sharpenb commited on
Commit
e962927
1 Parent(s): b47b65e

15ed9831af07af80d21903317007e38235feaeb74d5f3d8aa845141cadfcde1b

Browse files
Files changed (5) hide show
  1. README.md +3 -3
  2. config.json +2 -2
  3. model.safetensors +2 -2
  4. plots.png +0 -0
  5. smash_config.json +1 -1
README.md CHANGED
@@ -34,12 +34,12 @@ tags:
34
 
35
  ## Results
36
 
37
- Detailed efficiency metrics coming soon!
38
 
39
  **Frequently Asked Questions**
40
  - ***How does the compression work?*** The model is compressed with llm-int8.
41
  - ***How does the model quality change?*** The quality of the model output might vary compared to the base model.
42
- - ***How is the model efficiency evaluated?*** These results were obtained on Tesla V100-PCIE-32GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
43
  - ***What is the model format?*** We use safetensors.
44
  - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data.
45
  - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
@@ -61,7 +61,7 @@ You can run the smashed model with these steps:
61
  from transformers import AutoModelForCausalLM, AutoTokenizer
62
 
63
  model = AutoModelForCausalLM.from_pretrained("PrunaAI/smallcloudai-Refact-1_6B-fim-bnb-4bit-smashed",
64
- trust_remote_code=True)
65
  tokenizer = AutoTokenizer.from_pretrained("smallcloudai/Refact-1_6B-fim")
66
 
67
  input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"]
 
34
 
35
  ## Results
36
 
37
+ ![image info](./plots.png)
38
 
39
  **Frequently Asked Questions**
40
  - ***How does the compression work?*** The model is compressed with llm-int8.
41
  - ***How does the model quality change?*** The quality of the model output might vary compared to the base model.
42
+ - ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
43
  - ***What is the model format?*** We use safetensors.
44
  - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data.
45
  - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
 
61
  from transformers import AutoModelForCausalLM, AutoTokenizer
62
 
63
  model = AutoModelForCausalLM.from_pretrained("PrunaAI/smallcloudai-Refact-1_6B-fim-bnb-4bit-smashed",
64
+ trust_remote_code=True, device_map='auto')
65
  tokenizer = AutoTokenizer.from_pretrained("smallcloudai/Refact-1_6B-fim")
66
 
67
  input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"]
config.json CHANGED
@@ -1,5 +1,5 @@
1
  {
2
- "_name_or_path": "/tmp/tmp6wl2l8fw",
3
  "architectures": [
4
  "GPTRefactForCausalLM"
5
  ],
@@ -23,7 +23,7 @@
23
  "quantization_config": {
24
  "bnb_4bit_compute_dtype": "bfloat16",
25
  "bnb_4bit_quant_type": "fp4",
26
- "bnb_4bit_use_double_quant": true,
27
  "llm_int8_enable_fp32_cpu_offload": false,
28
  "llm_int8_has_fp16_weight": false,
29
  "llm_int8_skip_modules": [
 
1
  {
2
+ "_name_or_path": "/tmp/tmpnw_ig9vu",
3
  "architectures": [
4
  "GPTRefactForCausalLM"
5
  ],
 
23
  "quantization_config": {
24
  "bnb_4bit_compute_dtype": "bfloat16",
25
  "bnb_4bit_quant_type": "fp4",
26
+ "bnb_4bit_use_double_quant": false,
27
  "llm_int8_enable_fp32_cpu_offload": false,
28
  "llm_int8_has_fp16_weight": false,
29
  "llm_int8_skip_modules": [
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:8cbc76ff71e3f106f49d62a5eb522ada007fb1fdd7ecf0e1e0544d3afd643b3d
3
- size 1117786467
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ccd8b6118d37fca764aab8bc69502f3a6dfaf26293aacd1200132b8af5f1e619
3
+ size 1182115200
plots.png ADDED
smash_config.json CHANGED
@@ -8,7 +8,7 @@
8
  "compilers": "None",
9
  "task": "text_text_generation",
10
  "device": "cuda",
11
- "cache_dir": "/ceph/hdd/staff/charpent/.cache/models3kkp_gcv",
12
  "batch_size": 1,
13
  "model_name": "smallcloudai/Refact-1_6B-fim",
14
  "pruning_ratio": 0.0,
 
8
  "compilers": "None",
9
  "task": "text_text_generation",
10
  "device": "cuda",
11
+ "cache_dir": "/ceph/hdd/staff/charpent/.cache/models3r_l7out",
12
  "batch_size": 1,
13
  "model_name": "smallcloudai/Refact-1_6B-fim",
14
  "pruning_ratio": 0.0,