Upload folder using huggingface_hub (#2)

- 8755a4e02ada2f95305fa372799b595824ebf5318f61b9c24518a31e8180f1f5 (14ec4e498fb93f0bae275745291963138b9df727)
- 5c7e172668ac9267aa5bc62be75e72f9313e4c099794408ec549f8bf2ea11fb6 (50c7c8cad3c106c927b907d66f846cf68846d4c2)

README.md

@@ -34,7 +34,7 @@ tags:
 
 ## Results
 
-Detailed efficiency metrics coming soon!
+![image info](./plots.png)
 
 **Frequently Asked Questions**
 - ***How does the compression work?*** The model is compressed with llm-int8.
@@ -61,7 +61,7 @@ You can run the smashed model with these steps:
    from transformers import AutoModelForCausalLM, AutoTokenizer
 
     model = AutoModelForCausalLM.from_pretrained("PrunaAI/migtissera-SynthIA-7B-v1.3-bnb-4bit-smashed",
-                                                 trust_remote_code=True)
+                                                 trust_remote_code=True, device_map='auto')
     tokenizer = AutoTokenizer.from_pretrained("migtissera/SynthIA-7B-v1.3")
     
     input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"]

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "/tmp/tmpc133hvuk",
+  "_name_or_path": "/tmp/tmpupwcj58d",
   "architectures": [
     "MistralForCausalLM"
   ],
@@ -18,7 +18,7 @@
   "quantization_config": {
     "bnb_4bit_compute_dtype": "bfloat16",
     "bnb_4bit_quant_type": "fp4",
-    "bnb_4bit_use_double_quant": true,
+    "bnb_4bit_use_double_quant": false,
     "llm_int8_enable_fp32_cpu_offload": false,
     "llm_int8_has_fp16_weight": false,
     "llm_int8_skip_modules": [

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4f1e05aa03a23a51acf984567b746795caa16f373152aa7ec1d7c1e4b82adcab
-size 4125687606
+oid sha256:453ee3fc4e40f9f6da25bdec9d265914889aba012b677a06872e90e759e9e21f
+size 4450837496

smash_config.json

@@ -8,7 +8,7 @@
         "compilers": "None",
         "task": "text_text_generation",
         "device": "cuda",
-        "cache_dir": "/ceph/hdd/staff/charpent/.cache/models91pj28po",
+        "cache_dir": "/ceph/hdd/staff/charpent/.cache/modelschciutys",
         "batch_size": 1,
         "model_name": "migtissera/SynthIA-7B-v1.3",
         "pruning_ratio": 0.0,