Merge pull request oobabooga#295 from Zerogoki00/opt4-bit

Add support for quantized OPT models

Author: oobabooga

README.md

@@ -140,8 +140,9 @@ Optionally, you can use the following command-line flags:
 | `--cai-chat`  | Launch the web UI in chat mode with a style similar to Character.AI's. If the file `img_bot.png` or `img_bot.jpg` exists in the same folder as server.py, this image will be used as the bot's profile picture. Similarly, `img_me.png` or `img_me.jpg` will be used as your profile picture. |
 | `--cpu`       | Use the CPU to generate text.|
 | `--load-in-8bit`  | Load the model with 8-bit precision.|
-| `--load-in-4bit`  | Load the model with 4-bit precision. Currently only works with LLaMA.|
-| `--gptq-bits GPTQ_BITS`  |  Load a pre-quantized model with specified precision. 2, 3, 4 and 8 (bit) are supported. Currently only works with LLaMA. |
+| `--load-in-4bit`  | DEPRECATED: use `--gptq-bits 4` instead. |
+| `--gptq-bits GPTQ_BITS`  |  Load a pre-quantized model with specified precision. 2, 3, 4 and 8 (bit) are supported. Currently only works with LLaMA and OPT. |
+| `--gptq-model-type MODEL_TYPE`  |  Model type of pre-quantized model. Currently only LLaMa and OPT are supported. |
 | `--bf16`  | Load the model with bfloat16 precision. Requires NVIDIA Ampere GPU. |
 | `--auto-devices` | Automatically split the model across the available GPU(s) and CPU.|
 | `--disk` | If the model is too large for your GPU(s) and CPU combined, send the remaining layers to the disk. |

modules/quantized_LLaMA.py renamed to modules/GPTQ_loader.py

@@ -7,28 +7,40 @@
 import modules.shared as shared
 
 sys.path.insert(0, str(Path("repositories/GPTQ-for-LLaMa")))
-from llama import load_quant
+import llama
+import opt
 
 
-# 4-bit LLaMA
-def load_quantized_LLaMA(model_name):
-    if shared.args.load_in_4bit:
-        bits = 4
+def load_quantized(model_name):
+    if not shared.args.gptq_model_type:
+        # Try to determine model type from model name
+        model_type = model_name.split('-')[0].lower()
+        if model_type not in ('llama', 'opt'):
+            print("Can't determine model type from model name. Please specify it manually using --gptq-model-type "
+                  "argument")
+            exit()
     else:
-        bits = shared.args.gptq_bits
+        model_type = shared.args.gptq_model_type.lower()
+
+    if model_type == 'llama':
+        load_quant = llama.load_quant
+    elif model_type == 'opt':
+        load_quant = opt.load_quant
+    else:
+        print("Unknown pre-quantized model type specified. Only 'llama' and 'opt' are supported")
+        exit()
 
     path_to_model = Path(f'models/{model_name}')
-    pt_model = ''
     if path_to_model.name.lower().startswith('llama-7b'):
-        pt_model = f'llama-7b-{bits}bit.pt'
+        pt_model = f'llama-7b-{shared.args.gptq_bits}bit.pt'
     elif path_to_model.name.lower().startswith('llama-13b'):
-        pt_model = f'llama-13b-{bits}bit.pt'
+        pt_model = f'llama-13b-{shared.args.gptq_bits}bit.pt'
     elif path_to_model.name.lower().startswith('llama-30b'):
-        pt_model = f'llama-30b-{bits}bit.pt'
+        pt_model = f'llama-30b-{shared.args.gptq_bits}bit.pt'
     elif path_to_model.name.lower().startswith('llama-65b'):
-        pt_model = f'llama-65b-{bits}bit.pt'
+        pt_model = f'llama-65b-{shared.args.gptq_bits}bit.pt'
     else:
-        pt_model = f'{model_name}-{bits}bit.pt'
+        pt_model = f'{model_name}-{shared.args.gptq_bits}bit.pt'
 
     # Try to find the .pt both in models/ and in the subfolder
     pt_path = None
@@ -40,7 +52,7 @@ def load_quantized_LLaMA(model_name):
         print(f"Could not find {pt_model}, exiting...")
         exit()
 
-    model = load_quant(str(path_to_model), str(pt_path), bits)
+    model = load_quant(str(path_to_model), str(pt_path), shared.args.gptq_bits)
 
