Commit
•
280a0d2
1
Parent(s):
45fa4ff
Update config.json
Browse files- config.json +20 -2
config.json
CHANGED
|
@@ -23,10 +23,28 @@
|
|
| 23 |
"num_key_value_heads": 8,
|
| 24 |
"pad_token_id": 128004,
|
| 25 |
"pretraining_tp": 1,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 26 |
"rms_norm_eps": 1e-05,
|
| 27 |
"rope_scaling": {
|
| 28 |
-
"
|
| 29 |
-
"
|
|
|
|
|
|
|
|
|
|
| 30 |
},
|
| 31 |
"rope_theta": 500000.0,
|
| 32 |
"tie_word_embeddings": false,
|
|
|
|
| 23 |
"num_key_value_heads": 8,
|
| 24 |
"pad_token_id": 128004,
|
| 25 |
"pretraining_tp": 1,
|
| 26 |
+
"quantization_config": {
|
| 27 |
+
"_load_in_4bit": true,
|
| 28 |
+
"_load_in_8bit": false,
|
| 29 |
+
"bnb_4bit_compute_dtype": "bfloat16",
|
| 30 |
+
"bnb_4bit_quant_storage": "uint8",
|
| 31 |
+
"bnb_4bit_quant_type": "nf4",
|
| 32 |
+
"bnb_4bit_use_double_quant": true,
|
| 33 |
+
"llm_int8_enable_fp32_cpu_offload": false,
|
| 34 |
+
"llm_int8_has_fp16_weight": false,
|
| 35 |
+
"llm_int8_skip_modules": null,
|
| 36 |
+
"llm_int8_threshold": 6.0,
|
| 37 |
+
"load_in_4bit": true,
|
| 38 |
+
"load_in_8bit": false,
|
| 39 |
+
"quant_method": "bitsandbytes"
|
| 40 |
+
},
|
| 41 |
"rms_norm_eps": 1e-05,
|
| 42 |
"rope_scaling": {
|
| 43 |
+
"factor": 8.0,
|
| 44 |
+
"high_freq_factor": 4.0,
|
| 45 |
+
"low_freq_factor": 1.0,
|
| 46 |
+
"original_max_position_embeddings": 8192,
|
| 47 |
+
"rope_type": "llama3"
|
| 48 |
},
|
| 49 |
"rope_theta": 500000.0,
|
| 50 |
"tie_word_embeddings": false,
|