* wow argmax is so good
* 1 less line
* clean up and better variable names
* is this torch thing right...?
* add more tests
* slap a TODO on it
* clean ups
* prettier looking code and fix ceil mode test
* add return types and some docs
* ok that was a bad example since indices == value, just no example
* extra/gemm/max_matmul: start of custom kernels for GEMM
* add an unoptimized FP16/FP16 MMA example
* add slow 3-stage fp16 acc example
* add correct 3-stage pipeline with unswizzled/flat smem input (slow)
* add acc fp16 example with 3 stages and swizzle (no bank conflicts)
* add max version of NV fp16_fp16_fp16
* fix up comments and removed unused code in max variations
* add start of no_xor example
* fix to account for UOps to Ops
* train_shakespeare_char.py works
* move aten.where.self_out to tiny_backend_out
* fix memory leak
* corealize in the backward_hook
* Update backend.py
---------
Co-authored-by: George Hotz <72895+geohot@users.noreply.github.com>
* poc
* repeated values fail, sigh
* is this being timed out?
* fix up down names
* bitonic v2, does this run?
* bitonic v3, faster
* bitonic v3.1, faster
* bitonic v3.1.1, same speed unlucky
* support dim and indices
* bitonic v3.2, simpler code, TODO repeated indices
* bruv gimme green for once cmon
* cat (stack) implementation, slow but maybe one day when cat is fast meow
* revert to v3.2
* bitonic v4, who let the cats out edition
* clean up variable names
* figured out repeated indices :D
* ruff check --fix
* use sort for topk
* add Tensor.sort everywhere
* fix docs and add some types
* slightly better variable names
* am I doing torch inplace correctly?
* delegate sort to values_stable
* add a contig, faster first sort
* maybe don't test_inplace
---------
Co-authored-by: chenyu <chenyu@fastmail.com>
* np generates randoms
* hotfix: use generator for int dtype
* float32 as default dtype for float generator
* use np.float32 instead of stirng
* add dtype= to integers generator
* change import _to_np_dtype source
* add ability to ORT=1
* test_vs_ort
* useless f
* actually have benchmark take in modelproto for more flexibility in huggingface stuff
* ok runs
* good
* oops fix benchmark_onnx __main__
* 224 as default
* add ORT=1 option to huggingface_onnx
* use Tensor to get_input
* add abilty to do single onnx model testing
* better names
* merge properly...
* copy in onnx_helpers
* better
* decent script
* need to add debug tool first
* new limit usage
* why did narrowing_error come back..
* pretty decent
* revert validate change
* more ops bug fixes
* revert unnecessary changes
* fix InstanceNorm too
* remove op from O4
* minimize diff
* address old feedback
* unsure of this, just revert
* remove that assert
* working attention
* to_python_const Attention
* cant init from np constant so just do this
* final
* fix bug in attention
* attention clean ups
* add hard TODOs and REPOPATH and TRUNCATE envvar
* fix input_ids default value
* final
* fix scatter
* cleaner _prepare_quantize
* use new attention and tempfile for huggingface script
* more stats
* update
* remove outdated code
* big refactor to something usable by CI
* booooooom
* clean up
* update to using yaml as env var input
* add dry run
* try
* valid pad
* use argparser and fix gather bug
* ignore all yaml
* tiny bit more polish
* woah ignoring all yaml was not right
* typo
* decouple huggingface_onnx_run debug run with huggingface_onnx_download
* bug fix for downloading single model
* WOOOO ok much better
* oops argparse 'required' is an invalid argument for positionals
* oops argparse 'required' is an invalid argument for positionals
* add assert
* fix types
---------
Co-authored-by: chenyu <chenyu@fastmail.com>
* sqtt
* docs
* multi-device
* ProfileSQTTEvent
* exec update
* 256mb default
* don't let people hang their gpus
* bitfields from autogen
* asic info from mesa
* more bitfields from autogen
* SQTT_ITRACE_SE_MASK
---------
Co-authored-by: George Hotz <72895+geohot@users.noreply.github.com>
* fix leak, realize everything on torch optim step
* only realize a subset
---------
Co-authored-by: George Hotz <72895+geohot@users.noreply.github.com>
* add torch inplace tests
* first set of tests passing
* wrap all inplace funcs, add more tests
* fixes and wrap more functions
* fix all uint8 tests to avoid slow tests
* fix the one test
* another test, another fix
* and one more, works for ddp now
* something on contiguous, cleanup
---------
Co-authored-by: qazal <77887910+Qazalin@users.noreply.github.com>
* terrible but somewhat working impl
* linux behaves differently than macos?
