Remove llama2 7B config for 24 cores
Browse files
inference-cache-config/llama2.json
CHANGED
|
@@ -6,18 +6,6 @@
|
|
| 6 |
"num_cores": 2,
|
| 7 |
"auto_cast_type": "fp16"
|
| 8 |
},
|
| 9 |
-
{
|
| 10 |
-
"batch_size": 1,
|
| 11 |
-
"sequence_length": 4096,
|
| 12 |
-
"num_cores": 8,
|
| 13 |
-
"auto_cast_type": "fp16"
|
| 14 |
-
},
|
| 15 |
-
{
|
| 16 |
-
"batch_size": 1,
|
| 17 |
-
"sequence_length": 4096,
|
| 18 |
-
"num_cores": 24,
|
| 19 |
-
"auto_cast_type": "fp16"
|
| 20 |
-
},
|
| 21 |
{
|
| 22 |
"batch_size": 4,
|
| 23 |
"sequence_length": 4096,
|
|
@@ -30,47 +18,23 @@
|
|
| 30 |
"num_cores": 8,
|
| 31 |
"auto_cast_type": "fp16"
|
| 32 |
},
|
| 33 |
-
{
|
| 34 |
-
"batch_size": 4,
|
| 35 |
-
"sequence_length": 4096,
|
| 36 |
-
"num_cores": 24,
|
| 37 |
-
"auto_cast_type": "fp16"
|
| 38 |
-
},
|
| 39 |
{
|
| 40 |
"batch_size": 8,
|
| 41 |
"sequence_length": 4096,
|
| 42 |
"num_cores": 8,
|
| 43 |
"auto_cast_type": "fp16"
|
| 44 |
},
|
| 45 |
-
{
|
| 46 |
-
"batch_size": 8,
|
| 47 |
-
"sequence_length": 4096,
|
| 48 |
-
"num_cores": 24,
|
| 49 |
-
"auto_cast_type": "fp16"
|
| 50 |
-
},
|
| 51 |
{
|
| 52 |
"batch_size": 16,
|
| 53 |
"sequence_length": 4096,
|
| 54 |
"num_cores": 8,
|
| 55 |
"auto_cast_type": "fp16"
|
| 56 |
},
|
| 57 |
-
{
|
| 58 |
-
"batch_size": 16,
|
| 59 |
-
"sequence_length": 4096,
|
| 60 |
-
"num_cores": 24,
|
| 61 |
-
"auto_cast_type": "fp16"
|
| 62 |
-
},
|
| 63 |
{
|
| 64 |
"batch_size": 32,
|
| 65 |
"sequence_length": 4096,
|
| 66 |
"num_cores": 8,
|
| 67 |
"auto_cast_type": "fp16"
|
| 68 |
-
},
|
| 69 |
-
{
|
| 70 |
-
"batch_size": 32,
|
| 71 |
-
"sequence_length": 4096,
|
| 72 |
-
"num_cores": 24,
|
| 73 |
-
"auto_cast_type": "fp16"
|
| 74 |
}
|
| 75 |
],
|
| 76 |
"meta-llama/Llama-2-13b-chat-hf": [
|
|
|
|
| 6 |
"num_cores": 2,
|
| 7 |
"auto_cast_type": "fp16"
|
| 8 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 9 |
{
|
| 10 |
"batch_size": 4,
|
| 11 |
"sequence_length": 4096,
|
|
|
|
| 18 |
"num_cores": 8,
|
| 19 |
"auto_cast_type": "fp16"
|
| 20 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 21 |
{
|
| 22 |
"batch_size": 8,
|
| 23 |
"sequence_length": 4096,
|
| 24 |
"num_cores": 8,
|
| 25 |
"auto_cast_type": "fp16"
|
| 26 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 27 |
{
|
| 28 |
"batch_size": 16,
|
| 29 |
"sequence_length": 4096,
|
| 30 |
"num_cores": 8,
|
| 31 |
"auto_cast_type": "fp16"
|
| 32 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 33 |
{
|
| 34 |
"batch_size": 32,
|
| 35 |
"sequence_length": 4096,
|
| 36 |
"num_cores": 8,
|
| 37 |
"auto_cast_type": "fp16"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 38 |
}
|
| 39 |
],
|
| 40 |
"meta-llama/Llama-2-13b-chat-hf": [
|