Upload 13 files

README.md

@@ -1,3 +1,7 @@
 ---
-license: apache-2.0
+library_name: peft
 ---
+
+# LLaMa-Saiga-7b-Gofman
+
+[llama-7b-saiga-merged](https://huggingface.co/cwiz/llama-7b-saiga-merged) trained on [Igor Gofman](https://github.com/Shoe-Eye/gofman-digital-oracle) dataset.

adapter_config.json

@@ -0,0 +1,17 @@
+{
+  "base_model_name_or_path": "cwiz/llama-7b-saiga-merged",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "lora_alpha": 64,
+  "lora_dropout": 0.05,
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 32,
+  "revision": null,
+  "target_modules": ["q_proj", "v_proj"],
+  "task_type": "CAUSAL_LM"
+}

adapter_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ba35964fd471e7c77ada5454a43d5145fc28686948b1d32b09f3090b087f724
+size 67154893

generation_config.json

@@ -0,0 +1,13 @@
+{
+  "pad_token_id": 0,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "temperature": 0.8,
+  "top_p": 0.9,
+  "top_k": 40,
+  "do_sample": true,
+  "num_beams": 1,
+  "max_new_tokens": 1536,
+  "repetition_penalty": 1.1,
+  "no_repeat_ngram_size": 15
+}

huggingface-metadata.txt

@@ -0,0 +1,6 @@
+url: https://huggingface.co/IlyaGusev/saiga_7b_lora
+branch: main
+download date: 2023-07-09 17:43:52
+sha256sum:
+    eb512e1c4bbb3123c4b482e5f3a996a7b8fdb10a48a2799e15c01b4ced3a58cf adapter_model.bin
+    9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347 tokenizer.model

special_tokens_map.json

@@ -0,0 +1,7 @@
+{
+    "bos_token": "<s>",
+    "eos_token": "</s>",
+    "pad_token": "<unk>",
+    "sep_token": "<s>",
+    "unk_token": "<unk>"
+}

template.json

@@ -0,0 +1,7 @@
+{
+    "system_prompt": "Ты — Сайга, русскоязычный автоматический ассистент. Ты разговариваешь с людьми и помогаешь им.",
+    "message_template": "<s>{role}\n{content}</s>\n",
+    "start_token_id": 1,
+    "end_token_id": 2,
+    "bot_token_id": 9225
+}

tokenizer.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
+size 499723

tokenizer_config.json

@@ -0,0 +1,12 @@
+{
+    "tokenizer_class": "LlamaTokenizer",
+    "add_bos_token": false,
+    "add_eos_token": false,
+    "model_max_length": 2048,
+    "padding_side": "left",
+    "bos_token": "<s>",
+    "eos_token": "</s>",
+    "unk_token": "<unk>",
+    "clean_up_tokenization_spaces": false,
+    "special_tokens_map_file": "special_tokens_map.json"
+}

training_config.json

@@ -0,0 +1,34 @@
+{
+  "trainer": {
+    "evaluation_strategy": "steps",
+    "per_device_train_batch_size": 4,
+    "per_device_eval_batch_size": 4,
+    "gradient_accumulation_steps": 32,
+    "eval_steps": 50,
+    "save_steps": 50,
+    "logging_steps": 5,
+    "learning_rate": 0.0003,
+    "num_train_epochs": 5,
+    "lr_scheduler_type": "cosine",
+    "warmup_steps": 30,
+    "fp16": true,
+    "bf16": false,
+    "torch_compile": false,
+    "optim": "adamw_torch"
+  },
+  "lora": {
+    "r": 16,
+    "lora_alpha": 16,
+    "lora_dropout": 0.05,
+    "bias": "none",
+    "target_modules": ["q_proj", "v_proj", "k_proj", "o_proj"],
+    "task_type": "CAUSAL_LM"
+  },
+  "load_in_8bit": true,
+  "only_target_loss": true,
+  "mode": "chat",
+  "templates_path": "internal_prompts/saiga_v2.json",
+  "model_name": "models/llama-7b",
+  "model_type": "causal",
+  "max_tokens_count": 2000
+}

training_log.json

@@ -0,0 +1,15 @@
+{
+  "base_model_name": "llama-saiga",
+  "base_model_class": "LlamaForCausalLM",
+  "base_loaded_in_4bit": false,
+  "base_loaded_in_8bit": true,
+  "loss": 2.4306,
+  "learning_rate": 0.0,
+  "epoch": 3.0,
+  "current_steps": 4319,
+  "train_runtime": 3963.6329,
+  "train_samples_per_second": 4.36,
+  "train_steps_per_second": 0.034,
+  "total_flos": 1.7670854947110912e+17,
+  "train_loss": 2.5247063248245802
+}

training_parameters.json

@@ -0,0 +1,27 @@
+{
+  "lora_name": "llama-saiga-7b-gofman",
+  "always_override": true,
+  "save_steps": 0.0,
+  "micro_batch_size": 4,
+  "batch_size": 128,
+  "epochs": 3.0,
+  "learning_rate": "3e-4",
+  "lr_scheduler_type": "linear",
+  "lora_rank": 32,
+  "lora_alpha": 64,
+  "lora_dropout": 0.05,
+  "cutoff_len": 256,
+  "dataset": "None",
+  "eval_dataset": "None",
+  "format": "None",
+  "eval_steps": 100.0,
+  "raw_text_file": "gofman",
+  "overlap_len": 128,
+  "newline_favor_len": 128,
+  "higher_rank_limit": false,
+  "warmup_steps": 100.0,
+  "optimizer": "adamw_torch",
+  "hard_cut_string": "\\n\\n",
+  "train_only_after": "",
+  "stop_at_loss": 0
+}

training_prompt.json

@@ -0,0 +1,3 @@
+{
+  "template_type": "raw_text"
+}