* slightly better impl
* small clean up; haven't figured this out yet
* better
* torch has different behavior on linux and macos for duplicated values
* add sum docs
* fix test
* add torch return_type test
* add an exception test
* wrap_fxn instead, and move op lower in order
* better repeated values test
* rerun ci
* load llama3-1B to WEBGPU device
* include compile script for loading llama3 to WEBGPU
* parametrize max_context in build_transformer fxn
* jit_model with two different args sets
* compile for webgpu, split weights
* load model weight parts in browser
* export all tensors from initialized transformer
* run transformer inference in browser
* enable tiktoken with llama bpe in browser
* count total tokens on client with tiktoken.js
* full client-side chat streaming, eliminate server
* revert change that enabled jitting with 2 argsets
* llama without Variable or cache_kv, for webgpu
* have client use mask tokens / whole context
* cleanup staged weights
* add tiktoken.js build script, README
* export CLANG for Q6_k to float32 decompression
* fix and test exported CLANG code for Q6_k to fp32
* revert changes to jit and export_model
* isolate clang export
* test Q6_K to float32 decompression in browser
* gguf_load now also returns t_infos and data_start
* prepare llama-1B Q6_K gguf chunks for browser
* cache and decompress quantized llama in browser
* enable separate deployment of large files
* fix kv cache and symbolic with llama wgpu
* eliminate browser lag during decompression
* hash metadata and weight chunks
* delete obsolete indexeddb cache to free disk
* add progress bar, track model download/decompress
* refactor progress callback
* skip buffer hash verification for speed
* Display progress for entire loading scope
* Report page load errors to user
* actually display errors
* skip prompt tokens already seen by model
* skip prefilling with last assistant message tokens
* on page load tell user if webgpu not enabled
* push deployed URL root to window.history
* make note of bug sources with TODO items
* isolate bug in CLANG with BEAM=2
* remove clang_bug.py from diff
* decompress q6k to f32 on webgpu instead of clang
* remove unused code
* inter-weight decomp with larger wgpu kernels
* parallelize decompression submissions
* refactor dequantize scheduling
* add progress bar back
* fix bug
* temp fix for loading GGUF Q6_K to fp16 not fp32
* fix rendering of exported CLANG
* remove weight casts, sketch js functions for clang
* get symbolic vars from jit_cache for model export
* include symbolic vars in exported CLANG
* render js for clang transformer
* toggle clang/webgpu deployment; refactor decomp
* compile and render clang Q6_K->fp16 and int8 quant
* fix rendered clang for abs(fp16), to work in wasm
* simplify clang js wrapping
* run compiled clang in worker
* prepare llama weights in workers, q6k to int8/fp16
* tinychat on clang in browser, f32/int8 weights
* move wasm inference to (now flexible) worker
* don't load redundant embeddings
* modest wasm perf gain with compile flags
* set default backend, enable backend choice/backup
* render symbolic vars in exported WEBGPU
* quantize webgpu llama to int8/f32
* improve UX arising from rendered WEBGPU
* clean up webgpu launch
* new weights split: smaller chunks, tinygrad quant.
