Add vicuna prune 0.84 model

vicuna_prune84/2024-04-25-05-46-16/description.txt

@@ -0,0 +1,28 @@
+- Training Parameters: 
+  - base_model: lmsys/vicuna-7b-v1.3
+  - save_ckpt_log_name: llama_prune
+  - pruning_ratio: 0.84
+  - pruner_type: taylor
+  - temperature: 1.0
+  - top_p: 0.95
+  - max_seq_len: 128
+  - channel_wise: False
+  - block_wise: True
+  - layer_wise: False
+  - layer: 12
+  - block_attention_layer_start: 4
+  - block_attention_layer_end: 30
+  - block_mlp_layer_start: 4
+  - block_mlp_layer_end: 30
+  - iterative_steps: 1
+  - grouping_strategy: sum
+  - global_pruning: False
+  - taylor: param_first
+  - num_examples: 10
+  - device: cpu
+  - test_before_train: False
+  - eval_device: cuda
+  - test_after_train: True
+  - seed: 42
+  - save_model: True
+  - torch_version: 2.2

vicuna_prune84/2024-04-25-05-46-16/train.sh

@@ -0,0 +1 @@
+python hf_prune.py --pruning_ratio 0.84 --block_wise --block_mlp_layer_start 4 --block_mlp_layer_end 30 --block_attention_layer_start 4 --block_attention_layer_end 30 --pruner_type taylor --test_after_train --device cpu --eval_device cuda --save_ckpt_log_name llama_prune --save_model --base_model lmsys/vicuna-7b-v1.3

vicuna_prune84/2024-04-25-05-46-16/training.log

@@ -0,0 +1,42 @@
+2024-04-25 05:46:19 - INFO :       Use taylor pruner...
+2024-04-25 05:46:19 - INFO :       Pruning Attention Layer = [4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29]
+2024-04-25 05:46:19 - INFO :       Pruning MLP Layer = [4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29]
+2024-04-25 05:46:23 - INFO :       Start Pruning
+2024-04-25 05:46:33 - INFO :       Start Backwarding in iterative steps = 0...
+2024-04-25 05:46:41 - INFO :       Loss = 3.7534372806549072
+2024-04-25 05:47:03 - INFO :       After Iter 1/1, #parameters: 2366435328
+2024-04-25 05:47:03 - INFO :       #Param before: 6738415616, #Param after: 2366435328, Ratio = 35.1186%
+2024-04-25 05:47:13 - INFO :       
+==================Generation Results After Pruning================
+
+2024-04-25 05:47:14 - INFO :       <s> I believe the meaning of life is deeper still well deep within within within the as of within an intellect, I have yet my soul to the voudle's rain, Of the rain, ominaed thr.</s>
+2024-04-25 05:47:15 - INFO :       <s> Simply put, the theory of relativity states that 7 times on a whaphead kneseheadroard the asetese and whet etebpred foll of her be trandied by the chob and amb</s>
+2024-04-25 05:47:16 - INFO :       <s> Building a website can be done in 10 simple steps:
+Kin Reavfortunately to take the Soora in the
+Nus, Un’t She Da</s>
+2024-04-25 05:47:16 - INFO :       <s> Tweet: "I hate it when my phone battery dies."
+Sentiment: Negative
+###
+Tweet: "My day has been 👍"
+Sentiment: Positive
+###
+Tweet: "This is the link to the article"
+Sentiment: Neutral
+###
+Tweet: "This new music video was incredibile"
+Sentiment: Initial Ch</s>
+2024-04-25 05:47:16 - INFO :       <s> Translate English to French:
+
+sea otter => loutre de mer
+
+peppermint => menthe poivrée
+
+plush girafe => girafe peluche
+
+cheese => S, ’O</s>
+2024-04-25 05:47:16 - INFO :       
+==================Finish================
+
+2024-04-25 05:48:38 - INFO :       PPL after pruning: {'wikitext2': 3075.5837511620985, 'ptb': 3485.092969906799}
+2024-04-25 05:48:38 - INFO :       Memory Requirement: 4572.6494140625 MiB
+

vicuna_prune84/config.json

@@ -0,0 +1,23 @@
+{
+  "_name_or_path": "/home/ubuntu/model_weights/vicuna-7b-v1.3",
+  "architectures": [
+    "LlamaForCausalLM"
+  ],
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "initializer_range": 0.02,
+  "intermediate_size": 11008,
+  "max_position_embeddings": 2048,
+  "model_type": "llama",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 32,
+  "pad_token_id": 0,
+  "rms_norm_eps": 1e-06,
+  "tie_word_embeddings": false,
+  "torch_dtype": "float16",
+  "transformers_version": "4.28.1",
+  "use_cache": true,
+  "vocab_size": 32000
+}

vicuna_prune84/description.txt

@@ -0,0 +1,28 @@
+- Training Parameters: 
+  - base_model: lmsys/vicuna-7b-v1.3
+  - save_ckpt_log_name: llama_prune
+  - pruning_ratio: 0.84
+  - pruner_type: taylor
+  - temperature: 1.0
+  - top_p: 0.95
+  - max_seq_len: 128
+  - channel_wise: False
+  - block_wise: True
+  - layer_wise: False
+  - layer: 12
+  - block_attention_layer_start: 4
+  - block_attention_layer_end: 30
+  - block_mlp_layer_start: 4
+  - block_mlp_layer_end: 30
+  - iterative_steps: 1
+  - grouping_strategy: sum
+  - global_pruning: False
+  - taylor: param_first
+  - num_examples: 10
+  - device: cpu
+  - test_before_train: False
+  - eval_device: cuda
+  - test_after_train: True
+  - seed: 42
+  - save_model: True
+  - torch_version: 2.2

vicuna_prune84/generation_config.json

@@ -0,0 +1,7 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "pad_token_id": 0,
+  "transformers_version": "4.28.1"
+}

vicuna_prune84/pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:113fa87622ba6b8cb631d9ea2f108aef83a3a938841d0dc20ed5ffb489d26077
+size 4767303626

vicuna_prune84/special_tokens_map.json

@@ -0,0 +1,24 @@
+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": "<unk>",
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  }
+}