llama.cpp: cuBLAS error 15 at ggml-cuda.cu:7956: the requested functionality is not supported
git log | head -1
commit fe680e3d1080a765e5d3150ffd7bab189742898d
#Cuda Env
nvidia-smi --query-gpu=compute_cap --format=csv
compute_cap
5.2
nvidia-smi
Sat Dec 9 13:18:59 2023
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 545.29.06 Driver Version: 545.29.06 CUDA Version: 12.3 |
|-----------------------------------------+----------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+======================+======================|
| 0 NVIDIA GeForce GTX 980 Off | 00000000:01:00.0 Off | N/A |
| 28% 34C P8 13W / 180W | 6MiB / 4096MiB | 0% Default |
| | | N/A |
+-----------------------------------------+----------------------+----------------------+
+---------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=======================================================================================|
| 0 N/A N/A 1656 G /usr/lib/Xorg 3MiB |
+---------------------------------------------------------------------------------------+
#Command
lama.cpp/main -m ~/Downloads/stablelm-zephyr-3b.Q4_K_M.gguf -ngl 33 --color -c 4096 --temp 0.7 --repeat_penalty 1.1 -n -1 -p "<|user|>Tell me a story<|endoftext|>\n<|assistant|>"
#Output
Log start
main: build = 1620 (fe680e3)
main: built with cc (GCC) 13.2.1 20230801 for x86_64-pc-linux-gnu
main: seed = 1702147497
ggml_init_cublas: GGML_CUDA_FORCE_MMQ: no
ggml_init_cublas: CUDA_USE_TENSOR_CORES: yes
ggml_init_cublas: found 1 CUDA devices:
Device 0: NVIDIA GeForce GTX 980, compute capability 5.2
llama_model_loader: loaded meta data with 21 key-value pairs and 356 tensors from /home/evo/Downloads/stablelm-zephyr-3b.Q4_K_M.gguf (version GGUF V3 (latest))
llama_model_loader: - tensor 0: output.weight q6_K [ 2560, 50304, 1, 1 ]
llama_model_loader: - tensor 1: token_embd.weight q4_K [ 2560, 50304, 1, 1 ]
llama_model_loader: - tensor 2: blk.0.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 3: blk.0.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 4: blk.0.ffn_down.weight q6_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 5: blk.0.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 6: blk.0.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 7: blk.0.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 8: blk.0.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 9: blk.0.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 10: blk.0.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 11: blk.0.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 12: blk.0.attn_v.weight q6_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 13: blk.1.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 14: blk.1.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 15: blk.1.ffn_down.weight q6_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 16: blk.1.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 17: blk.1.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 18: blk.1.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 19: blk.1.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 20: blk.1.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 21: blk.1.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 22: blk.1.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 23: blk.1.attn_v.weight q6_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 24: blk.10.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 25: blk.10.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 26: blk.10.ffn_down.weight q6_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 27: blk.10.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 28: blk.10.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 29: blk.10.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 30: blk.10.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 31: blk.10.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 32: blk.10.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 33: blk.10.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 34: blk.10.attn_v.weight q6_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 35: blk.11.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 36: blk.11.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 37: blk.11.ffn_down.weight q6_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 38: blk.11.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 39: blk.11.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 40: blk.11.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 41: blk.11.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 42: blk.11.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 43: blk.11.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 44: blk.11.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 45: blk.11.attn_v.weight q6_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 46: blk.12.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 47: blk.12.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 48: blk.12.ffn_down.weight q4_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 49: blk.12.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 50: blk.12.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 51: blk.12.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 52: blk.12.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 53: blk.12.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 54: blk.12.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 55: blk.12.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 56: blk.12.attn_v.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 57: blk.13.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 58: blk.13.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 59: blk.13.ffn_down.weight q4_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 60: blk.13.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 61: blk.13.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 62: blk.13.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 63: blk.13.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 64: blk.13.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 65: blk.13.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 66: blk.13.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 67: blk.13.attn_v.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 68: blk.14.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 69: blk.14.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 70: blk.14.ffn_down.weight q6_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 71: blk.14.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 72: blk.14.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 73: blk.14.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 74: blk.14.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 75: blk.14.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 76: blk.14.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 77: blk.14.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 78: blk.14.attn_v.weight q6_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 79: blk.15.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 80: blk.15.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 81: blk.15.ffn_down.weight q4_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 82: blk.15.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 83: blk.15.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 84: blk.15.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 85: blk.15.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 86: blk.15.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 87: blk.15.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 88: blk.15.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 89: blk.15.attn_v.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 90: blk.16.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 91: blk.16.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 92: blk.16.ffn_down.weight q4_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 93: blk.16.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 94: blk.16.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 95: blk.16.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 96: blk.16.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 97: blk.16.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 98: blk.16.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 99: blk.16.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 100: blk.16.attn_v.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 101: blk.17.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 102: blk.17.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 103: blk.17.ffn_down.weight q6_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 104: blk.17.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 105: blk.17.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 106: blk.17.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 107: blk.17.