import os, json, hashlib, math from extra.export_model import export_model from examples.llama3 import build_transformer, Tokenizer from tinygrad.nn.state import get_state_dict, load_state_dict from tinygrad import Device, Variable, Tensor, dtypes, TinyJit from tinygrad.helpers import fetch, Context from tiktoken.load import load_tiktoken_bpe, dump_tiktoken_bpe def prepare_browser_chunks(model): # split weights into browser-friendly chunks state_dict = get_state_dict(model) del state_dict['output.weight'], state_dict['output.scale'] # same as tok_embeddings; ensures consistency with model export chunk_size = 16 * 1024 * 1024 # small chunks based on iphone browser constraints metadata = {} # We won't export cache_kv bytes (because we start inference on client at start_pos=0), but we will tell the client how big cache_kv needs to be t_infos = [(v.lazydata.base.realized.nbytes, k, v.dtype) for k,v in state_dict.items() if "cache_kv" not in k] empty_t_infos = [(v.lazydata.base.realized.nbytes, k, v.dtype) for k,v in state_dict.items() if "cache_kv" in k] split_t_infos = [] for size, name, dtype in t_infos: if size <= chunk_size: split_t_infos.append((size, name, dtype, ())) else: # split large weights into multiple parts for i in range(0, size, chunk_size): split_t_infos.append((min(chunk_size, size-i), f"{name}_part{math.ceil(i/chunk_size)}", dtype, (i, min(i+chunk_size, size)))) files = [] # pack weights into files with FFD bin packing split_t_infos = sorted(split_t_infos, reverse=True) for info in split_t_infos: placed = False for file in files: if sum(i[0] for i in file) + info[0] <= chunk_size: if info[3] and any(i[3] for i in file): continue # no two split tensors can touch the same file, due to wasm loading constraints file.append(info) placed = True break if not placed: files.append([info]) tinygrad_dtypes = {dtypes.float32: "float32", dtypes.float16: "float16", dtypes.int8: "int8", dtypes.int32: "int32"} for i, file in enumerate(files): cursor = 0 with open(os.path.join(os.path.dirname(__file__), f'./net_part{i}.chunk'), "wb+") as writer: for size, name, dtype, offsets in file: name, part_num = (name, 0) if "_part" not in name else (name.split("_part")[0], int(name.split("_part")[1])) default = {"parts": {}, "dtype": tinygrad_dtypes[dtype]} weight_metadata = metadata.get(name, default) weight_metadata["parts"][part_num] = {"file": i, "file_start_pos": cursor, "size": size} metadata[name] = weight_metadata data = bytes(state_dict[name].lazydata.base.realized.as_buffer()) data = data if not offsets else data[offsets[0]:offsets[1]] writer.write(data) cursor += size metadata.update({name: {"parts": {0: {"empty": True, "size": size}}, "dtype": tinygrad_dtypes[dtype]} for size, name, dtype in empty_t_infos}) for k in metadata: metadata[k]["parts"] = [part for part_num, part in sorted(metadata[k]["parts"].items(), key = lambda x: x[0])] cursor = 0 for i, part in enumerate(metadata[k]["parts"]): metadata[k]["parts"][i]["target_start_pos"] = cursor cursor += part["size"] metadata[k]["size"] = cursor # compute hashes, which client app will check to determine whether to update with new weights and/or detect integrity issues state_dict_hash = hashlib.sha256(json.dumps(metadata, sort_keys=True).encode("utf-8")).hexdigest() metadata = {"state_dict": metadata, "state_dict_hash": state_dict_hash, "files": []} hashes = set() for i in range(len(files)): with open(os.path.join(os.path.dirname(__file__), f'./net_part{i}.chunk'), "rb") as reader: hash = hashlib.sha256(reader.read()).hexdigest() hashes.add(hash) metadata["files"].append({"name": f'net_part{i}.chunk', "hash": hash}) if len(hashes) != len(files): print(f"WARNING: {len(files)} files were exported, but only {len(hashes)} are unique: something may have gone wrong") metadata_hash = hashlib.sha256(json.dumps(metadata, sort_keys=True).encode("utf-8")).hexdigest() metadata = {"metadata": metadata, "metadata_hash": metadata_hash} with