     # Multiple GPUs or GPU+CPU
     if shared.args.gpu_memory:

modules/models.py

@@ -1,6 +1,5 @@
 import json
 import os
-import sys
 import time
 import zipfile
 from pathlib import Path
@@ -35,14 +34,15 @@
     ds_config = generate_ds_config(shared.args.bf16, 1 * world_size, shared.args.nvme_offload_dir)
     dschf = HfDeepSpeedConfig(ds_config) # Keep this object alive for the Transformers integration
 
+
 def load_model(model_name):
     print(f"Loading {model_name}...")
     t0 = time.time()
 
     shared.is_RWKV = model_name.lower().startswith('rwkv-')
 
     # Default settings
-    if not any([shared.args.cpu, shared.args.load_in_8bit, shared.args.load_in_4bit, shared.args.gptq_bits > 0, shared.args.auto_devices, shared.args.disk, shared.args.gpu_memory is not None, shared.args.cpu_memory is not None, shared.args.deepspeed, shared.args.flexgen, shared.is_RWKV]):
+    if not any([shared.args.cpu, shared.args.load_in_8bit, shared.args.gptq_bits, shared.args.auto_devices, shared.args.disk, shared.args.gpu_memory is not None, shared.args.cpu_memory is not None, shared.args.deepspeed, shared.args.flexgen, shared.is_RWKV]):
         if any(size in shared.model_name.lower() for size in ('13b', '20b', '30b')):
             model = AutoModelForCausalLM.from_pretrained(Path(f"models/{shared.model_name}"), device_map='auto', load_in_8bit=True)
         else:
@@ -87,11 +87,11 @@ def load_model(model_name):
 
         return model, tokenizer
 
-    # 4-bit LLaMA
-    elif shared.args.gptq_bits > 0 or shared.args.load_in_4bit:
-        from modules.quantized_LLaMA import load_quantized_LLaMA
+    # Quantized model
+    elif shared.args.gptq_bits > 0:
+        from modules.GPTQ_loader import load_quantized
 
-        model = load_quantized_LLaMA(model_name)
+        model = load_quantized(model_name)
 
     # Custom
     else:

modules/shared.py

@@ -69,8 +69,9 @@ def str2bool(v):
 parser.add_argument('--cai-chat', action='store_true', help='Launch the web UI in chat mode with a style similar to Character.AI\'s. If the file img_bot.png or img_bot.jpg exists in the same folder as server.py, this image will be used as the bot\'s profile picture. Similarly, img_me.png or img_me.jpg will be used as your profile picture.')
 parser.add_argument('--cpu', action='store_true', help='Use the CPU to generate text.')
 parser.add_argument('--load-in-8bit', action='store_true', help='Load the model with 8-bit precision.')
-parser.add_argument('--load-in-4bit', action='store_true', help='Load the model with 4-bit precision. Currently only works with LLaMA.')
-parser.add_argument('--gptq-bits', type=int, default=0, help='Load a pre-quantized model with specified precision. 2, 3, 4 and 8bit are supported. Currently only works with LLaMA.')
+parser.add_argument('--load-in-4bit', action='store_true', help='DEPRECATED: use --gptq-bits 4 instead.')
+parser.add_argument('--gptq-bits', type=int, default=0, help='Load a pre-quantized model with specified precision. 2, 3, 4 and 8bit are supported. Currently only works with LLaMA and OPT.')
+parser.add_argument('--gptq-model-type', type=str, help='Model type of pre-quantized model. Currently only LLaMa and OPT are supported.')
 parser.add_argument('--bf16', action='store_true', help='Load the model with bfloat16 precision. Requires NVIDIA Ampere GPU.')
 parser.add_argument('--auto-devices', action='store_true', help='Automatically split the model across the available GPU(s) and CPU.')
 parser.add_argument('--disk', action='store_true', help='If the model is too large for your GPU(s) and CPU combined, send the remaining layers to the disk.')
@@ -95,3 +96,8 @@ def str2bool(v):
 parser.add_argument('--auto-launch', action='store_true', default=False, help='Open the web UI in the default browser upon launch.')
 parser.add_argument('--verbose', action='store_true', help='Print the prompts to the terminal.')
 args = parser.parse_args()
+
+# Provisional, this will be deleted later
+if args.load_in_4bit:
+    print("Warning: --load-in-4bit is deprecated and will be removed. Use --gptq-bits 4 instead.\n")
+    args.gptq_bits = 4