* switch webgpu inference to int8 quant
* remove unneeded clang decompression
* eliminate unneeded kv cache transfer to wasm
* use 1 worker for simplified clang decompression
* display launch errors
* refactor: stream load weight chunks to WebGPU
* show loading chunk completion
* quantize embeddings to int8
* test float16 as input for quantization
* webgpu: use f16 source, int8 embed, eliminate q6k
* simplify split weights prep: all from state_dict
* revert change to nn.state.gguf_load
* remove unneeded decompression from webgpu client
* remove unneeded code
* decrease dl chunks from 47 to 16 MiB
* improve stability of webgpu loading on mobile
* autodetect mobile, improve load stability
* refactor: progress closure
* refactor: one unified progress bar
* remove unneeded code
* revert changes to tinygrad core library
* enforce ios18.3 nerfed max buf size
* BEAM=3 webgpu
* cache integrity, mobile save throttling
* improve mobile UX - no autozoom on prompt box
* clang: int8 from f16, remove q6k
* reduce concurrent dls on mobile to 2 for stability
* refactor: wasm backend with stream loading
* prevent race between wasm load and indexedb save
* split wasm kernels into separate modules
* js wrapper for multiple wasm module inference
* revert multi-module wasm to single module
* make mobile wasm load more stable/fast
* refactor: copy weights into wasm without crashes
* fix bug in download queue; increase mobile dls
* refactor exported clang wrapper, split weights
* remove unnecessary code
* greatly improve int8 quant quality with rounding
* eliminate mobile throttling
* increase webgpu context to 4096 tokens
* export webgpu js functions
* enable separate hosted weights for mobile/pc
* enable prompt-thread switching during generation
* stop generation when max_context is reached
* show progress bar for prefill
* tell user if webgpu fails, while wasm loads
* make loading messages more concise
* update font
* revert changes to tinychat python app launch
* cleanup quantization, add scale_dtype param
* cleanup kv cache code
* cleanup compile code
* link tok_embeddings with output in webgpu export
* refactor: export_model webgpu: symbolic vars
* refactor: export_model weight loading
* forgot to commit export_model.py
* change CLANG to CPU
* deal with pylint incorrectly failing tests
* simplify f-strings for older CI python version
* fix pre-python3.12 parser errors
* [Int32Array] not Int32Array
* cleanup webgpu compile after refactor export_model
* refactor WASM export into export_model
* merge WebGPU/WASM compile scripts
* simplify max_contexts for local deployment
* fix parser issues and whitespace
* deduplicate variable defs for non-wasm clang export
* cleanup code
* cleanup compile scripts
* simplify wasm inference wrapping
* simplify webgpu symbolic vars export
* refactor: unify export of symbolic variables
* simplify WASM export
* simplify clang/wasm export
* update README and build scripts
* separate files for browser/python apps
* restore original python tinychat app files
* browser and python tinychats share assets
* minor cleanup
* isolate compile/export model
---------
Co-authored-by: George Hotz <72895+geohot@users.noreply.github.com>
* fix some torch tests
* fixup
* small change
* fixup
* fix test
* use default function
* add todo
* bunch of small changes
* fix tests
* more tests
* fix
* fix
* test fix
* simplify
* add DynamicDequantizeLinear and corresponding tests
* wow qlinearops are round away from zero
* this passes locally...
* again
* try
* try separate test
* round to even again
* also add QLinearMul
---------
Co-authored-by: chenyu <chenyu@fastmail.com>
* least_upper_float is at least default_float
en route for div rounding mode. dtype of true int division would change from int32 to default_float, which matches torch too.
* fix bert acc
* load llama3-1B to WEBGPU device
* include compile script for loading llama3 to WEBGPU
* parametrize max_context in build_transformer fxn
* jit_model with two different args sets
* compile for webgpu, split weights
* load model weight parts in browser
* export all tensors from initialized transformer
* run transformer inference in browser
* enable tiktoken with llama bpe in browser
* count total tokens on client with tiktoken.js
* full client-side chat streaming, eliminate server
* revert change that enabled jitting with 2 argsets
* llama without Variable or cache_kv, for webgpu
* have client use mask tokens / whole context
* cleanup staged weights
* add tiktoken.js build script, README
* export CLANG for Q6_k to float32 decompression
* fix and test exported CLANG code for Q6_k to fp32
* revert changes to jit and export_model
* isolate clang export
* test Q6_K to float32 decompression in browser
* gguf_load now also returns t_infos and data_start
* prepare llama-1B Q6_K gguf chunks for browser
* cache and decompress quantized llama in browser
* enable separate deployment of large files
* fix kv cache and symbolic with llama wgpu
* eliminate browser lag during decompression
* hash metadata and weight chunks
* delete obsolete indexeddb cache to free disk
* add progress bar, track model download/decompress
* refactor progress callback
* skip buffer hash verification for speed
* Display progress for entire loading scope
* Report page load errors to user
* actually display errors
* skip prompt tokens already seen by model
* skip prefilling with last assistant message tokens
* on page load tell user if webgpu not enabled
* push deployed URL root to window.history
* make note of bug sources with TODO items
* isolate bug in CLANG with BEAM=2
* remove clang_bug.py from diff
* decompress q6k to f32 on webgpu instead of clang
* remove unused code
* inter-weight decomp with larger wgpu kernels
* parallelize decompression submissions
* refactor dequantize scheduling
* add progress bar back
* fix bug
* temp fix for loading GGUF Q6_K to fp16 not fp32
* fix rendering of exported CLANG
* remove weight casts, sketch js functions for clang
* get symbolic vars from jit_cache for model export
* include symbolic vars in exported CLANG
* render js for clang transformer
* toggle clang/webgpu deployment; refactor decomp
* compile and render clang Q6_K->fp16 and int8 quant
* fix rendered clang for abs(fp16), to work in wasm
* simplify clang js wrapping
* run compiled clang in worker
* prepare llama weights in workers, q6k to int8/fp16
* tinychat on clang in browser, f32/int8 weights
* move wasm inference to (now flexible) worker
* don't load redundant embeddings
* modest wasm perf gain with compile flags
* set default backend, enable backend choice/backup
* render symbolic vars in exported WEBGPU
* quantize webgpu llama to int8/f32
* improve UX arising from rendered WEBGPU
* clean up webgpu launch
* new weights split: smaller chunks, tinygrad quant.
