Add config.json
Browse files- config.json +90 -0
config.json
ADDED
|
@@ -0,0 +1,90 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"version": "0.1.0",
|
| 3 |
+
"model_type": "llama",
|
| 4 |
+
"quantization": "w4a16g128",
|
| 5 |
+
"model_config": {
|
| 6 |
+
"hidden_size": 3072,
|
| 7 |
+
"intermediate_size": 8192,
|
| 8 |
+
"num_attention_heads": 24,
|
| 9 |
+
"num_hidden_layers": 28,
|
| 10 |
+
"rms_norm_eps": 1e-05,
|
| 11 |
+
"vocab_size": 128256,
|
| 12 |
+
"tie_word_embeddings": true,
|
| 13 |
+
"position_embedding_base": 500000.0,
|
| 14 |
+
"rope_scaling": {
|
| 15 |
+
"factor": 32.0,
|
| 16 |
+
"high_freq_factor": 4.0,
|
| 17 |
+
"low_freq_factor": 1.0,
|
| 18 |
+
"original_max_position_embeddings": 8192,
|
| 19 |
+
"rope_type": "llama3"
|
| 20 |
+
},
|
| 21 |
+
"context_window_size": 8192,
|
| 22 |
+
"prefill_chunk_size": 128,
|
| 23 |
+
"num_key_value_heads": 8,
|
| 24 |
+
"head_dim": 128,
|
| 25 |
+
"tensor_parallel_shards": 1,
|
| 26 |
+
"pipeline_parallel_stages": 1,
|
| 27 |
+
"max_batch_size": 128
|
| 28 |
+
},
|
| 29 |
+
"vocab_size": 128256,
|
| 30 |
+
"context_window_size": 8192,
|
| 31 |
+
"sliding_window_size": -1,
|
| 32 |
+
"prefill_chunk_size": 128,
|
| 33 |
+
"attention_sink_size": -1,
|
| 34 |
+
"tensor_parallel_shards": 1,
|
| 35 |
+
"pipeline_parallel_stages": 1,
|
| 36 |
+
"temperature": 0.6,
|
| 37 |
+
"presence_penalty": 0.0,
|
| 38 |
+
"frequency_penalty": 0.0,
|
| 39 |
+
"repetition_penalty": 1.0,
|
| 40 |
+
"top_p": 0.9,
|
| 41 |
+
"tokenizer_files": [
|
| 42 |
+
"tokenizer.json",
|
| 43 |
+
"tokenizer_config.json"
|
| 44 |
+
],
|
| 45 |
+
"tokenizer_info": {
|
| 46 |
+
"token_postproc_method": "byte_level",
|
| 47 |
+
"prepend_space_in_encode": false,
|
| 48 |
+
"strip_space_in_decode": false
|
| 49 |
+
},
|
| 50 |
+
"conv_template": {
|
| 51 |
+
"name": "llama-3_1",
|
| 52 |
+
"system_template": "<|start_header_id|>system<|end_header_id|>\n\n{system_message}<|eot_id|>",
|
| 53 |
+
"system_message": "You are a helpful, respectful and honest assistant.",
|
| 54 |
+
"system_prefix_token_ids": [
|
| 55 |
+
128000
|
| 56 |
+
],
|
| 57 |
+
"add_role_after_system_message": true,
|
| 58 |
+
"roles": {
|
| 59 |
+
"user": "<|start_header_id|>user",
|
| 60 |
+
"assistant": "<|start_header_id|>assistant",
|
| 61 |
+
"tool": "<|start_header_id|>ipython"
|
| 62 |
+
},
|
| 63 |
+
"role_templates": {
|
| 64 |
+
"user": "{user_message}",
|
| 65 |
+
"assistant": "{assistant_message}",
|
| 66 |
+
"tool": "{tool_message}"
|
| 67 |
+
},
|
| 68 |
+
"messages": [],
|
| 69 |
+
"seps": [
|
| 70 |
+
"<|eot_id|>"
|
| 71 |
+
],
|
| 72 |
+
"role_content_sep": "<|end_header_id|>\n\n",
|
| 73 |
+
"role_empty_sep": "<|end_header_id|>\n\n",
|
| 74 |
+
"stop_str": [],
|
| 75 |
+
"stop_token_ids": [
|
| 76 |
+
128001,
|
| 77 |
+
128008,
|
| 78 |
+
128009
|
| 79 |
+
],
|
| 80 |
+
"function_string": "",
|
| 81 |
+
"use_function_calling": false
|
| 82 |
+
},
|
| 83 |
+
"pad_token_id": 0,
|
| 84 |
+
"bos_token_id": 128000,
|
| 85 |
+
"eos_token_id": [
|
| 86 |
+
128001,
|
| 87 |
+
128008,
|
| 88 |
+
128009
|
| 89 |
+
]
|
| 90 |
+
}
|