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 108: blk.17.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 109: blk.17.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 110: blk.17.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 111: blk.17.attn_v.weight q6_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 112: blk.18.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 113: blk.18.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 114: blk.18.ffn_down.weight q4_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 115: blk.18.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 116: blk.18.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 117: blk.18.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 118: blk.18.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 119: blk.18.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 120: blk.18.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 121: blk.18.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 122: blk.18.attn_v.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 123: blk.19.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 124: blk.19.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 125: blk.19.ffn_down.weight q4_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 126: blk.19.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 127: blk.19.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 128: blk.19.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 129: blk.19.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 130: blk.19.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 131: blk.19.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 132: blk.19.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 133: blk.19.attn_v.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 134: blk.2.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 135: blk.2.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 136: blk.2.ffn_down.weight q6_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 137: blk.2.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 138: blk.2.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 139: blk.2.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 140: blk.2.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 141: blk.2.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 142: blk.2.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 143: blk.2.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 144: blk.2.attn_v.weight q6_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 145: blk.20.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 146: blk.20.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 147: blk.20.ffn_down.weight q4_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 148: blk.20.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 149: blk.20.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 150: blk.20.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 151: blk.20.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 152: blk.20.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 153: blk.20.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 154: blk.20.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 155: blk.20.attn_v.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 156: blk.21.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 157: blk.21.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 158: blk.21.ffn_down.weight q4_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 159: blk.21.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 160: blk.21.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 161: blk.21.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 162: blk.21.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 163: blk.21.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 164: blk.21.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 165: blk.21.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 166: blk.21.attn_v.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 167: blk.22.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 168: blk.22.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 169: blk.22.ffn_down.weight q6_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 170: blk.22.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 171: blk.22.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 172: blk.22.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 173: blk.22.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 174: blk.22.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 175: blk.22.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 176: blk.22.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 177: blk.22.attn_v.weight q6_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 178: blk.23.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 179: blk.23.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 180: blk.23.ffn_down.weight q4_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 181: blk.23.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 182: blk.23.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 183: blk.23.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 184: blk.23.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 185: blk.23.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 186: blk.23.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 187: blk.23.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 188: blk.23.attn_v.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 189: blk.24.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 190: blk.24.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 191: blk.24.ffn_down.weight q4_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 192: blk.24.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 193: blk.24.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 194: blk.24.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 195: blk.24.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 196: blk.24.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 197: blk.24.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 198: blk.24.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 199: blk.24.attn_v.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 200: blk.25.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 201: blk.25.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 202: blk.25.ffn_down.weight q6_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 203: blk.25.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 204: blk.25.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 205: blk.25.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 206: blk.25.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 207: blk.25.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 208: blk.25.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 209: blk.25.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 210: blk.25.attn_v.weight q6_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 211: blk.26.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 212: blk.26.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 213: blk.26.ffn_down.weight q4_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 214: blk.26.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 215: blk.26.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 216: blk.26.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 217: blk.26.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 218: blk.26.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 219: blk.26.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 220: blk.26.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 221: blk.26.attn_v.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 222: blk.27.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 223: blk.27.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 224: blk.27.ffn_down.weight q4_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 225: blk.27.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 226: blk.27.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 227: blk.27.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 228: blk.27.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 229: blk.27.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 230: blk.27.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 231: blk.27.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 232: blk.27.attn_v.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 233: blk.28.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 234: blk.28.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 235: blk.28.ffn_down.weight q6_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 236: blk.28.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 237: blk.28.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 238: blk.28.