* switch webgpu inference to int8 quant
* remove unneeded clang decompression
* eliminate unneeded kv cache transfer to wasm
* use 1 worker for simplified clang decompression
* display launch errors
* refactor: stream load weight chunks to WebGPU
* show loading chunk completion
* quantize embeddings to int8
* test float16 as input for quantization
* webgpu: use f16 source, int8 embed, eliminate q6k
* simplify split weights prep: all from state_dict
* revert change to nn.state.gguf_load
* remove unneeded decompression from webgpu client
* remove unneeded code
* decrease dl chunks from 47 to 16 MiB
* improve stability of webgpu loading on mobile
* autodetect mobile, improve load stability
* refactor: progress closure
* refactor: one unified progress bar
* remove unneeded code
* revert changes to tinygrad core library
* enforce ios18.3 nerfed max buf size
* BEAM=3 webgpu
* cache integrity, mobile save throttling
* improve mobile UX - no autozoom on prompt box
* clang: int8 from f16, remove q6k
* reduce concurrent dls on mobile to 2 for stability
* refactor: wasm backend with stream loading
* prevent race between wasm load and indexedb save
* split wasm kernels into separate modules
* js wrapper for multiple wasm module inference
* revert multi-module wasm to single module
* make mobile wasm load more stable/fast
* refactor: copy weights into wasm without crashes
* fix bug in download queue; increase mobile dls
* refactor exported clang wrapper, split weights
* remove unnecessary code
* greatly improve int8 quant quality with rounding
* eliminate mobile throttling
* increase webgpu context to 4096 tokens
* export webgpu js functions
* enable separate hosted weights for mobile/pc
* enable prompt-thread switching during generation
* stop generation when max_context is reached
* show progress bar for prefill
* tell user if webgpu fails, while wasm loads
* make loading messages more concise
* update font
* revert changes to tinychat python app launch
* cleanup quantization, add scale_dtype param
* cleanup kv cache code
* cleanup compile code
* link tok_embeddings with output in webgpu export
* refactor: export_model webgpu: symbolic vars
* refactor: export_model weight loading
* forgot to commit export_model.py
* change CLANG to CPU
* deal with pylint incorrectly failing tests
* simplify f-strings for older CI python version
* fix pre-python3.12 parser errors
* [Int32Array] not Int32Array
* cleanup webgpu compile after refactor export_model
* refactor WASM export into export_model
* merge WebGPU/WASM compile scripts
* simplify max_contexts for local deployment
* fix parser issues and whitespace
* deduplicate variable defs for non-wasm clang export
* cleanup code
* cleanup compile scripts
* simplify wasm inference wrapping
* simplify webgpu symbolic vars export
* refactor: unify export of symbolic variables
* simplify WASM export
* simplify clang/wasm export
* update README and build scripts
* separate files for browser/python apps
* restore original python tinychat app files
* browser and python tinychats share assets
* minor cleanup
* isolate diffs to llama files
* minor cleanup
* set default scale_dtype
* set default scale_dtype for NF4 quantization
* make quantization of tok_embeds optional
* match output with tok_embeds if not quantizing
* minor change
