issues/14312 ### What is the issue? ## What happened? When running both models locally with Ollama using the same quantization (`Q4_K_M`) and comparable context sizes, `qwen2.5vl:7b` consistently consumes significantly more total memory than `qwen3-vl:8b`. Observed behavior: - `qwen2.5vl:7b` - 8,192 context → ~15 GB - 16,384 context → ~17 GB - `qwen3-vl:8b` - 32,768 context → ~11 GB This occurs even though: - `qwen2.5vl:7b` has ~8.3B parameters - `qwen3-vl:8b` has ~8.8B parameters - Both use `Q4_K_M` quantization Additionally, doubling context for `qwen2.5vl:7b` (8k → 16k) increases memory by ~2 GB, which is significantly larger than what would be expected from KV cache growth alone based on published architecture parameters. From the model configs: Qwen2.5-VL-7B (text tower) - 28 layers - 4 KV heads - head dim = 128 Expected KV cache size at fp16: - ~0.44 GB @ 8k - ~0.88 GB @ 16k However, the observed memory increase (~2 GB when doubling context) suggests a ~4–5× multiplier over the theoretical KV size. This indicates that: - KV cache may be stored at higher precision (e.g., fp32), - KV may be duplicated or staged, - or there is unexpected additional runtime allocation for this model. ## What did you expect to happen? Given: - `qwen3-vl:8b` has more layers (36 vs 28), - more KV heads (8 vs 4), - larger hidden size (4096 vs 3584), The theoretical KV cache size for `qwen3-vl:8b` should be significantly larger than for `qwen2.5vl:7b` at the same context length. Therefore, total runtime memory usage for `qwen3-vl:8b` should be equal to or greater than `qwen2.5vl:7b`, not substantially smaller. The observed behavior contradicts expectations based on architecture-derived KV cache size and parameter count. Related issues that appear consistent with unexpected memory estimation or allocation behavior in Qwen 2.5 models: - https://github.com/ollama/ollama/issues/10163 - https://github.com/ollama/ollama/issues/13687 ### Relevant log output ```shell ## Relevant log output ollama ps output: qwen2.5vl:7b 5ced39dfa4ba 15 GB 100% GPU 8192 4 minutes from now qwen2.5vl:7b 5ced39dfa4ba 17 GB 12%/88% CPU/GPU 16384 4 minutes from now qwen3-vl:8b 901cae732162 11 GB 100% GPU 32768 4 minutes from now Model metadata: ollama show qwen3-vl:8b parameters 8.8B embedding length 4096 quantization Q4_K_M ollama show qwen2.5vl:7b parameters 8.3B embedding length 3584 quantization Q4_K_M ``` ### OS Windows ### GPU Nvidia ### CPU Intel ### Ollama version 0.16.2
GiteaMirror added the bug label 2026-04-12 22:10:36 -05:00
GiteaMirror commented 2026-04-12 22:10:38 -05:00
Author
Owner
Copy Link

@somera commented on GitHub (Feb 21, 2026):

Same problems after upgrade to ollama v0.16.3

Feb 21 11:11:08 AI-DEV-VM-Neptun ollama[805]: time=2026-02-21T11:11:08.524+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:12[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:12(16..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
Feb 21 11:11:09 AI-DEV-VM-Neptun ollama[805]: time=2026-02-21T11:11:09.728+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:11[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:11(17..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
Feb 21 11:11:10 AI-DEV-VM-Neptun ollama[805]: time=2026-02-21T11:11:10.938+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:10[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:10(18..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
Feb 21 11:11:12 AI-DEV-VM-Neptun ollama[805]: time=2026-02-21T11:11:12.133+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:9[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:9(19..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
Feb 21 11:11:13 AI-DEV-VM-Neptun ollama[805]: time=2026-02-21T11:11:13.341+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:8[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:8(20..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
Feb 21 11:11:14 AI-DEV-VM-Neptun ollama[805]: time=2026-02-21T11:11:14.541+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:7[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:7(21..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
Feb 21 11:11:15 AI-DEV-VM-Neptun ollama[805]: time=2026-02-21T11:11:15.744+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:6[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:6(22..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
Feb 21 11:11:16 AI-DEV-VM-Neptun ollama[805]: time=2026-02-21T11:11:16.940+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:5[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:5(23..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
Feb 21 11:11:18 AI-DEV-VM-Neptun ollama[805]: time=2026-02-21T11:11:18.151+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:4[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:4(24..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
Feb 21 11:11:19 AI-DEV-VM-Neptun ollama[805]: time=2026-02-21T11:11:19.355+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:3[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:3(25..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
Feb 21 11:11:20 AI-DEV-VM-Neptun ollama[805]: time=2026-02-21T11:11:20.570+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:2[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:2(26..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
Feb 21 11:11:21 AI-DEV-VM-Neptun ollama[805]: time=2026-02-21T11:11:21.769+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:1[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:1(27..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"

...

