Create generation_config.json
Browse files- generation_config.json +91 -0
generation_config.json
ADDED
|
@@ -0,0 +1,91 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 128000,
|
| 4 |
+
"eos_token_id": 128009,
|
| 5 |
+
"pad_token_id": 128001,
|
| 6 |
+
"transformers_version": "4.40.0",
|
| 7 |
+
"model_type": "llama",
|
| 8 |
+
"architectures": [
|
| 9 |
+
"LlamaForCausalLM"
|
| 10 |
+
],
|
| 11 |
+
"do_sample": true,
|
| 12 |
+
"temperature": 0.7,
|
| 13 |
+
"top_p": 0.9,
|
| 14 |
+
"top_k": 50,
|
| 15 |
+
"repetition_penalty": 1.1,
|
| 16 |
+
"max_length": 131072,
|
| 17 |
+
"max_new_tokens": 4096,
|
| 18 |
+
"min_length": 0,
|
| 19 |
+
"min_new_tokens": null,
|
| 20 |
+
"early_stopping": false,
|
| 21 |
+
"num_beams": 1,
|
| 22 |
+
"num_beam_groups": 1,
|
| 23 |
+
"diversity_penalty": 0.0,
|
| 24 |
+
"length_penalty": 1.0,
|
| 25 |
+
"no_repeat_ngram_size": 0,
|
| 26 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 27 |
+
"bad_words_ids": null,
|
| 28 |
+
"forced_bos_token_id": null,
|
| 29 |
+
"forced_eos_token_id": null,
|
| 30 |
+
"remove_invalid_values": false,
|
| 31 |
+
"exponential_decay_length_penalty": null,
|
| 32 |
+
"suppress_tokens": null,
|
| 33 |
+
"begin_suppress_tokens": null,
|
| 34 |
+
"forced_decoder_ids": null,
|
| 35 |
+
"num_return_sequences": 1,
|
| 36 |
+
"output_attentions": false,
|
| 37 |
+
"output_hidden_states": false,
|
| 38 |
+
"output_scores": false,
|
| 39 |
+
"return_dict_in_generate": true,
|
| 40 |
+
"use_cache": true,
|
| 41 |
+
"typical_p": 1.0,
|
| 42 |
+
"epsilon_cutoff": 0.0,
|
| 43 |
+
"eta_cutoff": 0.0,
|
| 44 |
+
"renormalize_logits": false,
|
| 45 |
+
"constraints": null,
|
| 46 |
+
"guidance_scale": null,
|
| 47 |
+
"low_memory": null,
|
| 48 |
+
"watermarking_config": null,
|
| 49 |
+
"presets": {
|
| 50 |
+
"creative": {
|
| 51 |
+
"temperature": 0.9,
|
| 52 |
+
"top_p": 0.95,
|
| 53 |
+
"top_k": 50,
|
| 54 |
+
"repetition_penalty": 1.1,
|
| 55 |
+
"description": "High creativity for stories, brainstorming, creative writing"
|
| 56 |
+
},
|
| 57 |
+
"precise": {
|
| 58 |
+
"temperature": 0.3,
|
| 59 |
+
"top_p": 0.85,
|
| 60 |
+
"top_k": 40,
|
| 61 |
+
"repetition_penalty": 1.15,
|
| 62 |
+
"description": "Low randomness for factual, technical, or code generation"
|
| 63 |
+
},
|
| 64 |
+
"balanced": {
|
| 65 |
+
"temperature": 0.7,
|
| 66 |
+
"top_p": 0.9,
|
| 67 |
+
"top_k": 50,
|
| 68 |
+
"repetition_penalty": 1.1,
|
| 69 |
+
"description": "Balanced for general purpose conversations"
|
| 70 |
+
},
|
| 71 |
+
"deterministic": {
|
| 72 |
+
"temperature": 0.0,
|
| 73 |
+
"top_p": 1.0,
|
| 74 |
+
"top_k": 1,
|
| 75 |
+
"repetition_penalty": 1.0,
|
| 76 |
+
"do_sample": false,
|
| 77 |
+
"description": "Fully deterministic output, same input = same output"
|
| 78 |
+
}
|
| 79 |
+
},
|
| 80 |
+
"stop_sequences": [
|
| 81 |
+
"<|im_end|>",
|
| 82 |
+
"<|endoftext|>",
|
| 83 |
+
"</s>"
|
| 84 |
+
],
|
| 85 |
+
"chat_format": {
|
| 86 |
+
"system": "<|im_start|>system\n{content}<|im_end|>\n",
|
| 87 |
+
"user": "<|im_start|>user\n{content}<|im_end|>\n",
|
| 88 |
+
"assistant": "<|im_start|>assistant\n{content}<|im_end|>\n",
|
| 89 |
+
"prompt_suffix": "<|im_start|>assistant\n"
|
| 90 |
+
}
|
| 91 |
+
}
|