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 239: blk.28.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 240: blk.28.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 241: blk.28.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 242: blk.28.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 243: blk.28.attn_v.weight q6_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 244: blk.29.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 245: blk.29.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 246: blk.29.ffn_down.weight q4_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 247: blk.29.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 248: blk.29.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 249: blk.29.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 250: blk.29.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 251: blk.29.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 252: blk.29.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 253: blk.29.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 254: blk.29.attn_v.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 255: blk.3.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 256: blk.3.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 257: blk.3.ffn_down.weight q4_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 258: blk.3.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 259: blk.3.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 260: blk.3.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 261: blk.3.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 262: blk.3.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 263: blk.3.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 264: blk.3.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 265: blk.3.attn_v.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 266: blk.30.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 267: blk.30.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 268: blk.30.ffn_down.weight q6_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 269: blk.30.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 270: blk.30.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 271: blk.30.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 272: blk.30.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 273: blk.30.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 274: blk.30.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 275: blk.30.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 276: blk.30.attn_v.weight q6_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 277: blk.31.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 278: blk.31.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 279: blk.31.ffn_down.weight q4_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 280: blk.31.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 281: blk.31.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 282: blk.31.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 283: blk.31.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 284: blk.31.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 285: blk.31.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 286: blk.31.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 287: blk.31.attn_v.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 288: blk.4.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 289: blk.4.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 290: blk.4.ffn_down.weight q4_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 291: blk.4.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 292: blk.4.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 293: blk.4.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 294: blk.4.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 295: blk.4.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 296: blk.4.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 297: blk.4.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 298: blk.4.attn_v.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 299: blk.5.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 300: blk.5.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 301: blk.5.ffn_down.weight q6_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 302: blk.5.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 303: blk.5.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 304: blk.5.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 305: blk.5.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 306: blk.5.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 307: blk.5.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 308: blk.5.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 309: blk.5.attn_v.weight q6_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 310: blk.6.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 311: blk.6.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 312: blk.6.ffn_down.weight q6_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 313: blk.6.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 314: blk.6.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 315: blk.6.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 316: blk.6.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 317: blk.6.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 318: blk.6.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 319: blk.6.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 320: blk.6.attn_v.weight q6_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 321: blk.7.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 322: blk.7.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 323: blk.7.ffn_down.weight q6_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 324: blk.7.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 325: blk.7.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 326: blk.7.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 327: blk.7.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 328: blk.7.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 329: blk.7.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 330: blk.7.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 331: blk.7.attn_v.weight q6_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 332: blk.8.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 333: blk.8.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 334: blk.8.ffn_down.weight q6_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 335: blk.8.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 336: blk.8.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 337: blk.8.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 338: blk.8.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 339: blk.8.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 340: blk.8.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 341: blk.8.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 342: blk.8.attn_v.weight q6_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 343: blk.9.attn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 344: blk.9.attn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 345: blk.9.ffn_down.weight q6_K [ 6912, 2560, 1, 1 ]
llama_model_loader: - tensor 346: blk.9.ffn_gate.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 347: blk.9.ffn_up.weight q4_K [ 2560, 6912, 1, 1 ]
llama_model_loader: - tensor 348: blk.9.ffn_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 349: blk.9.ffn_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 350: blk.9.attn_k.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 351: blk.9.attn_output.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 352: blk.9.attn_q.weight q4_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 353: blk.9.attn_v.weight q6_K [ 2560, 2560, 1, 1 ]
llama_model_loader: - tensor 354: output_norm.bias f32 [ 2560, 1, 1, 1 ]
llama_model_loader: - tensor 355: output_norm.weight f32 [ 2560, 1, 1, 1 ]
llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
llama_model_loader: - kv 0: general.architecture str = stablelm
llama_model_loader: - kv 1: general.name str = source
llama_model_loader: - kv 2: stablelm.context_length u32 = 4096
llama_model_loader: - kv 3: stablelm.embedding_length u32 = 2560
llama_model_loader: - kv 4: stablelm.block_count u32 = 32
llama_model_loader: - kv 5: stablelm.feed_forward_length u32 = 6912
llama_model_loader: - kv 6: stablelm.rope.dimension_count u32 = 20
llama_model_loader: - kv 7: stablelm.attention.head_count u32 = 32
llama_model_loader: - kv 8: stablelm.use_parallel_residual bool = true
llama_model_loader: - kv 9: stablelm.attention.layer_norm_epsilon f32 = 0.000010
llama_model_loader: - kv 10: tokenizer.ggml.model str = gpt2
llama_model_loader: - kv 11: tokenizer.ggml.tokens arr[str,50304] = ["<|endoftext|>", "<|padding|>", "!",...