open(os.path.join(os.path.dirname(__file__), f'./net_metadata.json'), "w") as writer: json.dump(metadata, writer, indent=4) return metadata def validate_model(model, tokenizer): prompt = "yo" toks = [tokenizer.bos_id] toks += [tokenizer.special_tokens["<|start_header_id|>"]] + tokenizer.encode("user") + [tokenizer.special_tokens["<|end_header_id|>"]] + tokenizer.encode("\n\n") toks += tokenizer.encode(prompt) + [tokenizer.special_tokens["<|eot_id|>"]] toks += [tokenizer.special_tokens["<|start_header_id|>"]] + tokenizer.encode("assistant") + [tokenizer.special_tokens["<|end_header_id|>"]] + tokenizer.encode("\n\n") start_pos = 0 run = TinyJit(model.forward) for tok in toks[:-1]: run(Tensor([[tok]]), Variable("start_pos", 0, model.max_context).bind(start_pos), 0.0, 0, 0.0, 0.0, 0.0).realize() start_pos += 1 tok = toks[-1] result = "" expected = "How's it going?" while True: tok = run(Tensor([[tok]]), Variable("start_pos", 0, model.max_context).bind(start_pos), 0.0, 0, 0.0, 0.0, 0.0).item() start_pos += 1 if tok in tokenizer.stop_tokens or len(result) > len(expected): break result += tokenizer.decode([tok]) assert result == expected, f"Model validation failed, expected output: {expected}, actual output: {result}" if __name__=="__main__": # Export BPE data for use with tiktoken.js tokenizer_path = fetch("https://huggingface.co/bofenghuang/Meta-Llama-3-8B/resolve/main/original/tokenizer.model", "tokenizer.model", subdir="llama3-1b-instruct") mergeable_ranks = load_tiktoken_bpe(str(tokenizer_path)) bpe_path = os.path.join(os.path.dirname(__file__), "llama3-2.tiktoken") dump_tiktoken_bpe(mergeable_ranks, bpe_path) tokenizer = Tokenizer(str(tokenizer_path)) model_path = fetch("https://huggingface.co/bartowski/Llama-3.2-1B-Instruct-GGUF/resolve/main/Llama-3.2-1B-Instruct-f16.gguf", "Llama-3.2-1B-Instruct-f16.gguf", subdir="llama3-1b-instruct") Tensor.no_grad = True max_context=1024 tok = 128000 TEMPERATURE, TOP_K, TOP_P, ALPHA_F, ALPHA_P = 0.95, 0, 0.0, 0.0, 0.0 start_pos = Variable("start_pos", 0, max_context).bind(0) model_input = lambda: [Tensor([[tok]]), start_pos, TEMPERATURE, TOP_K, TOP_P, ALPHA_F, ALPHA_P] Device.DEFAULT="CPU" model = build_transformer(model_path, model_size="1B", quantize="int8", scale_dtype=dtypes.float32, device=Device.DEFAULT, max_context=max_context) state_dict = get_state_dict(model) validate_model(model, tokenizer) model_name = "transformer" with Context(BEAM=3): cprog, js_wrapper = export_model(model, "wasm", *model_input(), model_name=model_name) # ensure consistency with exported weights js_wrapper = js_wrapper.replace("output.weight", "tok_embeddings.weight").replace("output.scale", "tok_embeddings.scale") with open(os.path.join(os.path.dirname(__file__), f"{model_name}.c"), "w") as f: f.write(cprog) with open(os.path.join(os.path.dirname(__file__), "net_clang.js"), "w") as f: f.write(js_wrapper) Device.DEFAULT="WEBGPU" # float16 is not yet supported for dawn/Vulkan/NVIDIA stack, see: https://issues.chromium.org/issues/42251215 # therefore for now, we used CLANG to quantize the float16 llama to int8 with float32 scales, then load to WEBGPU model = build_transformer(model_path, model_size="1B", quantize="int8", max_context=max_context, load_weights=False) load_state_dict(model, state_dict) # these were the same before load_state_dict model.output.weight, model.output.scale = model.tok_embeddings.weight, model.tok_embeddings.scale validate_model(model, tokenizer) metadata = prepare_browser_chunks(model) # export weights to disk with Context(BEAM=3): prg, input_sizes, output_sizes, state = export_model(model, "webgpu", *model_input(), model_name=model_name, stream_weights=True) # ensure consistency with exported weights prg = prg.replace("output.weight", "tok_embeddings.weight").replace("output.scale", "tok_embeddings.scale") with open(os.path.join(os.path.dirname(__file__), "net.js"), "w") as f: f.write(prg)