Feb 21 11:40:41 AI-DEV-VM-Neptun ollama[4269]: ggml_backend_cuda_buffer_type_alloc_buffer: allocating 37058.21 MiB on device 0: cudaMalloc failed: out of memory
Feb 21 11:40:41 AI-DEV-VM-Neptun ollama[4269]: ggml_gallocr_reserve_n_impl: failed to allocate CUDA0 buffer of size 38858347648
Feb 21 11:40:42 AI-DEV-VM-Neptun ollama[4269]: time=2026-02-21T11:40:42.039+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:alloc LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:27[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:27(1..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
Feb 21 11:40:45 AI-DEV-VM-Neptun ollama[4269]: ggml_backend_cuda_buffer_type_alloc_buffer: allocating 37272.58 MiB on device 0: cudaMalloc failed: out of memory
Feb 21 11:40:45 AI-DEV-VM-Neptun ollama[4269]: ggml_gallocr_reserve_n_impl: failed to allocate CUDA0 buffer of size 39083137024
Feb 21 11:40:45 AI-DEV-VM-Neptun ollama[4269]: time=2026-02-21T11:40:45.612+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:alloc LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:26[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:26(2..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
Feb 21 11:40:49 AI-DEV-VM-Neptun ollama[4269]: ggml_backend_cuda_buffer_type_alloc_buffer: allocating 37361.27 MiB on device 0: cudaMalloc failed: out of memory
Feb 21 11:40:49 AI-DEV-VM-Neptun ollama[4269]: ggml_gallocr_reserve_n_impl: failed to allocate CUDA0 buffer of size 39176134656
Feb 21 11:40:49 AI-DEV-VM-Neptun ollama[4269]: time=2026-02-21T11:40:49.365+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:alloc LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:25[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:25(3..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
<!-- gh-comment-id:3938555153 --> @somera commented on GitHub (Feb 21, 2026): Same problems after upgrade to ollama v0.16.3 ``` Feb 21 11:11:08 AI-DEV-VM-Neptun ollama[805]: time=2026-02-21T11:11:08.524+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:12[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:12(16..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}" Feb 21 11:11:09 AI-DEV-VM-Neptun ollama[805]: time=2026-02-21T11:11:09.728+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:11[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:11(17..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}" Feb 21 11:11:10 AI-DEV-VM-Neptun ollama[805]: time=2026-02-21T11:11:10.938+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:10[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:10(18..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}" Feb 21 11:11:12 AI-DEV-VM-Neptun ollama[805]: time=2026-02-21T11:11:12.133+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:9[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:9(19..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}" Feb 21 11:11:13 AI-DEV-VM-Neptun ollama[805]: time=2026-02-21T11:11:13.341+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:8[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:8(20..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}" Feb 21 11:11:14 AI-DEV-VM-Neptun ollama[805]: time=2026-02-21T11:11:14.541+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:7[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:7(21..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}" Feb 21 11:11:15 AI-DEV-VM-Neptun ollama[805]: time=2026-02-21T11:11:15.744+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:6[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:6(22..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}" Feb 21 11:11:16 AI-DEV-VM-Neptun ollama[805]: time=2026-02-21T11:11:16.940+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:5[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:5(23..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}" Feb 21 11:11:18 AI-DEV-VM-Neptun ollama[805]: time=2026-02-21T11:11:18.151+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:4[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:4(24..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}" Feb 21 11:11:19 AI-DEV-VM-Neptun ollama[805]: time=2026-02-21T11:11:19.355+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:3[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:3(25..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}" Feb 21 11:11:20 AI-DEV-VM-Neptun ollama[805]: time=2026-02-21T11:11:20.570+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:2[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:2(26..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}" Feb 21 11:11:21 AI-DEV-VM-Neptun ollama[805]: time=2026-02-21T11:11:21.769+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:1[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:1(27..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}" ... Feb 21 11:40:41 AI-DEV-VM-Neptun ollama[4269]: ggml_backend_cuda_buffer_type_alloc_buffer: allocating 37058.21 MiB on device 0: cudaMalloc failed: out of memory Feb 21 11:40:41 AI-DEV-VM-Neptun ollama[4269]: ggml_gallocr_reserve_n_impl: failed to allocate CUDA0 buffer of size 38858347648 Feb 21 11:40:42 AI-DEV-VM-Neptun ollama[4269]: time=2026-02-21T11:40:42.039+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:alloc LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:27[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:27(1..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}" Feb 21 11:40:45 AI-DEV-VM-Neptun ollama[4269]: ggml_backend_cuda_buffer_type_alloc_buffer: allocating 37272.58 MiB on device 0: cudaMalloc failed: out of memory Feb 21 11:40:45 AI-DEV-VM-Neptun ollama[4269]: ggml_gallocr_reserve_n_impl: failed to allocate CUDA0 buffer of size 39083137024 Feb 21 11:40:45 AI-DEV-VM-Neptun ollama[4269]: time=2026-02-21T11:40:45.612+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:alloc LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:26[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:26(2..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}" Feb 21 11:40:49 AI-DEV-VM-Neptun ollama[4269]: ggml_backend_cuda_buffer_type_alloc_buffer: allocating 37361.27 MiB on device 0: cudaMalloc failed: out of memory Feb 21 11:40:49 AI-DEV-VM-Neptun ollama[4269]: ggml_gallocr_reserve_n_impl: failed to allocate CUDA0 buffer of size 39176134656 Feb 21 11:40:49 AI-DEV-VM-Neptun ollama[4269]: time=2026-02-21T11:40:49.365+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:alloc LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:128000 KvCacheType: NumThreads:12 GPULayers:25[ID:GPU-bf11243f-f6da-11f0-94c9-4cc7a902f6ae Layers:25(3..27)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}" ```
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label bug
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#9313
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?