llama_model_loader: - kv 12: tokenizer.ggml.token_type arr[i32,50304] = [3, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
llama_model_loader: - kv 13: tokenizer.ggml.merges arr[str,50009] = ["Ġ Ġ", "Ġ t", "Ġ a", "h e", "i n...
llama_model_loader: - kv 14: tokenizer.ggml.bos_token_id u32 = 0
llama_model_loader: - kv 15: tokenizer.ggml.eos_token_id u32 = 0
llama_model_loader: - kv 16: tokenizer.ggml.unknown_token_id u32 = 0
llama_model_loader: - kv 17: tokenizer.ggml.padding_token_id u32 = 0
llama_model_loader: - kv 18: tokenizer.chat_template str = {% for message in messages %}\n{% if m...
llama_model_loader: - kv 19: general.quantization_version u32 = 2
llama_model_loader: - kv 20: general.file_type u32 = 15
llama_model_loader: - type f32: 130 tensors
llama_model_loader: - type q4_K: 193 tensors
llama_model_loader: - type q6_K: 33 tensors
llm_load_vocab: mismatch in special tokens definition ( 31/50304 vs 52/50304 ).
llm_load_print_meta: format = GGUF V3 (latest)
llm_load_print_meta: arch = stablelm
llm_load_print_meta: vocab type = BPE
llm_load_print_meta: n_vocab = 50304
llm_load_print_meta: n_merges = 50009
llm_load_print_meta: n_ctx_train = 4096
llm_load_print_meta: n_embd = 2560
llm_load_print_meta: n_head = 32
llm_load_print_meta: n_head_kv = 32
llm_load_print_meta: n_layer = 32
llm_load_print_meta: n_rot = 20
llm_load_print_meta: n_gqa = 1
llm_load_print_meta: f_norm_eps = 1.0e-05
llm_load_print_meta: f_norm_rms_eps = 0.0e+00
llm_load_print_meta: f_clamp_kqv = 0.0e+00
llm_load_print_meta: f_max_alibi_bias = 0.0e+00
llm_load_print_meta: n_ff = 6912
llm_load_print_meta: rope scaling = linear
llm_load_print_meta: freq_base_train = 10000.0
llm_load_print_meta: freq_scale_train = 1
llm_load_print_meta: n_yarn_orig_ctx = 4096
llm_load_print_meta: rope_finetuned = unknown
llm_load_print_meta: model type = 3B
llm_load_print_meta: model ftype = mostly Q4_K - Medium
llm_load_print_meta: model params = 2.80 B
llm_load_print_meta: model size = 1.59 GiB (4.88 BPW)
llm_load_print_meta: general.name = source
llm_load_print_meta: BOS token = 0 '<|endoftext|>'
llm_load_print_meta: EOS token = 0 '<|endoftext|>'
llm_load_print_meta: UNK token = 0 '<|endoftext|>'
llm_load_print_meta: PAD token = 0 '<|endoftext|>'
llm_load_print_meta: LF token = 128 'Ä'
llm_load_tensors: ggml ctx size = 0.13 MiB
llm_load_tensors: using CUDA for GPU acceleration
llm_load_tensors: mem required = 69.85 MiB
llm_load_tensors: offloading 32 repeating layers to GPU
llm_load_tensors: offloading non-repeating layers to GPU
llm_load_tensors: offloaded 33/33 layers to GPU
llm_load_tensors: VRAM used: 1558.02 MiB
............................................................................................
llama_new_context_with_model: n_ctx = 4096
llama_new_context_with_model: freq_base = 10000.0
llama_new_context_with_model: freq_scale = 1
llama_kv_cache_init: VRAM kv self = 1280.00 MB
llama_new_context_with_model: KV self size = 1280.00 MiB, K (f16): 640.00 MiB, V (f16): 640.00 MiB
llama_build_graph: non-view tensors processed: 741/741
llama_new_context_with_model: compute buffer total size = 282.07 MiB
llama_new_context_with_model: VRAM scratch buffer: 279.00 MiB
llama_new_context_with_model: total VRAM used: 3117.66 MiB (model: 1558.65 MiB, context: 1559.00 MiB)
system_info: n_threads = 4 / 8 | AVX = 1 | AVX2 = 1 | AVX512 = 0 | AVX512_VBMI = 0 | AVX512_VNNI = 0 | FMA = 1 | NEON = 0 | ARM_FMA = 0 | F16C = 1 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 1 | SSE3 = 1 | SSSE3 = 1 | VSX = 0 |
sampling:
repeat_last_n = 64, repeat_penalty = 1.100, frequency_penalty = 0.000, presence_penalty = 0.000
top_k = 40, tfs_z = 1.000, top_p = 0.950, min_p = 0.050, typical_p = 1.000, temp = 0.700
mirostat = 0, mirostat_lr = 0.100, mirostat_ent = 5.000
sampling order:
CFG -> Penalties -> top_k -> tfs_z -> typical_p -> top_p -> min_p -> temp
generate: n_ctx = 4096, n_batch = 512, n_predict = -1, n_keep = 0
<|user|>Tell me a story\n<|assistant|>
cuBLAS error 15 at ggml-cuda.cu:7956: the requested functionality is not supported
current device: 0
GGML_ASSERT: ggml-cuda.cu:7956: !"cuBLAS error"
[New LWP 31945]
[New LWP 31947]
[New LWP 31948]
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/usr/lib/libthread_db.so.1".
0x00007fb9f61181c7 in wait4 () from /usr/lib/libc.so.6
#0 0x00007fb9f61181c7 in wait4 () from /usr/lib/libc.so.6
#1 0x00005564738090cb in ggml_print_backtrace ()
#2 0x00005564738d2950 in ggml_cuda_mul_mat_mat_batched_cublas(ggml_tensor const*, ggml_tensor const*, ggml_tensor*) ()
#3 0x00005564738da9d0 in ggml_cuda_compute_forward ()
#4 0x000055647383644d in ggml_graph_compute_thread ()
#5 0x000055647383aa08 in ggml_graph_compute ()
#6 0x0000556473860079 in llama_decode_internal(llama_context&, llama_batch) ()
#7 0x0000556473860e06 in llama_decode ()
#8 0x00005564737fcfb6 in main ()
[Inferior 1 (process 31944) detached]
About this issue
- Original URL
- State: closed
- Created 7 months ago
- Comments: 20 (6 by maintainers)
I’ve been running into this
cuBLAS error 15issue for the past couple of weeks which I haven’t had time to identify.Something must have changed since I first encountered this. At that time, using
LLAMA_CUDA_FORCE_MMQdid not fix the issue for me, at least not without a major performance hit.Now, though, using
LLAMA_CUDA_FORCE_MMQdoes fix it.I tested this with
TheBloke/zephyr-7B-beta-GGUF’szephyr-7b-beta.Q5_K_M.ggufmodel and the CMake configuration: